Citation
Gender relationship power, and HIV testing in rural Malawi

Material Information

Title:
Gender relationship power, and HIV testing in rural Malawi
Creator:
Conroy, Amy Anne. ( author )
Place of Publication:
Denver, CO
Publisher:
University of Colorado Denver
Publication Date:
Language:
English
Physical Description:
1 electronic file (315 pages). : ;

Subjects

Subjects / Keywords:
HIV (Viruses) -- Malawi ( lcsh )
Relationship quality -- Malawi ( lcsh )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Review:
This dissertation considers how relationship power shapes men and women's decisions around HIV testing, which is the gateway to HIV/AIDS prevention, treatment, and care in sub-Saharan Africa. The study is situated in a high HIV prevalence community in southern Malawi where rapid expansions in HIV testing services provides new opportunities to learn one's HIV status. I focus on two public health concerns around HIV testing that have been inadequately studied at the relationship level: uptake of HIV testing services and disclosure to primary partners. To accomplish this, I use an iterative, mixed-methods design to: 1) develop a model of relationship power for Malawi, 2) test for associations between power and HIV testing behavior, and 3) explain, validate, or challenge these findings using qualitative data. Quantitative data on 466 couples come from Tsogolo La Thanzi, a study on reproduction and AIDS in southern Malawi. Qualitative data consist of 34 semi-structured interviews and 8 focus group discussions. Using the three social structures of the Theory of Gender and Power - labor, power, and cathexis - I develop and test a preliminary model of relationship power consisting of three domains: economic power, relationship violence, and relationship dominance. Factor analysis revealed two addition construct of power related to unity and discordance, which were validated as important using qualitative narratives. Next, I use this modified model of relationship power to test hypotheses around uptake and disclosure. Economic power, relationship dominance, unite and violence emerged as important predictors of testing uptake for men and women. In the disclosure models, economic power, violence, and unity were predictive. Contrary to my hypothesis, young people with higher levels of unity were less likely to test for HIV. This finding was explained by the widespread belief that HIV testing was reserved for times of trouble, such as infidelity, rather than for relationships filled with unity, love, and trust. Study conclusions highlight the need to facilitate theoretical and applied approaches to HIV testing that encompass the couple context - including aspects of unity. Universal HIV testing may provide a promising solution for couples to circumvent exceptionally difficult negotiations by deferring decisions to healthcare providers.
Thesis:
Thesis (Ph.D.)--University of Colorado Denver. Health and behavioral sciences
Bibliography:
Includes bibliographic references.
System Details:
System requirements: Adobe Reader.
General Note:
Department of Health and Behavioral Sciences
Statement of Responsibility:
by Amy Anne Conroy.

Record Information

Source Institution:
|University of Colorado Denver
Holding Location:
|Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
891212005 ( OCLC )
ocn891212005

Downloads

This item has the following downloads:


Full Text
GENDER, RELATIONSHIP POWER, AND HIV TESTING
IN RURAL MALAWI
by
AMY ANNE CONROY
B.S.E., University of Iowa, 2002
M.P.H., University of Colorado, 2008
A thesis submitted to the
Faculty of the Graduate School of the
University of Colorado in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
Health and Behavioral Sciences
2013


2013
AMY ANNE CONROY
ALL RIGHTS RESERVED


This thesis for the Doctor of Philosophy degree by
Amy Anne Conroy
has been approved for the
Health and Behavioral Sciences Program
by
Sheana S. Bull, Dissertation Chair
Sara E. Yeatman, Examination Chair
Jean N. Scandlyn
Jennifer J. Harman
June 7, 2013
11


Conroy, Amy Anne (Ph.D., Health and Behavioral Sciences)
Gender, Relationship Power, and HIV Testing in Rural Malawi
Thesis directed by Professor Sheana S. Bull
ABSTRACT
This dissertation considers how relationship power shapes men and womens decisions
around HIV testing, which is the gateway to HIV/AIDS prevention, treatment, and care in
sub-Saharan Africa. The study is situated in a high HIV prevalence community in southern
Malawi where rapid expansions in HIV testing services provide new opportunities to learn
ones HIV status. I focus on two public health concerns around HIV testing that have been
inadequately studied at the relationship level: uptake of HIV testing services and disclosure to
primary partners. To accomplish this, I use an iterative, mixed-methods design to: 1) develop
a model of relationship power for Malawi, 2) test for associations between power and HIV
testing behavior, and 3) explain, validate, or challenge these findings using qualitative data.
Quantitative data on 466 couples come from Tsogolo 1m Thansp, a study on reproduction and
AIDS in southern Malawi. Qualitative data consist of 34 semi-structured interviews and 8
focus group discussions. Using the three social structures of the Theory of Gender and
Powerlabor, power, and cathexisI develop and test a preliminary model of relationship
power consisting of three domains: economic power, relationship violence, and relationship
dominance. Factor analysis revealed two additional constructs of power related to unity and
discordance, which were validated as important using qualitative narratives. Next, I use this
modified model of relationship power to test hypotheses around uptake and disclosure.
Economic power, relationship dominance, unity, and violence emerged as important
predictors of testing uptake for men and women. In the disclosure models, economic power,
violence, and unity were predictive. Contrary to my hypothesis, young people with higher
levels of unity were less likely to test for HIV. This finding was explained by the widespread
belief that HIV testing was reserved for times of trouble, such as infidelity, rather than for
relationships filled with unity, love, and trust. Study conclusions highlight the need to
facilitate theoretical and applied approaches to HIV testing that encompass the couple
contextincluding aspects of unity. Universal HIV testing may provide a promising solution
for couples to circumvent exceptionally difficult negotiations by deferring decisions to
healthcare providers.
in


The form and content of this abstract are approved. I recommend its publication.
Approved: Sheana S. Bull
IV


DEDICATION
I dedicate this thesis to the Malawians who suffer unnecessarily from AIDS and to others
who live in a state of uncertainty about whether they will become infected with HIV. Your
warm hearts, incredible strength and resilience, and receptiveness to my research has given
me the fortitude to complete this dissertation and dedicate my professional life to AIDS
intervention research in Africa. I can only hope that one day my career will be forced to go
in a dramatically different direction after an effective cure or vaccine for AIDS is discovered
and made accessible to all.
v


ACKNOWLEGEMENT
First and foremost, I am deeply appreciative for a number of peopleboth locally and
abroad-who supported me during this endeavor. For guidance on this dissertation, I am
indebted to my committee, Sheana Bull, Sara Yeatman, Jean Scandlyn, and Jennifer Flarman
for their valuable feedback and support throughout the process. I would like to extend a
special thanks to Sara Yeatman for inviting me to Malawi back in 2009 and allowing me to
participate in the Tsogolo 1m Than^i (TLT) study. In addition, I would like to acknowledge my
fellow students in the Health and Behavioral Sciences program, Kate Dovel, Stacey
McKenna, Deanna McQuillan, Christine Tagliaferri Rael, and Jerry Wulff, for picking me up
during times of frustration and providing words of encouragement to keep me moving
towards my goals. In Malawi, I thank the dedicated research assistants who tirelessly
collected the semi-structured interview and focus group data for this study: Caroline
Augustine, Steven Kabvinga, Chisomo Kalogwile, Andrew Kanjirawaya, Zahra Khan, and
Thandizo Shaba. A special thanks goes out to three TLT staff members, in particular, who
supported me both logistically and through their friendships during two fieldwork trips to
Malawi: Hazel Namadingo, Abdallah Chilungo, and Sydney Lungu. Finally, this dissertation
would not have been possible without the love, support, and encouragement from my
parents, Thomas and Christine Conroy, my husband, TJ Moretto, and my daughter Chiara
who accompanied me to rural Malawi when she was only 6 months old. To all of you, I
express my deepest gratitude for your patience and for believing in me every step along the
way. Especially, thank you Chiara, for being my bright light, showing me what is most
vi


important in life, and challenging me everyday to be the best mother and scholar I can be. I
look forward to sharing with you my love of learning and committment to a better global
world.
Finally, I would like to acknowledge the generous funding sources that made this
dissertation possible. Data collected through the TLT study was funded by grants (R01-
HD058366 and R03-HD067099) from the National Institute of Child Health and Human
Development (NICHD). My time and research expenses were supported through a Ruth
Kirschstein pre-doctoral fellowship from the National Institute of Mental Health (NIMH)
(F31-MH093260), the Robinson Durst Scholarship through the University of Colorado
Denver Center for Global Health, and the Henry David Research Grant through the
American Psychological Foundation.
To all of you, I say Zikomo Kwambiri.
Vll


TABLE OF CONTENTS
CHAPTER
I. INTRODUCTION..................................................1
Research Aims.............................................2
The Importance of Couple Context..........................4
The Importance of Relationship Power......................6
Summary of Research Design................................8
Implications for Policy, Prevention, and Public Health....9
Chapter Outline..........................................10
II. BACKGROUND AND CONCEPTUAL FRAMEWORK..........................12
Gender, Relationship Power, and HIV/AIDS.................12
The Theory of Gender and Power...........................14
HIV Prevention and Testing in Sub-Saharan Africa.........18
The Social Ecological Model of Health....................21
Symbolic Interactionism and HIV Testing Behavior.........35
Study Hypotheses.........................................37
Study Innovation.........................................40
The Malawi Context.......................................42
III. METHODOLOGY.................................................53
Study Design Overview....................................53
The TUT Study............................................56
Vlll


Creation of the Analysis Datasets..............................58
Measure Development Phase......................................65
The Hypothesis-Testing Phase...................................77
The Interpretative Phase.......................................98
Ethical Considerations........................................114
IV. CONSTRUCTIONS OF GENDER AND POWER: TRADITION,
UNITY, AND RIGHTS.................................................117
Tradition, Unity, and Rights..................................118
Cathexis: Gender and Relationship Ideals......................121
The Sexual Division of Labor..................................129
The Sexual Division of Power..................................139
The Influence of Gender Policy on Relationship Power..........147
Discussion....................................................160
V. THE ASSOCIATION BETWEEN RELATIONSHIP POWER AND
HIV TESTING BEHAVIOR..............................................170
Conceptual Framework and Hypotheses...........................170
Summary of Analysis Approach..................................176
Characteristics of the Couple Sample..........................180
Gender Differences in Relationship Factors....................184
HIV Testing Histories.........................................187
Predictors of HIV Testing Uptake..............................190
Disclosure of HIV Test Results................................197
IX


Reliability of Disclosure Reports...............................200
Predictors of HIV Status Disclosure.............................203
Discussion......................................................208
Limitations.....................................................218
VI. WHEN THERE IS DOUBT IN THE HOUSE, THATS WHEN
YOU GO FOR TESTING: HIV TESTING PERCEPTIONS AND
EXPERIENCES.........................................................221
Summary of Research Methods.....................................221
The Symbolic Meaning of an HIV Test.............................222
The Relationship Ideal of Testing before Marriage.............224
The Incompatibility of Testing with Marriage....................227
Risk, Infidelity, and HIV Testing...............................230
Tradition, Rights, and Unity....................................233
Testing Together as a Workaround Strategy.......................238
Antenatal Testing as an Alternative to Couples VCT..............245
Relationship Dissolution as a Fallback Strategy.................248
Discussion......................................................250
Public Health Implications......................................253
VII. IMPLICATIONS AND FUTURE DIRECTIONS.................................259
Theoretical and Policy Implications.............................259
Final Conceptual Model for Relationship Power and HIV Testing..262
Future Directions...............................................268
x


Conclusions.............................................270
REFERENCES.........................................................272
APPENDIX
A. Relationship Power Scale Items..............................292
B. Semi-Structured Interview Guide.............................294
C. Focus Group Discussion Guide................................298
xi


LIST OF FIGURES
FIGURE
2.1 Example Lets Go Slogan Promoting HCT......................20
2.2 Conceptual Model for HIV Testing Behavior Hypotheses.........38
2.3 Balaka District of Southern Malawi...........................44
3.1 Timeline of Data Collection and Analysis, 2009-2013..........56
3.2 Overview of the VCT Infrastructure for TLT...................58
3.3 The Relationship Power Scale Development Stages..............71
3.4 Iterative Stages of Development for the Relationship Power
Model........................................................76
4.1 Re-Conceptualizing the Connells (1987) Theory of Gender and
Power.......................................................120
5.1 Conceptual Model of Relationship Power based on the Modified
TGP.........................................................172
5.2 Conceptual Framework for Main Predictor Variables and HIV
Testing Behavior............................................173
5.3 Percent of TLT Respondents Who Had Ever Tested for HIV at
Baseline....................................................187
5.4 TLT Respondents Who Had a New HIV Test Since the Previous
Wave........................................................189
5.5 Cumulative Number of New HIV Tests from Waves 4 to 7........190
7.1 Final Conceptual Model on Relationship Power and HIV Testing
Behavior....................................................263
xii


LIST OF TABLES
TABLE
3.1 Example data from the wave 3 individuals dataset..................59
3.2 Example data from the wave 3 couples dataset......................60
3.3 Example data from the final wave 3 couples dataset................61
3.4 Example data from the uptake dataset (long format)................63
3.5 Example data from the disclosure dataset (long format)............65
3.6 Final rotated factor pattern for preliminary relationship power
scale.............................................................68
3.7 Means, standard deviations, and reliability coefficients for the
preliminary relationship power measure............................69
3.8 Descriptive statistics for the relationship power subscales, TLT
wave 3 couples sample (N=932).....................................72
3.9 Standardized factor loadings and coefficient alphas for the power
scale items.......................................................73
3.10 Gender differences in relationship power factors..................74
3.11 Format and interpretation of the main predictor variables.........89
3.12 Independent and dependent variables for the two statistical models
of uptake and disclosure..........................................94
3.13 Mean value of unity at TLTs wave 3 and 5.........................97
3.14 Characteristics of the focus group respondents...................110
4.1 Focus group respondents perceptions of ideal relationships......128
4.2 What makes people feel powerful in their relationships?..........142
xiii


5.1 Independent and dependent variables for the two statistical models
of uptake and disclosure.........................................178
5.2 Selected characteristics of the baseline sample of couples, TLT
wave 3...........................................................182
5.3 Couple HIV status, TLT wave 4....................................186
5.4 HIV testing history by gender, TLT waves 3-7....................189
5.5 Odds ratios predicting a new HIV test among women and men,
TLT waves 4-7....................................................196
5.6 Disclosure of last HIV test to main partner, TLT wave 5..........198
5.7 Do respondents really disclose and do their partners believe their
reports?.........................................................202
5.8 Are respondents told the correct test result by their partners?.203
5.9 Odds ratios from logistic regression models predicting disclosure
among women and men, TLT wave 5..................................207
7.1 Theoretical assertions and supporting evidence found in
quantitative data................................................266
xiv


CHAPTER I
INTRODUCTION
In a dusty African village on the outskirts of the Balaka town in southern Malawi,
Caroline interviews a 23-year old married woman named Ruth about her relationship history.
As the interview unfolds, the two women begin to talk about HIV testing, a salient topic of
discussion in a setting where HIV prevalence ranks among the highest in the world. Caroline
asks Ruth, So when you went for HIV testing, did you tell husband or did you just go?
Ruth answers, I told him that there was no reason to be afraid of knowing how your blood
is but he refused to go, so I went by myself. He was saying that I was doubting myself.
When Caroline asks Ruth if she continues to plead with her husband to go for testing or if
she gave up trying, she responds:
No, I didnt give up, I always tell him. The first time I went I told him that you should also go so
that we can have proof that we are alright, but if you dont then I will be having doubts. Maybe my
husband is positive or negative. I will not be sure. The second time I went I told him the same, but to
no avail. He says that he cant go and that those who go for testing are doubting themselves. They
dont trust themselves. It pains me that he is not helping me and the children, (semi-structured
interview, female #9)
I present Ruths narrative not to bolster support for the powerless woman depiction that is
common in gender research and policy on AIDS in Africa (Esacove 2013; Higgins,
Hoffman, and Dworkin 2010), but instead to illustrate Ruths tremendous agency in her
persistent attempts to change her husbands position on HIV testing. Unlike Ruths
situation, not all Malawian women are as forthcoming and relentless with their partners
about testing. Nor do all husbands refuse to get tested with their wives. Her story does,
1


however, exemplify one of many different ways that wives and husbands negotiate HIV
testing with each other to leam their HIV status.
Research Aims
This dissertation seeks to better understand the relationship context around HIV testing
behavior, particularly decisions to get tested and disclose test results to spouses or primary
sexual partners. Such an examination of heterosexual partnerships requires attention to the
everyday relationship dynamics within young couples. In this dissertation, I examine the
pathways and mechanisms through which relationship factors, particularly power, influences
HIV testing behavior among young couples from rural Malawi. Three specific aims will be
addressed with this research:
(1) To explore quantitative measures of relationship power for the Malawi context (a)
and to evaluate their cultural relevance with qualitative data (b).
(2) To test whether relationship power influences two types of HIV testing behavior:
uptake of HIV testing services (a) and disclosure of test results to primary partners
(b).
(3) To understand the meaning of HIV testing within the context of a sexual
relationship.
To my surprise, I find that what unites the numbers and narrative in this dissertation is
the underlying construct of unity, that is, notions of love, trust, intimacy, communication, and
reciprocity. At first glance, the idea of unity appears antithetical to the term power, which has
been historically conceptualized as an imbalance between men and women. But to the
2


contrary, rural Malawians believe that these relationship qualities provide a source of power
for them in their relationships. This raises the question, if one has power or unity in their
relationship, is it still necessary to get tested for HIV? The qualitative data from this
dissertation show that the HIV test does not simply provide a medical diagnosis; instead,
people construct meaning around the act of testing and the test result itself within the
context of their everyday relationships. At times, this addition of symbolism around testing
conflicts with how HIV testing services were designed to function from a public health
perspective: to detect HIV infection as earliest as possible in the disease course.
This idea resonates in Ruths narrative above. She tries to convince her husband about
the diagnostic benefits of HIV testing, i.e., to learn how their bodies are, but her husband
claims that HIV testing is for those who suspect they are HIV positive or have engaged in
promiscuous sexual behavioras opposed to those who are faithful to their partner and
practice safe sex. According to her side of the story, both Ruths desire to test and her
husbands refusal make both members question the others HIV status and sexual
exclusivity, as expressed with the euphemism having doubt. Here, an HIV test becomes a
symbol of something larger and perhaps more important to some people; it becomes a
marker of their relationship status rather than just their HIV serostatus. Thus, an HIV test
signifies the absence of unity in the relationshipa characteristic that many couples idealize
and strive for in their partnerships.
3


The Importance of Couple Context
Sexual relationships comprise the social fabric of life in rural Malawi. They are the
intimate setting where people get married and children are bom, which in turn brings
tremendous social status and fulfillment to both men and women. Scholars have even stated
that in sub-Saharan Africa, marriage remains one of the most important individual
aspirations and social duties that an individual ever experiences (Smith and Mbakwem
2007). In a very poor setting complicated by seasonal famines and high unemployment rates,
relationships are not only critical for social identity, but for survival. The gendered social
norms dictating the division of labor prescribe how husbands and wives are supposed to
behave and contribute to the household well beingand it is the economic and social
support received through these arrangements that maintains peoples existence and
livelihoods. Yet, in sub-Saharan Africa, these ongoing heterosexual relationships are the
place where most new HIV infections occur (Dunkle et al. 2008; Maleta and Bowie 2010).
Therefore, young people face competing aspirations and needs related to marriage,
childbearing, death, and survival while navigating their risk for HIV infection at the same
time: a daunting task to say the least, which may force some people to prioritize what is most
important to them (Dionne, Gerland, and Watkins 2013).
Young couples are particularly susceptible to HIV infection as they undergo a series of
transitions from courtship to marriage and then later, to parenthood. Paradoxically, serious
relationships such as marriage in which high levels of intimacy and trust have been
established are thought to be more risky than causal, promiscuous encounters (Mkandawire-
4


Valhmu et al. 2013). This is for several reasons. As young couples begin to form their
families, condoms are no longer considered an acceptable preventive measure for
HIV/AIDS (Chimbiri 2007). In love marriages, where women rely on ideals of love and
intimacy to negotiate relationships with their husbands, condoms are seen as undermining
the very thing they wish to preserve (Smith 2006). Low condom use coupled with
extramarital partnershipsmore notable among men than women (Schatz 2005; Carpenter
et al. 1999; Lurie, Williams, Zuma, Mkaya-Mwaburi, Garnett et al. 2003)provide ample
opportunities for new HIV infections to develop during young adulthood.
Despite relatively high rates of HIV infection among young people who are just
beginning their sexual and reproductive lives, recent expansions in HIV testing and
counseling (referred to as HTC from a policy perspective, but also labeled as VCTor
voluntary HIV counseling and testingin the literature and among rural Malawians) services
throughout Malawi provide an important opportunity for both individuals and couples alike
to leam their HIV status and take action to protect their own and their partners health. HIV
testing plays a pivotal role in the public health response to the AIDS epidemic and is a vital
entry point for HIV prevention services, care, and treatment (VCT Efficacy Group 2000;
Painter 2001). Although access to HIV testing has generally improved throughout the
region, many people still do not know their HIV status and for those who do get tested,
levels of disclosure of HIV test results to sexual partners remain low (Obermeyer and
Osborn 2007). Young people confront difficult decisions of whether to be tested and if
tested, to disclose the results to their loved ones. They must therefore weigh the advantages
5


of learning ones HIV status with the potentially negative consequences of being diagnosed
with a life threateningand arguably stigmatizingdisease in a context where access to
antiretroviral therapy (ART) is improving, but still not guaranteed.
In their version of the Social Ecological Model, McLeroy and colleagues (1988)
contend that health behaviors are influenced by factors at five different levels: interpersonal
(individual level), intrapersonal (includes family and relationship levels), institutional,
community, and public policy (the last three categories comprise structural levels). With
regard to HIV/AIDS in sub-Saharan Africa, Catherine Campbell (2003) argues that the
overwhelming focus has been on factors that occur at the individual level and are related to
psychological and social considerations. Factors at the interpersonal level are frequently
overlooked despite the fact that HIV testing decisions are often made within the context of
the dyad, rather than in isolation. Understanding the couple context is important not only
for decision-making around HIV testing, but after testing when individuals leam their HIV
statusperhaps for the first timeand decide whether to disclose their results to partners.
Without more attention paid to the interpersonal level, the current state of research on HIV
testing remains inadequate. More research using both partners perspectives is needed to
more fully understand how the relationship context shapes HIV testing behavior.
The Importance of Relationship Power
Power as expressed through social interactions is not a new concept and has received
tremendous scholarly attention by academics from a variety of different disciplines including
anthropology, political science, social psychology, public health, and sociology (Connell
6


1987; Cromwell and Olson 1975; Blanc 2001). Riley (1997) observed that gender affects both
power to and power over. Power to refers to the ability to act, whereas power over
refers to the ability to assert wishes and goals in the face of opposition from another. Some
scholars have also stressed that it is not the absolute power of either couple member that
matters, but rather their power relative to each other (Blanc 2001). In relation to HIV
infection, Wingood and DiClemente (2000) describe power as having the ability to act or
change or having power over others. In this study, I consider a broad, sociological
perspective of gender-based relationship power (shortened to relationship power) to refer
to the socially constructed gender differences between men and women where gender
refers to the expectations and norms shared within a society about appropriate male and
female behavior, characteristics, and roles (Gupta 2000). This definition allows for more
flexibility to study the culturally rooted and more nuanced forms of power that go beyond
simplistic notions of male dominance and female submissiveness.
Of the studies that have directly measured relationship power, researchers have
conceptualized powerand its association with HIV/AIDSin many different ways
depending upon the context. In 2000, Julie Pulerwitz and colleagues developed one of the
first theoretically-based measures of relationship power using the Theory of Gender and
Power (Connell 1987; Wingood and DiClemente 2002) and Social Exchange Theory
(Emerson 1981) called the Sexual Relationship Power Scale (SRPS) (Pulerwitz, Gortmaker,
and Dejong 2000). The SRPS was developed and tested among a sample of Latina women in
the US. This single publication sparked a number of HIV-related studies in the West and
7


also in sub-Saharan Africa, as researchers attempted to adapt the scale to African samples
(Harrison et al. 2006; Pettifor et al. 2004; Jewkes et al. 2002; Dunkle et al. 2007). Pulerwitz et
al. (2000) defined relationship power as the ability to control a partners actions, act
independently, dominate decision-making, or engage in behavior against the other partners
wishes. In their adoption of Pulerwitz et al.s work to South Africa, Pettifor and colleagues
(2004) measured relationship power as a combination of two factors: relationship control
and recent experience of forced sex. One important limitation of most of these studies is
that they measured relationship power and its effects on behavior and health from the
female point of view, resulting in an incomplete representation of male and couple power
dynamics. In this dissertation, I contribute to the growing body of literature on relationship
power by addressing both couple members perspectives, thereby giving a voice to men who
are often silenced in this topic of inquiry.
Summary of Research Design
The present research uses a sequential mixed methods design with three complementary
phases: a measure development phase (qualitative and quantitative) to develop a measure and
conceptual model of relationship power, a hypothesis-testingphase (quantitative) to test whether
the measure of relationship power predicts HIV testing behavior, and an intepretativephase
(qualitative) to offer context for the quantitative findings through the use of grounded
theory. In the first phase, I develop a pilot measure of relationship power for the Malawi
context and then re-formulate it using a larger set of couple data collected as part of the


Tsogolo Lm Than^i (TLT)1 study. I bring in qualitative data to provide context for the measure
and to finalize the conceptual model of relationship power. In the next phase, I test for
associations between relationship power and HIV testing behavior. Here, I utilize
longitudinal couple data to test hypotheses related to two HIV testing behaviors: HIV
testing uptake over a 16-month period and disclosure of test results to primary sexual
partners. During the interpretive phase, I use qualitative data to explain, cross-validate, or
challenge the quantitative findings from the hypothesis-testing aim. I draw upon focus group
discussions (FGDs) and other sources of qualitative data such as semi-structured interviews
with young couples, detailed field notes, and informal interviews with young people, HIV
testing counselors, and village chiefs.
Implications for Policy, Prevention, and Public Health
The results of this research have important theoretical, scientific, and practical
implications for the field of public health and the social sciences, and for HIV testing policy
in sub-Saharan Africa. Several innovative features of this research advance the study of the
relationship context and HIV testing in sub-Saharan Africa. First, this research develops and
evaluates a new measure of relationship power that could be used by other HIV/AIDS
researchers working in Malawi and surrounding countries. This measure builds upon the
SRPS (Pulerwitz, Gortmaker, and Dejong 2000), which has been mostly applied to womens
risk for HIV/AIDS in sub-Saharan Africa (Pettifor et al. 2004; Jewkes et al. 2010; Shannon
1 Tsogolo la Thanzi (TLT) is a research project on reproduction and AIDS designed by
Jenny Trinitapoli and Sara Yeatman, and funded by grant (R01-HD058366, PI Trinitapoli)
from the National Institute of Child Health and Human Development (NICHD).
9


et al. 2012). This dissertation broadens existing measures of relationship power by
incorporating the perspectives of both men and women. Second, this study is the first to use
longitudinal couple data to explicitly study the association between relationship power and
HIV testing behavior (uptake and disclosure). Much of what is currently known about
relationship power and HIV relates to risk, not HIV testing behavior. Last, the findings
generate new theoretical perspectives on local constructions of relationship power in Malawi,
that is, theory that extends our current understandings of power and relationships.
This research also has practical implications. Information on the key factors that
influence health decision-making will generate far-reaching conclusions about relationship
factors that can be used to facilitate improved approaches for couple-oriented HIV testing
programs, which are largely absent throughout the country. In addition, new measures of
relationship power for the Malawi context will provide new opportunities to evaluate the
effectiveness of public health interventions aimed at improving gender relations and health
outcomes within couples.
Chapter Outline
The remaining chapters of this dissertation can be summarized as follows. In Chapter 2,
I describe the predominant theoretical perspectives used to understand gender and power
relations and how this intersects with the background literature on HIV testing. Specifically,
I outline the barriers to HIV testing uptake and disclosure in the African context. Chapter 3
presents the mixed methodology I use accomplish the specific aims of this study. As part of
this chapter, I provide the details on the measure of power developed for the Malawian
10


setting. Chapters 4 through 6 provide the main results of this dissertation. In Chapter 4, I
use qualitative data to understand what power means in the Malawian context and to
investigate how the measure of relationship power developed in Chapter 3 fits with local
constructions of power. In Chapter 5, I present the main quantitative findings for this
dissertation, specifically, the relationship power factors that influence uptake of HIV testing
and disclosure of HIV test results. In Chapter 6,1 transition back to the qualitative data to
explain the main quantitative findings from Chapter 5. While each results chapter has its
own discussion section that links to the findings from the previous chapter, I tie all the
findings together in Chapter 7. To conclude, I present the theoretical and public health
implications, and future directions of research.
11


CHAPTER II
BACKGROUND AND CONCEPTUAL FRAMEWORK
In this chapter, I begin with a broad examination of how gender and relationship power
have been studied in relation to HIV infection. I then describe what is currently known
and where the literature falls shortregarding the association between relationship power
and two types of HIV testing behavior: uptake of testing services and disclosure of test
results to sexual partners. As I review the literature, I incorporate discussions of how social
science theory has been and could be applied to inform research on HIV testing behavior. I
conclude with a description of the rural Malawian context: the historical, social, economic,
and cultural landscape, the HIV/AIDS epidemic, and the state of existing HIV prevention
programs and policies.
Gender, Relationship Power, and HIV/AIDS
On a June afternoon in 2009, we respectfully approached a couples home in the Mponda
village to conduct the fifth interview of the project. After formally greeting each member of
the household according to Malawian customs, our intention was to learn about the couples
relationship history. But their stories were already imprinted on their bodies and immediately
revealed through a single glance at the couples physical appearance. The man was tall, very
thin, and had what appeared to be sores on his face; the wife suffered from a bad cough,
later determined to be tuberculosis. In a private conversation, the wife explains how she was
previously a bar girl (akin to a prostitute in this case) and her current husband made her an
offer she couldnt refuse: to marry him and gain the chance to build a respectable life for
12


herself as a wife and mother. The husband was older, had several ex-wives, and his steady
job at a local factory afforded him the opportunity to have multiple sexual partners. Despite
the presence of these clear risk factors for HIV, it was only after the couple lost their
young child to AIDS that they decided to get tested themselves and were discovered to be
HIV positive.
Times are still dangerous for Africans living amid a generalized AIDS epidemic. The
common phrase in Africa that you are either affected or infected with HIV/AIDS reflects
the hard-hit reality that almost everyone knows someone currently living with HIV or who
has recently died of its complications. Despite valiant and well-intentioned efforts by global
health institutions to prevent new HIV infections, children, adolescents, and adults continue
to become infected at relatively high rates. Sub-Saharan Africa is home to approximately 22
million individuals currently living with the HIV virus (UNAIDS 2010). In 2009, an
estimated 1.8 million people were infected with HIV in this region alone. HIV incidence is
gradually declining in many regions of sub-Saharan Africa, but the number of people living
with HIV continues to rise, largely due to ART expansion (UNAIDS 2010).
Although estimates differ by country, women in sub-Saharan Africa now make up nearly
60% of all HIV infections (UNAIDS 2010). These gender disparities in HIV infection
persist across adults and young people (UNAIDS 2010; National Statistical Office & ORC
Macro 2011). Gender-based relationship power is one of the most widely cited reasons used
to explain gender disparities in infection rates. Many social epidemiologic studies from
southern Africa demonstrate that relationship power imbalances are associated with key risk
13


factors for HIV, including less condom use, increased number of sexual partners, coercive
sex, transactional sex, increased physical violence, and alcohol and drug abuse (Dunkle et al.
2004; Pettifor et al. 2004; Jewkes, Levin, and Penn-Kekana 2002; Blanc et al. 1996; Harrison
et al. 2006; Dunkle et al. 2007). In several studies, a direct association has been established
between power imbalances and HIV infection (Dunkle et al. 2004; Pettifor et al. 2004).
The Theory of Gender and Power
In 1987, R.W. Connell developed the first systematic conceptual framework for the
social analysis of gender called the Theory of Gender and Power or TGP for short
(Connell 1987). Wingood and DiClemente (2002; 2000) later adapted Connells theory to
specifically study womens social exposure and risk for HIV/AIDS. One unique feature of
the TGP is its potential to study HIV risk beyond the individual-level by examining the
broader context of relationships that perpetuate risk. The TGP proposes that power
inequities arise from three overlapping social structures that interact to generate different
exposures and risk factors for HIV/AIDS: the sexual division of labor, the sexual division of
power, and social norms related to gender (also known as cathexis). Because these three
social structures are so closely intertwined and often inseparable, it is difficult to label an
exposure or risk factor as the result of one structure over another; rather, it is often the
intersection of these factors together that creates a risky environment.
The sexual division of labor functions at the societal level through the allocation of men
and women to certain occupations (Wingood and DiClemente 2002). Women are often
relegated to jobs and responsibilities deemed to be womens work, such as those that exist
14


in the domestic sphere. The inequities resulting from the sexual division of labor are
manifested as economic exposures and socioeconomic risk factors for HIV/AIDS. In sub-
Saharan Africa, there is evidence to support the sexual division of labors role in HIV risk
for women. Traditional breadwinner-homemaker marriages constrain womens earning
potential and mobility while at the same time providing men with more access to wealth,
opportunities, and sex partners. A number of studies from the region support the view that
mens mobility is a risk factor for HIV (Voeten, Vissers, and Gregson 2009; Vissers et al.
2008; Kishamawe et al. 2006; Lurie, Williams, Zuma, Mkaya-Mwaburi, Garrett et al. 2003).
In addition, many African societies are organized according to patrilineal systems of
decent and inheritance that allowed men to control economic resources in the household
(Caldwell, Caldwell, and Orubuloye 1992). Resource theory (Foa and Foa 1980) helps to
elaborate on the TCPs structure of labor by extending the idea of economic dependence.
According to resource theory, women with less access to and control over resources as
compared to their partners become economically dependent on men. This dependence is
thought to limit their negotiating power over sex, potential to mitigate violence, and ability
to leave a risky relationship.
Although sex linked to subsistence is critical for the survival of many marginalized
African women (Leclerc-Madlala 2003; Wojcicki 2002), capital-led globalization has created a
new form of dependence in the form of luxury items that are becoming increasingly desired
by women from their more wealthy sexual partners (Hunter 2010, 2002; Epstein 2007;
Wamoyi et al. 2011). Receipt of money, gifts, or financial assistance from men, sometimes
15


referred to as material transfers, has been shown to decrease womens bargaining power in
the sexual realm, including condom use and frequency of sex (Luke et al. 2011; Dunkle et al.
2004; Luke 2006).
The TCPs sexual division of power is maintained by social mechanisms such as the
abuse of authority and control in relationships (Wingood and DiClemente 2002). Gender-
based violence is one manifestation of these power imbalances and an important risk factor
for HIV in sub-Saharan Africa (Dunkle et al. 2004; Jewkes, Levin, and Penn-Kekana 2003;
Jewkes et al. 2010). Multiple direct and indirect pathways have been proposed to link power
imbalances, womens experience of violence, and HIV infection (Jewkes et al. 2010). Direct
effects occur through violent, unprotected sexual encounters with abusive men who are
more likely to engage in risky sex and to be HIV infected themselves (Jewkes et al. 2009).
Indirectly, the threat of violence may prevent women from negotiating the circumstances of
sex, resulting in more frequent sex and less condom use (Jewkes et al. 2006; Pettifor et al.
2004). Large age and economic differences between partners such as those characterized as
sugar daddy relationships may also place adolescent women at increased risk for violence
(Luke 2003) and HIV infection (Kelly et al. 2001) through their limited power to negotiate
safe sexual behavior with older, wealthier men (Luke 2005; Luke and Kurz 2002).
According to the TGP, social norms surrounding gender and sexual behavior (or
cathexis) add a third dimension of risk for HIV. In many African societies, traditional
gender roles and socialization patterns implicitly or explicitly dictate what men and women
do and how they behave (Shettima 1998). A husbands right to punish his wife or demand
16


sex are often condoned and considered socially acceptable (Jewkes 2002). The presence of a
sexual double standard (Hunter 2010) that makes it more socially acceptable for men to
have extramarital partners also places women at increased risk for HIV through their
partners behaviors. Women, on the other hand, in order to meet the ideal qualities of a
respectful wife, may avoid topics that could create conflict in the household, for example,
condom use or extramarital affairs. Longitudinal studies have demonstrated that women are
more likely to become infected by their husbands, while men are more likely to become
infected through their own extramarital affairs (King et al. 1993; deZoysa, Sweat, and
Denison 1996; Heise and Elias 1995; McKenna et al. 1997).
Though typically not applied to men, mens health is similarly affected by the three
social structures of gender and power. There are many pathways to HIV infection for men in
sub-Saharan Africa; I highlight just a few examples. Through the sexual division of labor,
dominant masculine ideologies surrounding fatherhood and the provider role keeps men out
of the household and in the labor force. Mens mobility and freedom in the public space
facilitates their engagement in HIV risk behaviors, particularly, having multiple concurrent
sexual partners. Through the sexual division of power, male dominance over the timing and
circumstances of sex directly places men at risk for HIV as well. Finally, through the third
structure of social norms and affective attachments, mens desire for multiple sexual partners
and the elevated social status that comes with this practice creates exposures to HIV
infection (Swidler and Watkins 2007; Smith 2009).
17


HIV Prevention and Testing in Sub-Saharan Africa
This dissertation attempts to translate what we know about the power/sexual risk nexus
to non-sexual behavior related to HIV testing (namely, uptake and disclosure). Before
outlining how we might apply this knowledge to HIV testing, I will first discuss the
predominant approaches to HIV prevention in this region and why HIV testing is an
important area of inquiry.
Current HIV prevention strategies center on the ABC approach or Abstinence, Be
faithful, and Condoms. For young adolescent couples, abstinence has been promoted as an
idealistic HIV prevention strategy by forgoing or delaying sex until marriage. While by far
the most effective strategy to prevent HIV/AIDS when practiced, its high failure rate has
often been attributed to a morality-driven agenda that lacks rigorous scientific support
(Santelli et al. 2006). For those who fail in A, or abstinence, albeit under circumstances
that make it difficult to be successful, B encourages people to stay with one sex partner. If
neither A nor B is feasible or possible, consistent condom use or C, is recommended
as a fallback strategy. However, for young couples just starting their families, condoms
interfere with childbearing intentions. In addition, public health messages continue to
perpetuate the association of condoms with high-risk or causal sex, thus limiting their use
within serious partnerships if people perceive them to interfere with intimacy and trust.
Condoms can also diminish womens sexual pleasure and interfere with intimacy (Higgins
and Hirsch 2008; Chimbiri 2007; Tavory and Swidler 2009). As an alternative to condoms,
married couples are advised to remain faithful to each other through widely disseminated
18


public messages on fidelity (for example, Ugandas zero grazing campaign). The objective
here is to minimize exposure to sexual networks that ostensibly act as superhighways to
transmit HIV infectiona phenomenon referred to as concurrency (Epstein 2007).
While the predominant prevention approach for HIV/AIDS in sub-Saharan Africa
emphasizes the ABCs, the continued spread of the disease signifies that knowledge alone
is not sufficient and thus other approaches are warranted. Policy makers and international
health experts have promoted HIV testing and counseling as one such alternative for HIV
prevention (CDC 2006; UNAIDS 2006; WHO 2007). Two main terms are used to refer to
HIV testing: HTC (HIV testing and counseling) and VCT (voluntary HIV counseling
and testing)2. The term HTC recently replaced the term VCT in official AIDS policy,
although much of literature that I reference in this dissertation reflects on VCT. From a
public health standpoint, HTC serves to reduce HIV transmission through its
complementary risk reduction counseling and timely access to care, treatment, and other
HIV prevention services (Painter 2001; VCT Efficacy Group 2000).
Tremendous efforts have been made to increase the number of people who know their
HIV status in sub-Saharan Africa. Public health slogans such as Know Your Status
2 In the past, the term VCT was used to refer to walk-in testing services located within
health facilities or as standalone testing sitesor client-initiated testing. The Ministry of
Health in Malawi recently replaced the term VCT with HTC to reflect revised standards set
forth by the WHO and UNAIDS. Provider-initiated HIV testing and counseling is now
considered the international standard of care. Health care providers are encouraged to offer
an HIV test paired with complementary risk reduction counseling to their patients during all
medical encounters (WHO, 2011). Therefore, the term HTC encompasses both provider-
initiated and client-initiated services offered at a wide range of testing sites including mobile
clinics, the home, and at the workplace.
19


(featured in Figure 2.1) are omnipresent in Malawi, carefully placed in high-traffic public
spaces and outside HTC clinics in an attempt to normalize the act of testing and promote
the individual right, freedom, and entitlement to knowledge of HIV status. The marketing of
HIV testing is clear, concise, and appeals directly to testing consumers as if the act of testing
is as simple as just showing up (i.e., Lets go!).
Figure 2.1: Example Lets Go Slogan Promoting HCT
(Photograph taken by Amy Conroy in Cape Maclear, Malawi, 2009)
Despite substantial economic and human resources dedicated to promoting HIV testing
in sub-Saharan Africa, testing programs continue to fall short. An extensive body of
literature from sub-Saharan Africa cites three main areas of concern: uptake of testing is
relatively low in many settings (Denison et al. 2008; Obermeyer and Osborn 2007),
disclosure of test results to primary partners is less than ideal (Obermeyer and Osborn 2007),
and sexual behavior change, particularly among those who test negative, is modest at best
(Shelton 2008; Yeatman 2007). This dissertation focuses on the first two areas of concern:
20


uptake and disclosure.
Recent estimates based on surveys in 12 high-burden countries in sub-Saharan Africa
indicate that a median of 12% of men and 10% of women in the general population have
been tested for HIV and received the results (WHO 2007). If this is accurate, then
improving the use of HIV testing services would require an understanding of the barriers
and facilitators to uptake in the general population. The benefits of HIV testing are also
predicated on the assumption that couples first disclose their HIV status to each other.
Disclosure may reduce HIV transmission by increasing awareness and decreasing risky
sexual behavior (Medley et al. 2004). It is difficult to generalize about disclosure rates at the
population level. Most studies utilize convenience samples of HIV clinic patients or pregnant
women (Kilewo et al. 2001; Antelman, Fawzi, and Kaaya 2001). However, aggregate
estimates from multiple countries show that an average of 52% (range: 16-86%) of HIV
positive women disclosed their status to sexual partners (WHO 2003). Recent studies that
include menboth HIV positive and negativeindicate that upwards of 70% of men from
Malawi, South Africa, and Kenya disclosed their status to sexual partners (Anglewicz and
Chintsanya 2011; Katz et al. 2009).
The Social Ecological Model of Health
The Social Ecological Model of health provides a useful framework for examining how
determinants of HIV testing behavior (uptake and disclosure) have been studied and where
gaps in the literature remain. According to the Social Ecological Model, multiple levels of
factors influence health behaviors including intrapersonal, interpersonal, institutional,
21


community, and public policy (Sallis, Owen, and Fisher 2008). In the discussion that follows,
I present how factors at the individual (intrapersonal) level and relationship (interpersonal)
level have influenced HIV testing behavior. I also consider structural level factors related to
economics, but only as they are manifested within the context of the relationship. I
hypothesize that a lack of understanding of and consideration for relationship factors,
specifically power, contributes to limited HIV testing success at the programmatic level.
What It Takes to Know Your Status in Africa
The Individual Level
At the individual level, a variety of different psychosocial factors are thought to
influence the uptake of HIV testing services in sub-Saharan Africa. The most widely
documented barriers include stigma and discrimination (Berendes and Rimal 2011;
Kalichman and Simbayi 2003; Weiser et al. 2006; Hutchinson and Mahlalela 2006),
confidentiality concerns (Bwambale et al. 2008; Weiser et al. 2006), and low HIV-related
knowledge3 (Jean et al. 2012; Gage and Ali 2005; Berendes and Rimal 2011).
One particularly important individual-level factor is perceived risk. This construct is one of
the most widely applied components of the Health Belief Model or HBM (Becker 1974),
especially for studying HIV/AIDS. The underlying idea is that beliefs about the likelihood of
contracting a disease or condition will motivate people to adopt a certain preventative
3 Though knowledge is cited as a barrier to testing in recent studies, it is important to
point out that it is generally understood that a lack of HIV-related knowledge is no longer
a major driver of the AIDS epidemic in Africa. People are well aware of AIDS, how the
virus it is transmitted, and the precautionary measures to help avoid HIV infectionthough
this knowledge is not perfect and nuanced understandings still exist.
22


behavior (Champion and Skinner 2008). Indeed, previous research demonstrates that low
perceived risk for HIV may limit uptake of testing (deGraft-Johnson et al. 2005; Weiser et al.
2006; Creel and Rimal 2012; Sambisa, Curtis, and Mishra 2012). There exists some
inconsistency, however, about the direction of this association; other studies have found that
individuals who perceive themselves to be at high risk for HIV may refuse testing out of fear
of the consequences (MacPherson, Corbett et al. 2012; Pool, Nyanzi, and Whitworth 2001).
Because the HBM construct of perceived risk has been such an influential variable in the HIV
testing literature, I consider it in my quantitative analysis. By doing so, I also hope to resolve
some of the inconsistency regarding direction of the association between perceived risk and
HIV testing uptake.
Less research has focused on the relationship context of risk perception. This is despite
the fact that evaluation of risk for HIV is often made in a dyadic context, that is, risk
assessments are based not only on self, but also on partners sexual backgrounds. In Malawi,
women are most worried about getting HIV from unfaithful husbands, while men are most
worried about getting HIV from their extramarital partners (Smith and Watkins 2005). There
is mixed evidence on whether perceived partner infidelity is positively or negatively
associated with testing uptake. Several studies from South Africa and Zimbabwe
demonstrate that women who suspect or know that their partners have other sexual partners
are more likely to get tested for HIV (Luseno and Wechsberg 2009; Morin et al. 2006). In
contrast, a study from Malawi finds no association between perceptions of a partners
infidelity and ever having tested for HIV (deGraft-Johnson et al. 2005). In this dissertation,
23


I extend the research that focuses mostly on individual-level perceived risk to include a
partners perceived risk as a predictor of HIV testing uptake.
The Relationship Level
Though the literature centers on individual-level determinants, an investigation of the
relationship level is critical in order to better understand the social environment in which
HIV testing decisions are made. A broader body of health literature posits that serious
intimate relationships play an important role in couple health and consequently, there have
been growing calls to study how partners mutually influence each others behaviors (Lewis et
al. 2006). The Theory of Interdependence suggests that behaviors within dyads are
interdependent; each couple member has a certain amount of influence over the interaction
they have together (Kelley and Thibalt 1978). Therefore, an understanding of the effects of
couple interactions on health requires that both partners perspectives be taken into account.
Unlike certain behavioral outcomes such as sex that cannot occur without the participation
of both partners, decisions to test for HIV and disclose test results can, in theory, be made
independently (unless the behavior under investigation is couples testing, which clearly
requires both couple members). However, the literature suggests that HIV testing behavior
does not occur in isolation; men and women make HIV testing decisions using information
about their relationship (Luseno and Wechsberg 2009; Morin et al. 2006). Thus, I posit that
the Theory of Interdependence can still provide the rationale for why it is important to
examine how partnership context influences individual decisions to test or disclose.
24


The Theory of Gender and Power (TGP) may provide a useful theoretical lens to study
how the relationship contextparticularly powershapes HIV testing behavior (uptake and
disclosure). To briefly recap, the TGP argues that three social structures, sexual division of
labor.; sexual division of power, and cathexis or social norms around gender, shape sexual risk for
HIV/AIDS. Of the studies that have examined the association between relationship power
and HIV testing uptake, many have centered on male control over their female partners
decisions to test. The three social structures of the TGP are implicitly involved here,
although they are rarelyif everdirectly referred to in the literature. Studies suggest that
decision-making power imbalances related to the sexual division of power constrain womens
use of testing services. For example, a common reason provided by women who refuse
testing is the need to discuss the issue with their husband or because the husband refused
testing himself (Kranzer et al. 2009; Dahl et al. 2008; Perez et al. 2006; Baiden et al. 2005).
Issues related to cathexis may also play a role; qualitative research suggests that a double
standard may exist around HIV testing such that women need to request permission from
their husbands, but men are free to make testing decisions on their own (Maman, Hogan,
and Kilonza 2001). Finally, womens reported obligation to seek permission from their
partners may be reinforced by the sexual division of labor through male control over economic
resources. For example, Morin and colleagues (2006) note that women inform their
husbands before testing in order to obtain money to travel to testing sites. More importantly,
this reinforces womens inability to access money and resources in the future.
25


Wingood and DiClemente (2000) argue how being involved in an abusive relationship
creates physical exposures to HIV through the sexual division of power. The authors provide an
example of how Latina women in the US who feared a partners anger in response to
requests to use condoms were less likely to use them (Marin et al. 1993). The same logic
parallels existing research on use of HIV testing services for women in sub-Saharan Africa.
Worry about physical abuse has been found to hinder the ability to test if individuals suspect
that they could be HIV positive (MacPherson, Corbett et al. 2012; Pool, Nyanzi, and
Whitworth 2001). The TGP structure of cathexis is inextricably linked to these manifestations
at the division of power. At the societal level, the structure of cathexis dictates appropriate
sexual behavior for women and reinforces cultural taboos regarding female sexuality
(Wingood and DiClemente 2000). Therefore, women may suffer abuse after testing for
reasons related to cathexis, for example, because they failed to inform their partner of their
intentions to test and thus violated social norms around male permission or because the act
of testing implies guilt from infidelity or past promiscuous behaviorregardless of whether
women engaged in these behaviors.
Women have also reported fears of divorce or partner abandonment as barriers to
testing (Mlay, Lugina, and Becker 2008; Irungu et al. 2008). I suspect that these findings
operate through the pathway of financial resources. Through the sexual division of labor,
women generally have less access to economic resources than men. According to the theory,
women who are unemployed or underemployed may be forced to rely on their male partners
for economic assistance, which in turn limits their agency in the relationship (Foa and Foa
26


1980; Wingood and DiClemente 2000). As the theory implies, women with less access to
economic resources may be more susceptible to the effects of relationship dissolution should
they test positive, especially if they perceive few alternatives to the current relationship.
Thus, the risk of losing a relationshipand the financial support that accompanies itmay
overshadow any of the perceived benefits of HIV testing. But also, womens economic
dependence on men may limit their negotiating power over HIV testing if their partner is
reluctant to test.
Up until this point, I have focused my theory and review of the literature and relevant
theory on relationship factors that may influence womens uptake of HIV testing, but what
may explain mens behaviors and how might this differ from womens? While the TGP
proposes that three social structures limit womens uptake of HIV testing, these same social
structures provide men with a surplus of power which provide men with a surplus of power
may ultimately prevent them from testing as well.
During an informal interview in 2011 with a male VCT counselor named Joseph, I
frankly asked him if and why men were scared of testing.
Men are more reluctant to come in for testing. Joseph says they are running away from their
responsibilities. He also says that men are often blamed (he said victims in English) for
transmitting HIV to women through extramarital affairs. Consequently, many will refuse to get tested
out of fear. It takes a long time for women to convince their husbands to get tested as well. When a
married woman comes in for testing, Clinic staff will invite the husband to come too by going out
to the home using maps of the area and then asking them to come in for VCT. Wives blame the
husbands for bringing in HIV since they are at home with the children while the men are free to move
around with other women. (Interview and field notes dated October 18, 2011)
I previously suggested that womens fear of being blamed for infidelity might limit their
uptake of HIV testing (i.e., cathexis). In the above quote, Joseph suggests that it is the
women who are doing the blaming, not the menhence, men become victims. Men are
27


very much aware of the global AIDS discourse that faults men for higher rates of HIV
infection among women; this emphasis on female vulnerability to HIV/AIDS may have
unintended consequences on mens testing behavior.
Anthony Simpsons (2009) research on masculinities and AIDS in Zambiaa country
that shares its borders with Malawidocuments that men are indeed having extramarital
affairs and are well-informed about the risks of having these relationships. But in order to
circumvent an admission of guilt that comes with the act of going for HIV testing, men will
avoid it even if they are clearly dying of AIDS (Simpson 2009). In fact, there may be real
consequences for men who are found to be cheating on their wives during an AIDS
epidemic. Evidence from the region shows a growing intolerance for mens sexual
indiscretions. Several studies from Malawi point out that women will bring in marriage
mediators, confront his mistresses directly, and may ultimately leave a partner who refuses to
reform his sexual behavior (Watkins 2004; Schatz 2005). In nearby Uganda, new legislation
allows wives to divorce husbands for infidelity (Parikh 2009). Taken together, this research
indicates that men may internalize some of the same worries around infidelity accusations
and divorce when making decisions to testand later, to disclose.
Beyond these affective attachments that link HIV infection to infidelity (aspects of
cathexis), questions remain around how the other social structures of TGP limit or facilitate
mens rates of HIV testing. Through the sexual division of labor.; mens breadwinner role and
the income that accompanies it may buffer the potential negative consequences of testing
positive, thus making testing seem like a more risk-worthy venture. At the same time, being
28


employed means that men have less time to wait in long lines at crowded health centers to
receive testing at the cost of losing out on wages. Of the fewer studies that include men, the
findings show that migration for work (Weiser et al. 2006), logistical barriers around VCT
including access (Bwambale et al. 2008), and testing service characteristics (Kranzer et al.
2009; Hutchinson and Mahlalela 2006) may be important factors for mens uptake.
Through the intersection of the sexual division of power and cathexis, men involved in male-
dominated relationships who perhaps adhere to stronger patriarchal ideals around
masculinity may be more likely to avoid overly feminized health care spaces. A large body of
literature on masculinities has established a relationship between lower rates of health service
utilization and poorer health among men who adopt these gender ideals. In Zimbabwe,
Skovdal and colleagues (2011) discovered that hegemonic masculinities that require men to
be strong, informed, resilient, disease-free, and highly sexual and economically productive,
may serve as a barrier to HIV testing. Here, these ideals of masculinity conflict with the
good patient perspective that expects patients to be concerned about their health, regularly
seek care at what men largely perceive to be female-dominated spaces, take instructions from
health care providers, and engage in health promoting behaviors. In Malawi and Uganda,
Izugbara and colleagues (2009) found that young males resisted testing because they
perceived the act to signify a lack of self-confidence and vulnerability to HIVtraits that
conflicted with their male youth identity. To conclude, men face similar social barriers
around gender and power that limits their uptake of HIV testing, however, their position has
largely been neglected in this body of research.
29


What it Takes to Disclose Your HIV Status
The Individual Level
It is widely acknowledged that individuals face great difficulties when making decisions to
disclose their HIV status. Decisions to disclose test results to sexual partners are often made
using a complex calculus that involves weighing the many advantages and disadvantages of
the behavior. In the literature, two particular constructs of the HBM have been widely
studied with regard to disclosure: the perceived barriers (or costs) and the perceived
benefits. The HBM constructs offer explanations of behavior in relation to the threat of a
particular disease or health conditionor in this case, the perceived outcomes of sharing
personal health information with sexual partners.
Given that decisions to test for HIV often include whether individuals anticipate
disclosure, many of the same barriers and benefits of testing uptake apply to disclosure. At
the individual level, demographic barriers include lower socio-economic status (SES), being
unmarried, and younger age (Anglewicz and Chintsanya 2011; King et al. 2008; Antelman et
al. 2001; Farquhar et al. 2004; Wong et al. 2009). Combined, the findings suggest that these
individuals may be more vulnerable to the negative consequences of disclosure: simply put,
they have more to lose than their older, married, and higher SES counterparts. Risky sexual
behavior such as having unprotected sex and multiple sexual partners has also been cited to
constraint disclosure among HIV positive samples (Antelman et al. 2001; King et al. 2008),
perhaps because these individuals expect to be blamed for immoral sexual behavior.
30


HIV serostatus is another individual-level factor thought to play a role in decisions to
disclose test results with sexual partners (Maman, Hogan, and Kilonza 2001). Unsurprisingly,
HIV negative individuals are more likely to disclose than HIV positive individuals
(Anglewicz 2008). Katz and colleagues (2009) found that HIV positive men had a more
difficult time disclosing to their partners than HIV positive women. Research on stigma and
discrimination in South Africa shows that HIV positive individuals are more susceptible to
negative social and economic consequences such as loss of employment, increased poverty,
and rejection by family members and friends (Simbayi et al. 2007). While HIV status at the
individual level is highly relevant, little research has explored it within the context of the
relationshipthat is, whether or not HIV status couple concordance (both partners are
either HIV positive or HIV negative) and discordance (one partner is HIV positive, one
partner is HIV negative) plays a role in decisions to disclose to sexual partners.
The Relationship Level
While a variety of relationship-level barriers are hypothesized to hinder HIV disclosure,
the main obstacles center on fear: fear of stigma, fear of relationship dissolution and the
accompanying loss of economic support, and fear of physical and sexual violence (Medley et
al. 2004; Kilewo et al. 2001; Maman, Hogan, and Kilonza 2001; Maman et al. 2003; Farquhar
et al. 2004). Fear-based barriers are more salient for women than men, which reflect the
unequal and limited power that many women have around HIV infection (Maman, Hogan,
and Kilonza 2001). In Malawi, a study on couples found that AIDS-related stigma was a
more salient barrier to disclosure for women than men (Anglewicz and Chintsanya 2011).
31


Similar to HIV testing uptake, disclosure-related behavior can be explained by the three
social structures of the TGP. Regarding cathexis, some qualitative research suggests that
women who choose to disclose a positive test result to their spouses are blamed for bringing
HIV into the family (Lugalla et al. 2008), especially if men were not informed that their
partners went for testing (Maman, Hogan, and Kilonza 2001). At the sexual division of power,
other studies show that women who disclosed their HIV test results experienced negative
outcomes such as violence (Maman et al. 2002; Maman et al. 2001; Maman et al. 2003). In a
study from Tanzania, Maman et al. (2002) reported that partner violence was 10 times
greater among HIV-infected women as compared with their uninfected counterparts.
However, there are conflicting accounts about whether women experience high levels of
adverse consequences after disclosure. In fact, many studies report that violence is a rare
outcome of disclosure for women (Medley et al. 2004; Desgrees-Du-Lou 2005; Keogh et al.
1994; Vissers et al. 2008) and tends to occur more in serodiscordant couples (Maman et al.
2003). Interestingly, in Tanzania, less partner violence was reported among women who
disclosed their serostatus to their partners than among women who did not (Maman et al.
2001). It is possible that individuals who tested and then later disclosed are a biased sample
since they tested for HIV in the first place, as compared to those in violent relationships
who delay or forgo testing altogether out of fear. Regardless, these studies suggest that
violence after disclosure is situational and depends on the relationship context.
Inequities resulting from imbalances at the sexual division of labor may also shape womens
decisions to discloseespecially if the woman is the HIV positive partner. Fears of loss of
32


economic support due to partner abandonment may drive some women to hide their status
from their partners. These fears may be justified. In Uganda, discordant couples with an
HIV positive woman were more likely to dissolve than seroconcordant or serodiscordant
(positive male, negative female) couples (Porter et al. 2004). It is important to point out that
adverse consequences may be biased if women who suspect that their partners will react
negatively are less likely to disclose in the first place.
Limitations of the TGP
Several limitations of the TGP are noteworthy. First, the TGP is rooted in the so-called
vulnerability paradigm, which overemphasizes womens lack of power in society and their
inability to protect themselves from HIV/AIDS. Globally, contemporary AIDS discourse
perpetuates portrayals of women as biologically and socially vulnerable to HIV, victims of
mens abuse, and innocent bystanders in the global HIV/AIDS pandemic (Higgins,
Hoffman, and Dworkin 2010). Not all women are victims, nor are all men overly dominant
and controlling. These popular depictions of female vulnerability serve an important and
valid purpose by reinforcing the evidence that women are indeed disadvantaged by
HIV/AIDS; however, at the same time, may restrict our understanding womens agency and
resilience.
Second, the female vulnerability paradigm is historically grounded in western feminism
and centers around individual human rights and freedoms that are defined and socially
celebrated in the Westa setting very different from rural Africa where the model is
frequently applied. It is possible that womens power in Malawi may be overlooked because
33


of the ways it is expressed within the context of culture. From a western perspective, it may
appear that women succumb to their partners dominance; yet, in reality, they may be using
backstage techniques (Goffman 1959) to maintain control over their lives. Sociologists
working in rural Malawi have started to tap into the subtleties of female agency and how this
agency is invoked to avoid HIV infection. For example, researchers have found that wives
sometimes use subtle and gendered communication to encourage fidelity in their marriages
(Watkins, Rutenberg, and Wilkinson 1997; Zulu and Chepngeno 2003) and other locally-
formulated strategies, including approaching husbands about their behavior, bringing in
marriage mediators, confronting his mistresses, and leaving a partner who refuses to reform
(Watkins 2004; Schatz 2005). By relying solely on the TGPand its female vulnerability
orientationwe may miss the important graduations of female agency that could be
harnessed to improve HIV/AIDS interventions among others who are truly powerless.
Third, a female vulnerability emphasis detracts attention away from heterosexual mens
social disadvantages and needs around HIV. It assumes that men are active transmitters of
HIV infection, but not active agents of prevention (Higgins, Hoffman, and Dworkin 2010).
In response to this, there have been growing calls for new scholarship on masculinity and
the ways in which mens own limited life choices relate to global patterns of power (Hirsch
et al. 2009).
To summarize, the TCPs underlying emphasis on female vulnerability restricts our
understanding womens agency and resilience, mens position, and the broader dyadic
context for HIV risk. In this dissertation, I hope to illuminate areas of nuance regarding the
34


gender/power/HIV testing nexus in order to move this body of research in new directions
that consider these important gaps. To accomplish this aim, I invoke the strengths of
grounded theory as a method to more fully capture the construct of relationship power and
its intersection with HIV testing. By listening to the perspectives of rural Malawians who are
the true cultural experts on these issues, I hope to broaden our knowledge of how
relationship factors intersect with the act of HIV testing.
Symbolic Interactionism and HIV Testing Behavior
As one could imagine, the act of testing is not a straightforward process; HIV/AIDS
and its interventions carry symbolic meanings that are deeply embedded within the social
and cultural milieu and strongly shape peoples decisions to test. According to Herbert
Blumers (1969) perspective on social interactionism: (1) people act on the basis of meanings
that things have for them; (2) these meanings derive from social interaction; and (3) these
meanings are modified by their interpretations in practice. The basic research assumption
consistent with this view is that if one desires to understand human interactions and how
they shape behaviors, one needs to study peoples experiences as they perceive them (Jeon
2004).
By doing so, a symbolic interactionism perspective gives agency to the rural Malawians
living amongst an AIDS epidemic rather than assuming they are mindless beings to be
fashioned by western AIDS policy. Indeed, research on HIV/AIDS interventions designed
in the West but implemented in Africa shows that local interpretations and responses may
differ substantially from how these programs are intended to function from a public health
35


perspective (Kaler and Watkins 2010; Angotti, Dionne, and Gaydosh 2010; Tavory and
Swidler 2009). In rural Malawi, Kaler and Watkins (2010) use qualitative data to understand
local perceptions of HIV testing and find that the reluctance to test is connected to the
perception that testing inevitably leads to a positive diagnosis and subsequent death. Rural
Malawians describe how testing is analogous to asking God about the day you will die and
many would prefer not to live with this uncertainty (Kaler and Watkins 2010).
The disconnection between AIDS policy and local responses resonates in the condom
literature as well. In love marriages, where women rely on ideals of love and intimacy to
negotiate relationships with their husbands, condoms are seen as undermining the very thing
they wish to preserve (Chimbiri 2007; Tavory and Swidler 2009). Given their strong
association with high-risk sex and immorality, condom use within serious partnerships
symbolizes a relationship characterized by mistrust, instability, and immoral sexual behavior
(Smith 2009). Bringing up condoms unexpectedly raises concerns about a partners HIV
status or faithfulness and symbolizes a partnership filled with uncertainty. The rationale is
simple and logical: why use condoms with a partner you trust and love? I suspect that the
same premise might be true with HIV testing.
Other scholars have used a symbolic interactionist approach to examine HIV testing
within the context of a relationship. Kathryn Rhine (2009), for example, discovered that an
HIV test is not simply a tool employed to measure immunological malfunction. She says, A
diagnosis also illuminates a set of social facts. The virtues and fears embedded in the act of
taking a test are related to larger questions of how families and relationships might change in
36


light of a positive result Deborah Lupton and colleagues (1995) add that HIV test results
have meanings that are tied to relationships, faithfulness, and trust. Individuals often take
HIV tests when they are ending a relationship or starting a new one and thus testing serves
to mark these important life transitions. In Tanzania, Maman and colleagues (2001) found
that couples used testing as a means to reaffirm ones commitment to the relationship. Thus,
for some people, the act of testing may be more of a symbolic gesture than an attempt to
confirm perceptions of risk. Through a symbolic interactionist approach, I hope to
understand how beliefs, perceptions, and experiences of HIV testing created through social
interactions at the relationship level ultimately shape peoples reactions to HIV testing
programs in Malawi. This approach will supply locally rooted evidence for the applicability
of the TGP and illuminate areas of contradiction, confirmation, and further exploration.
Study Hypotheses
This dissertation seeks to investigate how relationship power may act as a barrier or
facilitator to HIV testing behavior (uptake and disclosure) in southern Malawi. Based on the
TGP and background literature, I hypothesize that relationship power will consist of the
following domains: socio-economic power within the relationship, decision-making
dominance, and relationship violenceeach of which will be independently associated with
the two outcomes of HIV testing behavior. For both women and men, having less socio-
economic power as compared to a partner (sexual division of labor), being in a male-dominated
relationship (sexual division of power), and having a history of relationship violence (sexual
division of power) will indicate low power in the relationship. Given the background theory and
37


literature, I hypothesize that the following conceptual model will explain HIV testing
behavior within rural Malawian couples (see Figure 2.2).
Figure 2.2: Conceptual Model for HIV Testing Behavior Hypotheses
Hypothesis #1
I hypothesize that each power variable will be associated with uptake of future HIV
testing over a 16-month period. In addition, I hypothesize that the perceived risk construct
from the HBM will be associated with HIV testing uptake. Four sub-hypotheses will be
tested:
A. Socio-economic inequality (division of laborj: Individuals in a lower socio-economic position
relative to their partners will be less likely to get tested since they will have stronger fears
around divorce or abandonmentand loss of financial supportthat may come with
38


HIV testing. With less economic power, these individuals may also be in a more
disadvantaged position to negotiate testing with their partners.
B. Relationship dominance (division of power)'. Women in a male-dominated relationship as
compared to an egalitarian or female-dominated relationship will be less likely to test for
HIV due to male control over testing decision-making. Male-dominance will also be
negatively associated with testing among men. These men may be more likely to adhere
to traditional beliefs about masculinity and therefore feel disinclined to test.
C. Relationship violence (division of power): Having a history of relationship violence (physical
and sexual) is a proxy for fear of abuse, which has been shown to be a barrier to testing.
Individuals in violent relationships will therefore be less likely to test for HIV. Having a
history of violence (sexual or physical) could also operate through the pathway of risk,
thereby decreasing the likelihood of testing (assuming people are higher risk are less
likely to test).
D. Perceived risk (self and partner). For women, those who believe they are at higher risk for
HIV will fear the negative consequences of testing and therefore be less likely to test.
For men, those who believe they are at higher risk for HIV will fear being blamed for
infidelity and therefore be less likely to test.
Hypothesis #2
I hypothesize that each power variable will also be associated with HIV testing disclosure
to a primary sexual partner. In addition, I hypothesize that perceived HIV status
39


concordance will be associated with HIV testing disclosure. Four sub-hypotheses will be
tested:
A. S ocio-economic inequality (division of laborj: Individuals in a lower socio-economic position
relative to their partners will be less likely to disclose their test results since they will
more likely to fear divorce or abandonmentand loss of financial supportas a
consequence of disclosing test results.
B. Relationship dominance (division of power): Being in a male-dominated relationship as
compared to an egalitarian relationship will make women less likely to disclose out of
fear of the violence that is associated with dominance.
C. Relationship violence (division of powerj: Having a history of relationship (physical and sexual)
violence will reinforce fears of abuse, thereby decreasing the likelihood of HIV test
disclosure. Even if they tested negative, individuals in violent relationships may still
avoid disclosure if they did not inform their partners of their plans to test.
D. Perceived HIV status concordance: HIV positive individuals will be less likely to disclose their
test results if they perceive their partners to be HIV negative as compared to HIV
negative individuals who also perceive their partners to be HIV negative.
Study Innovation
This study advances our knowledge of HIV testing uptake and status disclosure in sub-
Saharan Africa in the following ways.
1. At one-time testis not enough to prevent HIV. Despite the fact that HIV testing rollouts have
increased the number of people who know their HIV status, a large majority of studies
40


on HIV testing in sub-Saharan Africa examine whether people have ever tested or
recently tested for HIV as if it is a one-time event (Berendes and Rimal 2011; deGraft-
Johnson et al. 2005; Irungu et al. 2008; Jean et al. 2012; MacPhail et al. 2007; Weiser et
al. 2006). Reliance on cross-sectional data where predictor variables and testing history
are collected alongside each other makes it is difficult to draw conclusions about
causation or temporality. Additionally, a single HIV test is not enough to prevent HIV,
especially in a setting where dense sexual networks of concurrent partnerships are
common and riskier than serial monogamy (Morris and Kretzschmar 1997; Morris and
Kretzschmar 2000). As more people begin to learn their status and as testing shifts to a
more normative practice, it is likely that a substantial proportion of HIV testing clients
will have tested before (Bradley et al. 2011). Thus, it becomes imperative to study the
factors that influence regular uptake of HIV testing throughout young peoples
reproductive years. This dissertation will examine a 16-month period of testing activity.
2. Studies that use data from both partners strengthen the evidence. Given the important role of
marriage on couple health, there have been growing calls to study how spouses mutually
influence each others behaviors using data from both partners (Lewis et al. 2006). A
couples dataset allows for the ability to predict respondent outcomes (i.e., the
dependent variable) using both the respondent and their partners independent
variablesyielding more information that if the respondent was studied in isolation.
3. Men receive less attention in the HIV testing research, particularly on disclosure, from sub-Saharan
Africa. This may be for a number of reasons and partially attributed to the ease of access
41


to pregnant women through clinic recruiting and the feminization of the AIDS
epidemic. Lower rates of testing among men in Malawi (National Statistical Office &
ORC Macro 2005, 2011) underscores the need to focus on issues that are pertinent to
men as well.
4. Studies on relationship-level factors and HIV testing in Malawi are generally lacking. Decisions to
test and disclose are often made in a dyadic context. Therefore, improvements to
testing services requires a better understanding of the relationship-level barriers and
motivating factors that affect whether people get tested and disclose their results to
sexual partners.
5. Few studies examine factors that influence disclosure amongpecple who test negative. From an HIV
prevention perspective, it is certainly understandable that greater attention would be
directed towards HIV positive people who choose to disclose. However, an
understanding of the drivers of disclosure among HIV negative people is also important
so that young people can more accurately assess their HIV risk. Furthermore, increasing
the rates of disclosure among those who are negative is important for establishing
positive relationship behaviors that may facilitate disclosure of a positive test result if
the time comes in the future. Thus, an understanding of the relationship factors that
influence disclosure of a negative status is also warranted.
The Malawi Context
Present-day Malawi is a small, landlocked country located in southeastern Africa whose
geography is largely structured by Lake Malawi, a fresh water lake that spans almost the
42


entire length of the country. Like its neighboring countries, Malawi has a long history of
colonialism. From 1891 to 1964, the British ruled Malawi as a protectorate state under the
name Nyasaland. In the early 1960s, Malawi achieved political independence from Britain
and became a one-party state with Dr. Hastings Banda as its first President. Bandas regime
came to an end in 1994 when the country entered a period of multi-party politics, which
interacted with the influence of donor aid and ultimately resulted in the failure of the private
market (Peters, 1997).
Today, Malawi has a population of 13.1 million (National Statistical Office & ORC
Macro 2011) and a Gross Domestic Product (GDP) per capita of $893 USD (The World
Bank 2011). Lilongwe and Blantyre are its two major cities and the majority of the
population resides in rural areas. More than 68% of Malawians over the age of five are
considered literate, with lower rates among women than among men (National Statistical
Office & ORC Macro 2011). Girls are significantly less likely to complete primary school
and go on to attend secondary or higher education as compared to boys, in part, due to early
childbearing. At current fertility levels, a Malawian woman would bear an average of 5.7
children over the course of her lifetime (National Statistical Office & ORC Macro 2011).
The economy of Malawi is primarily driven by agriculture, which accounts for 30% of the
countrys GDP (National Statistical Office & ORC Macro 2011). Nonetheless, at the
national level, the economy depends on substantial inflows of economic assistance from the
International Monetary Fund, World Bank, and individual donor nations.
43


The country is divided into three regions (north, central, and south) and 27 districts.
This dissertation was conducted in the Balaka district of southern Malawi (as indicated by
Figure 2.3). Balaka has a hot, dry climate given its lower elevation and distance from Lake
Malawi. Residents of the surrounding Balaka district villages travel to the Balaka boma (or
district center) for commerce and health services. The Balaka boma consists of a thriving
market, several grocery stores, banks, and bars, a soccer field, several regional non-
governmental organization (NGO) offices, the main Balaka district hospital, a combination
of public and private health clinics, and numerous churches and mosquesincluding a large
Catholic church that attracts many pilgrims and visitors from across the country.
Figure 2.3: Balaka District of Southern Malawi
44


The three regions of Malawi differ greatly with regard to socio-economic status,
ethnicity, religion, level of polygyny, lineage system, and pattern of residence. For both
genders, education levels are higher in the southern region than in the central region, but
lower in comparison to the north (National Statistical Office & ORC Macro 2011). The
major ethnic groups in the northern region are the Tumbuka, Ngoni, and Tonga. The Chewa
and Ngoni tribes predominantly populate the central region. The southern region is home to
the Yao, Lomwe, Sena, and Manganja peoples (Zulu 1996). The north is predominantly
Protestant, the central region is a combination of Protestants and Catholics, and the south is
predominantly Muslim. At the national level, about 86% of women and 84% of men are
Christians, while 13% of women and 12% of men are Muslims (National Statistical Office &
ORC Macro 2011). Polygamous unions are more common in the Christian north than in the
predominantly Muslim south (14% vs. 6%) (National Statistical Office & ORC Macro 2011).
In general, the southern region follows a matrilineal/matrilocal orientation where men
physically move in with their wives families after marriage as compared to the
predominantly patrilineal/patrilocal north (Chimbiri 2007; Peters 1997). In matrilineal areas,
villages are typically organized into clusters of compounds made up of houses consisting of
matrilineal relatives. For example, groups of sisters live together with their respective
husbands and children. Flowever, residence patterns are not always uniformly distributed by
region for reasons related to migration, temporary employment, and scarcity of land (Peters
2010). In matrilineal families, women determine the lineage structure, however, inheritance is
still passed through the maternal uncle who owns and controls the inherited property (Phiri
45


1983). The southern region of Malawi has higher rates of divorce as compared to other
regions with around 33% of all marriages ending before their 5th year anniversary (Reniers
2003), which may be partially attributed to its matrilineal marriage system.
Gender Relations and HIV/AIDS in Malawi
Matrilineal succession and inheritance provides women with considerable authority,
particularly over land, however, this factor alone does not guarantee gender equality. Most of
the major external influences on the region over the past 200 years have come from
patrilineal and patriarchal groups (Peters 1997). These groups include, but are not limited to,
British colonialism, Christian missionaries, and more recently, the international AIDS
enterprise (refer to Peters 2010 for more detail on the former two groups of influence).
During the colonialism period, Europeans obtained large plots of land from local chiefs at a
very low cost and forcibly recruited local laborerssome of whom had previously resided
on the landto tend to their agricultural estates in return for rent (1997). With them,
colonialists brought their conceptions of patriarchy and forced this gender hierarchy upon
their newly designated constituents. Pauline Peters writes, estate owners assumed and
promoted men as the natural holders of land and heads of households, so they assigned
authority to men (1997). Yet, in southern Malawi, this conflicted with the existing
matrilineal organization of the resident population and led to expansive population growth
as daughters married and were joined by their new husbands and future children on the
estates. Such rapid population growth could not be supported by the lands agricultural
production and resulted in famine in many parts of the Shire Highlands in southern Malawi.
46


Also in southern Malawi, Kaler (2001) discusses how patriarchal traditions were invented not
necessarily through time-honored practices, but rather through the interactions between
local chiefs and colonial administrators. Colonial administrators sought to define local
customs in southern Africa as something that paralleled British law. Chiefs saw that it was
advantageous for them to maintain the perception that tradition was responsible for
womens subservience to men and for male elders and chiefs to hold power over land and
other resources. The colonialist roots of patriarchy are likely to affect modem gender
relations in Malawi. Demographic data from 2010 show that approximately 72% of
households are considered headed by men (National Statistical Office & ORC Macro 2011).
The upper ranks of village chiefs are mostly male (Peters 2010). Similarly, higher-level
positions within government, civil service, and private corporations are overwhelmingly male
(with the exception of Malawis current President Joyce Banda, who served as Vice President
at the time President Bingu wa Mutharika died in office in 2012 and therefore became his
legitimate successor).
Across sub-Saharan Africa, adult HIV prevalence rates range from 0.1% in Comoros to
26% in Swaziland (UNAIDS 2010). Malawi has some of the highest rates of HIV infection
in this region, with around 11% of all adults of reproductive age infected (National Statistical
Office & ORC Macro 2011). Of the 42 sub-Saharan African countries with data, only eight
have higher adult HIV prevalence rates than Malawi (UNAIDS 2010). All eight countries are
located in southern Africa. In Malawi, HIV prevalence among adult women is higher than
their male counterparts (13% of women vs. 8% of men) (National Statistical Office & ORC
47


Macro 2011). Young men and women aged 15-24 also have high rates of HIV infection.
Approximately 3% and 5% of young people aged 15-19 and 20-24, respectively, tested
positive for HIV in 2010 (National Statistical Office & ORC Macro 2011). The southern
region of Malawi has the highest rates of HIV infection in the country, with an estimated
15% of its reproductive age population infected (National Statistical Office & ORC Macro
2011).
In Malawi, the connection between relationship power and HIV/AIDS is conflicted by
two competing bodies of literature with varying degrees of support for the vulnerability
paradigm. In one set of research, several qualitative studies on Malawian womens risk for
HIV argue that women do have the agency necessary to navigate the HIV epidemic and are
not just helpless victims of gender inequality (Schatz 2005; Tawfik and Watkins 2007). Here,
women invoke HIV prevention strategies that are considered locally, instead of globally,
appropriate for their everyday realities. For example, spouses use subtle and gender-specific
communication strategies to encourage fidelity in their marriages (Watkins, Rutenberg, and
Wilkinson 1997; Zulu and Chepngeno 2003). Other researchers argue that women draw
upon their social resources to protect themselves from HIV/AIDS including bringing in
marriage mediators, confronting male partners mistresses directly, and leaving a partner who
refuses to reform (Schatz 2005; Watkins 2004). Some evidence from research on pre-marital
partnerships illustrates how young women continuously evaluate a partners risk for HIV
using their social networks and then formulate or terminate relationships to minimize their
chance of infection (Poulin 2007).
48


In sharp contrast, a second set of research suggests that womens decision-making
power around HIV/AIDS is severely limited at the societal level. In rural areas, it has been
documented that married women dominate domestic areas around cooking and childbearing
while their husbands control domains like money and sex (Mbweza, Norr, and McElmurry
2008). While power is most likely expressed through multiple domains, it is important to
point out that female-dominated domains such as cooking are less important for HIV
transmission than male-dominated domains of sex. Womens power in Malawi likely varies,
but still some women report that they have little control over their relationships such as the
ability to choose their husbands, to bear children, and to have sex or not (Rankin 2001;
Lindgren, Rankin, and Rankin 2005). In fact, a recent demographic report featuring a
nationally representative sample shows that 25% and 28% of Malawian women reported
ever experiencing sexual and physical violence, respectively (National Statistical Office &
ORC Macro 2011).
HIV Testing and Treatment in Malawi
HIV testing and counseling (HTC) first became available in Malawi during the 1990s,
although such services were widely inaccessible to the majority of rural Malawians. Starting
in 2004, the Malawi Ministry of Health received external donor funding to support free HIV
testing services in the main district hospitals and rural clinics (Ministry of Health [Malawi]
2005, 2006). Around the same time period, Malawi expanded its prevention-of-mother-to-
child transmission (PMTCT) services by offering routine HIV testing of both mother and
child. In 2003, the government mandated routine HIV testing of all pregnant women
49


through antenatal care clinics (Office of the President and Cabinet and National AIDS
Commission [Malawi] 2003). Today, HTC is offered through integrated heath services such
as antenatal care and at standalone testing centers, clients homes, and workplace sites
among other venues. As of 2010, 73% of women and 53% of men of reproductive age had
ever been tested for HIV, reflecting a significant increase in the number of people who
know their status from previous years (National Statistical Office & ORC Macro 2005,
2011).
Many of Malawian women are now tested though opt-out, provider-initiated testing
during antenatal care, which is generally perceived to be compulsory (Angotti, Dionne, and
Gaydosh 2010). Some research suggests that rates of antenatal care testing among expectant
mothers who receive antenatal care may be as high as 99% (Weir, Hoffman, and Muula
2008). Few data are available on the prevalence of reoccurring HIV testing in Malawi;
however, I suspect that the young people in this study will likely have been tested multiple
times over the course of 16 months as they negotiate the circumstances of their relationships
and begin to have children.
Across the region, including Malawi, couples-based VCT (or CVCT) has gained little
momentum. In their randomized controlled trial of CVCT in Tanzania, Becker and
colleagues (2010) had to stop enrollment prematurely because rates of acceptance among the
intervention arm (CVCT) were significantly lower than in the control arm (VCT), thus
posing ethical issues. Low levels of couple involvement in VCT have been supported
elsewhere, including in Nairobi and Lusaka (Farquhar et al. 2004; Semrau et al. 2005). For
50


women, factors associated with reluctance towards couples testing include male dominance
over testing decision making and fears of the negative consequences of testing positive, such
as divorce or violence (Maman, Hogan, and Kilonza 2001; Kranzer et al. 2009; Njau et al.
2011). Although notably, Becker and colleagues demonstrated that for couples who did test
together, rates of marital dissolution and domestic violence were substantially lower than
those who tested individually (2010). Men, on the other hand, may be reluctant to test with
their partners because they feel embarrassed about being seen in the clinic or because of a
heightened sense of risk related to extramarital relationships (Gipson et al. 2010; Njau et al.
2011).
Home-based CVCT provides an alternative solution by circumventing public testing
venues or so-called womens places. In Malawi, some research suggests that home-based
testing is perceived to have many benefits: it is confidential, convenient, and credible
(Angotti et al. 2009). Indeed, a study on Likoma Island in Malawi found that rates of home-
based VCT were exceptionally high, especially among poorer households (Helleringer et al.
2009). Home-based couples testing may offer new opportunities for overcoming barriers
related to clinic-based testing (Njau et al. 2011).
The key benefit of HIV testing, however, depends upon reliable access to ART. In
Malawi, ART first became available on a fee basis (for pay) in public hospitals in 2000.
Although at the time, only a small group of patients could afford them (Van Oosterhout et
al. 2007). During early rollout, clinical shortcomings, drug supply interruptions, and
difficulties determining eligibility imposed limitations on quality of care and access to ART.
51


With the assistance of donor funding, ART was scaled up between 2004 and 2006 at large
hospitals and offered free of charge. Adults were eligible if they were HIV positive and
WHO clinical stage 3 or 4, or had a CD4 count less than 200 cells/L (Ministry of Health
[Malawi] 2003). Over this period, the number of patients on ART within the public sector
increased from approximately 4,000 to 60,000 (Lowrance et al. 2007). In 2009, almost
200,000 people were taking ART through 377 health facilities (Ministry of Health [Malawi]
2009). The WHO recently estimated that around 48% of adults in immediate need of ART
currently receive treatment in Malawi (WHO 2010). Estimates from Malawi using self-
reported data from 2010 show that between 24% and 62% of women who tested positive
for HIV are currently taking ART (National Statistical Office & ORC Macro 2011).
However, recently, a new policy was introduced in Malawi referred to as Option B+, which
allows all pregnant women who test HIV positive to be placed on ART for life regardless of
their CD4 counts or clinical stage (Schouten et al. 2011). This policy is expected to
dramatically increase the number of women on ART. For men, coverage estimates range
from 16% to 60% (MDHS, 2010). Overall, challenges still remain for ART programs in
Malawi in order to overcome clinical problems with ART staging and eligibility, continuity of
care after testing positive, and staff shortages (MacPherson, Lalloo et al. 2012; Makwiza et al.
2009; McGrath et al. 2010). Nonetheless, the ART program is Malawi is widely heralded as a
success in the region (Harries et al. 2011).
52


CHAPTER III
METHODOLOGY
In this chapter, I outline the research design and methods I used to accomplish the
following goals: (1) to investigate appropriate quantitative measures of relationship power
for the Malawi context (a) and to evaluate their cultural relevance with qualitative data (b);
(2) to test whether relationship power influences the use of HIV testing services (a) and
disclosure of test results to primary partners (b); and (3) to understand the meaning of HIV
testing within the context of a sexual relationship.
Study Design Overview
The present research used a sequential mixed methods design with three complementary
phases: a measure development phase (qualitative and quantitative) to develop a measure and
conceptual model of relationship power, a hypothesis-testingphase (quantitative) to test whether
the measure of relationship power predicts HIV testing behavior, and an intepretativephase
(qualitative) to offer context for the quantitative findings. In the first phase, I developed a
pilot measure of relationship power for the Malawi context and then re-formulated it using a
larger set of couple data collected as part of the TLT study. Then, I applied grounded theory
(Glaser and Strauss 1967) to a set of qualitative data in order to understand relationship
power from rural Malawians perspectives. This process allowed me to confirm or challenge
the quantitative measures of power and to uncover other areas of nuance. In the next phase,
I tested for associations between the final model of relationship power and HIV testing
behavior. Here, I utilized longitudinal couple data to test hypotheses related to relationship
53


power and two HIV testing behaviors: HIV testing uptake over time and disclosure of test
results to a primary sexual partner. During the interpretive phase, I again applied a grounded
theory approach to qualitative data to explain, cross-validate, or challenge the quantitative
findings from the hypothesis-testing aim. During the fall of 2011,1 spent two months in the
Balaka district of southern Malawi conducting focus group discussions (FGDs) with men
and women and gathering other contextual data. Previously collected couple interview data
from 2009 supplemented the FGD data as necessary.
I used a mixed methods design for important reasons that added significant value to
the study. As I have already discussed, relationship power is a complicated construct that
carries multiple meanings at many different levels of influence, which vary not only
according to geographical locality, but also over time. As such, understandings of power and
its hypothesized associations with HIV testing are likely to benefit from multiple,
complementary research methods including surveys, focus group discussions, couple
interviews, detailed field notes from several trips to Malawi, and participant observation,
including informal conversations with Malawians about issues related to power and HIV
testing. Survey measures of power are limited in capturing the depth, complexity, and
meanings behind relationship power. Not only could qualitative interviews provide insight
into the culturally rooted meanings of relationship power, interviews may uncover other
underlying constructs or variables that have not yet been identified through previous survey
research. Informal data gathering allowed me to observe the more nuanced forms of power
that may or may not arise during more formal interviews or FGDs. While the qualitative
54


results are helpful to identify new associations, they are not generalizable (at least in the
statistical sense) to the larger population of young people at risk for HIV. In addition to
playing off the strengths and weaknesses of multiple methods, I could also show if the
results were congruent, that is, the findings were not simply due to a methodological artifact
(Morgan 1998).
I characterized this study as a sequential mixed methods study where one method
precedes another as opposed to simultaneous data collection. Throughout the study, I
constantly compared and contrasted the various forms of information in order to formulate
and revise my understandings of relationship power and HIV testing. The model of
relationship power finalized in the measurement phase was used in the hypothesis-testing
phase. Preliminary findings from the hypothesis-testing phase informed the line of
questioning I used in the FGDs. Focus group discussions, field notes, and the previous
qualitative couple interviews I collected were examined together and used to reflect back on
what was learned through the survey data. Figure 3.1 provides a timeline for when the data
were collected and analyzed.
55


Developed preliminary
measure of power and pilot
tested it on 254 individuals
Began the hypothesis-
testing phase using TLT
data
Conducted 8 focus group
discussions in Malawi
'
July 2009
y ' <
July 2011 September 2011
June 2009 May 2011 January 2012-April 2013
I t 1
Conducted semi-structured Finalized the Used the semi-structured interviews
interviews on power with 34 and focus groups to reflect on the
pOWci IllcaSUXc L
coupled individuals measure of power; completed
hypothesis testing phase; used
qualitative data to explain the
quantitative findings
Figure 3.1: Timeline of Data Collection and Analysis, 2009-2013
The TLT Study
The quantitative data for the measure development and hypothesis-testing aims came
from Tsologo Tm Than^i (TLT; Healthy Futures in Chichewa). The overall objective of the
TLT study was to collect and analyze new data on the reproductive goals and behaviors of
young adults in Malawi within the context of HIV testing and counseling. Longitudinal
survey data were collected at quarterly intervals over a period of approximately three years
(for a total of 8 waves). Trained interviewers, matched by gender to the respondents,
administered questionnaires through face-to-face interviews conducted in private rooms at
56


the TLT research center in Balaka. The survey data included information on reproductive
goals and behaviors, fertility preferences, contraceptive use, and sexual behavior.
Previously administered TLT sampling procedures were as follows. A random sample of
1500 women aged 15-24 was selected from a household listing, which included people who
lived within 7 kilometers of the Balaka town center (boma). Women were given enumerated
incentive tokens for each partner that they named (husbands and boyfriends) and asked to
give the tokens to their partners, redeemable for 500 Malawi Kwatcha (around $3.25 USD)
upon completion of the interview. Women could initially recruit up to three partners;
however, since it was expected that some relationships would dissolve over the 3-year study
period, new partners could be enrolled at each successive interview. In addition to the
random sample of women, a smaller random sample of men aged 15-24 was also recruited
and enrolled in order to be able to make generalizations about men. Combined,
approximately 2500 individuals (500 couples) were enrolled in the study, although these
figures vary slightly with each wave due to attrition and partnership dissolution or formation.
However, the final couples sample (details to follow) only included these randomly selected
men if they were partners of the random sample of women.
Women and their partners were randomly assigned to receive HIV testing in order to
assess causal impacts of changes in the knowledge of HIV status on reproductive and
prevention goals. Once baseline enrollment was completed, the women were assigned to
three equal study groups consisting of 500 women and their male partners. Group 1 received
regular HIV testing every four months. Group 2 received an HIV test at the end of the first
57


year and then again at the end of the study. Group 3 only received an HIV test at the end of
the study. Figure 3.2 illustrates the timing and allocation of VCT over 8 waves of data
collection.
Group 1 VCT VCT VCT VCT VCT VCT VCT VCT
Group 2
Power VCT
Module
Power
Module
VCT
Group 3
VCT
Figure 3.2: Overview of the VCT Infrastructure for TLT
Creation of the Analysis Datasets
In waves 3 and 5, a special module on relationship power was added to the TLT
partnership survey (as shown in Figure 3.2). Respondents were asked the relationship power
statements if they reported a current serious sexual partner including a spouse, live-in
partner, steady boyfriend/girlfriend, or new boyfriend/girlfriend. If a respondent had more
than one currently ongoing sexual relationship, the relationship power questions were asked
with regard to the most serious relationship, ascertained with the question: Of the sexual
partners that you specified earlier, which one are you closest to? However, if the
respondent was married, their spouse automatically served as the reference partner even if
other extramarital relationships had been ongoing for quite some time.
A couple dataset was created for respondents who answered the power questions about
each other at wave 3 using a separate database linking women and their male partners. The
following steps were taken:
58


1. Respondents completed a baseline questionnaire eliciting demographic and other
information when they first enrolled in the study. At each wave, respondents completed
a wave-specific questionnaire and partnership questionnaire. Respondents who entered
the study at a subsequent wave only completed the baseline questionnaire at that wave
and then at the next wave, completed the wave-specific and partnership questionnaires.
Note that at wave 1, respondents completed all questionnaires. All respondents who
entered the study at wave 2 and completed baseline questionnaires (N=199) were
appended to respondents who participated at wave 1 (N=2,496). This created a sample
with complete baseline data by wave 3 (N=2,695).
2. This dataset was merged with respondents who participated in wave 3 (N=2,462). Of
these respondents, a total of 1,510 respondents were eligible (reported an ongoing sexual
relationship) and completed the power statements. This dataset was called as the wave 3
individuals dataset. Each individual was represented only once in this dataset with a
unique individual identifier. Example data are illustrated in Table 3.1.
Table 3.1: Example data from the wave 3 individuals dataset
Respondent ID Gender Birth year History of forced sex History of physical abuse
110266 2 1990 0 0
110307 2 1994
110311 2 1993 0 0
110420 2 1989 1 0
110467 2 1990 0 0
59


3. A second dataset called the wave 3 couples dataset was used to identify a set of
ongoing couples at wave 3. The couples dataset contained the following information: a
unique respondent ID for each individual, a unique couple ID, wave number (1 to 8),
two variables corresponding to the status of the relationship reported by both partners at
that wave, and a variable indicating when the relationship was first formed. The process
below resulted in a list of couples who completed the power statements at wave 3.
a. To create a baseline set of couples for wave 3, records for other waves were
dropped (i.e., the wave number does not equal 3).
b. Couples at wave 3 in which either member reported that the relationship was
dissolved, unknown, or missing were dropped. A code of 3 indicated an
ongoing relationship.
c. Couples who were first matched at wave 3 were dropped since new partners
would not have completed the wave 3 and partnership questionnaires (only
baseline). This process resulted in 493 couples. Example data are illustrated in
Table 3.2.
Table 3.2: Example data from the wave 3 couples dataset
Partner 1 Partner 2 Couple ID First matched (wave #)
110116 6101161 1001 1
110167 6101671 1002 1
110311 6103111 1003 1
110420 6104201 1004 1
111883 6118831 1009 1
60


d. The list of couples was transformed from wide to long format so that each row
represented an individual (no duplicates were present). The wave 3 couples
dataset was then merged with the wave 3 individuals dataset using the
respondent ID. Respondents who completed the power statements were
dropped if they were not matched in the wave 3 couples dataset by using an
internal Stata merge variable. Individuals could have completed the power
statements in reference to a past partner or to a partner who never showed up at
wave 3 to participate in the study. This process resulted in 466 couples or 932
individuals. This dataset was labeled as the final wave 3 couples dataset.
Example data are illustrated in Table 3.3.
Table 3.3: Example data from the final wave 3 couples dataset
Respondent ID Couple ID Gender Birth year First wave History of forced sex History of physical abuse
110167 1002 Female 1992 1 0 0
6101671 1002 Male 1989 1 0 0
110311 1003 Female 1993 1 0 0
6103111 1003 Male 1990 1 0 0
e. The data were transformed back into wide format in order to create additional
couple level variables and then back-transformed again into long format to run
the study analyses.
61


4. For the longitudinal couple dataset used to test the hypotheses related to testing uptake,
the final wave 3 couples dataset (N=932 respondents) was merged with questionnaire
data from waves 4, 5, 6, and 7. This dataset was called the uptake dataset. New
respondents who enrolled at each subsequent wave were not included since they were
not present in the final wave 3 couples dataset. Each of the 932 individuals (466
couples) could have up to 5 records (some respondents did not participate at every
subsequent wave). The following steps were taken:
a. The final wave 3 couples dataset was merged with the wave 4 questionnaire
and partnership data. Variables that were collected at each wave, e.g., perceived
likelihood of being infected with HIV, were renamed with a suffix corresponding
to the wave so they were not replaced during the merge (e.g., risk_w3, risk_w4,
etc.). This resulted in a new combined dataset linking the 466 couples from wave
3 with their corresponding data for wave 4. A similar process was completed
using wave 5, 6, and 7 data.
b. In order to get the data from a wide format (e.g., each row contains an
individual) to a long format (e.g., each row contains a wave), the data were
transformed using the variable suffix w for all longitudinal variables. Time
invariant variables (e.g., birth year, gender, etc.) were also carried over into the
uptake dataset. An internal time variable corresponding to each of the seven
waves was also created in this transformation. Note that since the main predictor
variables (perceived risk, power, etc.) from waves 1 and 2 were intentionally not
62


carried over during earlier data merges, the value of time at these waves is
displayed as missing (indicated by a . in Table 3.4) for all respondents.
Example data are illustrated in Table 3.4.
Table 3.4: Example data from the uptake dataset (long format)
Resp ID Couple ID Gender Birth Year Time Perceived HIV Risk
110167 1002 Female 1992 1
110167 1002 Female 1992 2
110167 1002 Female 1992 3 0
110167 1002 Female 1992 4 5
110167 1002 Female 1992 5 5
110167 1002 Female 1992 6 0
110167 1002 Female 1992 7 0
6101671 1002 Male 1989 1
6101671 1002 Male 1989 2
6101671 1002 Male 1989 3 0
6101671 1002 Male 1989 4 0
6101671 1002 Male 1989 5 0
6101671 1002 Male 1989 6 0
6101671 1002 Male 1989 7 0
5. For the dataset used to test the disclosure hypotheses, the final wave 3 couples dataset
(N=932 respondents) was merged with questionnaire data from waves 4 and 5. This
dataset is called the disclosure dataset (long format). The following additional steps
were taken.
a. At wave 5, respondents from the baseline set of respondents (N=932) were
dropped if they did not participate in the study at wave 5 (N=65) and had never
tested for HIVas reported at wave 5 (N=64). This process resulted in 803
63


respondents. Given that the analysis was stratified by gender, respondents were
retained even if their partner did not meet the above criteria (as shown in Table
3.5).
b. Given that there was a time lapse between wave 3 and 5, it was important to
verify that respondents at wave 5 were reporting disclosure information in
relation to their partner from wave 3. For some respondents, their marital status
changed over the period from wave 3 to 5. Of the 803 eligible respondents at
wave 5, 21 respondents had divorced, 6 respondents had separated, 2
respondents had lost their spouses, and 20 respondents became married. Two
additional respondents whose spouses had died were dropped since they could
not possibly report on disclosure to their partner at wave 5.
c. For those who reported being divorced or separated at wave 5, many still
reported that they had disclosed their results to their spouse. I assumed that
those who reported being divorced at wave 5 reported disclosure to a spouse
whom they were married to in wave 3. The chances of divorcing a spouse from
wave 3 and remarrying another by wave 5over a short eight-month period
would be rare. But for those who were newly married by wave 5, it was necessary
to ensure that they had married the same partner from wave 3. Otherwise, a
respondent could have reported on two different partners at wave 3 and 5. For
those 20 respondents, I manually checked the couple database from wave 5 to
verify that the relationship was indeed ongoing. All of these relationships were
64


ongoing at wave 5, indicating that respondents had married their partner from
wave 3with one exception. The couple database indicated that one of the
relationships had dissolved at wave 5, however, the female respondent was not
paired with another partner whom she had supposedly married. This woman and
her partner from wave 3 were dropped from the analysis given the uncertainty
that they may not have been reporting on each other at wave 5. This process
reduced the sample size to 799 (432 women, 367 men). Example data are
illustrated in Table 3.5.
Table 3.5: Example data from the disclosure dataset (long format)
Respid Couple ID Gender HIV status at wave 4 Disclosure to partner at wave 5
6103111 1003 Male 0 0
110420 1004 Female 0 1
6104201 1004 Male 0 0
111956 1010 Female 0 1
6119561 1010 Male 0 0
Measure Development Phase
Design and Rationale
For specific aim la, I re-formulated a measure of relationship power that was
previously developed for the Malawi context during a pilot study in 2009. The pilot data
demonstrated the feasibility of a Malawian relationship power scale, however, more research
was needed to confirm the psychometric properties of the scale using a larger sample with a
65


more diverse set of power experiences (Conroy 2010). I had expected that such a sample
would minimize the impact of random error on reliability estimates (Singleton and Straits
2005) and improve the scales reliability. First, I describe the process used to create the
power measure, starting from the qualitative interviews conducted in 2009.
Power Measure Development
In 2009,1 conducted semi-structured qualitative interviews with 34 coupled men and
women simultaneously, but separately, using trained Malawian research assistants. The
interviews elicited multiple dimensions of relationship power based on Connells (1987)
Theory of Gender and Power (e.g., relationship control, economic dependence, decision-
making dominance, and social norms). The qualitative data were analyzed to create a
preliminary pool of statements on power. Additional details of the semi-structured
interviews can be found later in this chapter.
Face validity was addressed by consulting with academic scholars, Malawian key
informants, and TLT interviewer staff and then the power statements were added, deleted,
or reworded accordingly. The power statements were translated from English to Chichewa
and reverse translated by two separate individuals unfamiliar with the study in order to
ensure sentence meaning was preserved. Items were both positively worded (e.g., My
partner shows that they care about me.) and negatively worded (e.g., My partner punishes
me when he/she is really angry with me.). Cognitive interviews (Tanur 1992) asking
respondents to think aloud as they responded to the power statements were administered
to a separate convenience sample of young adults (n=8) in order to detect comprehension
66


and translation problems. This process resulted in 31 power statements. Response choices
were based on a 4-point Likert scale (4=strongly disagree, 3=disagree, 2=agree, or
1=strongly agree).
The power statements were administered through face-to-face interviews using a small
pilot sample of 254 individuals. The sample was drawn from the same six target villages used
in the qualitative phase. Study participants were between the ages of 18 and 45 years and had
a primary sexual partner. Research assistants started at the village chiefs home, usually
centrally located within the village, and approached every third compound to recruit
respondents. Interviews took place in a quiet, private location usually near the respondents
primary residence.
An initial exploratory factor analysis with oblique rotation was performed to reduce the
set of items down and identify the underlying constructs. Exploratory factor analysis is
appropriate when you have obtained measures on a number of variables and want to identify
the number and nature of the underlying factors that are responsible for the covariation in
the data (Hatcher 1994). A scree plot suggested four meaningful factors so only these factors
were retained. All items receiving a factor loading of less than 0.30 were dropped. The
results showed that four items loaded on the first factor, which was subsequently labeled
autonomy. Four items loaded on a second factor, which was labeled communication.
Four items loaded on the third factor, which was labeled love and trust. Finally, four items
loaded on a fourth factor, which was labeled relationship dominance. Table 3.6 contains
67


the final factor pattern for the preliminary relationship power scale. Refer to Appendix A for
the full wording of the 16 items.
Table 3.6: Final rotated factor pattern for preliminary relationship power scale
Item3 F actor 1: Autonomy Factor 2: C ommunic ation Factor 3: Love and Trust Factor 4: Relationship Dominance
Would never leave 0.48b -0.19 0.11 0.11
In trouble if partner left 0.41 0.02 -0.01 0.23
Would leave if really bad 0.53 0.06 -0.01 0.02
Could find another partner 0.54 0.08 0.17 -0.07
Discuss matters together 0.02 0.39 0.09 0.09
Partner cares about me -0.19 0.35 0.25 0.31
Talk to partner about affair 0.04 0.65 0.00 -0.06
Consult advisors if problems 0.04 0.61 -0.20 -0.06
Helps me with needs -0.17 -0.09 0.43 0.06
Able to initiate sex 0.15 0.06 0.49 0.03
Able to buy expensive items 0.15 -0.07 0.53 -0.06
Have own money 0.11 -0.10 0.44 -0.18
Partner punishes me 0.08 -0.04 -0.12 0.58
Partner chooses relatives side -0.06 0.02 -0.06 0.41
Partner having an affair -0.17 0.02 0.07 0.36
Partner might beat me 0.14 -0.12 -0.01 0.53
Proportion of variance 0.40 0.39 0.32 0.32
a Scale items have been summarized into shorter descriptions for readability. Refer to Appendix
A for actual scale items.
bFactor loadings greater than 0.30 are in bold print.
The 16-item power scale accounted for the majority of variation in responses. The
proportion of variance was determined to be 40%, 39%, 32%, and 32% for the autonomy,
communication, love and trust, and relationship dominance factors, respectively (see Table
3.6). The reliability of the scale was determined by computing Cronbachs alpha for the
entire power scale and for each of the four factors separately (Cronbach and Meehl 1955).
The overall power scale demonstrated moderate reliability of 0.58, which falls within the
68


recommended range of 0.50 to 0.60 for early stages of research (Nunnally 1967). Scale
reliability was also addressed by gender since future researchers may desire to study men and
womens responses to the scale items separately (see Table 3.7). Reliability of the overall
scale was similar for men and women with coefficient alphas of 0.55 and 0.57, respectively.
Reliability estimates were 0.59, 0.59, 0.57, and 0.54 for the autonomy, communication, love
and trust, and relationship dominance subscales, respectively.
Table 3.7 also presents the mean factor scores computed from a range of 1 (strongly
agree) to 4 (strongly disagree) for each of the subscales. Mean factors scores were computed
for the entire sample and by gender. Higher mean factor scores (i.e., more likely to strongly
disagree with scale items) are indicative of higher relationship power. Mean factor scores
were higher for men on all the subscales with the exception of communication, suggesting
that women had higher relationship power in this domain as compared to men.
Table 3.7: Means, standard deviations, and reliability coefficients for the preliminary
relationship power measure
Overall (n=254) Men (n=127) Women (n= 127)
Factor Mean 95% Cl Alpha Mean 95% Cl Alpha Mean 95% Cl Alpha
Dependence/Autonomy 2.22 2.13-2.32 0.59 2.55 2.43-2.68 0.57 1.90 1.78-2.00 0.43
Communication 3.57 3.51-3.63 0.59 3.46 3.39-3.54 0.49 3.68 3.59-3.76 0.66
Love and Trust 2.77 2.68-2.86 0.57 3.16 3.07-3.25 0.28 2.38 2.25-2.51 0.46
Relationship Dominance 2.82 2.73-2.91 0.54 3.03 2.92-3.14 0.47 2.61 2.47-2.75 0.54
Mean refers to the mean value of the factor score. The factor score was computed by taking the sum of the
row (where l=Strongly Agree, 2=Agree, 3=Disagree, and 4=Strongly Disagree for each scale item) divided by
the number of scale items answered for the row. Positively worded items were reverse coded prior to
calculating the row value.
69


The preliminary relationship power scale items were then placed on TLTs wave 3 and
wave 5 partnership surveys. In order to confirm the underlying constructs using a larger,
more representative sample, I conducted a second exploratory factor analysis with oblique
rotation using the final wave 3 couple dataset (N=932 individuals). I applied rotation to
force variables to load more strongly on a given factor, thus making it easier to interpret the
data. Oblique rotation was used since factors are most likely to be correlated with each other,
thus yielding a more accurate representation of the data (Adock 2006; Hatcher 1994).
However, both oblique and orthogonal rotations yielded the same factor structure. Factors
were retained if eigenvalues were greater than one and as suggested by the scree plot. Scale
items were retained if factor loadings were greater than 0.40. I then performed two separate
factor analyses by gender to see if similar factor patterns were present for men and women.
Next, I tested for differences in the final factors and item scores between men and
women using two-group mean comparison / tests. I computed Cronbachs alpha for each
subscale by gender (Cronbach and Meehl 1955). Refer to Figure 3.3 for a visual
representation of the steps taken to develop the relationship power measure.
70


*
$
M
*
5
Conducted semi-structured interviews with 34 men and women
Developed a preliminary pool of power scale items
Conducted cognitive interviews
Revised items and developed a pool of 31 items
Pilot tested the items on a sample of 254 individuals
* Conducted exploratory factor analysis to group and reduce items
Placed final 17 scale items on TLT's wave 3 survey
Conducted second factor analysis using larger sample of 932 individuals
Figure 3.3: The Relationship Power Scale Development Stages
The Final Relationship Power Measure
Three items loaded on a first factor and 4 items loaded on a second factor, which were
named unity (My partner shows they care about me; When I need my
partners assistance, he/she is there to help me; and My partner and I discuss important
matters together) and discordance (If my partner was really angry
with me, he/she might beat me; My partner punishes me when he/she is angry with me;
When I disagree with my partners relatives, my partner chooses their side over mine; My
partner is probably having sex with someone else).
Table 3.8 presents the distribution and descriptive statistics for the relationship power
subscales and corresponding items (total sample). Missing data were negligible. All seven
scale items ranged in value from 1 to 4. For the three unity items, the majority of
71


respondents either agreed or strongly agreed with the statements (upwards of 92%).
Responses were more equally distributed between agreement and disagreement for the
discordance items, particularly for punishes me and might beat me, and sides with
relatives. For the having an affair item, the majority of respondents disagreed or strongly
disagreed with the statement (85.7%).
The final scores for the subscales were calculated using the mean of the items. Higher
scores meant more agreement with the statements. For the unity subscale, a mean of
approximately 3.77 indicates that on average, people either agreed or strongly agreed with
the items. For the discordance subscale, a mean of approximately 1.98 indicates that on
average, people disagreed with the items (meaning less discordance).
Table 3.8 also provides data on normality by subscale and by individual items. Severe
non-normality was defined as > | 3 | for skew or > | 8 | for kurtosis (Kline 2005). Using these
criteria, the mean scores for the unity factor were considered approaching a non-normal
distribution (skewness=-2.28; kurtosis=9.59). The mean scores for discordance were
considered normally distributed (skewness=0.43; kurtosis=2.50).
Table 3.8: Descriptive statistics for the relationship power subscales, TLT wave 3
couples sample (N=932)
Scale itemb Strongly Disagree Response (%)a Disagree Agree Strongly Agree Missing data Mean (SD) Skewness Kurtosis
Unity factor 3.77 (0.41) -2.28 9.59
Cares about me 0.3 1.2 10.1 88.4 0 3.87 (0.40) -3.4 16.49
Helps me 1.9 6.1 19.3 72.6 0 3.63 (0.69) -1.91 6.24
Discuss together 0.4 2.0 14.4 83.2 0 3.80 (0.47) -2.64 10.69
Discordance factor 1.98 (0.71) 0.43 2.51
Punishes me 33.8 24.1 22.1 20.0 0 2.28 (1.13) 0.26 1.66
Might beat me 50.6 19.6 14.8 15.0 1 1.94 (1.12) 0.76 2.09
Sides with relatives 37.6 31.7 16.0 14.4 3 2.08 (1.05) 0.59 2.1
Having an affair 58.9 26.8 8.6 5.7 0 1.61 (0.87) 1.37 4.01
a 1Strongly Disgree, 2Disagree, 3Agree, 4Strongly Agree. b Refer to Appendix A for full wording of the items.
72


Table 3.9 presents the standardized factor loadings and reliability coefficients for the
subscales and items. For the total sample, factor loadings for unity items ranged from 0.60 to
0.62, with women loading higher on the subscale than men. The discordance item factor
loadings were generally lower than those for unity and ranged from 0.52 to 0.56with the
exception of the having an affair item that had a loading of approximately 0.37. For all
four discordance items, mens loadings were slightly higher than womens loadings.
These differences in loadings were also reflected in the reliability coefficients of the
subscales for men and women. The Cronbachs alpha for unity was 0.65 (women: 0.74; men:
0.53) and for discordance was 0.60 (women: 0.56; men: 0.64). While the latter three reliability
coefficients fell below the recommended cutoff of 0.70 (Nunnally and Bernstein 1994), they
still exhibited normal reliability values for early stage research (Nunnally 1967). Table 3.9
contains the factor loadings and reliability coefficients for each subscale.
Table 3.9: Standardized factor loadings and coefficient alphas for the power scale
items
Scale item Total Unitv Women Men Total Discordance Women Men
Cares about me 0.62 0.70 0.51
Flelps me 0.61 0.71 0.49
Discuss together 0.60 0.63 0.59
Punishes me 0.56 0.52 0.63
Might beat me 0.53 0.52 0.53
Sides with relatives 0.52 0.51 0.55
Flaving an affair 0.37 0.32 0.38
Coefficient alpha 0.65 0.74 0.53 0.60 0.56 0.64
Refer to Appendix A for full wording of scale
73


Gender differences were examined for the final two factors and scale items (range: 1-4).
Table 3.10 presents the mean scores, standard deviations, and statistical differences in the
measures by gender. The mean score for unity was almost exactly the same for men and
women (3.77). Two-group / tests revealed that men and womens responses were not
statistically different from each other for the unity subscale or for the individual unity items.
For the discordance subscale, higher scores meant more discordance (less power). Mean
scores for the discordance factor were higher for women (2.12) than men (1.83) suggesting
that women were more likely to experience discordance in their relationships. Two-group /
tests revealed that men and womens responses were statistically different from each other
for the discordance subscale or for two of the discordance statements: If my partner were
angry with me, he/she might beat me and My partner is probably having an affair. Figure
3.4 illustrates the iterative process used to develop the final model of relationship power that
was used in the hypothesis-testing phase.
Table 3.10: Gender differences in relationship power factors
Scale itenT,b Total Women Men /test
Mean (SD) Mean (SD) Mean (SD) p value
Unity factor 3.77 (0.41) 3.77 (0.45) 3.77 (0.37) 0.937
Cares about me 3.87 (0.40) 3.86 (0.44) 3.88 (0.35) 0.460
Helps me 3.63 (0.69) 3.65 (0.67) 3.60 (0.70) 0.253
Discuss together 3.80 (0.47) 3.78 (0.51) 3.82 (0.43) 0.214
Discordance factor 1.98 (0.71) 2.12 (0.72) 1.83 (0.66) 0.000
Punishes me 2.28 (1.13) 2.35 (1.15) 2.22 (1.11) 0.088
Might beat me 1.94 (1.12) 2.32 (1.19) 1.56 (0.89) 0.000
Sides with relatives 2.08 (1.05) 2.07 (1.10) 2.08 (1.00) 0.924
Having affair 1.61 (0.87) 1.74 (0.91) 1.48 (0.80) 0.000
11Strongly Disgree, 2Disagree, 3Agree, 4Strongly Agree. b Refer to Appendix A
for full wording of the items.
74


In the hypothesis-testing phase that follows, I ultimately decided to drop the discordance
factor in the statistical analyses. This was for several reasons. First, there appears to be two
or more underlying constructs being measured in the discordance scale, including couple
disagreement or conflict (or the punitive aspects of power such as being able to beat or
punish a partner) and perceptions of a partners infidelity. The item related to perceptions of
a partners infidelity had significantly lower factor loadings than the other three items and
was conceptually distinct from the other three items. Hatcher (1994) argues that in order for
a factor to be retained, one of the criteria is that all the variables share a conceptual meaning.
Second, in my experience trying to publish the results of the scale, I learned that acceptable
minimum levels of scale reliability for high quality journals range from 0.65-0.70. While the
discordance subscale reliability was not terrible, it was not ideal either. Rather than pursing a
path that I will eventually deviate from when trying to publish selections from this
dissertation, I chose to examine several of the underlying constructs as single-item measures,
particularly, physical violence and perceived partner infidelity (as will be discussed). The
findings above related to the discordance subscale were intentionally documented herein to
demonstrate the process of creating a new scaleand the decisions that needed to be made
along the wayin order to provide other researchers with a possible starting point for a
measure of discordance for the Malawi context.
75


A Priori Model of Relationship Power Model of Relationship Power After
using the TGP Preliminary Factor Analysis
Model of Relationship Power After
Final Factor Analysis
Relationship Relationship
Violence Dominance
Economic
inequality
Figure 3.4: Iterative Stages of Development for the Relationship Power Model


The Hypothesis-Testing Phase
Design and Rationale
For specific aim 2, the objective was to test whether relationship power influences the
use of HIV testing services and disclosure of test results to primary partnersin accordance
with the study hypotheses documented in Chapter 2. My initial intention was to conduct a
dyadic analysis for the hypothesis-testing phase of this dissertation. As I started to analyze
the quantitative TLT data, preliminary findings changed my original analysis approach. I first
computed the Intraclass Correlation (ICC) for the unity and discordance measures, which is
the first step in a dyadic analysis to evaluate non-independence or the degree of similarity
between two members of a dyad on the same variable (Kenny, Kashy, and Cook 2006). The
ICC was computed through a large one-way ANOVA using a unique couple identifier as the
grouping variable. The ICC for unity showed that only 9% of the variance in unity was
explained by the dyad, F(465, 466) = 1.20, ^=0.025. The ICC for discordance showed that
only 6% of the variance in discordance was explained by the dyad, F(465, 466) = 1.12,y=0.12.
What do ICCs tell us? The idea behind evaluating a measures ICC is to assess how
similar two dyad members responses are to the same measurea measure that is
hypothesized to involve an interaction between the two people. With unity, for example, I
would expect that if a wife reports higher levels of communication, her husband would also
be more likely to report this, though certainly it would not be extraordinary for couple
members to report divergent responses. For discordance, which includes perceptions of a
partners infidelity and physical violence, I would expect that if couples had high levels of
77


reciprocal violence and mutual mistrust, they would provide similar responses to each other.
However, the ICC for discordance showed that couple members responses were even more
independent than for unity. Reciprocal violence only occurred in 1% of couples (or in 5.4%
of all physically violent couples). Similarly, only 3% of all couples both reported that they
believed there partners were having an affair. Thus, the majority of the time if one partner
was suspected of cheating, the other partner was not suspected of the same behavior.
The ICC values can be used to inform a particular analytic approach. Data with high
levels of non-independence would be best suited for a dyadic-level analysis. According to
Kenny and colleagues (2006), a dyadic level analysis controls for the non-independent nature
of two individuals providing more similar information on one measure than two individuals
who are not part of the same couple. Individual-level data that are actually hierarchical such
as data from dyads may bias estimates when non-independence is present (Kenny, Kashy,
and Cook 2006).
Another way to conceptualize the ICC is as a proportion of the total variance in a
measure that is explained by the dyad. For unity, only 9% of the total variance (within
couples and between couples) is explained by the dyadthe ICC for discordance is even less
(and also non-significant). There is generally a lack of agreement about how much non-
independence is enough to warrant a dyadic level analysis. Cohen (1998) defines 0.50 as a
large ICC, 0.30 as a medium ICC, and 0.10 as a small ICC. Dagne et al. (2002) argues that if
there is little variance between couples and the ICCs are close to zero, then it is possible to
examine the data at the individual level.
78


For the hypothesis-testing phase of this dissertation, I chose to analyze the data at the
individual level rather than at the dyadic level given that the couple dataat least the main
predictor variables for this study-were largely independent. Multiple possible reasons may
explain why the ICCs were found to be so low. One reason may relate to statistics. Given
that the majority of the sample agreed or strongly agreed with the unity statements, there
may not have been enough variation overall to show differences between couples. In other
words, if most people in the entire sample reported that they agree with a given statement
on a 4-point Likert scale, then the ICC could be low because the two dyad members
responses were not more similar than any other two people in the entire sample.
Now that I have outlined the rationale for analyzing the data at the individual level, I will
justify the general approach I used to testing the hypotheses related to testing uptake and
disclosure. For the HIV testing uptake models, I used longitudinal testing history data
collected over a 16-month period (wave 4, 5, 6, and 7). This allowed me to follow
respondents over time to see if relationship power, measured at wave 3, had an effect on
testing over each 4-month interval. Longitudinal analyses are generally thought to be
superior to cross-sectional analyses. Cross-sectional studies are inherently limited at
demonstrating causal linkages since directionality of cause and effect cannot be evaluated
(Singleton and Straits 2005). If the data were captured at the same time, it would be difficult
to determine whether having an HIV test changed power or whether power changed after
getting tested. The background literature presented in Chapter 2 suggests that both
directions are possible.
79


For the disclosure models, I used the power measures captured at wave 3 to predict
disclosure at wave 5 in order to help establish temporality and widen the time gap between
when power and disclosure was ascertained. For both analyses (HIV testing uptake and
disclosure), I stratified the analysis by gender because I perceived that the pathways would
differ between men and women. Though the data are largely independent, gender
stratification provided an additional level of protection against any bias from non-
independent data.
Sample Selection
Two separate datasets were used to test the hypotheses related to: 1) HIV testing uptake;
and 2) disclosure to primary partners. For the HIV testing uptake models, I used the baseline
sample of couples from wave 3 who completed the power module questions (N=932). Here,
I refer back to the uptake dataset discussed earlier. For several reasons, I chose to use the
couple dataset instead of a cross-sectional dataset of all respondents who completed the
power module. First, use of the couple dataset ensured that only legitimate and serious
couples answered the power questions about their partner, potentially increasing the validity
of the data (e.g., respondents were not reporting on a phantom partner). Second, because I
used data from both couple members as predictors, it was necessary to link the couple
members to determine their partners information.
The baseline sample was followed from wave 3 to 7 in order to obtain HIV testing
histories over a 16 month time period. Limiting the analysis to only TLTs group 3 (refer to
Figure 3.2)who had no testing until the end of the studywould have minimized the
80


effects of TLT testing on regular testing behavior; however, this would have dramatically
reduced the sample size, which was already halved through use of gender stratification.
Therefore, I used all three TLT study groups in order to make use of a larger sample. Even
though group 1 received an HIV test through TLT at each wave and group 2 received an
HIV test at wave 4,1 still expected that respondents could have obtained an HIV test
outside of the TLT study. Respondents were not told when they would be tested and thus if
they desired to learn their status, knowledge of the TLT testing schedule may have had little
impact on the decision to wait to be tested at TLT. I also controlled for cumulative number
of previous TLT tests at each wave in order to account for any differences between the
regular TLT testers (group 1), the occasional testers (group 2), and the non-testers (group 3).
For HIV test disclosure, I started with the baseline set of couples identified at wave 3
(N=932). Here, I refer back to the disclosure dataset discussed earlier. Respondents were
not eligible for the analysis if they were never tested for HIV or if they did not participate in
wave 5. When attempting to understand factors that influence disclosure of HIV test results,
it was important to consider HIV status in the analysis since it had been noted as a key
predictor of disclosure in the literature. At wave 4, approximately two-thirds of the entire
TLT sample received an HIV test as part of TLT, allowing me to know the HIV status of
many respondents. At wave 5, the TLT questionnaires asked respondents about whether
they shared the results of their last HIV test with spouses or sexual partners. Thus, it was
possible to determine whether those tested through TLT at wave 4 disclosed their test
results by wave 5. Since group 3 tested outside of TLT (if they tested at all), I could compare
81


the rates and predictors of disclosure between groups 1 and 2 to determine if the same
associations held true.
Measures
Table 3.11 outlines the format and interpretation for each of the main predictor variables
(minus control variables). Table 3.12 contains a summary of the all independent and
dependent variables used in the two statistical models for HIV testing uptake and disclosure.
Socio-Demographic Characteristics
Several individual-level variables were included in all multivariate models as statistical
controls: age, years of education, household economic status, and marital status. Age and
years of education were modeled as continuous variables. Up to and including 8 years of
education was considered primary school, 9 to 12 years was considered secondary school,
and greater than 12 years was considered tertiary school. An index of 9 common household
goods (bicycle, television, bed with mattress, radio, land line/mobile phone, motorcycle,
animal-drawn cart, car/truck, or Bible/Koran) was used to approximate household
economic status.
At wave 3, respondents had to have an ongoing sexual relationship to be included in the
couple dataset. However, over the 16-month period, it was expected that for some, their
marital status would change. For the HIV testing uptake models, marital status was
considered a 4-level categorical variable consisting of the following states:
married/cohabitating, separated/divorced, widowed, or unmarried. Marital status was
considered a time-varying predictor and included at each wave.
82


For the disclosure models, a binary variable was used to capture marital status
(married/cohabitating or unmarried). I expected that married couples would be more likely
to share test results due to emotional closeness and difficulties associated with hiding test
results from their spouses. I created a hybrid measure of marital status (married or
unmarried) using the marital status measures from waves 3 and 5. Respondents were still
considered married even if they had reported separating or divorcing at wave 5 since most
reported that they had disclosed to their spouse in the disclosure questionthus, reflecting a
ongoing connection with their ex-spouse. Respondents whose marital status had changed
from unmarried at wave 3 to married at wave 5 were considered married in the analysis.
Relationship Characteristics
Relationship duration and shared children were included for descriptive purposes, but
not in the analytical models. Relationship duration was computed by subtracting the date of
the survey from the date the respondent first started spending time with their partner. For
shared children, respondents were asked how many living children they have with their
partner at wave 1. Given that the sample was young and the mean number of shared
children was around 1, a binary variable was created to indicate whether a couple had at least
one child together. Discrepancies were sometimes noted between couple members account
of relationship duration and whether or not the couple had children together. For
relationship duration, the average between female and male partners reports was calculated
and replaced individual reports. For shared children, the female couple members report was
used, with the assumption that the mothers report might be more accurate.
83


Relationship Power
As shown in Figure 3.4, five variables or constructs were conceptualized as relationship
power: 1) socio-economic inequalities, 2) relationship dominance, 3) relationship violence, 4)
unity, and 5) discordance. As previously discussed, the discordance construct was re-
formulated as single item measures of relationship violence and mistrust/suspected partner
infidelity.
Socio-economic inequality. Three variables were created to measure socio-economic
inequality between partners: age inequality, education inequality, and employment inequaliy. Given
that men normally marry when they are on average of 3 years older than women (MDHS,
2011), I considered an age gap of 5 to be a meaningful measure of age inequality (National
Statistical Office & ORC Macro 2011). Thus, age inequaliy was captured as a binary variable
where 0 referred to less than or equal to 5 years age difference and 1 referred to greater than
5 years difference. Education inequaliy was captured as a three-level categorical variable where
0 referred to equivalency in education for partners, 1 referred to higher male education, and
2 referred to higher female education. Respondents were asked to specify their occupation
and then asked if the work was piecework, temporary employment, or a steady job. Women
who specified that their occupation was housewife were not asked the second question on
employment type. Using the responses for women and their partners, I created a four-level
categorical variable for employment inequaliy where 0 referred to both unemployed, 1 referred
to man employed, woman unemployed, and 2 referred to woman employed, man
unemployed, and 3 referred to both employed.
84


Relationship dominance. Relationship dominance was measured by asking respondents: In
your relationship, who would you say is generally in charge? with answer choices
respondent, equal control, or partner. Since less than 2% of women and less than 1% of men
responded that their relationship was female-dominated, a binary variable was created where
0 referred to egalitarian or female-dominated and 1 referred to male-dominated.
Relationship unity. As previously discussed, the factor analysis resulted in two subscales
named unity and discordance. The three-item unity subscale was included in all statistical
models (My partner shows they care about me; When I need my
partners assistance, he/she is there to help me; and My partner and I discuss important
matters together). TLT interviewers asked respondents whether they strongly agreed (1),
agreed (2), disagreed (3), or strongly disagreed (4) with these statements. Responses were
reverse scored so that higher mean scores meant more unity in the relationship. The
discordance subscale was dropped for lower reliability and other reasons, but re-formulated
using three single-item measures: relationship violence (physicaland sexual) and mistrust/perceived
partner infidelity.
Relationship violence (.sexual andphysical). Respondents were asked if they were victims of
sexual and physical violence (but not if they were perpetrators) in relation to the reference
partner noted in TLTs power module. The format and wording of these two measures came
from the study conducted by Pulerwitz and colleagues (2000), who used these measures to
assess construct validity for the original Sexual Relationship Power Scale (SRPS). Sexual
85


Full Text
C. Relationship violence (division of power): Having a history of relationship violence (physical
and sexual) is a proxy for fear of abuse, which has been shown to be a barrier to testing.
Individuals in violent relationships will therefore be less likely to test for HIV. Having a
history of violence (sexual or physical) could also operate through the pathway of risk,
thereby decreasing the likelihood of testing (assuming people are higher risk are less
likely to test).
D. Unity (cathexis/social norms). Unity will be positively associated with HIV testing uptake
among men and women. Aspects of unity such as communication, reciprocity, and love
will foster a more supportive environment for couples to discuss testing and get tested.
Therefore, individuals in relationships with higher levels of unity will be more likely to
test for HIV.
E. Mistrust/partner infidelity. This variable emerged as an underlying construct of the
discordance subscale. In Chapter 4, trust was perceived as an important social norm
related to unity (cathexis); thus more trust, higher power. As such, mistrust (lack of trust)
will be negatively associated with the act of testing.
F. Perceived risk: Perceived risk of self and partner will be negatively associated with getting
tested for HIV. For both men and women, those who believe they are at higher risk for
HIV will fear being blamed for infidelity and therefore less likely to test.
174



PAGE 1

GENDER, RELATIONSHIP POWER, AND HIV TESTING IN RURAL MALAWI by AMY ANNE CONROY B.S.E., University of Iowa, 2002 M.P.H., University of Colorado, 2008 A thesis submitted to the Faculty of the Graduate School of the University of Colorado i n partial fulfillment of the requirements for the degree of Doctor of Philosophy Health and Behavioral Sciences 2013

PAGE 2

! 2013 AMY ANNE CONROY ALL RIGHTS RESERVED

PAGE 3

ii This thesis for the Doctor of Philosophy degree by Amy Anne Conroy has been approved for the Health and Behavioral Sciences Program b y Sheana S. Bull Dissertation Chair Sara E. Yeatman Examination Chair Jean N. Scandlyn Jennifer J. Harman June 7, 2013

PAGE 4

iii Conroy, Amy Anne (Ph.D., Health and Behavioral Sciences) Gender Relationship P ower, and HIV Testing in Rural Malawi Thesis directed by Professor Sheana S. Bull ABSTRACT This dissertation considers how relationship power shapes men and women's decisions around HIV testing, which is the gateway to HIV/AIDS preven tion, treatment, and care in sub Saharan Africa. The study is situated in a high HIV prevalence community in southern Malawi where rapid expansions in HIV testing services provide new opportunities to learn one's HIV status. I focus on two public health co ncerns around HIV testing that have been inadequately studied at the relationship level: uptake of HIV testing services and disclosure to primary partners. To accomplish this, I use an iterative, mixed methods design to: 1) develop a model of relationship power for Malawi, 2) test for associations between power and HIV testing behavior, and 3) explain, validate, or challenge these findings using qualitative data. Quantitative data on 466 couples come from Tsogolo La Thanzi a study on reproduction and AIDS in southern Malawi. Qualita tive data consist of 34 semi structured interviews and 8 focus group discussions. Using the three social structures of the Theory of Gender and Power labor, power, and cathexis I develop and test a preliminary model of relationsh ip power consisting of three domains: economic power, relationship violence, and relationship dominance. Factor analysis revealed two additional constructs of power related to unity and discordance which were validated as important using qualitative narra tives. Next, I use this modified model of relationship power to test hypotheses around uptake and disclosure Economic power, relationship dominance, unity, and violence emerged as important predictors of testing uptake for men and women. In the disclosure models, economic power, violence, and unity were predictive. Contrary to my hypothesis, young people with higher levels of unity were less likely to test for HIV. This finding was explained by the widespread belief that HIV testing was reserved for times of trouble, such as infidelity, rather than for relationships filled with unity, love, and trust. Study conclusions highlight the need to facilitate theoretical and applied approaches to HIV testing that encompass the couple context including aspects of un ity. Universal HIV testing may provide a promising solution for couples to circumvent exceptionally difficult negotiations by deferring decisions to healthcare providers.

PAGE 5

iv The form and content of this abstract are approved. I recommend its publication. Approved: Sheana S. Bull

PAGE 6

v DEDICATION I dedicate this thesis to the Malawians who suffer unnecess arily from AIDS and to others who live in a state of uncertainty about whether they will become infected with HIV Your warm hearts incredible strength and resilience and receptiveness to my research has given me the fortitude to complete this dissert ation and dedicate my professional life to AIDS intervention research in Africa I can only hope that one day my c areer will be forced to go in a dramatica lly different direction after a n effective cure or vaccine for AIDS is discov ered and made accessible to all.

PAGE 7

vi ACKNOWLEGEMENT First and foremost I am deeply appreciative for a number of people both locally and ab road who supported me during this endeavor For guidance on this dissertation, I am indebted to my committee, Sheana Bull, Sara Yeatman, Jean Scandlyn and Jennifer Harman for their valuable feedback and support throughout the process I would like to extend a special thanks to Sara Yeatman for in viting me to Malawi back in 2009 and allowing me to participate in the Tsogolo La Thanzi (TLT) study I n addition, I would like to acknowledge my fellow students in the Health and Behavioral Sciences program, Kate Dovel, Stacey McKenna, Deanna McQuillan, C hristine Tagliaferri Rael, and Jerry Wulff, f or picking me up during times of frustration and providing words of encouragement to keep me movin g towards my goals I n Malawi I thank the dedicated research assistants who tirelessly collected the semi struct ured interview and focus group data for this study: Caroline Augustine, Steven Kabvinga, Chisomo Kalogwile, Andrew Kanjirawaya, Zahra Khan, and Thandizo Shaba. A special thanks goes out to three TLT staff members in particular, who supported me both logis tically and through their friendships during two fieldwork trips to Malawi: Hazel Namadingo, Abdallah Chilungo, and Sydney Lungu Finally, this dissertation would not have been possible without the love, support, and encouragement from my parent s, Thomas a nd Christine Conroy, my husband, TJ Moretto and my daugh t er Chiara who accompanied me to rural Malawi when she was only 6 months old. To all of you, I express my deep est gratitude for your patience and for believing in me every step along the way. Especia lly, t hank you Chiara, for being my "bright light" showing me what is most

PAGE 8

vii important in life and challenging me everyday to be the best mother and scholar I can be I look forward to sharing with you my love of learning and committment to a better global world. Finally I would like to acknowledge the generous funding sources that made this dissertation possible. Data collected through the TLT study was funded by grants (R01 HD058366 and R03 HD067099) from the National Institute of Child Health and Human Development (NICHD) My time and research expenses were supported through a Ruth Kirschstein pre doctoral fellowship from the National Institute of Mental Health (NIMH) (F31 MH093260), the Robinson Durst Scholarship through the University of Colorado Denv er Center for Global Health, and the Henry David Research Grant through the American Psychological Foundation. To all of you, I say Zikomo Kwambiri.

PAGE 9

viii TABLE OF CONTENTS CHAPTER I. INTRODUCTION ................................ ................................ ................................ .. 1 Research Aims ................................ ................................ ................................ ... 2 The Importance of Couple Context ................................ ............................... 4 The Importance of Relationship Power ................................ ........................ 6 Summary of Research Design ................................ ................................ ......... 8 Implications for Policy, Prevention, and Public Health .............................. 9 Chapter Outline ................................ ................................ ............................... 10 II. BACKGROUND AND CONCEPTUAL FRAMEWORK ........................... 12 Gender, Relationship Power, and HIV/AIDS ................................ ........... 12 The Theory of Gender and Power ................................ ............................... 14 HIV Prevention and Testing in Sub Saharan Africa ................................ 18 The Social Ecological Model of Health ................................ ....................... 21 Symbolic Interactionism and HIV Testing Behavior ................................ 35 Study Hypotheses ................................ ................................ ............................ 37 Study Innovation ................................ ................................ ............................. 40 The Malawi Context ................................ ................................ ........................ 42 III. METHODOLOGY ................................ ................................ ................................ 53 Study D esign Overview ................................ ................................ .................. 53 The TLT Study ................................ ................................ ................................ 56

PAGE 10

ix Creation of the Analysis Datasets ................................ ................................ 58 Measure Development Phase ................................ ................................ ........ 65 The Hypothesis Testing Phase ................................ ................................ ...... 77 The Interpretative Phase ................................ ................................ ................ 98 Ethical Considerations ................................ ................................ .................. 114 IV. CONSTRUCTIONS OF GENDER AND POWER: TRADITION, UNITY, AND RIGHTS ................................ ................................ ...................... 117 Tradition, Un ity, and Rights ................................ ................................ ........ 118 Cathexis: Gender and Relationship Ideals ................................ ................. 121 The Sexual Division of Labor ................................ ................................ ..... 129 The Sexual Division of Power ................................ ................................ ..... 139 The I nfluence of Gender Policy on Relationship Power ........................ 147 Discussion ................................ ................................ ................................ ...... 160 V. THE ASSOCIATION BETWEEN RELATIONSHIP POWER AND HIV TESTING BEHAVIOR ................................ ................................ ............. 170 Conceptual Framework and Hypotheses ................................ .................. 170 Summary of Analysis Approach ................................ ................................ 176 Characteristics of the Couple Sample ................................ ........................ 180 Gender Differences in Relationship Factors ................................ ............ 184 HIV Testing Histories ................................ ................................ .................. 187 Predictors of HIV Testing Uptake ................................ ............................. 190 Disclosure of HIV Test Results ................................ ................................ .. 197

PAGE 11

x Reliability of Disclosure Reports ................................ ................................ 200 Predictors of HIV Status Disclosure ................................ .......................... 203 Discussion ................................ ................................ ................................ ...... 208 Limitations ................................ ................................ ................................ ...... 218 VI. "WHEN THERE IS DOUBT IN THE HOUSE, THAT'S WHEN YOU GO FOR TESTING": HIV TESTING PERCEPTIONS AND EXPERIENCES ................................ ................................ ................................ ... 221 Summary of Research Methods ................................ ................................ .. 221 The Symbolic Meaning of an HIV Test ................................ .................... 222 The Relationship Ideal of Testing before Marriage' ............................... 224 The Incompatibility of Testing with Marriage ................................ .......... 227 Risk, Infidelity, and HIV Testing ................................ ................................ 230 Tradition, Rights, and Unity ................................ ................................ ........ 233 Testing Together as a Workaround Strategy ................................ ............. 238 Antenatal Testing as an Alternative to Couples VCT ............................. 245 Relationship Dissolution as a Fallback Strategy ................................ ....... 248 Discussion ................................ ................................ ................................ ...... 250 Public Health Implications ................................ ................................ .......... 253 VII. IMPLICATIONS AND FUTURE DIRECTIONS ................................ ....... 259 Theoretical and Policy Implications ................................ ........................... 259 Final Conceptual Model for Relationship Power and HIV T esting ...... 262 Future Directions ................................ ................................ .......................... 268

PAGE 12

xi Conclusions ................................ ................................ ................................ .... 270 REFERENCES ................................ ................................ ................................ ................... 272 APPENDIX A. Relationship Power Scale Items ................................ ................................ ......... 292 B. Semi Structured Interview Guide ................................ ................................ ...... 294 C. Focus Group Discussion Guide ................................ ................................ ......... 298

PAGE 13

xii LIST OF FIGURES FIGURE 2.1 Example "Let's Go" Slogan Promoting HCT ................................ ...... 20 2.2 Conceptual Model for HIV Testing Behavior Hypotheses ................ 38 2.3 Balaka District of Southern Malawi ................................ ........................ 44 3.1 Timeline of Data Collection and Analysis, 2009 2013 ......................... 56 3.2 Overview of the VCT Infrastructure for TLT ................................ ...... 58 3.3 The Relationship Power Scale Development Stages ............................ 71 3.4 Iterative Stages of Development for the Relationship Power Model ................................ ................................ ................................ ........... 7 6 4.1 Re Conceptualizing the Connell's (1987) Theory of Gender and Power ................................ ................................ ................................ ......... 120 5.1 Conceptual Model of Relationship Power based on the Modified TGP ................................ ................................ ................................ ........... 172 5.2 Conceptual Framework for Main Predictor Variables and HIV Testing Behavior ................................ ................................ ...................... 173 5.3 Percen t of TLT Respondents Who Had Ever Tested for HIV at Baseline ................................ ................................ ................................ ...... 187 5.4 TLT Respondents Who Had a New HIV Test Since the Previous Wave ................................ ................................ ................................ .......... 189 5.5 Cumulative Number of New HIV Tests from Waves 4 to 7 ........... 190 7.1 Final Conceptual Model on Relationship Power and HIV Testing Behavior ................................ ................................ ................................ .... 263

PAGE 14

xiii LIST OF TABLES TABLE 3.1 Example data from the wave 3 individuals datas et .............................. 59 3 .2 Example data from the wave 3 couples dataset ................................ .... 60 3.3 Example data from the final wave 3 couples dataset ........................... 61 3.4 Example data from the uptake dataset (long fo rmat) .......................... 63 3.5 Example data from the disclosure dataset (long format) .................... 65 3.6 Final rotated factor pattern for preliminary relationship power scale ................................ ................................ ................................ .............. 68 3.7 Means, stan dard deviations, and reliability coefficients for the preliminary relationship power measure ................................ ................ 69 3.8 Descriptive statistics for the relationship power subscales, TLT wave 3 couples sample (N=932) ................................ ............................. 7 2 3.9 Standardized factor loadings and coefficient alphas for the power scale items ................................ ................................ ................................ ... 73 3.10 Gender differences in relationship power factors ................................ 74 3.11 Format and interpretation of the main predictor variables ................. 89 3.12 Independent and dependent variables for the two statistical models of uptake and disclosure ................................ ................................ ........... 94 3.13 Mean value of unity at TLT's wave 3 and 5 ................................ .......... 97 3.14 Characteristics of the focus group respondents ................................ .. 110 4.1 Focus group respondents' perceptions of ideal relationships ........... 128 4 .2 What makes people feel powerful in their relationships? .................. 142

PAGE 15

xiv 5.1 Independent and dependent variables for the two statistical models of uptake and disclosure ................................ ................................ ......... 178 5.2 Selected characteristics of the baseline sample of couples, TLT wave 3 ................................ ................................ ................................ ........ 182 5.3 Couple HIV status, TLT wave 4 ................................ ........................... 186 5.4 HIV testing history by gender, TLT waves 3 7 ................................ ... 189 5.5 Odds ratios predicting a new HIV test among women and men, TLT waves 4 7 ................................ ................................ ......................... 196 5.6 Disclosure of last HIV test to main partner, TLT wave 5 ................ 198 5.7 Do respondents really disclose and do their partners believe their reports? ................................ ................................ ................................ ...... 202 5.8 Are respondents told the correct test result by their partners? ........ 203 5.9 Odds ratios from logistic reg ression models predicting disclosure among women and men, TLT wave 5 ................................ .................. 207 7.1 Theoretical assertions and supporting evidence found in quantitative data ................................ ................................ ....................... 266

PAGE 16

1 CHAPTER I INTRODUCTION In a dusty African village on the outsk irts of the Balaka town in southern Malawi, Caroline interviews a 23 year old married woman named Ruth about her relationship history As the interview unfolds the two women begin to talk about HIV testing, a salient topic of discussion in a setting where HIV prevalence ranks among the highest in the world Caroline asks Ruth "So when you went for HIV testing, did you tell husband or did you just go?" Ruth answers, "I told him that there was no reason to be afraid of knowing how your blood is but he refus ed to go, so I went by myself. He was saying that I was doubting myself." When Caroline asks Ruth if she continues to plead with her husband to go for testing or if she gave up trying, she responds: No, I didn't give up, I always tell him. The fir st time I went I told him that you should also go so that we can have proof that we are alright, but if you don 't then I will be having doubts Maybe my husband is positive or negative. I will not be sure. The second time I went I told him the same, but to no av ail. He says that he can't go and that those who go for testing are doubting themselves. They don't trust themselves. It pains me that he is not helpin g me and the children. (semi structured interview, female #9) I present Ruth's narrative not to bolster support for the powerless woman depiction that is common in gender research and policy on AIDS in Africa (Esacov e 2013; Higgins, Hoffman, and Dworkin 2010) but instead to illustrate Ruth's tremendous agency in her persistent attempts to change her husband's position on HIV testing Unlike Ruth's situation, not all Malawian women are as forthcoming and relentless w ith their partners about testing. Nor do all husbands refuse to get tested with their wives. Her story does,

PAGE 17

2 however, exemplify one of many different ways that wives and husbands negotiate HIV testing with each other to learn their H IV status. Research Aims This dissertation seeks to better understand the relationship context around HIV testing behavior, particularly decisions to get tested and disclose test results to spouses or primary sexual partners. Such an examination of heter osexual partnerships requires attention to the everyday relationship dynamics within young couples. In this dissertation, I examine the pathways and mechanisms through which relationship factors, particularly power, influence s HIV testing behavior among yo ung couples from rural Malawi. Three specific aims will be addressed with this research: (1) To explore quantitative measures of relationship power for the Malawi context (a) and to evaluate their cultural relevance with qualitative data (b) (2) To test whether relationship power influence s two ty pes of HIV testing behavior: uptake of HIV testing services (a) and disclosure of test results to primary partners (b) (3) To understand the meaning of HIV testing within the context of a sexual relationship. To my surp rise, I find that what unites the numbers and narrative in this dissertation is the underlying construct of unity : that is, notions of love, trust, intimacy, communication, and reciprocity. At first glance, the idea of unity appear s antithetical to the ter m power which has been historically conceptualized as an imbalance between men and women But to the

PAGE 18

3 co ntrary, rural Malawians believe that these relationship qualities provide a source of power for them in their relationships. This raises the question, i f one has power or unity in their relationship, is it still necessary to get tested for HIV? The qualitative data from this dissertation show that the HIV test does not simply provide a medical diagnosis; instead, people construct meaning around the act of testing and the test result itself within the context of their everyday relationships. At times, th is addition of symboli sm around testing conflicts with how HIV testing services were designed to function from a public health perspective: to detect HIV in fection as earliest as possible in the disease course. This idea resonates in Ruth's narrative above. She tries to convince her husband about the diagnostic benefits of HIV testing, i.e., to "learn how their bodies are ," but her husband claims that HIV te sting is for those who suspect they are HIV positive or have engaged in promiscuous sexual behavior as opposed to those who are faithful to their partner and practice safe sex. According to her side of the story, both Ruth's desire to test and her husband' s refusal make both members question the other's HIV status and sexual exclusivity as expressed with the euphemism "having doubt ." Here, an HIV test becomes a symbol of something larger and perhaps more important to some people; it becomes a marker of the ir relationship status rather than just their HIV serostatus Thus, an HIV test signifies the absence of unity in the relationship a characteristic that many couples idealize and strive for in their partnerships.

PAGE 19

4 The Importance of Couple Context Sexual relationships comprise the social fabric of life in rural Malawi. They are the intimate setting where people get married and children are born, which in turn bring s tremendous social status and fulfillment to both men and women. Scholars have even stated t hat in sub Saharan Africa, "marriage remains one of the most important individual aspirations and social duties that an individual ever expe riences" (Smith and Mbakwem 2007) In a very poor setting complicated by seasonal famines and high unemployment rates, relationshi ps are not only critical for social identity, but for survival. The gendered social norms dictating the division of labor prescribe how husbands and wives are supposed to behave and contribute to the household well being and it is the economic and social s upport received through these arrangements that maintains people's existence and livelihoods. Ye t in sub Saharan Africa, these ongoing heterosexual relationships are the place where most new HIV infections occur (Dunkle et al. 2008; Maleta and Bowie 2010) Therefore, young people face competing aspirations and needs related to marriage, childbearing, death, and survival while navigating their risk for HIV infection at the same time: a daunting task to say the least, which may force some people to prioritize what is most important to them (Dionne, Gerland, and Watkins 2013) Young couples are particularly susceptible to HIV infection as they undergo a series of transitions from courtship to marriage and then later, to parenthood. Paradoxically, serious relationships such as marriage in which high levels of intimacy and trust have be en established are thought to be more risky than causal, promiscuous encounters (Mkandawire

PAGE 20

5 Valhmu et al. 2013) T his is for several reasons. As young couples begin to form their families, condoms are no longer consi dered an acceptable preventive measure for HIV/AIDS (Chimbiri 2007) In love marriages where women rely on ideals of love and intimacy to negotiate relationships with their husbands, condoms are seen as undermining the very thing they wish to preserv e (Smith 2006) Low condom use coupled with extramarital partnerships more notable among men than women (Schatz 200 5; Carpenter et al. 1999; Lurie, Williams, Zuma, Mkaya Mwaburi, Garnett et al. 2003) provide ample opportunities for new HIV infections to develop during young adulthood. Despite relatively high rates of HIV infection among young people who are just beg inning their sexual and reproductive lives, recent expansions in HIV testing and counseling (referred to as "HTC" from a policy perspective, but also labeled as "VCT" or voluntary HIV counseling and testing in the literature and among rural Malawians ) serv ices throughout Malawi provide an important opportunity for both individuals and couples alike to learn their HIV status and take action to protect their own and their partner's health. H IV testing plays a pivotal role in the public health response to the AIDS epidemic and is a vital entry point for HIV prevention services, care, and treatment (VCT Efficacy Group 2000; Painter 2001) Although access to H IV testing has generally improved throughout the region, many people still do not kno w their HIV status and for those who do get tested, levels of disclosure of HIV test results to sexual partners remain low (Obermeyer and Osborn 2007) Young pe ople confront difficult decisions of whether to be tested and if tested, to disclose the results to their loved ones They must therefore weigh the advantages

PAGE 21

6 of learning one's HIV status with the potential ly negative consequences of being diagnosed with a life threatening and arguably stigmatizing disease in a context where access to antiretroviral therapy (ART) is improving but still not guaranteed. In their version of the Social Ecological M odel, McLeroy and colleagues (1988) conten d that health behaviors are influenced by factors at five different levels: interpersonal (individual level), intrapersonal (includes family and relationship levels), institutional, community, and public policy (the last three categories comprise "structur al" levels) With regard to HIV/AIDS in sub Saharan Africa Catherine Campbell (2003) argues that t he overwhelming focus has been on factors th at occur at the individual level and are related to psychological and social considerations Factors at the interpersonal level are frequently overlooked despite the fact that HIV testing decisions are often made within the context of the dyad rather than in isolation Understanding the couple context is important not only for decision making around HIV testing, but after testing when individuals learn their HIV status perhaps for the first time and decide whether to disclose their results to partners. Wit hout more attention paid to the interpersonal level, the current state of research on HIV testing remains inadequate More research using both partners' perspectives is needed to more fully understand how the relationship context shapes HIV testing behavio r. The Importance of Relationship Power Power as expressed through social interaction s is not a new concept and has received tremendous scholarly attention by academics from a variety of different disciplines including anthropology, political science, soc ial psychology, public health an d sociology (Connell

PAGE 22

7 1987; Cromwell and Olson 1975; Blanc 2001) Riley (1997) observed that gender affects both "power to" and "power over." "Power to" refers to the ability to act, whereas "power over" refers to the ability to assert wishes and goals in the face of op position from another. Some scholars have also stressed that it is not the absolute power of either couple member that matters but rather their power relative to each other (Blanc 2001) In relation to HIV infection, Wingood and DiClemente (2000) describe power as "having the ability to act or change or having power over others ." In this study, I consider a broad sociological perspective of gender based relationship power (shortened to "relationship power") to refer to the socially constructed gender differences between men and women where "gender" refers to the expectations and norms shared within a society about appropriate male and female behavior, characteristics, and roles (Gupta 2000) This definition allows for more flexibility to study the culturally rooted and more nuanced forms of power that go beyond simplistic not ions of male dominance and female submissiveness. Of the studies that have directly measured relationship power, researchers have conceptualized power and its association with HIV/AIDS in many different ways depending upon the context. In 2000, Julie Pule rwitz and colleagues developed one of the first theoretically based measure s of relationship power using the Theory of Gender and Power (Connell 1987; Wingood and DiClemente 2002) and Social Exchange Theory (Emerson 1981) called the Sexual Relationship Power Scale (SRP S) (Pulerwit z, Gortmaker, and DeJong 2000) The SRPS was developed and tested among a sample of Latina women in the US. This single publication sparked a number of HIV related studies in the West and

PAGE 23

8 also in sub Saharan Africa, as researchers attempted to adapt the s cale to African samples (Harrison et al. 2006; Pettifor et al. 2004; Jewkes et al. 2002; Dunkle et al. 2007) Pulerwitz et al. (2000) defined relationship power as the ability to control a partner's actions, a ct independently, dominate decision making, or engage in behavior against the other partner's wishes. In their adoption of Pulerwitz et al.'s work to South Africa, Pettifor and colleagues (2004) measured relationship power as a combination of two factors: relationship control and recent experience of forced sex. One important limitation of most of these studies is that they measured relationship power and its effects on behavior a nd health from the female point of view, resulting in an incomplete representation of male and couple power dynamics. In this dissertation, I contribute to the growing body of literature on relationship power by addressing both couple members' perspectives thereby giving a voice to men who are often silenced in this topic of inquiry. Summary of Research Design The present research uses a sequential mixed methods design with three complementary phases: a measur e development phase (qualitative and quantitat ive) to develop a measure and conceptual model of relationship power, a hypothesis testing phase (quantitative) to test whether the measure of relationship power predicts HIV testing behavior and an interpretative phase (qualitative) to offer context for the quantitative findings through the use of grounded theory In the first phase, I develop a pilot measure of relationship power for the Malawi context and then re formulate it using a larger set of couple data collected as part of the

PAGE 24

9 Tsogolo La Thanzi ( TLT ) 1 study. I bring in qualitative data to provide context f or the measure and to finalize the conceptual model of relationship power. In the next phase, I test for associations between relationship power and HIV testing behavior. Here, I utilize longitu dinal couple data to test hypotheses related to two HIV testing behaviors: HIV testing uptake over a 16 month period and disclosure of test results to primary sexual partner s During the interpretive phase, I use qualitative data to explain, cross validate or challenge the quantitative findings from the hypothesis testing aim. I draw upon focus group discussions (FGDs) and other source s of qualitative data such as semi structured interviews with young couples detailed field notes, and informal interviews with young people, HIV testing counselors, and village chiefs. Implications for Policy, Prevention, and Public Health The results of this research have important theoretical, scientific, and pract ical implications for the field of public health and the s ocial sciences, and for HIV testing policy in sub Saharan Africa. Several innovative features of this research advance the study of the relationship context and HIV testing in sub Saharan Africa. First, this research develops and evaluates a new measure of relationship power that could be used by other HIV/AIDS researchers working in Malawi and surrounding countries This measure builds upon the SRPS (Pulerwitz, Gortmaker, and DeJong 2000) which has been mostly applied to women 's risk for HIV/AIDS in sub Saharan Afri ca (Pettifor et al. 2004; Jewkes et al. 2010; Shannon 1 Tsogolo la Thanzi (TLT) is a research project on reproduction and AIDS designed by Jenny Trinitapoli and Sara Yeatman, and funded by grant (R01 HD058366, PI Trinitapoli) from the National Institute of Child Health and Human Development (NICHD).

PAGE 25

10 et al. 2012) This dissertation broadens existing measures of relationship power by incorporating the perspectives of both men and women. Second, this study is the first to use longitudinal couple data to explicitly study the association between relationship power and HIV testing behavior (uptake and disclosure) Much of what is currently known about relationship power and HIV relates to risk, not HIV testing beh avior. Last, the findings generate new theoretica l perspectives on local constructions of relationship power in Malawi that is, theory that extends our current understanding s of power and relationships This research also has practical implications. Info rmation on the key factors that influence health decision making will generate far reaching conclusions about relationship factors that can be used to facilitate improved approaches for couple oriented HIV testing programs, which are largely absent through out the country. In addition, new measures of relationship power for the Malawi context will provide new opportunities to evaluate the effectiveness of public health interventions aimed at improving gender relations and health outcomes within couples. Cha pter Outline The remaining chapters of this dissertation can be summarized as follows. In Chapter 2, I describe the predominant theoretical perspectives used to understand gender and power relations and how this intersects with the background literature on HIV testing. Specifically, I outline the barriers to HIV testing uptake and disclosure in the African context. Chapter 3 presents the mixed method ology I use accomplish the specific aims of this study. As part of this chapter, I provide the details on the measure of power developed for the Malawian

PAGE 26

11 setting. Chapters 4 through 6 provide the main results of this dissertation. In Chapter 4, I use qualitative data to understand what power means in the Malawian context and to investigate how the measure of rela tionship power developed in Chapter 3 fits with local constructions of power. In Chapter 5, I present the main quantitative findings for this dissertation, specifically, the relationship power factors that influence uptake of HIV testing and disclosure of HIV test results. In Chapter 6, I transition back to the qualitative data to explain the main quantitative findings from Chapter 5. While each results chapter has its own discussion section that links to the findings from the previous chapter, I tie all th e findings together in Chapter 7. To conclude, I present the theoretical and public health implications, and future directions of research.

PAGE 27

12 CHAPTER II BACKGROUND AND CONCEPTUAL FRAMEWORK In this chapter, I begin with a broad examination of how gender and relationship power have been studied in relation to HIV infection. I then describe what is currently known and where the literature falls short regarding the association between relationship power and two types of HIV testing behavior: uptake of testin g services and disclosure of test results to sexual partners. As I review the literature, I incorporate discussions of how social science theory has been and could be applied to inform research on HIV t esting behavior I conclude with a description of the rural Malawian context: the historical, social, economic, and cultural landscape the HIV/AIDS epidemic, and the state of existing HIV prevention programs and policies. Gender, Relationship Power, and HIV/AIDS On a June afternoon in 2009, we respectfully a pproached a couple's home in the Mponda village to conduct the fifth interview of the project. After formally greeting each member of the household according to Malawian customs, our intention was to learn about the couple's relationship history. But their stories were already imprinted on their bodies and immediately revealed through a single glance at the couple's physical appearance. The man was tall, very thin, and had what appeared to be sores on his face; the wife suffered from a bad cough, later dete rmined to be tuberculosis. In a private conversation, the wife explains how she was previously a bar girl (akin to a prostitute in this case ) and her current husband made her an offer she couldn't refuse: to marry him and gain the chance to build a respect able life for

PAGE 28

13 herself as a wife and mother. The husband was older, had several ex wives, and his steady job at a local factory afforded him the opportunity to have multiple sexual partners. Despite the presence of these clear "risk factors" for HIV, it was only after the couple lost their young child to AIDS that they decided to get tested themselves and were discovered to be HIV positive. Times are still dangerous for Africans living amid a generalized AIDS epidemic. The common phrase in Africa that "you a re either affected or infected with HIV/AIDS" reflects the hard hit reality that almost everyone knows someone currently living with HIV or who has recently died of its complications. Despite valiant and well intentioned efforts by global health institutio ns to prevent new HIV infections, children, adolescents, and adults continue to become infected at relatively high rates. Sub Saharan Africa is home to a pproximately 22 million individuals currently living with the HIV virus (UNAIDS 2010) In 2009, an estimated 1.8 million people were infected with HIV in this region alone HIV incidence is gradually declining in many regions of sub Saharan Africa, but the number of people living with HIV continues to ris e, largely due to ART expansion (UNAIDS 2010) Although estimates differ by country, women in sub Saharan Africa now make up nearly 60% of all HIV infections (UNAIDS 2010) These gender disparities in HIV infection persist across adults an d y oung people (UNAIDS 2010; National Statistical Office & ORC Macro 2011) Gender based relationship power is one of the most widely cited reasons used to explain gender disparities in infection rates. Many social epidemiologic studies from southern Africa demonstrate that relationship power imbalances are associated with key risk

PAGE 29

14 factors for HIV, including less condom use, increased number of sexual partners, coercive sex, transactional sex, increased physical violence, and alcohol and drug abuse (Dunkle et al. 2004; Pettifor et al. 2004; Jewkes, Levin, and Penn Kekana 2002; Blanc et al. 1996; Harrison et al. 2006; Dunkle et al. 2007) In several studies, a direct association has been establishe d between power imbalances and HIV infection (Dunkle et al. 2004; Pettifor et al. 2004) The Theory of Gender and Power In 1987, R.W. Connell developed the first systematic conceptual framework for the social analysis of gend er called the Theory of Gender and Power or "TGP" for short (Connell 1987) Wingood and DiClemente (2002; 2000) later adapted Connell's theory to specifically st udy women's social exposure and risk for HIV/AIDS. One unique feature of the TGP is its potential to study HIV risk beyond the individual level by examining the broader context of relationships that perpetuate risk. The TGP proposes that power inequities a rise from three overlapping social structures that interact to generate different exposures and risk factors for HIV/AIDS: the sexual division of labor, the sexual division of power, and social norms related to gender (also known as "cathexis"). Because th ese three social structures are so closely intertwined and often inseparable, it is difficult to label an exposure or risk factor as the result of one structure over another ; rather, it is often the intersection of these factors together that creates a ris ky environment. The sexual division of labor functions at the societal level through the allocation of men and women to certain occupations (Wingood and DiClemente 2002) Women are often relegated to jobs and responsibilities deemed to be "women's work", such as those that exist

PAGE 30

15 in the domestic sphere. The inequities resulting from the sexual division of labor are manifested as economic exposures and socioeconomic risk factors for HIV/AIDS. In sub Saharan Africa, t here is evidence to support the sexual division of labor's role in HIV ris k for women. T raditional breadwinner homemaker marriages constrain women's earning potential and mobility while at the same time providing men with more access to wealth, opportunities, and sex part ners. A number of studies from the region support the view that men's mobility is a risk factor for HIV (Voeten, Vissers, and Gregson 2009; Vissers et al. 2008; Kishamawe et al. 2006; Lurie, Williams, Zuma, Mkaya Mwaburi, Garrett et al. 2003) In addition many African s ocieties are organized according to patrilineal systems of decent and inheritan ce that allowed men to control economic resources in the household (Caldwell, Caldwell, and Orubuloye 1992) Resource theory (Foa and Foa 1980) helps to elaborate on the TGP's structure of labor by extending the idea of economic dependence. According to resource theory women with less access to and control over resources as compared to their partners become economi cal ly dependent on men. This dependence is thought to limit their negotiating power over sex, potential to mitigate violence, and ability to leave a risky relationship. Although sex linked to subsistence is critical for the survival of many marginalized A frican women (Leclerc Madlala 2003; Wojcicki 2002 ) capital led globalization has created a new form of dependence in the form of luxury items that are becoming increasingly desired by women from their more wealthy sexual partners (Hunter 2010, 2002; Epstein 2007 ; Wamoyi et al. 2011) Receipt of money, gifts, or financial assistance from men, sometimes

PAGE 31

16 referred to as material transfers, has been shown to decrease women's bargaining power in the sexual realm, including condom use and frequency of sex (Luke et al. 2011; Dunkle et al. 2004; Luke 2006) The TGP's sexual division of power is maintained by social mechanisms such as the abuse of authority and control in relationships (Wingood and DiClemente 2002) Gender based violence is one manifestation of these power imbalances and an important risk factor for HIV in sub Saharan Africa (Dunkle et al. 2004; Jewkes, Levin, and Penn Kekana 2003; Jewkes et al. 2010) Multiple direct and indirect pathways have been prop osed to link power imbalances, women's experience of violence, and HIV infection (Jewkes et al. 2010) Direct effects occur through viole nt, unprotected sexual encounters with abusive men who are more likely to engage in risky sex and to be HIV infected themselves (Jewkes et al. 2009) Indirectly, the threat of violence may prevent women from negotiating the circumstances of sex, resulting in more frequent sex and less condom use (Jewkes et al. 2006; Pettifor et al. 2004) Large age and economic differences between partners such as those characterized as "sugar daddy" relationships may also place adolescent women at increased risk for violence (Luke 2003) and HIV infection (Kelly et al. 2001) through their limited power to negotiate safe sexual behavior with older, wealthier men (Luke 2005; Luke and Kurz 2002) According to the TGP, s ocial norms surrounding gender and sexual behavior (or "cathexis") add a third dimen sion of risk for HIV. In many African societies, traditional gender roles and socialization patterns implicitly or explicitly dictate what men and women do and how they behave (Shettima 1998) A husband's right to punish his wife or demand

PAGE 32

17 sex are often condoned and considered socially acceptable (Jewkes 2002) The presence of a "sexual double standard" (Hunter 2010) that makes it more socially acceptable for men to have extramarital p artners also place s women at increased risk for HIV through their partners' behaviors. Women, on the other hand, in order to meet the ideal qualities of a respectful wife, may avoid topics that could cr eate conflict in the household, for example, condom us e or extramar ital affairs Longitudinal studies have demonstrated that women are more likely to become infected by their husbands, while men are more likely to become infected through their own extramarital affairs (King et al. 1993; deZoysa, Sweat, and Denison 1996; Heise and Elias 1995; McKenna et al. 1997) Though typically not applied to men, men's health is similarly affected by the three social structures of gender and power. There are many pathways to HIV in fection for men in sub Saharan Africa ; I highlight just a few examples. Throug h the sexual division of labor, dominant masculine ideologies surrounding fatherhood and the provider role keeps m en out of the household and in the labor force. Men's mobility a nd freedom in the public space facilitates the ir engagement in HIV risk behaviors particularly, having multiple concurrent sexual partners Through the sexual division of power, male dominance over the timing and circumstances of sex directly places men a t risk for HIV as well. Finally, through the third structure of social norms and affective attachments, men's desire for multiple sexual partners and the elevated social status that comes with this practice creates exposures to HIV infection (Swidler and Watkins 2007; Smith 2009)

PAGE 33

18 HIV Prevention and Testing in Sub Saharan Africa This dissertation attempts to translate what we know about the power/ sexual risk nexus to non sexual behavior related to HIV testing ( namely, uptake and disc losure). Before outlining how we might apply this knowledge to HIV testing I will first discuss the predominant approaches to HIV prevention in this region and why HIV testin g is an important area of inquiry. Current HIV prevention strategies center on t he "ABC" approach or A bstinence, B e faithful, and C ondoms For young adolescent couples, abstinence has been promoted as an idealistic HIV prevention strategy by forg o ing or delaying sex until marriage. While by far the most effective strategy to prevent H IV/AIDS when practiced, its high failure rate has often been attributed to a morality driven agenda that lacks rigorous scientific support (Santelli et al. 2006) For those who fail in "A", or abstinence, albeit under circumstances that make it difficult to be successful, "B" encourages people to stay with one sex partner. If neither "A" nor "B" is feasible or possible, consistent condom use or "C", is recommended as a fallback strategy. However, for young couples just starting their families, condoms interfere with childbearing intentions. In addition, public health messages continue to perpetuate the association of condoms with high risk or causal sex, t hus limiting their use within serious partnerships if people perceive them to interfere with intimacy and trust Condoms can also diminish women's sexual pleasure and interfere with intimacy (Higgins and Hirsch 20 08; Chimbiri 2007; Tavory and Swidler 2009) As an alternative to condoms, married couples are advised to remain faithful to each other through widely disseminated

PAGE 34

19 public messages on fidelity (for example, Uganda's "zero grazing" campaign). The objective here is to minimize exposure to sexual networks that ostensibly act as superhighways to transmit HIV infection a phenomenon referred to as concurrency (Epstein 2007) While the predominant prevention approach for HIV/AIDS in sub Saharan Africa emphasizes the "ABCs", the continued spread of the disease signifies that knowledge alone is not sufficient and thus other approaches are warranted. P olicy makers and international health experts have promoted HIV testing and counseling as one such alternative for HIV prevention (CDC 2006; UNAIDS 2006; WHO 2007) Two main terms are used to refer to HIV testing: "HTC" (HIV testing and counseling) and "VCT" (volun tary HIV counseling and testing) 2 The term HTC recently replaced the term VCT in official AIDS policy, although much of literature that I reference in this dissertation reflects on VCT From a public health standpoint, HTC serves to reduce HIV transmissio n through its complementary risk reduction counseling and timely access to care, treatment, and other HIV prevention services (Painter 2001; VCT Efficacy Group 2000) Tremendous efforts have been made to increase the number of people w ho know their HIV status in sub Saharan Africa. Public health slogans such as "Know Your Status" 2 In the past, the term VCT was used to refer to walk in testing servic es located within health facilities or as standalone testing sites or client initiated testing. The Ministry of Health in Malawi recently replaced the term VCT with HTC to reflect revised standards set forth by the WHO and UNAIDS. Provider initiated HIV te sting and counseling is now considered the international standard of care. Health care providers are encouraged to offer an HIV test paired with complementary risk reduction counseling to their patients during a ll medical encounter s (WHO, 2011). Therefore, the term HTC encompasses both provider initiated and client initiated services offered at a wide range of testing sites including mobile clinics, the home, and at the workplace.

PAGE 35

20 (featured in Figure 2.1) are o mnipresent in Malawi carefully placed in high traffic public spaces and outside HTC clinics in an attempt to normalize the act o f testing and promote the individual right, freedom, and entitle ment to knowledge of HIV status The marketing of HIV testing is clear, concise, and appeals directly to testing consumers as if the act of testing is as simple as just showing up (i.e., "Let' s go!"). Figure 2.1 : Example "Let's Go" Slogan Promoting HCT (Photograph taken by Amy Conroy in Cape Maclear, Malawi, 2009) Despite substantial economic and human resources dedicated to promoting HIV testing in sub Saharan Africa, testing programs con tinue to fall short. An extensive body of literature from sub Saharan Africa cites three main areas of concern: uptake of testing is relatively low in many settings (Denison et al. 2008; Obermeyer and Osborn 2007) disclosure of test results to primary partners is less than ideal (Obermeyer and Osborn 2007) and sexual behavior change particularly among those who test negative is modest at best (Shelton 2008; Yeatman 2007) This dissertation focuses on the first two areas of con cern:

PAGE 36

21 uptake and disclosure. Recent estimates based on surveys in 12 high burden countries in sub Saharan Afric a indicate that a median of 12% of men and 10% of women in the general population have been tested for HIV and received the results (WHO 2007) If this is accurate then improving the use of HIV testing services would requi re an understanding of the barriers and facilitators to uptake in the general population. The benefits of HIV testing are also predicated on the assumption that couples first disclose their HIV status to each other. Disclosure may reduce HIV transmission b y increasing awareness and decreasing risky sexual behavior (Medley et al. 2004) I t is difficult to generalize about disclosure rates at the population level Most studies utilize convenience samples of HIV clinic patients or pregnant women (Kilewo et al. 2001; Antelman, Fawzi, and Kaaya 2001) However, a ggregate estimates from multiple countries show that an average of 52% (range: 16 86%) of HIV positive women disclosed their status to sexual partners (WHO 2003) Recent studies that include men both HIV positiv e and negative indicate that upwards of 70 % of men from Malawi, South Africa, and Kenya disclosed their status to sexual partners (Anglewicz and Chintsanya 2011; Katz et al. 2009) The Social Ecological Model of Health The S ocial Ecological M odel of heal th provide s a useful framework for examining how determinants of HIV testing behavior (uptake and disclosure) have been studied and where gaps in the literature remain. According to the Social Ecological M odel multiple levels of factors in fluence health b ehaviors including intrapersonal, interpersonal, institutional,

PAGE 37

22 community, and public policy (Sallis Owen, and Fisher 2008) In the discussion that follows, I present how factors at the individual (intrapersonal) level and relationship (interpersonal) level have influenced HIV testing behavior. I also consider structural level factors related to econom ics, but only as they are manifested within the context of the relationship. I hypothesize that a lack of understanding of and consideration for relationship factors, specifically power, contributes to limited HIV testing success at the programmatic level. What It Takes to "Know Your Status" in Africa The Individual Level At the individual level, a variety of different psychosocial factors are thought to influence the uptake of HIV testing services in sub Saharan Africa. The most widely documented barrier s include stigma and discrimination (Berendes and Rimal 2011; Kalichman and Simbayi 2003; Weiser et al. 2006; Hutchinson and Mahlalela 2006) confidentiality concerns (Bwambale et al. 2008 ; Weiser et al. 2006) and low HIV related knowledge 3 (Jean et al. 2012; Gage and Ali 2005; Berendes and Rimal 2011) One particularly important individual leve l factor is perceived risk This construct is one of th e most widely applied components of the Health Belief Model or HBM (Becker 1974) especially for studying HIV/AIDS The underlying idea is that beliefs about the likelihood of contracting a disease or condition will motivate people to adopt a certain preventative 3 Though knowledge is cited as a barrier to testing in recent studies, it is important to point out that it is generally understood that a lack of "HIV related knowledge" is no longer a major driver of the AIDS epidemic in Africa. People are well aware of AIDS, how the virus it is transmitted, and the precautionary measures to hel p avoid HIV infection though this knowledge is not perfect and nuanced understandings still exist.

PAGE 38

23 behavior (Champion and Skinner 2008) Indeed previous research demonstrates that low perceived risk for HIV may limit uptake of testing (deGraft Johnson et al. 2005; Weiser et al. 2006; Creel and Rimal 2012; Sambisa, Curtis, and Mishra 2012) There exists s ome inconsistency, however, about the direction of this association; other studies have foun d that individuals who perceive themselves to be at high risk for HIV may refuse testing out of fear of the consequences (MacPherson, Corbett et al. 2012; Pool, Nyanzi, and Whitworth 2001) Because the HBM construct of perceived risk has been such an influential variable in t he HIV testing literature I consider it in my quantitative analysis By doing so, I al so hope to resolve some of the inconsistency regardin g direction of the association between perceived risk and HIV testing uptake. Less research has focused on the rela tionship context of risk perception. This is despite the fact that e valuation of risk for HIV is often made in a dyadic context, that is, risk assessments are based not only on self, but also on partners' sexual backgrounds. In Malawi, women are most worri ed about getting HIV from unfaithful husbands while men are most worried about getting HIV from their extramarital partners (Smith and Watkins 2005) T here is mixed evidence on whether perceived partner infidelity is positively or negatively assoc iated with testing uptake Several studies from South Africa and Zimbabwe demonstrate that women who suspect or know that their partners have other sexual partners are more likely to get tested for HIV (Luseno and Wechsberg 2009; Morin et al. 2006) In contrast, a study from Malawi finds no asso ciation between perceptions of a partner's infidelity and ever having tested for HIV (deGraft Johnson et al. 2005) In this dissertation,

PAGE 39

24 I extend the research that focuses mostly on individual le vel perceived risk to include a partner's perceived risk as a predictor of HIV testing uptake. The Relationship Level Though the literature centers on individual level determinants, an investigation of the relationship level is critical in order to bette r understand the social environment in which HIV testing decisions are made. A broader body of health literature posits that s erious intimate relationships play an important role i n couple health and consequently, t here have been growing calls to study how partners mutually influence each other's behaviors (Lewis et al. 2006) The Theory of Interdependence suggests tha t behaviors within dyads are interdependent ; each couple member has a certain amount of influence over the interaction they have together (Kelley and Thibalt 1978) Therefor e an understand ing of the effects of couple interactions on health requires that both partner's perspectives be taken into account Unlike certain behavioral outcomes such as sex that cannot occur without the participation of both partners, decision s to t est for HIV and disclose test results can, in theory, be made independently ( unless t he behavior under investigation is couples testing, which clearly requires both couple members ) However, the liter ature suggests that HIV testing behavior does not occur in isolation ; men and women make HIV testing decisions using information about their relatio nship (Luseno and Wechsberg 2009; Morin et al. 2006) Thus, I posit that the Theory of Interdependence can still provide t he rationale for why it is important to examine how partnership context influences individual decisions to test or disclose.

PAGE 40

25 T he Theory of Gender and Power (TGP) may provide a useful theoretical lens to study how the relationship context particularly pow er s hapes HIV testing behavior (uptake and disclosure) To briefly recap the TGP argues that three social structures, sexual division of labor sexual division of power, and cat h exis or social norms around gender, shape sexual risk for HIV/AIDS. Of the st udies that have examined the association between relationship power and HIV testing uptake, many have centered on male control over their female partners' decisions to test. The three social structures of the TGP are implicitly involved here although they are rarely if ever directly referred to in the literature S tudies suggest that decision making power i mbalances r elated to the sexual division of power constrain women's use of testing services. For example, a common reason provided by women who refuse t esting is the need to discuss the issue with their husband or because the husband refused testing himself (Kranzer et al. 2009; Dahl et al. 2008; Perez et al. 2006; Baiden et al. 2005) Issues related to cathexis m ay also play a role; qualitative research suggests that a double standard may exist around HIV testing such that women need to request permission from their husbands but men are free to make testing decisions on their own (Maman, Hogan, and Kilonza 2001) Finally, w omen's reported obligation to seek permission from their partners may be reinforced by the sexual division of labor through male control over economic resources. For example, Morin and colleagues (2006) note that women inform their husba nds before testing in order to obtain money to travel to testing sites. More importantly, this reinforces women's inability to access money and resources in the future.

PAGE 41

26 Wingood and DiC lemente (200 0) argue how being involved in an abusive relationship creates physical exposures to HIV through the sexual division of power The authors provide an example of how Latina women in the US who feared a partner's anger in response to requests to use condoms were less likely to use them (Marin et al. 1993) The same logic parallels existing research on use of HIV testing services for women in sub Saharan Africa. Worry about physical abuse has been found to hinder the ability to test if individuals suspect that they could be HIV positive (MacPherson, Corbett et al. 2012; Pool, Nyanzi, and Whitworth 2001) The TGP structure of cathexis is inextricably lin ked to these manifestations at the division of power. At the societal level, the structure of cathexis dictates appropriate sexual behavior for women and reinforces cultural taboos regarding female sexuality (Wing ood and DiClemente 2000) Therefore, w omen may suffer abuse after testing for reasons related to cathexis for example, because they failed to inform their partner of their intentions to test and thus violate d social norms around male permission or becaus e the act of testing implies guilt from infidelity or past promiscuous behavior regardless of whether women engaged in these behaviors. Women have also reported fears of divorce or partner abandonment as barriers to testing (Mlay, Lugina, and Becker 2008; Irungu et al. 2008) I suspect that t hese findings operate through the pathway of financial resources. Through the sexual division of labor women generally have less access to economic resources than men. According to the theory, women who are unemployed or underemployed may be forced to rely on their male partners for economic assistance which in turn limits their agency in the relationship (Foa and Foa

PAGE 42

27 1980; Wingood and DiClemente 2000) As the theory implies, w omen with less access to economic re sources may be more susceptible to the effects of relationship dissolution should they test positive, especially if they perceive few alternatives to the current relationship. Thus, the risk of losing a relationship and the financial support that accompani es it may overshadow any of the perceived benefits of HIV testing. But also, women's economic dependence on men may limit their negotiating power over HIV testing if their partner is reluctant to test. Up until this point, I have focused my theory and rev iew of the literature and relevant theory on relationship factors that may influence women's uptake of HIV testing, but what may explain men's behaviors and how might this differ from women 's ? While the TGP propose s that three social structures limit women 's uptake of HIV testing, these same social structures prov ide men with a surplus of power which provide men with a surplus of power may ultimately prevent them from testing as well. During an informal interview in 2011 with a male VCT counselor named Jos eph, I frankly asked him if and why men were scared of testing. Men are more reluctant to come in for testing. Joseph says they are "running away from their responsibilities ." He also says that men are often blamed (he said "victims" in English ) for trans mitting HIV to women through extramarital affairs. Consequently, many will refuse to get tested out of fear. It takes a long time for women to convince their husbands to get tested as well. When a married woman comes in for testing, Clinic staff will "invi te" the husband to come too by going out to the home using maps of the area and then asking them to come in for VCT. Wives blame the husbands for bringing in HIV since they are at home with the children while the men are free to move around with other wome n. (Interview and field notes dated October 18, 2011) I previously suggested that women 's fear of being bla med for infidelity might limit their uptake of HIV testing (i.e., cathexis) In the above quote, Joseph suggests that it is the women who are doing t he blaming not the men hence, men become "victims" M en are

PAGE 43

28 very much aware of the global AIDS discourse that faults men for high er rates of HIV infection among women; this emphasis on female vulnerability to HIV/AIDS may have unintended consequences on m en's testing behavior. Anthony Simpson's (2009) research on masculinities and AIDS in Zambia a country that shares its borders with Malawi documents that men are indeed having extramarital affairs and are well informed about the risks of having these relationships. But in order to circumvent an admission of guilt that comes with the act of going for HIV testing, men will avoid it even if they are clearly dy ing of AIDS (Simpson 2009) In fact t here may be real consequences for men who are found to be cheating on their wives during an AIDS epidemic. Evidence from the region s hows a growing intolerance for men's sexual indiscretions Several s tudies from Malawi point out that women will bring in marriage mediators, confront his mistresses directly, and may ultimately leave a partner who refuses to reform his sexual behavior (Watkins 2004; Schatz 2005) In nearby Uganda, new legislation allows wives to divorce husbands for infidelity (Parikh 2009) Taken together, this research indicates that men may internalize some of the same worries around infidelity accusations and divorce when m aking decisions to test and later, to disclose. Beyond these affective attachments that link HIV infection to infidelity ( aspects of cathexis ) q uestions remain around how the other social structures of T GP limit or facilitat e men's rates of HIV testing. Through the sexual division of labor men's breadwinner role and the income that accompanies it may buffer the potential negative consequen ces of testing positive, thus making testing seem like a more risk worthy venture. At t he same time, being

PAGE 44

29 employed means that men have less time to wait in long lines at crowded health centers to receive testing at the cost of losing out on wages. Of the fewer studies that include men the findings show that migration for work (Weiser et al. 2006) logistical barriers a round VCT including access (Bwambale et al. 2008) and testing service characteristics (Kranzer et al. 2009; Hutchinson and Mahlalela 20 06) may be important factors for men's uptake. Through the intersection of the sexual division of power and cathexis men involved in male dominated relationships who perhaps adhere to stronger patriarchal ideals around masculinity may be more likely to avoid overly feminized health care spaces. A large body of literature on masculinities has established a relationship between lower rates of health service utilization and poorer health among men who a dopt these gender ideals. In Zimbabwe, Skovdal and coll eagues (2011) discovered t hat hegemonic masculinities that require men to be strong, informed, resilient, disease free, and highly sexual and economically productive may serve as a barrier to HIV testing Here, t hese ideals of masculinity conflict with the good patient' perspective that expects patients to be concerned about their health, regularly seek care at what men largely perceive to be female dominated spaces, take instructions from health care providers, and engage in health promoting behaviors. In Malawi and Uganda, Izugbara and colleagues (2009) found that young males resisted test ing because they perceived the act to signify a lack of self confidence and vulnerability to HIV traits that conflicted with their male youth identity. To conclude, men face similar social barriers around gender and po wer that limits their uptake of HIV testing, however their position has largely been neglected in this body of research.

PAGE 45

30 What it Takes to Disclose Your HIV Status The Individual Level It is widely acknowledged that individuals face great difficulties when making decisions to disclose their HIV status. Decisions to disclose test results to sexual partners are often made using a complex calculus that involves weighing the many advantages and disadvantages of the behavior. In the literature, two particula r constructs of the HBM have been widely studied with regard to disclosure: the perceived barriers (or "costs") and the perceived benefits. The HBM constructs offer explanations of behavior in relation to the threat of a particular disease or health condit ion or in this case, the perceived outcomes of sharing personal health information with sexual partners. G iven that decisions to test for HIV often include whether individuals anticipate disclosure, many of the same barriers and benefits of testing uptake apply to disclosure. At the individual level, demographic barriers include lower socio economic status (SES), being unmarried, and younger age (Anglewicz and Chintsanya 2011; King et al. 2008; Antelman et al. 2001; Farquhar et al. 2004; Wong et al. 2009) Combined, the findings suggest that these individuals may be more vulnerable to the negative consequences of disclosure: simply put, they have more to lose than their older, married, and higher SES counterparts R isky sexual behavior such as having unprotected sex and multiple sexual partners has also been cited to constraint disclosure among HIV positive samples (Antelman et al. 2001; King et al. 2008) perhaps because these individuals expect to be blamed for immoral sexual behavior.

PAGE 46

31 HIV serostatus is another individual level factor thoug ht to play a role in decisions to disclose test results with sexual partners (Maman, Hogan, and Kilonza 2001) Unsurprisingly, HIV negative individuals are more likely to disclose than HIV positive individuals (Anglewicz 2008) Katz and colleagues (2009) found that HIV positive men had a more difficult time disclosing to their partners than HIV positive women. Research on stigma and discrimination in South Africa shows that HIV positive individuals are more susceptible to negative social an d economic consequences such as loss of employment, increased poverty, and rejection by family members and friends (Simbayi et a l. 2007) While HIV status at t he individual level is highly relevant little research has explored it within the context of the relationship that is, whether or not HIV status couple concordance (both partners are either HIV positive or HIV negative) and discordance (one partner is HIV positive, one partner is HIV negative) plays a role in decisions to disclose to sexual partners. The Relationship Level While a variety of relationship level barriers are hypothesized to hinder HIV disclosure, the main ob stacles center on fear: fear of stigma, fear of relationship dissolution and the accompanying loss of economic support, and fear of physical and sexual violence (Medley et al. 2004; Kilewo et al. 2001; Maman, Hogan, and Kilonza 2001; Maman et al. 2003; Farquhar et al. 2004) Fear based barriers are more salient for women than men, which reflect the unequal and limited power that many women have around HIV infection (Maman, Hogan, and Kilonza 2001) In Malawi a study on couples found that AIDS related stigma was a more sali ent barrier to disclosure for women than men (Anglewicz and Chintsanya 2011)

PAGE 47

32 Similar to HIV testing uptake, disclosure related behavior can be explained by the three social structures of the TGP. Regarding cathexis, s ome qualitative research suggests that women who choose to disclose a positive test result to their spouses are blamed for bringing HIV into the family (Lugalla et al. 2008) especially if men were not informed that their partners went for testing (Maman, Hogan, and Kilonza 2001) At the sexual divi sion of power o ther studies show that women who disclosed their HIV test results experienced negative outcomes such as violence (Maman et al. 2002; Maman et al. 2001; Maman et al. 2003) In a study from Tanzania, Maman et al. (2002) reported that partner violence was 10 times greater among HIV infected women as compared with their uninfected counterparts. However, t here are conflicting accounts about whether women experience high levels of adverse consequences after disclosure. In fact, many studies report that violence is a rare outcome of disc losure for women (Medley et al. 2004; Desgrees Du Lou 2005; Keogh et al. 1994; Vissers et al. 2008) and tends to occur more in serodiscordant couples (Maman et al. 2003) Interestingly, in Tanzania, less partner violence was reported among women who disclosed their serostatus to their partners tha n among women who did not (Maman et al. 2001) It is possible that individuals who test ed and then later disclose d are a biased sample since they tested for HIV in the first place as comp ared to those in violent relationship s who delay or forgo testing altogether out of fear Regardless, these studies suggest that violence after disclosure is situational and depends on the relationship context. Inequities resulting from imbalances at the sexual division of labor may also shape women's decisions to discl ose especially if the woman is the HIV positive partner. Fears of loss of

PAGE 48

33 economic support due to partner abandonment may drive some women to hide their status from their partners These fears may be justified. In Uganda, d iscordant couples with an HIV pos itive woman were more likely to dissolve than seroconcordant or serodiscordant (positive male, negative female) couples (Porter et al. 2004) It is important to point out that adverse consequences may be biased if women who suspect that their partners will react negatively are less likely to disclose in the first place. Limitations of the TGP Several limitation s of the TGP are noteworthy First, th e TGP is rooted in the so called vulnerability paradigm which overemphasizes women's lack of power in society and their in ability to pr otect themselves from HIV/AIDS. Globally, c on temporary AIDS discourse perpetuate s portrayal s of women as biologically and socially vulnerable to HIV, victims of men's abuse, and innocent bystanders in the global HIV/AIDS pandemic (Higgins, Hoffman, and Dworkin 2010) Not all women are victims, nor are all men o verly dominant an d controlling. These popular depictions of female vulnerability serve an important and valid purpose by reinforcing the evidence t hat women are indeed disadvantaged by HIV/AIDS ; h owever, at the same time, may restrict our understanding women's agency and r esilience. Second, the female vulnerability paradigm is historically grounded in western feminism and centers around individual human rights and freedoms that are defined and socially celebrated in the West a setting very different from rural Africa where the model is frequently applied. It is possible that women's power in Malawi may be overlooked because

PAGE 49

34 of the ways it is expressed within the context of culture. From a western perspective, it may appear that women succumb to their partner's dominance; ye t in reality, they may be using "backstage" techniques (Goffman 1959) to maintain control over their lives Sociologists working in rural Malawi have started to tap into the subtleties of female agency and how this agency is invoked to avoid HIV infection. For example, researchers have found that wives sometimes use subtle and gendered communication to encourage fidelity in their marriages (Watkins, Rutenberg, and Wilkinson 1997; Zulu and Chepngeno 2003) and other locally formulated strategies including approaching husbands about their behavior, bringing in marriage mediators, confronting his mistresses, and leaving a partner who refuses to reform (Watkins 2004; Schatz 2005) By relying solely on the TGP and its female vulne rability orientation we may miss the important graduations of female agency that could be harnessed to improve HI V/AIDS interventions among others who are truly powerless. Third, a female vulnerability emphasis detract s attention away from heterosexual m en's social disadvantages and needs around HIV. It assumes that men are active transmitters of HIV infection but not active agents of prevention (Higgins, Hoffman, and Dworkin 2010) In response to this there have been growing calls for new scholarship on masculinity and the ways in which men's own limited life choices relate to global patterns of power (Hirsch et al. 2009) To summarize the TGP's underlying emphasis on female vulnerability restricts our understanding women's ag enc y and resilience, men's position and the broader dyadic context for HIV risk. In this dissertation, I hope to illuminate areas of nuance regarding the

PAGE 50

35 gender/power/HIV testing nexus in order to move this body of research in ne w directions that consider these important gaps To accomplish this aim I invoke the strengths of grounded theory as a method to more fully capture the construct of relationship power and its intersection with HIV testing By listening to the perspectives of rural Malawians who a re the true cultural experts on these issues, I hope to broaden our knowledge of how relationship factors intersect with the act of HIV te sting. Symbolic Interactionism and HIV Testing Behavior As one could imagine the act of testing is not a straightfo rwa rd process ; HIV/AIDS and its interventions carry symbolic meanings that are deeply embedded within the social and cultural milieu and strongly shape people's decisions to test. According to Herbert Blumer's (1969) perspective on social interactionism: (1) people act on the basis of meanings that things have for them; (2) these meanings derive from social interaction; and (3) these meanings are modified by their interpretations in practice. The basic research assumption consistent with this view is that if one desires to understand human interactions and how they shape behaviors, one needs to study people's experiences as they perceive them (Jeon 2004) By doing so, a symbolic interactionism perspective gives agency to the rural Malawians living amongst an AIDS epidemic r ather than assuming they are mindless beings to be fashioned by western AIDS policy. Indeed, research on HIV/AIDS interventions designed in the West but implemented in Africa shows that local interpretations and responses may differ substantially from how these programs are intended to function from a public health

PAGE 51

36 perspective (Kaler and Watkins 2010; Angotti, Dionne, and Gaydosh 2010; Tavory and Swidler 2009) In rural Malawi, Kaler and Watkins (2010) use qualitative data to u nderstand local perceptions of HIV testing and find that the reluctance to test is connected to the perception that testing inevitably leads to a positive diagnosis and subsequent death. Rural Malawians describe how testing is analogous to "asking God abou t the day you will die" and many would prefer not to live with this un certainty (Kaler and Watkins 2010) The disconnection between AIDS policy and local responses resonates in the condom literature as well. In love marriages, where women rely on ideals of love and intimacy to negot iate relationships with their husbands, condoms are seen as undermining the very thing they wish to preserve (Chimbiri 2007; Tavory and Swidler 2009) Given their strong association with high risk sex and immorality, condom use within serious partnerships symbolizes a relationship characterized by mistrust, instabi lity, and immoral sexual behavior (Smith 2009) Bringing up condoms unexpectedly raises concerns about a partner's HIV status or faithfulness and symbolizes a partnership filled with uncertainty. The rationale is simple and logical: why use condoms with a partner you trust and love? I suspect that the same prem ise might be true with HIV testing. Other scholars have used a symbolic interactionist approach to examine HIV testing within the context of a relationship. Kathryn Rhine (2009) for example, discovered that an HIV test is not simply a tool employed to measu re immunological malfunction. She says, "A diagnosis also illuminates a set of social facts. The virtues and fears embedded in the act of taking a test are related to larger questions of how families and relationships might change in

PAGE 52

37 light of a positive re sult." Deborah Lupton and colleagues (1995) add that HIV test results have meanings that are tied to relationships, faithfulness, and trust. Individuals often take HIV tests when they a re ending a relationship or starting a new one and thus testing serves to mark these important life transitions. In Tanzania, Maman and colleagues (2001) found that couples used testing as a means to reaffirm one's commitment to the relationship. Thus, for some people, the act of testing may be more of a symbolic gesture than an attempt to confirm perceptions of risk. Through a symbolic interactionist approach, I hope to understand how beliefs, perceptions, and experiences of HIV tes ting created through social interactions at the relationship level ultimately shape people's reactions to HIV testing programs in Malawi. This approach will supply locally rooted evidence for the applicability of the TGP and illuminate areas of contradicti on, confirmation, and further exploration. Study Hypotheses This dissertation seeks to investigate how relationship power may act as a barrier or facil itator to HIV testing beha vior (uptake and disclosure) in southern Malawi. Based on the TGP and backgrou nd literature, I hypothesize that relationship power will consist of the following domains: socio economic power within the relationship, decision making dominance and relationship violence each of which will be independently associated with the two outco me s of HIV testing behavior. For both women and men having less socio economic power as compared to a partner ( sexual division of labor ), being in a male dominated relationship ( sexual division of power ) and having a history of relationship violence ( sex ual division of power ) will indicate low power in the relationship. Given the background theory and

PAGE 53

38 literature, I hypothesize that the following conceptual model will explain HIV testing behavior within rural Malawian couples (see Figure 2. 2 ). Figure 2. 2 : Conceptual Model for HIV Testing Behavior Hypotheses Hypothesis #1 I hypothesize that each power variable will be associated with uptake of future HIV testing over a 16 month period. In addition, I hypothesize that the perceived risk construct from th e HBM will be associated with HIV testing uptake. Four sub hypotheses will be tested: A. Socio e c onomic inequality (division of labor) : Individuals in a lower socio economic position relative to their partners will be less likely to get tested since they wil l have stronger fears around divorce or abandonment and loss of financial support that may come with

PAGE 54

39 HIV testing. With less economic power, these individuals may also be in a more disadvantaged position to negotiate testing with their partners. B. Relationsh ip dominance (division of power) : Women in a male dominated relationship as compared to an egalitarian or female dominated relationship will be less likely to test for HIV due to male control over testing decision making. Male dominance will also be negati vely associated with testing among men. These men may be more likely to adhere to traditional beliefs about masculinity and therefore feel disinclined to test. C. Relationship v iolence (division of power): Having a history of relationship violence (physical and sexual) is a proxy for fear of abuse, which has been shown to be a barrier to testing. Individuals in violent relationship s will therefore be less likely to test for HIV. Having a history of violence (sexual or physical) could also operate through the pathway of risk, thereby decreasing the likelihood of testing (assuming people are higher risk are less likely to test). D. Perceived risk (self and partner). For women, those who believe they are at higher risk for HIV will fear the negative consequences of testing and therefore be less likely to test. For men, those who believe they are at higher risk for HIV will fear being blamed for infidelity and therefore be less likely to test. Hypothesis #2 I hypothesize that each power variable will also be associ ated with HIV testing disclosure to a primary sexual partner. In addition, I hypothesize that perceived HIV status

PAGE 55

40 concordance will be associated with HIV testing disclosure. Four sub hypotheses will be tested: A. Socio economic inequality (division of labor ) : Individuals in a lower socio economic position relative to their partners will be less likely to disclose their test results since they will more likely to fear divorce or abandonment and loss of financial support as a consequence of disclosing test res ults. B. Relationship dominance (division of power): Being in a male dominated relationship as compared to an egalitarian relationship will make women less likely to disclose out of fear of the violence that is associated with dominance. C. Relationship v iolen ce (division of power): Having a history of relationship (physical and sexual) violence will reinforce fears of abuse, thereby decreasi ng the likelihood of HIV test disclosure Even if they tested negative, individuals in violent relationships may still av oid disclosure if they did not inform their partners of their plans to test. D. Perceived HIV status concordance: HIV positive individuals will be less likely to disclose their test results if they perceive their partners to be HIV negative as compared to H IV negative individuals who also perceive their partners to be HIV negative. Study Innovation This study advances our knowledge of HIV testing uptake and status disclosure in sub Sahar an Africa in the following ways. 1. A one time test is not enough to pre vent HIV. Despite the fact that HIV testing rollouts have increased the number of people who know their HIV status, a large majority of studies

PAGE 56

41 on HIV testing in sub Saharan Africa examine whether people have ever tested or recently tested for HIV as if it is a one time event (Berendes and Rimal 2011; deGraft Johnson et al. 2005; Irungu et al. 2008; Jean et al. 2012; MacPhail et al. 2007; Weiser et al. 2006) Reliance on cross sectional data where predictor variable s and testing history are collected alongside each other makes it is difficult to draw conclusions about causation or temporality. Additionally, a single HIV test is not enough to prevent HIV, especially in a setting where dense sexual networks of concurre nt partnerships are common and riskier than serial monogamy (Morris and Kretzschmar 1997; Morris and Kretzschmar 2000) As more people begin to learn their status and as testing shifts to a more normative practice, it is likely that a substantial proportion of HIV testing clients will have tested before (Bradley et al. 2011) Thus, it becomes im perative to study the factors that influence regular uptake of HIV testing throughout young people's reproductive years. This dissertation will examine a 16 month period of testing activity. 2. Studies that use data from both partners strengthen the evidence. Given the important role of marriage on couple health, there have been growing calls to study how spouses mutually influence each other's behaviors using data from both partners (Lewis et al. 2006) A couples' dataset allows for the ability to predict respondent outcomes (i.e., the dependent variable) using both the respondent and their partner's independent variables yielding more information that if the respondent was studied in isolation. 3. Men receive less attention in the HIV testing research, particularly on disclosure, from sub Saharan Africa. This may be for a number of reasons and partially attributed to the ea se of access

PAGE 57

42 to pregnant women through clinic recruiting and the feminization of the AIDS epidemic. Lower rates of testing among men in Malawi (National Statistical Office & ORC Macr o 2005, 2011) underscores the n eed to focus on issues that are pertinent to men as well. 4. Studies on relationship level factors and HIV testing in Malawi are generally lacking. Decisions to test and disclose are often made in a dyadic context. Therefore, improvements to testing services requires a better understanding of the relationship level barriers and motivating factors that affect whether people get tested and disclose their results to sexual partners. 5. Few studies ex amine factors that influence disc losure among people who test negative. From an HIV prevention perspective, it is certainly understandable that greater attention would be directed towards HIV positive people who choose to disclose. However, an understanding of the drivers of disclosure am ong HIV negative people is also important so that young people can more accurately assess their HIV risk. Furthermore, increasing the rates of disclosure among those who are negative is important for establishing positive relationship behaviors that may fa cilitate disclosure of a positive test result if the time comes in the future. Thus, an understanding of the relationship factors that influence disclosure of a negative status is also warranted. The Malawi Context Present day Malawi is a small, landloc ked country located in southeastern Africa whose geography is largely structured by Lake Malawi, a fresh water lake that spans almost the

PAGE 58

43 entire length of the country Like its neighboring countries, Malawi has a long history of colonialism. From 1891 to 1 964, the British ruled Malawi as a protectorate state under the name Nyasaland. In the early 1960s, Malawi achieved p olit ical independence from Britain and became a one party state with Dr. Hastings Banda as its first President. Banda's regime came to an e nd in 1994 when the country entered a period of multi party politics which interacted with the influence of donor aid and ultimately resulted in the failure of the private market (Peters, 1997). Today, Malawi has a population of 13.1 million (National Statistical Office & ORC Macro 2011) and a Gross Domestic Product (GDP) per capita of $893 USD (The World Bank 2011) Lilongwe and Blantyre are its two major cities and the majority of the population resides in rural areas. Mo re than 68% of Malawians over the age of five are considered literate, with lower rates among women than among men (National Statistical O ffice & ORC Macro 2011) Girls are significantly less likely to complete primary school and go on to attend secondary or higher education as compared to boys in part, due to early childbearing At current fertility levels, a Malawian woman would bear an average of 5.7 children over the course of her lifetime (National Statistical Office & ORC Macro 2011) T he economy of Malawi is primaril y driven by agriculture, which accounts for 30% of the country's GDP (National Statistical Office & ORC Macro 2011) Nonetheless, at the national level, the economy depends on substantial inflows of economic assistance from the International Monetary Fund, World Bank, and individual donor nations.

PAGE 59

44 The country is divided into three regions (north, central, and south) and 27 districts. This dissertation was conducted in the Balaka district of southern Malawi (as indicated by Figure 2.3) Balaka has a hot, dry climate given its lower elevation and distance from Lake Malawi. Residents of the surrounding Balaka district villages travel to the Ba laka boma (or district center) for commerce and health services. The Balaka boma consists of a thriving market, several grocery stores, banks, and bars, a soccer field, several regional non governmental organization (NGO) offices, the main Balaka district hospital, a combination of public and private health clinics, and numerous churches and mosques including a large Catholic church that attracts many pilgrims and visitors from across the country. Figure 2.3 : Balaka District of Southern Malawi

PAGE 60

45 The th ree regions of Malawi differ greatly with regard to socio economic status, ethnicity, religion, level of polygyny, lineage system, and pattern of residence. For both genders, education levels are higher in the southern region than in the central region, bu t lower in comparison to the north (National Statistical Office & ORC Macro 2011) The major ethnic groups in the northern region are the Tumbuka, Ngoni, and Tonga. The Chewa and Ngoni tribes predominantly populate the central region. The southern region is home to the Yao, Lomwe, Sena, and Mang'anja peoples (Zu lu 1996) The north is predominantly Protestant, the central region is a combination of Protestants and Catholics, and the south is predominantly Muslim. At the national level, about 86% of women and 84% of men are Christians, while 13% of women and 12% o f men are Muslims (National Statistical Office & ORC Macro 2011) Polygamous unions are more common in the Christian north than in the pr edominantly Muslim south (14% vs. 6%) (National Statistical Office & ORC Macro 2011) In general, the southern region follows a matrilin eal/ matri local orientation where men physically move in with their wives' families after marriage as compared to the predominantly patrilineal/ patri local north (Chimbiri 2007; Peters 1997) In matrilineal areas, villages are typically org anized into clusters of compounds made up of houses consisting of matrilineal relatives. For example, groups of sisters live together with their respective husbands and children However, residence patterns are not always uniform ly distributed by region fo r reasons related to migration, temporary employment, and scarcity of land (Peters 2010) In matrilineal families, women determine the lineage structure, however, i nheritance is still passed through th e maternal uncle who owns and controls the inherited property (Phiri

PAGE 61

46 1983) The southern region of Malawi has higher rates of divorce as compared to other regions with around 33% of all marriages ending before their 5 th year anniversary (Reniers 2003) which may be partially attributed to its matrilineal marriage system. Gender Relations and HIV/AIDS in Malawi M atrilineal succession and inheritance pro vides women with considerable authority, particularly over land, however, this factor alone does not guarantee gender equality M ost of the major external influences on the region over the past 200 years have come from patrilineal and patriarchal groups (Peters 1997) These groups include, but are not l imited to, British colonialism Christian missionaries, and more recently, the international AIDS enterprise (refer to Peters 2010 for more detail on the former two groups of influence) During the colonialism period, Europeans obtained large plots of land from local chiefs at a very low cost and forcibly recruited l ocal laborers some of whom had previously resided on the land t o tend to the ir agricultural estates in return for rent (1997) With th em, colonialists brought their conceptions of patriarchy and forced this gender hierarchy upon their newly designated constituents. Pauline Peters writes, "estate owners assumed and promoted men as the natural holders of land and heads of households so th ey assigned authority to men" (1997). Yet, in southern Malawi, this conflicted with the existing matrilineal organization of the resident population and led to expansive population growth as daughters married and were joined by their new husbands and futur e children on the estates. Such rapid population growth could not be supported by the land's agricultural production and resulted in famine in many parts of the Shire Highlands in southern Malawi.

PAGE 62

47 Also i n s outhern Malawi, Kaler (2001) discusses how patriarchal traditions were in vented not necessarily through time honored practice s but rather through the interactions between local chiefs and colonial administrators. Colonial adminis trators sought to define local customs in southern Africa as something that paralleled British law. Chiefs saw that it was advantageous for them to maintain the perception that tradition was responsible for women's subservience to men and for male elders and chiefs to hold power over land and other resources. The colonialist roots of patriarchy are like l y to a ffect modern gender relations in Malawi. Demographic d ata from 2010 show that a pproximately 72% of households are considered headed by men (National Statistical Office & ORC Macro 2011) The upper ranks of village chiefs are mostly male (Peters 2010) Similarly, higher level positions within government, civil service, and private corporations are overwhelmingly mal e (with the exception of Malawi 's current President Joyce B anda, who served as Vice President at the time President Bingu wa Mutharika di ed in office in 2012 and therefore became his legiti mate successor ). Across sub S aharan Africa, adult HIV prevalence rates ra nge from 0.1% in Comoros to 26 % in Swaziland (UNAIDS 2010) Malawi has some of the highest rates of HIV infection in this region, with around 11% of all adults of reproductive age infected (National Statistical Office & ORC Macro 2011) Of the 42 sub Saharan African countries with data, only eight have higher adult HIV prevalence rates than Malawi (UNAIDS 2010) All eight countries are located in southern Africa. In Malawi, HIV prevalence among adult women is higher t han their male counterparts ( 13% of women vs. 8% of men) (National Statistical Office & ORC

PAGE 63

48 Macro 2011) Young men and women aged 15 24 a lso have high rates of HIV infection Approximately 3% and 5% of young people aged 15 19 and 20 24, respectively, tested positive for HIV in 2010 (National Statistical Office & ORC Macro 2011) The s outhern region of Malawi has the highest rates of HIV infection in the country, with an estimated 15 % of its reprod uctive age population infected (National Statistical Office & ORC Macro 2011) In Malawi, the connection between relationship power and HIV/AIDS is conflicted by two competing bodies of literature with varying degree s of support for the vulnerability paradigm. In one set of research, several qualitative studies on Malawian women's risk for HIV argue that women do have the agency necessary to navigate the HIV epidemic and are not just helpless victims of gender inequal ity (Schatz 2005; Tawfik and Watkins 2007) Here, women invoke HIV p revention strategies that are considered locally, instead of globally, appropriate for their everyday realities. For example, spouses use subtle and gender specific communication strategies to encourage fidelity in their marriages (Watkins, Rutenberg, and Wilkinson 1997; Zulu and Chepngeno 2003) Other researchers argue that women draw upon their social resources to protect themse lves from HIV/AIDS including bringing in marriage mediators, confronting male partners' mistresses directly, and leaving a partner who refuses to reform (Schatz 2005; Watkins 2004) Some evidence from research on pre marital partnerships illustrates how young women continuously evaluate a partner's risk for HIV using their social networks and then formulate or terminate relationships to mini mize their chance of infection (Poulin 2007)

PAGE 64

49 In sharp contrast, a second set of research suggests that women's decision making power around HIV/AIDS is severely limited at the societal level. In rural areas, it has been documented that married women dominate domestic areas arou nd cooking and childbearing while their husbands control domains like money and sex (Mbweza, Norr, and McElmurry 2008) While power is most likely expressed through multiple domains, it is important to point out that female dominated domains such as cooking are less i mportant for HIV transmission than male dominated domains of sex. Women's power in Malawi likely varies, but still some w omen report that they have little control over their relationships such as the ability to choose their husbands, to bear children, and to have sex or not (Rankin 2001; Lindgren, Rankin, and Rankin 2005) In fact, a recent demographic report featuring a nationally representat ive sample shows that 25% and 28% of Malawian women reported ever experiencing sexual and physical violence, respectively (National Statis tical Office & ORC Macro 2011) HIV Testing and Treatment in Malawi HIV testing and counseling (HTC) first became available in Malawi during the 1990's, although such services were widely inaccessible to the majority of rural Malawians. Starting in 2004, the Malawi Ministry of Health received external donor funding to support free HIV testing services in the main district hospitals and rural clinics (Ministry of Health [Malawi] 2005, 2006) Around the same time period, Malawi expanded its prevention of mother to c hild transmission (PMTCT) services by offering routine HIV testing of both mother and child. In 2003, the government mandated routine HIV testing of all pregnant women

PAGE 65

50 through antenatal care clinics (Office of the President and Cabinet and National AIDS Commission [Malawi] 2003) Today, HTC is offered throu gh integrated heath services such as antenatal care and at standalone testing centers, clients' homes, and workplace sites among other venues. As of 2010, 73% of women and 53% of men of reproductive age had eve r been tested for HIV reflecting a significan t increase in the number of people who know their status from previous years (National Statistical Office & ORC Macro 2005, 2011) Many of Malawian women are now tested though opt out, provider initiated testing during antenatal care, which is generally perceived to be compulsory (Angotti, Dionne, and Gaydosh 2010) Some research suggests that rates of antenatal care testing among expecta nt mothers who receive antenatal care may be as high as 99% (Weir, Hoffman, and Muula 2008) Few data are available on the prevalence of re occurring HI V testing in Malawi; however, I suspect that the young peop le in this study will likely have been tested multiple times over the course of 16 months as they negotiate the circumstances of their relationships and begin to have children. Across the region, including Malawi, couples based VCT (or CVCT) has gained little momentum. In their randomized controlled trial of CVCT in Tanzania, Becker and colleagues (2010) had to stop enrollment prematurely because rates of acceptance among the intervention arm (CVCT) were significantly lower than in the control arm (VCT), thus posing ethical issues. Low levels of couple involvement in VCT have been supported elsewhere, includi ng in Na irobi and Lusaka (Farquhar et al. 2004; Semrau et al. 2005) For

PAGE 66

51 women, factors associated with reluctance towards couples testing include male dominance over testing decision making and fears of the negative consequences of testing positive, such as divorce or vio lence (Maman, Hogan, and Kilonza 2001; Kranzer et al. 2009; Njau et al. 2011) Although notably, Becker and colleagues demonstrated that for couples who did test together, rates of marital dissolution and domestic violence were substantially lower than those who tested individu ally (2010) Men, on the other hand, may be reluctant to test with their partners because they feel embarras sed about being seen in the clinic or because of a heightened sense of risk related to extramarital relationships (Gipson et al. 2010; Njau et al. 2011) Home based CVCT provides an alternative solution by circumventing public testing venues or so called "women's places ." In Malawi, some research su ggests that home based testing is perceived to have many benefits: it is confidential, convenient, and credible (Angotti et al. 2009) Indeed, a study on Likoma Island in Malawi found that rates of home based VCT were exceptionally high, especially among poorer households (He lleringer et al. 2009) H ome based couples testing may offer new opportunities for overcoming barriers related to clinic based testing (Njau et al. 2011) The key benefit of HIV testing, howev er, depends up on reliable access to ART In Malawi, ART first became available on a fee basis ( for pay ) in public hospitals in 2000. Although at the time, only a small group of patients could afford them (Van Oosterhout et al. 2007) During early rollout, clinical shortcomings, drug supply interruptions, and difficulties determining eligibility imposed limitations on quality of care and access to ART.

PAGE 67

52 With the assistance of donor funding, ART was scaled up between 2004 and 2006 at large hospitals and offered free of charge. Adults were eligible i f they were HIV positive and WHO clinical stage 3 or 4, or had a CD4 count less than 200 cells/L (Ministry of Health [Malawi] 2003) Over this period, the number of patients on ART within the public sector increased from approximately 4,000 to 60,000 (Lowrance et al. 2007) In 2009, almost 200,000 people were taking ART through 377 health facilities (Ministry of Health [Malawi] 2009) The WHO recently estimated that around 48% of adults in immediate need of ART currently receive treatment in Malawi (WHO 2010) E stimates from Ma lawi using self reported data from 2010 show that between 24% and 62% of women who tested positive for HIV are currently taking ART (Natio nal Statistical Office & ORC Macro 2011) However, recently, a new policy was introduced in Malawi referred to as Option B+, which allows all pregnant women who test HIV positive to be placed on ART for life regardless of their CD4 counts or clinical stag e ( Schouten et al. 2011) This policy is expected to dramatically increase the number of women on ART. For men, coverage estimates range from 16% to 60% (MDHS, 2010). Overall, c hallenges still remain for ART programs in Malawi in order to overcome clinical p roblems with ART staging and eligibility, continuity of care after testing positive, and staff shortages (MacPherson, Lalloo et al. 2012; Makwiza et al. 2009; McGrath et al. 2010) Nonetheless, the ART program is M alawi is widely heralded as a success in the region (Harries et al. 2011)

PAGE 68

53 CHAPTER III METHODOLOGY In this chapter, I outline the research design and methods I used to accomplish the following goals: (1) to investigate appropriate quantitative measures of relationship power for the Malawi context (a) and to evaluate their cultural relevance with qualitative data (b) ; (2) to test whether relationship power infl uence s the use of HIV testing services (a) and disclosure of test results to primary partners (b) ; and (3) to understand the meaning of HIV testing within the context of a sexual relationship. Study Design Overview The present research used a sequential mixed methods design with three complementary phases: a measure development phase (qualitative and quantitative) to develop a measure and conceptual model of relationship power, a hypothesis testing phase (quantitative) to test whether the measure of rela tionship power predicts HIV testing behavior, and an interpretative phase (qualitative) to offer context for the quantitative findings. In the first phase, I develop ed a pilot measure of relationship power for the Malawi context and then re formulate d it u sing a larger set of couple data collected as part of the TLT study. Then, I applied grounded theory (Glaser and Strauss 1967) to a set of qualitative da ta in order to understand relationship power from rural Malawians' perspectives This process allow ed me to confirm or challenge the quantitative measure s of power and to uncover other areas of nuance In the next phase, I test ed for associations between t he final model of relationship power and HIV testing behavior. Here, I utilize d longitudinal couple data to test hypotheses related to relationship

PAGE 69

54 power and two HIV testing behaviors: HIV testing uptake over time and disclosure of test results to a primar y sexual partner. During the interpretive phase, I again applied a grounded theory approach to qualitative data to explain, cross validate, or challenge the quantitative findings from the hypothesis testing aim. During the fall of 2011, I spent two months in the Balaka district of southern Malawi conduct ing focus group discussions ( FGDs) with men and women and gathering o ther contextual data Previously collected couple interview data from 2009 supplemented the FGD data as necessary. I used a mixed method s design for important reasons that added significant value to the study. As I have already discussed, relationship power is a complicated construct that carries multiple meanings at many different levels of influence, which vary not only according to geog raphical locality, but also over time. As such, understandings of power and its hypothesized associations with HIV testing are likely to benefit from multiple, complementary researc h methods including surveys, focu s group discussions couple interviews, de tailed field notes from several trips to Malawi, and participant observation, includi ng informal conversations with Malawians about issues related to power and HIV testing Survey measures of power are limited in capturing the depth, complexity, and meanin gs behind relationship power. Not only could qualitative interviews provide insight into the culturally rooted meanings of relationship power, interviews may uncover other underlying constructs or variables that have not yet been identified through previou s survey research. Informal data gathering allowed me to observe the more nuanced forms of power that may or may not arise during more formal interviews or FGDs. While the qualitative

PAGE 70

55 results are helpful to identify new associations, they are not generaliz able (at least in the statistical sense) to the larger population of young people at risk for HIV. In addition to playing off the strengths and weaknesses of multiple methods, I could also show if the results were cong ruent, that is, the findings were not simply due to a methodological artifact (Morgan 1998) I characterized this study as a sequential mixed methods study where one method precedes another as opposed to simultaneous data collection. Throughout the study, I constantly compared and contrasted the various forms of information in order to formulate and revise my understandings of relationship power and HIV testing. The model of relationship power finalized in the measurement phase was used in the hypothesis testing phase. Preliminary findings from the hypot hesis testing phase informed the line of questioning I use d in the FGDs. Focus group discussions, field notes, and the previous qualitative couple interviews I collected were examined together and used to reflect back on what was learned through the survey data. Figure 3.1 provides a timeline for when the data w ere collected and analyzed.

PAGE 71

56 Figure 3.1: Timeline of Data Collection and Analysis, 2009 2013 The TLT Study The quantitative data for the measure development and hyp othesis testing aims came from Tsologo La Thanzi (TLT; "Healthy Futures" in Chichewa). The overall objective of the TLT study was to collect and analyze new data on the reproductive goals and behaviors of young adults in Malawi within the context of HIV te sting and counseling. Longitudinal survey data were collected at quarterly intervals over a period of approximately three years (for a total of 8 waves). Trained interviewers matched by gender to the respondents administered questionnaires through face t o face interviews conducted in private rooms at June 2009 Conducted semi-structured interviews on power with 34 coupled individuals July 2009 Developed preliminary measure of power and pilot tested it on 254 individuals May 2011 Began the hypothesistesting phase using TLT data Conducted 8 focus group discussions in Malawi September 2011 Finalized the power measure July 2011 Used the semi-structured interviews and focus groups to reflect on the measure of power; completed hypothesis testing phase; used qualitative data to explain the quantitative findings January 2012-April 2013

PAGE 72

57 the TLT research center in Balaka. The survey data included information on reproductive goals and behaviors, fertility preferences, contraceptive use, and sexual behavior. Previously administered TLT sampli ng procedures were as follows. A random sample of 1500 women aged 15 24 was selected from a household listing, which included people who lived within 7 kilometers of the Balaka town center ( boma) Women were given enumerated incentive tokens for each partn er that they name d (husbands and boyfriends) and asked to give the tokens to their partners, redeemable for 500 Malawi Kwatcha (around $3.25 USD) upon completion of the interview. Women could initially recruit up to three partners; however, since it was ex pected that some relationships would dissolve over the 3 year study period, new partners could be enrolle d at each successive interview. In addition to the random sample of women, a smaller random sample of men aged 15 24 was also recruited and enrolled in order to be able to make generalizations about men. Combined, a pproximately 2500 individuals (500 couples) were enrolled in the study, although these figures vary slightly with each wave due to attrition and partnership dissolution or formation. However, the final couples sample (details to follow) only included these randomly selected men if they were partners of the random sample of women. Women and their partners were randomly assigned to receive HIV testing in order to assess causal impacts of changes in the knowledge of HIV status on reproductive and prevention goals. Once baseline enrollment was completed, the women were assigned to three equal study groups consisting of 500 women and their male partners. Group 1 received regular HIV testing every fou r months. Group 2 received an HIV test at the end of the first

PAGE 73

58 year and then again at the end of the study. Group 3 only received an HIV test at the end of the study. Figure 3.2 illustrates the timing and allocation of VCT over 8 waves of data collection. Figure 3.2 : Overview of the VCT Infrastructure for TLT Creation of the Analysis Datasets In waves 3 and 5, a special module on relationship power was added to the TLT partnership survey (as sho wn in Figure 3.2 ). Respondents were asked the relationship power statements if they reported a current serious sexual partner including a spouse, live in partner, steady boyfriend/girlfriend, or new boyfriend/girlfriend. If a respondent had more than one currently ongoing sexual relationship, the relationship pow er questions were asked with regard to the most serious relationship, ascertained with the question: "Of the sexual partners that you specified earlier, which one are you closest to?" However, if the respondent was married, their spouse automatically serve d as the reference partner even if other extramarita l relationships had been ongoing for quite some time. A couple dataset was created for respondents who answered the power questions about each other at wave 3 using a separate database linking women and their male partners. The following steps were taken:

PAGE 74

59 1. Respondents completed a baseline questionnaire eliciting demographic and other information when they first enrolled in the study. At each wave, respondents completed a wave specific questionnaire and pa rtnership questionnaire. Respondents who entered the study at a subsequent wave only completed the baseline questionnaire at that wave and then at the next wave, completed the wave specific and partnership questionnaires. Note that at wave 1, respondents c ompleted all questionnaires. All respondents who entered the study at wave 2 and completed baseline questionnaires (N=199) were appended to respondents who participated at wave 1 (N=2,496). This created a sample with complete baseline data by wave 3 (N=2, 695). 2. This dataset was merged with respondents who participated in wave 3 (N=2,462). Of these respondents, a total of 1,510 respondents were eligible (reported an ongoing sexual relationship) and completed the power statements. This dataset was called as t he wave 3 individuals dataset. Each individual was represented only once in this dataset with a unique individual identifier. Example data are illustrated in Table 3.1. Table 3.1: Example data from the wave 3 individuals dataset

PAGE 75

60 3. A second dataset call ed the wave 3 couples dataset was used to identify a set of ongoing couples at wave 3. The couples dataset contained the following information: a unique respondent ID for each individual, a unique couple ID, wave number (1 to 8), two variables correspondin g to the status of the relationship reported by both partners at that wave, and a variable indicating when the relationship was first formed. The process below resulted in a list of couples who completed the power statements at wave 3. a. To create a baselin e set of couples for wave 3, records for other waves were dropped (i.e., the wave number does not equal 3). b. Couples at wave 3 in which either member reported that the relationship was dissolved, unknown, or missing were dropped. A code of "3" indicated a n ongoing relationship. c. Couples who were first matched at wave 3 were dropped since new partners would not have completed the wave 3 and partnership questionnaires (only baseline). This process resulted in 493 couples. Example data are illustrated in Table 3.2. Table 3.2: Example data from the wave 3 couples dataset

PAGE 76

61 d. The list of couples was transformed from wide to long fo rmat so that each row represented an individual (no duplicates were present). The wave 3 couples dataset was then merged with the wave 3 individuals dataset using the respondent ID. Respondents who completed the power statements were dropped if they were not matched in the wave 3 couples dataset by using an internal Stata "merge" variable. Individuals could have completed the power s tatements in reference to a past partner or to a partner who never showed up at wave 3 to participate in the study. This process resulted in 466 couples or 932 individuals. This dataset was labeled as the final wave 3 couples dataset Example data are illu strated in Table 3.3. Table 3.3: Example data from the final wave 3 couples dataset e. The data w ere transformed back into wide format in order to create additional couple level variables and then back transformed again into long format to run the study ana lyses.

PAGE 77

62 4. For the longitudinal couple dataset used to test the hypotheses related to testing uptake, the final wave 3 couples dataset (N=932 respondents) was merged with questionnaire data from waves 4, 5, 6, and 7. This dataset was called the uptake dataset New respondents who enrolled at each subsequent wave were not included since they were not present in the final wave 3 couples dataset Each of the 932 individuals (466 couples) could have up to 5 records (some respondents did not participate at every su bsequent wave). The following steps were taken: a. The final wave 3 couples dataset was merged with the wave 4 questionnaire and partnership data. Variables that were collected at each wave, e.g., perceived likelihood of being infected with HIV, were renamed with a suffix corresponding to the wave so they were not replaced during the merge (e.g., risk_w3, risk_w4, etc.). This resulted in a new combined dataset linking the 466 couples from wave 3 with their corresponding data for wave 4. A similar process was c ompleted using wave 5, 6, and 7 data. b. In order to get the data from a wide format (e.g., each row contains an individual) to a long format (e.g., each row contains a wave), the data w ere transformed using the variable suffix "w" for all longitudinal varia bles. Time invariant variables (e.g., birth year, gender, etc.) were also carried over into the uptake dataset An internal "time" variable corresponding to each of the seven waves was also created in this transformation. Note that since the main predicto r variables (perceived risk, power, etc.) from waves 1 and 2 were intentionally not

PAGE 78

63 carried over during earlier data merges, the value of "time" at these waves is displayed as "missing" (indicated by a "." in Table 3.4) for all respondents. Example data ar e illustrated in Table 3.4. Table 3.4: Example data from the uptake dataset (long format) 5. For the dataset used to test the disclosure hypotheses, the final wave 3 couples dataset (N=932 respondents) was merged with questionnaire data from waves 4 and 5. This dataset is called the disclosure dataset (long format) The following additional steps were taken. a. At wave 5, respondents from the baseline set of respondents (N=932) were dropped if they did not participate in the study at wave 5 (N=65) and had neve r tested for HIV as reported at wave 5 (N=64). This process resulted in 803

PAGE 79

64 respondents. Given that the analysis was stratified by gender, respondents were retained even if their partner did not meet the above criteria (as show n in Table 3.5). b. Given that there was a time lapse between wave 3 and 5, it was important to verify that respondents at wave 5 were reporting disclosure information in relation to their partner from wave 3. For some respondents, their marital status changed over the period from wave 3 to 5. Of the 803 eligible respondents at wave 5, 21 respondents had divorced, 6 respondents had separated, 2 respondents had lost their spouses, and 20 respondents became mar ried. Two additional respondents whose spouses had died were dropped since they could not possibly report on disclosure to their partner at wave 5. c. For those who reported being divorced or separated at wave 5, many still reported that they had disclosed their results to their spouse. I assumed that those who reported being divorced a t wave 5 reported disclosure to a spouse whom they were married to in wave 3. The chances of divorcing a spouse from wave 3 and remarrying another by wave 5 over a short eight month period would be rare. But for those who were newly married by wave 5, it w as necessary to ensure that they had married the same partner from wave 3. Otherwise, a respondent could have reported on two different partners at wave 3 and 5. For those 20 respondents, I manually checked the couple database from wave 5 to verify that th e relationship was indeed ongoing. All of these relationships were

PAGE 80

65 ongoing at wave 5, indicating that respondents had married their partner from wave 3 with one exception. The couple database indicated that one of the relationships had "dissolved" at wave 5, however, the female respondent was not paired with another partner whom she had supposedly married. This woman and her partner from wave 3 were dropped from the analysis given the uncertainty that they may not have been reporting on each other at wave 5 This process reduced the sample size to 799 (432 women, 367 men). Example data are illustrated in Table 3.5. Table 3.5: Example data from the disclosure dataset (long format) Measure Development Phase Design and Rationale For specific aim 1 a I re formulated a measure of relationship power that was previously developed for the Malawi context during a pilot study in 2009. The pilot data demonstrated the feasibility of a Malawian relationship power scale, however, more research was needed to co nfirm the psychometric properties of the scale using a larger sample with a !"#$%& ()*$+"',-' ."/&"0' 1,2'#343*#'43' 546"'7' -%#8+)#*0"'3)' $403/"0'43' 546"'9' !"#$"""% "##$% &'()% #% #% ""#*+#% "##*% ,)-'()% #% "% !"#*+#"% "##*% &'()% #% #% """./!% "#"#% ,)-'()% #% "% !""./!"% "#"#% &'()% #% #%

PAGE 81

66 more diverse set of power experiences (Conroy 2010) I had expected that such a sample would minimize the impact of random error on reliability estimates (Singleton and Straits 2005) and improve the scale's reliability. First, I describe the process used to create the power measure, starting from the qualitative interviews conducted in 2009. Power Measure Development In 2009, I conducted semi structured qualitative interviews with 34 coupled men and women simultaneously, but separately, using trained Malawian research assistants. The interviews elicited multiple dimensions of relationship power based on Connell's (1987) Theory of Gender and P ower (e.g., relationship control, economic dependence, decision making dominance, and social norms). The qualitative data were analyzed to create a preliminary pool o f statements on power. Additional details of the semi structured interviews can be found later in this chapter. Face validity was addressed by consulting with academic scholars, Malawian key informants, and TLT interviewer staff and then the power stateme nts were added, deleted, or reworded accordingly. The power statements were translated from English to Chichewa and reverse translated by two separate individuals unfamiliar with the study in order to ensure sentence meaning was preserved. Items were both positively worded (e.g., My partner shows that they care about me.") and negatively worded (e.g., My partner punishes me when he/she is really angry with me."). Cognitive interviews (Tanur 1992) asking respondents to "think aloud" as they responded to the power statements were administered to a separate convenience sample of y oung adults (n=8) in order to detect comprehension

PAGE 82

67 and translation problems. This process resulted in 31 power statements. Response choices were based on a 4 point Likert scale (4=strongly disagree, 3=disagree, 2=agree, or 1=strongly agree). The power sta tements were administered through face to face interviews using a small pilot sample of 254 individuals. The sample was drawn from the same six target village s used in the qualitative phase Study participants were between the ages of 18 and 45 years and h ad a primary sexual partner. Research assistants started at the village chief's home, usually centrally located within the village and approached every third compound to recruit respondents. Interviews took place in a quiet, private location usually near the respondent's primary residence. An initial exploratory factor analysis with oblique rotation was performed to reduce the set of items down and identify the underlying constructs. Exploratory factor analysis is appropriate when you have obtained measure s on a number of variables and want to identify the number and nature of the underlying factors that are responsible for the covariation in the data (Hatcher 1994) A scree plot suggested four meaningful factors so only these factors were retained. All items receiving a factor loading of less than 0.3 0 were dropped. The results showed that four items loaded on the first factor, which was subsequently labeled "autonomy ." Four items loaded on a second factor, which was labeled "communication ." Four items loaded on the third factor, which was labeled "lov e and trust ." Finally, four items loaded on a fourth factor, which was labeled "relationship dominance ." Table 3. 6 contains

PAGE 83

68 the final factor pattern for the preliminary relationship power scale Refer to A ppendix A for the full wording of the 16 items. Ta ble 3.6: Final rotated factor pattern for p reliminary r elationship power s cale The 16 item power scale accounted for the majority of variation in responses. The proportion of variance was determined to be 40%, 39%, 32%, and 32% for the autonomy, commu nication, love and trust, and relationship dominance factors, respectively (see Table 3.6 ). The r eliability of the scale was determined by computing Cronbach's alpha for the entire power scale and for each of the four factors separately (Cronbach and Meehl 1955) The overall power sca le demonstrated moderate reliability of 0.58, which falls within the Item a Factor 1: Autonomy Factor 2: Communication Factor 3: Love and Trust Factor 4: Relationship Dominance Would never leave 0.48 b -0.19 0.11 0.11 In trouble if partner left 0.41 0.02 -0.01 0.23 Would leave if really bad 0.53 0.06 -0.01 0.02 Could find another partner 0.54 0.08 0.17 -0.07 Discuss matters together 0.02 0.39 0.09 0.09 Partner cares about me -0.19 0.35 0.25 0.31 Talk to partner about affair 0.04 0.65 0.00 -0.06 Consult advisors if problems 0.04 0.61 -0.20 -0.06 Helps me with needs -0.17 -0.09 0.43 0.06 Able to initiate sex 0.15 0.06 0.49 0.03 Able to buy expensive items 0.15 -0.07 0.53 -0.06 Have own money 0.11 -0.10 0.44 -0.18 Partner punishes me 0.08 -0.04 -0.12 0.58 Partner chooses relatives side -0.06 0.02 -0.06 0.41 Partner having an affair -0.17 0.02 0.07 0.36 Partner might beat me 0.14 -0.12 -0.01 0.53 Proportion of variance 0.40 0.39 0.32 0.32 a Scale items have been summarized into shorter descriptions for readability. Refer to Appendix A for actual scale items. b Factor loadings greater than 0.30 are in bold print.

PAGE 84

69 recommended range of 0.50 to 0.60 for early stages of research (Nunnally 1967) Scale reliability was also addressed by gender since future researchers may desire to study men and women's responses to the scale items separately (see Table 3 .7 ). Reliability of the overall scale was similar for men and women with coefficient alphas of 0.55 and 0.57, respectively. Reliability estimates were 0.59, 0.59, 0.57, and 0.54 for the autonomy, communication, love and trust, and relationship dominance subscales, respectively. Table 3.7 also presents the mean factor scores computed from a range of 1 (strongly agree) to 4 (strongly disagree) for each of the subscales. Mean factors s cores were computed for the entire sample and by gender Higher mean factor scores (i.e., more likely to strongly disagree with scale items) are indicative of higher relationship power. Mean factor scores were higher for men on all the subscales with the e xception of communication, suggesting that women had higher relationship power in this domain as compared to men. Table 3.7: Means, standard deviations, and reliability c oefficients for the p reliminary r elationship power m easure Factor Mean 95% CI Alpha Mean 95% CI Alpha Mean 95% CI Alpha Dependence/Autonomy 2.22 2.13-2.32 0.59 2.55 2.43-2.68 0.57 1.90 1.78-2.00 0.43 Communication 3.57 3.51-3.63 0.59 3.46 3.39-3.54 0.49 3.68 3.59-3.76 0.66 Love and Trust 2.77 2.68-2.86 0.57 3.16 3.07-3.25 0.28 2.38 2.25-2.51 0.46 Relationship Dominance 2.82 2.73-2.91 0.54 3.03 2.92-3.14 0.47 2.61 2.47-2.75 0.54 Overall (n=254) Men (n=127) Women (n=127) a Mean refers to the mean value of the factor score. The factor score was computed by taking the sum of the row (where 1=Strongly Agree, 2=Agree, 3=Disagree, and 4=Strongly Disagree for each scale item) divided by the number of scale items answered for the row. Positively worded items were reverse coded prior to calculating the row value.

PAGE 85

70 The preliminary relatio nship power scale items were then placed on TLT's wave 3 and wave 5 partnership surveys. In order to confirm the underlying constructs using a larger, more representative sample, I conducted a second exploratory factor analysis with oblique rotation using the final wave 3 couple dataset (N=932 individuals). I applied rotation to force variables to load more strongly on a given factor, thus making it easier to interpret the data. Oblique rotation was used since factors are most likely to be correlated with e ach other, thus yielding a more accurate representation of the data (Adock 2006; Hatcher 1994) However, both oblique and orthogonal rotations yielded the same factor structure. Factors were retained if eigenvalues were greater than one and as suggested by the scree plot. Scale items were retained if factor loadings were greater than 0.40. I then performed two separate factor analyses by gender to see if similar factor patterns were present for men and women. Next, I tested for differences in the final factors and item scores between men and women using two group mean comparison t tests. I computed Cronbach's alpha for each subscale by gender (Cronbach and Meehl 1955) Refer to Figure 3. 3 for a visual representation of the steps taken to develo p the relationship power measure.

PAGE 86

71 Figure 3.3 : The Relationship Power Scale Development Stages The Final Relationship Power Measure Three items loaded on a first factor and 4 items loaded on a second factor, which were named unity ("My partner shows they care about me"; "When I need my partner's assistance, he/she is there to help me"; and "My partner and I discuss important matters together") and discordance ("If my partner was really angry with me, he/she might beat me"; "My partner punishes me whe n he/she is angry with me"; "When I disagree with my partner's relatives, my partner chooses their side over mine"; "My partner is probably having sex with someone else"). Table 3. 8 presents the distribution and descriptive statistics for the relationship power subscales and corresponding items (total sample). Missing data were negligible. All seven scale items ranged in value from 1 to 4. For the three unity items, the majority of

PAGE 87

72 respondents either agreed or strongly agreed with the statements (upwards o f 92%). Responses were more equally distributed between agreement and disagreement for the discordance items, particularly for "punishes me" and "might beat me", and "sides with relatives ." For the "having an affair" item, the majority of respondents disag reed or strongly disagreed with the statement (85.7%). The final scores for the subscales were calculated using the mean of the items. Higher scores meant more agreement with the statements. For the unity subscale, a mean of approximately 3.77 indicates t hat on average, people either agreed or strongly agreed with the items. For the discordance subscale, a mean of approximately 1.98 indicates that on average, people disagreed with the items (meaning less discordance). Table 3. 8 also provides data on norma lity by subscale and by individual items. Severe non normality was defined as >|3| for skew or >|8| for kurtosis (Kline 2005) Using these criteria, the mean scores for the unity factor were considered approaching a non normal distribution (skewness= 2.28; kurtosis=9.59). The mean scores for discordance were considered normally distributed (skewness=0.43; kurtosis=2.50). Table 3. 8 : D escriptive statistics for the relationship power subscales, TLT wave 3 couples sample (N=932) Scale item b Strongly Disagree Disagree Agree Strongly Agree Missing data Mean (SD) Skewness Kurtosis Unity factor 3.77 (0.41) -2.28 9.59 Cares about me 0.3 1.2 10.1 88.4 0 3.87 (0.40) -3.4 16.49 Helps me 1.9 6.1 19.3 72.6 0 3.63 (0.69) -1.91 6.24 Discuss together 0.4 2.0 14.4 83.2 0 3.80 (0.47) -2.64 10.69 Discordance factor 1.98 (0.71) 0.43 2.51 Punishes me 33.8 24.1 22.1 20.0 0 2.28 (1.13) 0.26 1.66 Might beat me 50.6 19.6 14.8 15.0 1 1.94 (1.12) 0.76 2.09 Sides with relatives 37.6 31.7 16.0 14.4 3 2.08 (1.05) 0.59 2.1 Having an affair 58.9 26.8 8.6 5.7 0 1.61 (0.87) 1.37 4.01 a 1=Strongly Disgree, 2=Disagree, 3=Agree, 4=Strongly Agree. b Refer to Appendix A for full wording of the items. Response (%) a

PAGE 88

73 Table 3. 9 presents the standardized factor loadings and reliability coefficients for the subscales and items. For the total sample, factor loadings for unity items ranged from 0.60 to 0.62, with women loading higher on the subscale than men. The discordance item factor loadings were generally lower than those for unity and ranged from 0.52 to 0.56 with the exception of the "having an affair" item that had a loading of approximately 0.37. For all four discordance items, men's loading s were slightly higher than women's loadings. These differences in loadings were also reflected in the reliability coefficients of the subscales for men and women. The Cronbach's alpha for unity was 0.65 (women: 0.74; men: 0.53) and for discordance was 0 .60 (women: 0.56; men: 0.64). While the latter three reliability coefficients fell below the recommended cutoff of 0.70 (Nunnally and Bernstein 1994) they still exhibited normal relia bility values for early stage research (Nunnally 1967) Table 3. 9 contains the factor loadings and reliability coefficients for each subscale. Table 3. 9 : Standardized factor loadings and c oefficient alphas for the power scale items Scale item Total Women Men Total Women Men Cares about me 0.62 0.70 0.51 Helps me 0.61 0.71 0.49 Discuss together 0.60 0.63 0.59 Punishes me 0.56 0.52 0.63 Might beat me 0.53 0.52 0.53 Sides with relatives 0.52 0.51 0.55 Having an affair 0.37 0.32 0.38 Coefficient alpha 0.65 0.74 0.53 0.60 0.56 0.64 Refer to Appendix A for full wording of scale items. Unity Discordance

PAGE 89

74 Gender differences were examined for the final two factors and scale items (range: 1 4). Table 3. 10 presents the mean scores, standard deviations, and statistical differences in the measures by gender. The mean score for unity was almost exactly t he same for men and women (3.77). Two group t tests revealed that men and women's responses were not statistically different from each other for the unity subscale or for the indiv idual unity items For the discordance subscale, h igher scores meant more di scordance (less power) Mean scores for the discordance factor were higher for women (2.12) than men (1.83) suggesting that women were more likely to experience discordance in their relationships. Two group t tests revealed that men and women's responses w ere statistically different from each other for the discordance subscale or for two of the discordance statements: "If my partner were angry with me, he/she might beat me" and "My partner is probably having an affair ." Figure 3.4 illustrates the iterative process used to develop the final model of relationship power that was used in the hypothesis testing phase. Table 3. 10 : Gender differences in relationship power factors Scale item a,b Total Women Men t test Mean (SD) Mean (SD) Mean (SD) p value Unity factor 3.77 (0.41) 3.77 (0.45) 3.77 (0.37) 0.937 Cares about me 3.87 (0.40) 3.86 (0.44) 3.88 (0.35) 0.460 Helps me 3.63 (0.69) 3.65 (0.67) 3.60 (0.70) 0.253 Discuss together 3.80 (0.47) 3.78 (0.51) 3.82 (0.43) 0.214 Discordance factor 1.98 (0.71) 2.12 (0.72) 1.83 (0.66) 0.000 Punishes me 2.28 (1.13) 2.35 (1.15) 2.22 (1.11) 0.088 Might beat me 1.94 (1.12) 2.32 (1.19) 1.56 (0.89) 0.000 Sides with relatives 2.08 (1.05) 2.07 (1.10) 2.08 (1.00) 0.924 Having affair 1.61 (0.87) 1.74 (0.91) 1.48 (0.80) 0.000 a 1=Strongly Disgree, 2=Disagree, 3=Agree, 4=Strongly Agree. b Refer to Appendix A for full wording of the items.

PAGE 90

75 In the hypothesis testing phase that follows, I ultimately decided to dr op the discordance factor in the statistical analyses. This was for several reasons. First, there appears to be two or more underlying constructs being measured in the discordance scale, including couple disagreement or conflict (or the punitive aspects of power such as being able to beat or punish a partner) and perceptions of a partner's infidelity. The item related to perceptions of a partner's infidelity had significantly lower factor loadings than the other three items and was conceptually distinct fro m the other three items. Hatcher (1994) argues that in order for a factor to be retained, one of the criteria is that all the variables share a conceptual meaning. Second, in my experience trying to publish the results of the scale, I learned that acceptable minim um levels of scale reliability for high quality journals range from 0.65 0.70. While the discordance subscale reliability was not terrible, it was not ideal either. Rather than pursing a path that I will eventually deviate from when trying to publish selec tions from this dissertation, I chose to examine several of the underlying constructs as single item measures, particularly, physical violence and perceived partner infidelity (as will be discussed). The findings above related to the discordance subscale w ere intentionally documented herein to demonstrate the process of creating a new scale and the decisions that n eeded to be made along the way in order to provide other researchers with a possible starting point for a measure of disc ordance for the Malawi c ontext.

PAGE 91

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igure 3.4: Iterative Stages of Development for the Relationship Power Model

PAGE 92

77 The Hypothesis T esting Phase Design and Rationale For specific aim 2, the objective was to test whether relationship power influence s the use of HIV testing services and disclosure of test resu lts to primary partners in accordance with the stu dy hypotheses documented in Chapter 2. My initial intention was to conduct a dyadic analysis for the hypothesis testing phase of this dissertation As I started to analyze the quantitative TLT data, preliminary findings changed my original analysis approac h. I first computed the Intraclass Correlation (ICC) for the unity and discordance measures, which is the first step in a dyadic analysis to evaluate non independence or the degree of similarity between two members of a dyad on the same variable (Kenny, Kashy, and Cook 2006) The ICC was computed through a large one way ANOVA using a unique couple identifier as the grouping variable. The ICC for unity showed that onl y 9% of the variance in unity was explained by the dyad, F (465, 466)=1.20, p =0.025. The ICC for discordance showed that only 6% o f the variance in discordance was explained by the dyad, F (465, 466)=1.12, p =0.12. What do ICCs tell us? The idea behind ev aluating a measure's ICC is to assess how similar two dyad members' responses are to the same measure a measure that is hypothesized to involve an interaction between the two people. With unity, for example, I would expect that if a wife reports higher lev els of communication, her husband would also be more likely to re port this, though certainly it would not be extraordinary for couple members to report divergent responses. For discordance, which includes perceptions of a partner's infidelity and physical violence, I would expect that if couples had high levels of

PAGE 93

78 reciprocal viol ence and mutual mistrust, they w ould provide similar responses to each other. However, the ICC for discordance showed that couple member's responses were even more independent than for unity Reciprocal violence only occurred in 1% of couples (or in 5.4% of all physically violent couples). Similarly, only 3% of all couples both reported that they believed there partners were having an affair. Thus, the majority of the time if one par tner was suspected of cheating the other partner was not suspected of the same behavior. The ICC values can be used to inform a particular analytic approach. Data with high levels of non independence would be best suited for a dyadic level analysis. Accor ding to Kenny and colleagues (2006) a dyadic level analysis controls for the non independent nature of two individuals providing more similar in formation on one measure than two individuals who are not part of the same couple. Individual level data that are actually hierarchical such as data from dyads may bias estimates when non independence is present (Kenny, Kashy, and Cook 2006) Another way to conceptualize the ICC is as a proport ion of the total variance in a measure that is explained by the dyad. For unity, only 9% of the total variance (within co uples and between couples) is explained by the dyad the ICC for discordance is even less (and also non significant). There is generally a lack of agreement about "how much" non independence is enough to warrant a dyadic level analysis. Cohen (1998) defines 0.50 as a large ICC, 0.30 as a medium ICC, and 0.10 as a small ICC. Dagne et al. (2002) argues that if there is little variance between coup les and the ICCs are close to zero, then it is possible to examine the data at the individual level.

PAGE 94

79 For the hypothesis testing phase of this dissertation, I chose to analyze the data at the individual level rather than at the dyadic level given that the c ouple data at least the main predictor variables for this study were largely independent. Multiple possible reasons may explain why the ICCs were found to be so low. One reason may relate to statistics. Given that the majority of the sample agreed or stron gly agreed with the unity statements, there may not have been enough variation overall to show differences between couples. In other words, if most people in the entire sample report ed that they "agree" with a given statement on a 4 point Likert scale, the n the ICC could be low because the two dyad members' responses were not more similar than any other two people in the entire sample. Now that I have outlined the rationale for analyzing the data at th e individual level I will justify the general approach I used to testing the hypotheses related to testing uptake and disclosure. For the HIV testing uptake models, I used longitudinal testing history data collected over a 16 month period (wave 4, 5, 6, and 7). This allowed me to follow respondents over time to see if relationship power, measured at wave 3, had an effect on test ing over each 4 month interval. Longitudinal analyses are generally thought to be superior to cross sectional analyses. C ross sectional studies are inherently limited at demonstrating c ausal linkages since directionality of cause and effect cannot be evaluated (Singleton and Straits 2005) If the data were captured at the same time, it would be difficult to determine whether having an HIV test changed power or whether power changed after getting tested. The background literature presented in Chapter 2 suggests that both directions are possible.

PAGE 95

80 For the disclosure m odels, I used the p ower measures captured at wave 3 to predict disclosure at wave 5 in order to help establish temporality and widen the time gap between w hen power and disclosure was ascertained. For both analyses (HIV testing uptake and disclosure), I stratified the analys is by gender because I perceived that the pathways would differ between men and women. Though the data are largely independent, gender stratification provided an additional level of protection again st any bias from non independent data. Sample Selection T wo separate datasets were used to test the hypotheses related to: 1) HIV testing uptake; and 2) disclosure to primary partners. For the HIV testing uptake models, I used the baseline sample of couples from wave 3 who completed the power module questions (N =932). Here, I refer back to the uptake dataset discussed earlier. For several reasons, I chose to use the couple dataset instead of a cross sectional dataset of all respondents who completed the power module First, use of the couple dataset ensured that only legitimate and serious couples answered the power questions about their partner, potentially increasing the validity of the data (e.g., respondents were not reporting on a phantom partner). Second, because I used data from both c ouple members as predi ctors, it was necessary to link the couple members to determine their partner's information. The baseline sample was followed from wave 3 to 7 in order to obtain HIV testing histories over a 16 month time period. Limiting the analysis to only TL T's group 3 (refer to Figure 3.2 ) who had no testing until the end of the study would have minimized the

PAGE 96

81 effects of TLT testi ng on regular testing behavior; h owever, this would have dramatically reduced the sample size, which was already halved through use of gender stratification. Therefore, I used all three TLT study groups in order to make use of a larger sample. Even though group 1 received an HIV test through TLT at each wave and group 2 received an HIV test at wave 4, I still expected that respondents could hav e obtained an HIV test outside of the TLT study. Respondents were not told when they would be tested and thus if they desired to learn their status, knowledge of the TLT testing schedule may have had little impact on the decision to wait to be tested at TL T. I also controlled for cumulative number of previous TLT tests at each wave in order to account for any differences between the regular TLT testers (group 1), the occasional testers (group 2), and the non testers (group 3). For HIV test disclosure, I s tarted with the baseline set of couples identified at wave 3 (N=932). Here, I refer back to the disclosure dataset discussed earlier. Respondents were not eligible for the analysis if they were never tested for HIV or if they did not participate in wave 5. When attempting to understand factors that influence disclosure of HIV test results, it was important to consider HIV status in the analysis since it had been noted as a key predictor of disclosure in the literature. At wave 4, approximately two thirds of the entire TLT sample rece ived an HIV test as part of TLT, allowing me to know the HIV status of many respondents. At wave 5, the TLT questionnaires asked respondents about whether they shared the results of their last HIV test with spouses or sexual part ners. Thus, it was possible to determine whether those tested through TLT at wave 4 disclosed their test results by wave 5. Since group 3 tested outside of TLT (if they teste d at all), I could compare

PAGE 97

82 the rates and predictors of disclosure between groups 1 and 2 to determine if the same associations held true. Measures Table 3.11 outlines the format and interpretation for each of the main predictor vari ables (minus control variables). Table 3. 12 contains a summary of the all independent and dependent vari ables used in the two statistical models for HIV testing uptake and disclosure Socio D emographic Characteristics Several individual level variables were included in all multivariate models as statistical controls: age, years of education, household econ omic status, and marital status. Age and years of education were modeled as continuous variables. Up to and including 8 years of education was considered primary school, 9 to 12 years was considered secondary scho ol, and greater than 12 years was considere d tertiary school. An index of 9 common household goods (bicycle, television, bed with mattress, radio, land line/mobile phone, motorcycle, animal drawn cart, car/truck, or Bible/Koran) was used to approximate household economic status. At wave 3, respond ents had to have an ongoing sexual relationship to be included in the couple dataset. However, over the 16 month period, it was expected that for some, their marital status would change. For the HIV testing uptake models, marital status was considered a 4 level categorical variable consisting of the following states: married/cohabitating, separated/divorced, widowed, or unmarried. Marital status was considered a time varying predictor and included at each wave.

PAGE 98

83 For the disclosure models, a binary variable was used to capture marital status (married/cohabitating or unmarried). I expected that married couples would be more likely to share test results due to emotional closeness and difficulties associated with hiding test results from their spouses. I created a hybrid measure of marital status (married or unmarried) using the marital status measures from waves 3 and 5. Respondents were still considered married even if they had reported separating or divorcing at wave 5 since most reported that they had disclos ed to their spouse in the disclosure question thus, reflecting a ongoing connection with their ex spouse. Respondents whose marital status had changed from unmarried at wave 3 to married at wave 5 were considered married in the analysis. Relationship C har acteristics Relationship duration and shared children were included for descriptive purposes, but not in the analytical models. Relationship duration was computed by subtracting the date of the survey from the date the respondent first started spending ti me with their partner. For shared children, respondents were asked how many living children they have with their partner at wave 1. Given that the sample was young and the mean number of shared children was around 1, a binary variable was created to indica te whether a couple had at least one child together. Discrepancies were sometimes noted between couple members' account of relationship duration and whether or not the couple had children together. For relationship duration, the average between female and male partner's reports was calculated and replaced individual reports. For shared children, the female couple member's report was used, with the assumption that the mother's report might be more accurate.

PAGE 99

84 Re lationship Power As shown in Figure 3.4, five variables or constructs were conceptualized as relationship pow er: 1) socio economic inequalities 2) relationship dominance, 3 ) relationship violence, 4 ) unity, and 5) discordance As previously discussed, the discordance construct was re formulated as si ngle item measures of relationship violence and mistrust/suspected partner infidelity. Socio economic inequality. Three variables were created to measure socio economic inequality between partners: age inequality, education inequality, and employment inequ ality. Given that me n normally marry when they are o n average of 3 years older than women (MDHS, 2011), I considered an age gap of 5 to be a meaningful measure of age inequality (National Statistical Office & ORC Macro 2011) Thus, age inequality was captured as a binary variable where 0 referred to less than or equal to 5 years age difference and 1 referred to greater than 5 years differen ce. Education inequality was captured as a three level categorical variable where 0 referred to equivalency in education for partners, 1 referred to higher male education, and 2 referred to higher female education. Respondents were asked to specify their o ccupation and then asked if the work was piecework, temporary employment, or a steady job. Women who specified that their occupation was housewife' were not asked the second question on employment type. Using the responses for women and their partners, I created a four level categorical variable for employment inequality where 0 referred to both unemployed, 1 referred to man employed woman unemployed, and 2 referred to woman employed, man unemployed and 3 referred to both employed

PAGE 100

85 Relationship dominan ce. Relationship dominance was measured by asking respondents: "In your relationship, who would you say is generally in charge?" with answer choices respondent, equal control, or partner. Since less than 2% of women and less than 1% of men responded that t heir relationship was female dominated, a binary variable was created where 0 referred to egalitarian or female dominated and 1 referred to male dominated. Relationship unity. As previously discussed, the factor analysis resulted in two subscales named un ity and discordance The three item unity subscale was included in all statistical models ("My partner shows they care about me"; "When I need my partners assistance, he/she is there to help me"; and "My partner and I discuss important matters together"). TLT interviewers asked respondents whether they strongly agreed (1), agreed (2), disagreed (3), or strongly disagreed (4) with these statement s Responses were reverse scored so that higher mean scores meant more unity in the relationship. The discordance subscale was dropped for lower reliability and other reasons but re formulated using three single item measures: relationship violence (physical and sexual) and mistrust/perceived partner infidelity. Relationship violence (sexual and physical). Responde nts were asked if they were victims of sexual and physical violence (but not if they were perpetrators) in relation to the reference partner noted in TLT's power module. The format and wording of these two measures came from the study conducted by Pulerwit z and colleagues (2000), who used these measures to assess construct validity for the original Sexual Relationship Power Scale (SRPS). Sexual

PAGE 101

86 violence was captured as a binary variable that asked respondents if their partner has ever forced them to have se x when they did not want to. I asked my research assistants to clarify the translation of the term "forced ." I found that the measure of se xual IPV does not refer to rape ( i.e., being physically forced against one's will to have sex ), but rather to verbal pressure to have sex when one does not want to. The term "sexual coercion" may more accurately reflect the measure of sexual violence Physical violence was captured with a binary variable that asked respondents if their partner ever hurt them by beating t hem. While multiple forms of physical abuse are possible such as hitting, kicking, or punching, the question on physical abuse in this study was restricted to "beating" in order to reflect the predominant local term used to describe physical abuse in Malaw i. Mistrust (perceived partner infidelity) Perceived infidelity of a partner was measured with the statement: "My partner is probably having sex with someone else ." TLT interviewers asked respondents whether they strongly agreed (1), agreed (2), disagreed (3), or strongly disagreed (4) with this statement. I next created a binary variable for perceived partner infidelity by collapsing the strongly agreed/agreed (set to 1) and strongly disagreed/disagreed (set to 0) response categories. Perceived Risk Per ceived risk for HIV of self and of partner were included as predictors of HIV testing uptake and modeled as categorical variables. Perceived risk of self was first captured with the statement: "Pick the number of beans that reflect how likely it is that yo u are infected with HIV now ." The respondent could select up to 10 beans. I created a 5 level categorical

PAGE 102

87 variable where 0 equaled no likelihood, 1 equaled low likelihood (1 4 beans), 2 equaled a medium likelihood (5 beans), 3 equaled high likelihood (6 9 beans), and 4 equaled certain likelihood. Rather tha n collapse the initial response options into low, medium, and high, I specifically wanted to include "no likelihood of infection" and "certain likelihood of infection" as separate categories to account fo r actual knowledge of HIV status (i.e., respondents who presumably knew their HIV status would theoretically be less likely to test). Perceived risk of partner was captured with the statement: What is the likelihood that your partner is currently infect ed with HIV?" Response options included no likelihood, low, medium, high, and "I know she/he is" (infected with HIV). I created a 3 level categorical variable for perceived risk of partner, where 0 referred to no or low likelihood of infection, 1 referred to medium likelihood of infection, and 2 referred to high or certain likelihood of infection. For the statistical models, I later collapsed the medium likelihood with the high/certain category given the low number of "medium" responses. HIV Status (self a nd couple) By wave 5, approximately two thirds of the sample was tested for HIV via TLT. The TLT survey also included a question about the respondent's likelihood of being HIV positive collected at each wave (i.e., perceived risk of self) I considered re spondents to be HIV positive if they tested positive through TLT (up to and including at wave 4) or if they indicated that they were certain they were HIV positive at wave 5 (i.e., reported a 10 on a

PAGE 103

88 scale of 0 to 10 for perceived likelihood of HIV infecti on). Thus, HIV status was a categorical variable wh ere 0 referred to HIV negative and 1 referred to HIV positive. It was expected that respondents would make decisions to disclose based on whether they perceived their partner's HIV status to be the same or different from their own HIV status. Therefore, in the disclosure models, I included a measure of concordance between the perceived status of a partner and their own HIV status. I used the HIV status variable and the perceived risk of partner variable ( ascertained at wave 4) above to create a new 4 level categorical variable called perceived HIV status concordance, where 0 referred to HIV negative respondent, partner perceived to be HIV negative/low risk', 1 referred to HIV positive respondent, partner perceived to be HIV positive/high risk', 2 referred to HIV positive respondent, partner perceived to be HIV negative/low risk', and 3 referred to HIV negative respondent, partner perceived to be HIV positive/high risk'.

PAGE 104

89 Table 3.11: Format and i nterpr etation of the main p redictor v ariables Variable name Categories /Range Interpretation Age inequality 0, 1 0=Partners within 5 years age difference 1=Male partner at least 5 years older Education inequality 0, 1, 2 0=Equal education 1=Male higher educati on 2=Female higher education Employment inequality 0, 1, 2 0=Both unemployed 1=Male employed, female unemployed 2=Female employed, male unemployed 3=Both employed Relationship dominance 0, 1 0=Female dominated/egalitarian 1=Male dominated Unity 1 4 High er scores mean more unity power Physical violence 0, 1 0=No history of physical violence in the relationship 1=History of physical violence in the relationship Sexual violence 0, 1 0=No history of sexual violence in the relationship 1=History of sexual v iolence in the relationship Mistrust 0, 1 0=Disagreed/strongly disagreed partner is having affair 1=Agreed/strongly agreed partner is having affair Perceived risk (self) 0, 1, 2, 3, 4 0=No likelihood of HIV infection 1=Low likelihood of HIV infection 2=M edium likelihood of HIV infection 3=High likelihood of HIV infection 4=Certain likelihood of HIV infection Perceived risk (partner) 0, 1 0=No/low likelihood partner is infected with HIV 1=Medium/high/certain likelihood partner is infected with HIV HIV st atus 0, 1 0=HIV negative 1=HIV positive Partner HIV status concordance 0, 1, 2, 3 0= HIV negative respondent, partner perceived to be HIV negative/low risk 1= HIV positive respondent, partner perceived to be HIV positive/high risk 2= HIV positive responden t, partner perceived to be HIV negative/low risk 3=HIV negative respondent, partner perceived to be HIV positive/high risk

PAGE 105

90 C ontrol Variables Three additional control variables were included in the multivariate models for HIV testing uptake: new antena tal care HIV test previous TLT testing and previous external testing (all modeled as time varying predictors). Since this study focused on relationship factors that influenced use of client initiated HIV testing, a control variable was included in the mu ltivariate models for women in order to adjust for those who tested as part of antenatal care and thus would be less likely to seek HIV testing elsewhere. At each wave, pregnant women completed a separate pregnancy questionnaire that asked if they had star ted antenatal care and if yes, whether or not they had been tested for HIV through their antenatal care. Women were not asked the date of the antenatal care HIV test. Pregnant women who started antenatal care very early could have reported they had receive d an antenatal care HIV test at two sequential waves over an 8 month study period. Thus, I considered it a new antenatal care test for that particular wave if they did not report in the previous wave that they had tested for HIV via antenatal care. Take, f or example, a pregnant woman who reported that she had been tested for HIV via antenatal care at wave 5. I did not consider her test a new antenatal care test for wave 5 if she also reported that she received an antena tal care test at wave 4. This logic al so assumes that women were only tested once through antenatal care and that a miscarriage did not occur between waves (e.g., reported antenatal care testing at wave 4, then miscarried and got pregnant again, and reported antenatal care testing at wave 7 ).

PAGE 106

91 The multivariate models also included a control variable for previous TLT testing Since I used all respondents' HIV testing histories collected at waves 4 through wave 7, approximately two thirds of people were tested through the TLT study at wave 4 and another one third were tested at waves 4 through 7. All respondents were included in this study because even if they were tested as part of TLT, they may have also tested outside of the study. Given that this study used a subsample of the entire wave 3 pop ulation to begin with (i.e., those who were in serious sexual relationships), statistical power would have been significantly reduced if respondents who tested via TLT were excluded from the analysis. Therefore, a continuous variable was included in the mo dels to control for the number of previous TLT tests (cumulative) at each wave. Respondents were considered to have tested via TLT if they were allocated to group 1 or group 2 (for wave 4 only) and had an HIV test result on file for that wave. I also exp ected previous external HIV testing would influence future uptake of HIV testing outside of TLT Therefore, I included a continuous variable called previous external testing in the models to control for the number of previous external tests (cumulative) at each wave. In the multivariate disclosure models, it was expected that disclosure rates would vary based on whether respondents tested through TLT at wave 4 or not. Therefore, an additional variable was included as a statistical control in the disclosur e models: TLT test at wave 4. This binary variable indicated whether a respondent's last HIV test was conducted at

PAGE 107

92 TLT's wave 4 (yes/no). Respondents were considered to have tested at wave 4 if they were allocated to TLT group 1 and 2 and if they had an HI V test result on file 4 Dependent Variables: New HIV Test and Status Disclosure At each wave, respondents were asked: "When were you last tested for HIV?" If they had previously tested, responden ts specified the date of their last HIV test (month and ye ar) and whether it was conducted through TLT or at a local health care center. A binary variable was created to track whether respondents received a new HIV test (outside of the TLT study) at waves 4 through 7 using the survey date of previous wave and las t test date of current wave. If the "last test date" was greater than "previous survey date" and the test was reportedly conducted outside of TLT, then it was considered a new HIV test for that wave. If a respondent was missing for a particular wave, they were not included in the analysis at that wave only. Respondents were specifically asked, "The last time you tested, to whom if anyone did you tell your results?" Response options included spouse/long term partner, other sexual partner, a relative, etc. A binary variable was created to measure whether a respondent disclosed the results of their last HIV test to either their spouse (if married) or to a primary sexual partner (if unmarried). For unmarried respondents who disclosed to their sexual 4 Initially, I considered respondents to have tested at wave 4 if they reported at wave 5 that their last test was at TLT, however, this appears to have mis sed a number of respondents who actually did test at TLT's wave 4. Of the 932 individuals in the couples dataset, 489 of them tested at wave 4 through TLT yet only 410 reported that their last test was at TLT. Some of these respondents may have simply misr eported where the last HIV test occurred while others may have tested again after TLT.

PAGE 108

93 partner, i t was not possible to co nfirm whether this individual was the same sexual partner in the couple dataset (though this was likely to be the case ).

PAGE 109

94 Variable Included in model? Type Wave(s) Included in model? Type Wave(s) Independent variables Marital status X Categorical 4-7 X Binary 3, 5 Age X Continuous 3 X Continuous 3 Years of education X Continuous 3 X Continuous 3 Household goods X Continuous 3 X Continuous 3 Age inequality X Binary 3 X Binary 3 Education inequality X Categorical 3 X Categorical 3 Income inequality X Categorical 3 X Categorical 3 Unity X Continuous 3 X Continuous 3 Relationship dominance X Binary 3 X Binary 3 Physical violence X Binary 3 Sexual violence X Binary 3 Perceived risk (self) X Categorical 4-7 Perceived risk (partner) X Binary 3 HIV status concordance X Categorical 4 Perceived partner infidelity X Binary 3 New antenatal care HIV test X Binary 4-7 HIV testing uptake models HIV test disclosure models Table 3.12: Independent and dependent variables for the two s tatistical models of uptake and disclosure

PAGE 110

95 Variable Included in models? Type Wave(s) Included in models? Type Wave(s) Independent variables Previous number of TLT tests X Continuous 4-7 Previous number of external tests X Continuous 4-7 Tested at TLT's wave 4 X Binary 4 Dependent variables New HIV test (since previous wave) X Binary 4-7 Disclosure to sexual partner X Binary 5 HIV testing uptake models HIV testing disclosure models Table 3.12, continued

PAGE 111

96 Statistical Analysis For repeated events data such as the binary variable of new HIV test' collecte d at each wave in this study, Allison (2 004) suggests the use of fixed and random effects models each of which has their own set of advantages and disadvantages. Peterson (2004) adds that the choice of model depends more on the underlying study objectives. Fixed effects models are used to a ddress how individual change on a given predictor (e.g., marital status) affects a change in the dependent variable over time. On the other hand, random effects models are tilted more towards differences between individuals, such as those who perceive them selves to be at high risk for HIV and those who do not. Fixed effects models are advantageous in the sense that each individual serves as their own control and it is possible to control for both observed and unobserved variables that remain constant over time (Peterse n 2004) However, the main drawback to fixed effects models is the inability to estimate the effects of any variables that do not vary over time (Allison 2004) Individuals with no across time variation in any of the predictor variables would not contribute at all in the estimation and may reduce the sample s ize considerably (Petersen 2004) In the present investigation, some of the predictor variables did not change over time. For example, data pertaining to the unity variable was also collected at wave 5 when the power module was included in the partnership survey. A tw o sample t test for the mean differences showed that unity was not significantly different in wave 3 and 5 (see Table 3. 13 below). I also calculated the percentage of respondents who provided the same response to each of the three unity items at both wave 3 and 5 (N=840). Approximately

PAGE 112

97 80.3%, 74.9%, and 60.8% of respondents provided identical responses for the "cares for me", "discuss together", and "helps me" items at both waves. Since unity is one of the most important variables used to predict HIV testin g uptake in this study, fixed effects models would not have been able to produce reliable estimates for this variable. Table 3. 13 : Mean value of unity at TLT's wave 3 and 5 Therefore, random effects models were used to test for associations between t he predictor variables and HIV testing uptake using the xtlogit function in Stata 11.2. All variables, with the exception of marital status, perceived risk, antenatal care testing for HIV, previous TLT testing, and previous external testing, were modeled a s time invariant predictor or control variables (as illustrated in Table 3. 12 by the "wave" column ). For the disclosure models, multivariate logistic regression was used to test for associations between relationship power and status disclosure using the logistic function in Stata 11.2. Even though multiple waves of data were used (e.g., predictor variables were from wave 3, HIV status was from wave 4, and disclosure was from wave 5), only a single time point was examined for the disclosure dependent varia ble thus making conventional regression an appropriate analytic approach. The disclosure models were also stratified by gender. Wave Mean of unity Standard deviation 3 3.773 0.398 5 3.767 0.426 A two-sample ttest showed that the mean difference between the two measures of unity was not significantly different from zero ( p = 0.70 )

PAGE 113

98 The Interpretative Phase Design and Rationale I used two main sets of qualitative data, namely semi structured interviews and focus group discussions (FGDs) to carry out specific aim 1b (to evaluate the cultural re levance of the measure of power) and specific aim 3 (to understand the meaning of HIV testing within the context of a sexual relationship) As part of these objective s the qualitative data also allowed me to verify, challenge, and explain the quantitative results from the factor analysis and from the hypothesis testing phase on HIV testing behavior. I use d grounded theory as a method to accomplish these qualitative o bjectives (Strauss and Corbin 1998; Glaser and Strauss 1967) Starks and T rinidad (2007) outline three qualitative approaches to data that can be used in health research: phenomenology, discourse analysis, and grounded theory. Discourse analysis focuses specifically on the use of language through a caref ul analysis of speech which was not my intention, although categories of codes that emerged did resemble narratives or discourses I chose groun ded theory over phenomenology for the following reasons. In contrast to phenomenology which seeks to understan d a particular phenomenon of interest by sampling subjects who have directly experienced this phenomenon, grounded theory serves to develop an explanatory theory of basic social processes using a range of different experiences (this will become evident thr ough the discussion of my sampling approach). Grounded theory originates from the field of sociology, specifical ly from symbolic interactionism (Starks and Trinidad 2007) Briefly, symbolic interactionism posits that meaning is negotiated and unde rstood through actions

PAGE 114

99 with others in social processes (Jeon 2004; Blu mer 1969) Because I was interested in the relationship context and how men and women's interactions create d new meanings and responses to HIV testing programs symbolic interactionis m was an appropriate approach For specific aim 1b I use d the TGP to frame constructions of power in the Malawi context, but also invoked grounded theory as a method to uncover areas of nuance In this respect, my approach to specific aim 1b was primarily deductive since I used th e TGP However, the approach was complemente d by an inductive perspective that allow ed Malawians' voice s to emerge organically from the data. The need for this complementary inductive approach is especially important given the limitations of the TGP (as pointed out in Chapter 2) and its primary appl ication in western settings ; b y relying solely on the TGP, I could miss important dimensions of power relevant to the Malawi context. A second limitation of the TGP is its rather stagnant prescription of gender relations as if they are stagnant and not sh aped by geography As sociologists have long argued, "gender" is a fluid concept that is socially constructed across time and space (Butler 2003; deBeauvoir 1989; Benedict 1959; Mead 1935) As such, people do not t ake on one particular gender identity for life but rather reconstruct their gender in response to the changing world around them. Gender norms are constantly evolving as people interact with larger social structures and with each other, and then respond ac cordingly. By listening to the voice of Malawians themselves, I hope d to capture the evolution and dynamic nature of gender and power that could not be revealed by relying on the TGP alone. For specific aim 3 I took a more purist inductive approach usi ng grounded theory and

PAGE 115

100 particularly symbolic interactionism as a method for understanding HIV testing behavior. With this type of analysis, I could then question whether my deductive hypothesis testing results were rooted in the actual experiences and per ceptions of Malawians. I used the semi structured interviews to document individuals' experiences getting tested for HIV within the context of their relationships and used the focus group discussions to capture broader social norms and perceptions of HIV t esting. For both of the specific aims above, I also drew upon other informal sources of qualitative data such as field notes from participant observation, interviews with VCT counselors and village chiefs, debriefing sessions with my Malawian informants, and casual conversations with other young Malawians. Rather than go back and conduct another set of semi structured interviews to follow up on interesting ideas or areas that were incomplete in the interviews, I used the FGDs as a way to fill these gaps in knowledge. Group discussions are also a useful method for observing how people interact and converse with one another on issues (Schensul 1999) albeit in a semi structured rese arch setting. As such, I use the focus group discussions as a way to explore interactions between participants on perceptions of power and HIV testing. I found the FGDs to be much richer and more colorful than the one on one individual interviews and there fore, I focused my analysis on the FGDs and used the semi structured interviews to provide backup support.

PAGE 116

101 The Couple Interviews Sampling and R ecruitment Purposive sampling was used to select three distinct geographical areas in order to obtain a diverse set of relationship experiences among participants: 1) Balaka boma or the villages surrounding Balaka town; 2) a trading center village; and 3) an isolated rural village. After selecting six villages as the final catchment area, a random sample of dating and married women (n=90) aged 18 to 25 were selected from each village using age, sex, and marital status data from a recently collected demographic household listing of the Balaka district. After obtaining permission from the local village authorities, we approached women at their homes. Sexual part ners were recruited through the sample of women. Since we did not have addresses for respondents, we relied heavily on the village chiefs to help us track down respondents in their villages. If the target sample member could not be found or the sample member was ineligible during the time of recruitment, the interviewer moved onto an alternate sample member listed for that particular individual. Training Prior to data collection, hired Malawian research assist ants attended a two day training session conducted by myself on topics related to qualitative research, interviewing techniques such as probing and listening, the interview guide, consent for participation, ethical issues, and transcription and translation Research assistants were trained to conduct flexible and unobtrusive interviews, so that fuller responses were elicited and new themes could emerge spontaneously. Research assistants were also re trained on qualitative interviewing

PAGE 117

102 throughout the data co llection process using completed transcripts from previous interviews. During training, research assistants were asked to provide feedback on confusing, unclear, and culturally irrelevant interview questions and their feedback was incorporated into the int erview guide. Semi Structured Interview G uide A semi structured interview guide was developed using constructs from the Theory of Gender and Power (Connell 1987; Wingood and DiClemente 2000) and background literature on gender and relationship power (Perez et al. 2006; Pulerwitz, Gortmaker, and DeJong 2000; Pulerwitz et al. 2002; Dunkle et al. 2004; Pettifor et al. 2004; M bweza, Norr, and McElmurry 2008) Respondents were asked to share their perceptions of ideal relationships but also asked about their relationship histories and experiences with regard to their current sp ouse or primary sexual partner if not married The first half of the interview covered topics such as relationship characteristics, gender roles and ideals, dependence/autonomy, control, decision making, spousal communication, love, and trust. For example, the following question was asked on decision maki ng: How do you and your partner decide on important things in your relationship? Probes were used to extract more detailed information from respondents abou t their relationship dynamics. Examples of probes included: What types of decisions do you have more say in, what types of things does your partner have more say in? Do you need to consult with your partner on certain types of decisions? The second half of the interview focused specifically on experiences with HIV testing. For example, respondents were a sked if they had ever been tested for HIV (but not

PAGE 118

103 about the test results), if their relationship changed after testing, and if yes, how so. Refer to Appendix B for the semi structured interview guide. Semi S tructured I nterviews In order to minimize soc ial desirability bias and help respondents feel more comfortable providing sensitive information about themselves two research assistants were matched by sex with the respondents. Another PhD student conducting fieldwork in Malawi had recommended several res earch assistants after interviewing them for her own work but found that they did not speak the local language of her study population and thus she could not hire them. After interviewing several possible candidates for the job, I selected two candidat es who had some college level training, demonstrated strong English, and who m I thought were personable enough to establish a good rapport with the respondents. The research assistants fell within the age range of the sample population (female interviewer, age 21; male interviewer, age 25). While the couple interviews were conducted at the same time, partners were interviewed separately in a private location chosen by the respondent such as under a tree, on a straw mat near the back of the house, or on th e veranda. The interviews lasted approximately 45 80 minutes. The interviews were audio taped with digital recorders. Each respondent received four hotel sized washing soaps as a gift for participating in the study. Semi structured interviews were conduc ted with 34 women and their male partners (17 couples). Of the 34 respondents, 12 respondents were in dating relationships and the remaining were married. The average age of the men was 23 years whereas the women were

PAGE 119

104 slightly younger with an average age of 21 years. Of the 17 couples, 8 resided in Balaka town villages, 5 resided in a rural village, and 4 resided in a trading center village. Translation and T ranscri ption After each interview, research assistants wrote a summary of the interview and debri efed me on the highlights of the interview. In the summaries, the research assistants noted their overall impression of the respondent and discrepancies in the conversation that may indicate respondent bias. Research assistants simultaneously translated an d transcribed their respective interviews from Chichewa to English immediately following the interview. Interviews were translated word for word. I reviewed the transcripts for clarity as they were completed and asked the research assistants to explain lan guage that was unclear. During the transcription reviews, I noted areas where additional clarification was needed from the research assistants on language, events, and cultural practices or beliefs. As I reviewed each successive interview, I noted in my ow n summaries any new information that is learned and how the information compared or differed from previously conducted interviews. The Focus Group Discussions Sampling a nd R ecruitment Participants were eligible to participate if they fell within the age range of 16 to 24 years and were considered maturated 5 U nmarried minors could be placed at considerable 5 While participants were advised to talk in general terms and share their opinions but not personal experiences, we anticipated that sensitive information could inadve rtently be disclosed to the group and create confidentiality risks. Therefore, we limited participants to maturated adults, i.e., if they were under 18, they had to be married. To the contrary, I found

PAGE 120

1 05 risks if they disclosed details on their secret relationships to the group and this information later became public. Individuals were considered matur ated adults if: 1) they were married and/or ; 2) over the age of 18 years. After obtaining permission from the local village authorities, we recruited focus group respondents from the same three geographical areas in the Balaka district as we did for the semi structured interviews. Respondents were approached by a project staff member at their homes, explained the details of the study and their responsibilities, and asked to participate. If they agreed, respondents were given appointment cards indicating t he time and location of the focus group discussion. In order to minimize confidentiality breaches and social desirability bias due to knowing a fellow focus group member, respondents in each focus group were selected from different geographical areas to th e extent possible. During recruitment, I had expected that some respondents would forget about their appointments or change their minds about participating and thus, we over sampled in order to allow for this (in order for a scheduled FGD to take place, w e needed at least five but no more than eight respondents). However to my surprise, most respondents showed up to participate sometimes very late, which I learne d was typical on "African time", and most likely motivated by the expectation that they would b e co mpensated for their time. We selected three sites to conduct the interviews, each with varying degrees of formality and levels of comfort. that most of the time people shared experiences about others from their home villages and family/friends as opposed to themselves.

PAGE 121

106 1) A teacher development center (referred to locally as a "TDC") located in the Balaka town. This building was cent rally located near a secondary school and was equipped with chairs and tables, and had intermittent electricity. 2) An under five immunization clinic that doubled as a village preschool. This small concrete structure had a large open space with two small pri vate rooms and was located in a rural trading center village. Respondents huddled along a cement wall ledge on the building's perimeter while the facilitator sat in the middle of the group on an empty bottle cart. 3) An agricultural field under a large baob ab tree. This site was located in an isolated, rural farming village. We had hopes of conducting the interviews outside of the chief's house on large straw mats, however, local men were brewing beer and talking loudly so we were forced to move to the field for more privacy. Sample S ize Regarding the number of FGDs to be conducted in this study, a number of researchers recommend conducting two group discussions for each variable of concern in order to ensure that the discussions adequately capture most asp ects of the topic (Schensul 1999) This study used two stratifying variables (e.g., gender and martial status), which indicated a minimum of two FGDs for men and two FGDs for wom en. The following calculations were based on Schensul's (1999) recommendations for the number of FGDs:

PAGE 122

107 Gender = 2 x (male, female) Marital Status = 2 x (married single) Total number of FGDs: 2 x 2 = 4 However, I decided to double the number of recommended FGDs to eight in order to ensure that saturation points would be met and to account for the fact that the quality of some interviews would be better than other s depending upon group dynamics. I also expected that as the facilitators gained more experience with each subsequent interview, the data quality would improve. The FG D G uide Individuals were asked to share their perceptions, attitudes, and opinions on r elationships among people like the mselves. They d id not need to be personally involved in a sexual relationship to be able to talk generally about relationship power. The focus group guide was divided into two sections. The first section asked questions on general power dynamics within married and dating couples such as what does it mean to be head of the household? What makes a man (and a woman) feel powerful in their relationships? How does power change after marriage? What happens when married couples di sagree or even cheat? The second half of the guide presented a series of vignettes on HIV testing that involved a hypothetical couple named "Lucy" and "Promise ." For example, in one scenario, Lucy was worried about getting HIV from her husband and responde nts were asked if she should go for testing on her own or get her partner's permission first, and why. Refer to Appendix C for more details on the FGD guide.

PAGE 123

108 Training I hired two English speaking Malawian research assistants to facilitate the FGDs in Chi chewa, who were matched by gender to the respondents. The two research assistants were trained to conduct flexible and unobtrusive group interviews so that fuller responses could be elicited and new themes were allowed to emerge spontaneously. Before start ing, we pilot tested the FGD guide on groups of TLT staff members in order to finalize the instrument and train the facilitators at the same time. The male interviewer was 30 years old and came highly recommended to me by the TLT staff. He had a college d egree in economics from a Malawian university, was very bright, and possessed strong leadership skills. The female interviewer was hired through word of mouth. I initially had heard about her from another PhD student who thought she was very articulate and had impeccable English. She was 19 years old, came from a relatively wealthy family in Balaka, and had college level training and certificates. Although the female interviewer was very capable and confident in her abilities, she had the difficult and chal lenging job of trying to get shy, village women to talk openly in a group setting. Certainly some women were boisterous and eager to participate, but many remained quiet and unsure about their responses (as if there was a right or wrong answer), and had to be continuously probed for their thoughts. The men, on the oth er hand, were very loquacious and verbose in their responses.

PAGE 124

109 Focus Group D iscussions One research assistant facilitated the interview, while the other person acted as the "observer" by as sisting the facilitator with paperwork and refreshments, taking notes, and creating interview summaries to be used in a debriefing session on the most salient discussion points, problematic questions, and how new information differed from the previous focu s groups. As the interviews took place, I recorded notes on body language, interactions between FGD members, and respondent characteristics. We conducted eight focus group discussions consisting of 7 8 participants each for a total of 62 respondents. Focus groups were stratified by gender and marital status in order to make participants feel more comfortable sharing their opinions, perceptions, and understandings of power and HIV testing with their peers: two groups were married women, two groups were marri ed men, two groups were single women, and two groups were single men. The focus group discussions lasted between 75 and 120 minutes. The set of transcripts consisted of 283 single spaced pages. Table 3. 14 shows the demographic characteristics of the FGD sa mple. Half of the sample was married. The average age of the men and women was approximately 21 and 20 years, respectively. The sample was split almost evenly between having a primary school and secondary school education. Approximately half of the respond ents resided in the Balaka town villages, 26% resided in a large trading center village, and 24% resided in the rural villages.

PAGE 125

110 Table 3. 14 : Characteristics of the focus group respondents Variable Men (n=32) N(%) Women (n=30) N(%) Total (n=62) N(%) Mean age 21.25 19.8 20.5 Marital status Married 16(50) 15(50) 31(50) Single 16(50) 15(50) 31(50) Education Primary school or less 16(50) 13(43) 29(47) Some secondary school 15(47) 17(57) 32(52) Higher education 1(0) 0(0) 1(0) Resident village Balaka town villages 16(50) 15(50) 31(50) Trading center village 8(25) 8(27) 16(26) Rural villages 8(25) 7(23) 15(24) Translation and T ranscript ion Audio recorded focus group discussions were immediately translated from Chichewa into English and typ ed into electronic format for review. T he focus group recordings were translated word for word. Short and long pauses, laughter, and other non verbal gestures were included in parentheses in the transcripts These non verbal gestures were important to incl ude because they indicated instances where respondents were struggling with a question (i.e., a long pause). I interpreted laughter as a positive sign that respondents felt comfortable with each other and were enjoying themselves enough to bring a little b it of humor into the discussions. During transcription, focus group respondents were assigned a number based on the first time they spoke in the interview. Respondents who could not be identified by the transcriptionist were labeled with a question mark (i .e., "Man ?" instead of "Man #2"). If respondents were speaking at the same time (overlapping speech) and it was not possible to distinguish what each pe rson was saying, the transcriptionist placed the phrase "cross talk"

PAGE 126

111 in square brackets immediately aft er the last identifiable speaker's text and then picked up with the next audible speaker. There were also times when it was not possible to decipher what was said on the audiotapes, for example, when it was raining outside or there was too much echo in the room. In these cases, the transcriptionist identified segments of the tape that were unintelligible and indicated this in square brackets in the transcripts. As each transcript was completed and I reviewed it, we all sat down to discuss translation issues where better probes could have been used, and the larger themes found in the focus group discussion. Qualitative Data Analysis With the semi structured interviews, I imported the data into the Atlas.ti software for analysis. After significant coding of the interviews, I realized that the software inhibited my ability to intuitively analyze the data; I felt too distracted by the laborious process of coding instead of focusing on what was going on in the data. Therefore with the focus groups, I decided to develop my own data analysis method by coding the data in Microsoft W ord, organizing the coded data in binders that I could easily flip through to compare codes acros s and within the FGDs, and used colored sticky notes to identify key passages and themes. Yet the same overall process was us ed for both analysis strategies; I utilized a series of steps outlined by the grounded theory approach (S trauss and Corbin 1998) to analyze the qualitative data.

PAGE 127

112 Open C oding Descriptive codes and categories of themes, concepts, experiences, and processes were created prior to data analysis using the preliminary reviews of the data as described above, the interview guides, and relevant background literature. Codes refer to labels for assigning units of meaning to the information collected as part of the study (Miles and Huberman 1994) An initial set of codes included labels such as "qualities of an ideal partner", "meanings of love", "sources of power", an d "HIV testing decision making." Coding began as an open coding process, in which a priori code s were assigned to text in a systematic manner by examining the transcripts line by line or by set of lines. As I read each passage, I asked myself the questions "what is this about?" and "what is being reference d here?" As the coding continued existing c odes were modified or deleted, and new codes (such as in vivo codes ) were continuously added until all the transcripts were coded. Axial C oding After open coding, axial coding was used to specify the relationship between codes and to group codes into cate gories and sub categories. Categories are "concepts that stand for a phenomenon" and subcategories are "concepts that pertain to a category, giving it further clarification and specification" (Strauss and Corbin 1998) Selective S oding The final coding step was selective coding, defined as "the process of integrating and refining categories" (Strauss and Corbin 1998) In this stage, I organized the codes around a central, unifying core category that reflected the ma in actions and events described in the

PAGE 128

113 semi structured interview s and FGDs. Through selective coding, the categories were integrated and developed into theory around what relationship power means in Malawi and its connection to HIV testing. Memo W riting I used the codes to develop theoretical memos that compare d and contrast ed and summarize d phenomena, groups, and individuals. A memo is "the theorizing write up of ideas about codes and their relationships as they strike the analyst while codingit can be a sentence, a paragraph, or a few pagesit exhausts the analys t's momentary ideation based on data with perhaps a little conceptual elaboration" (Glaser 1978) Data M atrices I used the methods outlined by Miles and Huberman (1994) to develop data reduction matrices that summarize d and display ed key concepts/themes by demographic characteristics such as gender and marital status. A matrix allows the researcher to conceptually piece together fragmented text in one place in order to reduce a complicated data set to a manageable size (Ulin et al. 2002) Data P resentation Modest changes were made to the transcripts to improve readability. For con fusing words or phrases that mad e sense to me, but would be considered foreign to others not familiar with Malawian English I inserted brackets with short explanations for clarification. Given the difficulties in ass igning names or pseudonyms to 62 different focus group respondents, I choose to leave the respondents' identif ication as numbers; I realize that this

PAGE 129

114 may be awkward to read and decontexualize peoples' stories. With the semi structured interviews, I could more easily label respondents by providing each of the 34 respondents with a pseu donym (e.g., "Ruth" and "James") in order to give life and context to their stories. Ethical Considerations This study received approval from the Colorado Multiple Institutional Review Board or COMIRB (protocol numbers 10 758 and 10 861). The TLT study received approval from Penn State University (protocol number 31397) and the National Health Sciences Research Committee (NHSRC) in Malawi (protocol number 558). All TLT data were de identified prior to my analysis and thus I had no access to respondent na mes or personal identifiers. For the qualitative data, many of the same ethical considerations were taken into account during data collection for both the semi structured interviews and focus group discussions. In what follows, I briefly summarize some of the key ethical considerations for the focus group discussions. During recruitment of the focus groups, respondents were informed that the FGDs would involve questions about gender and power i n sexual relationships and were asked if they felt comfortable talking about th ese issues in a group. Respondents were told that they did not need to provide intimate details of their own sexual relationships but rather to talk more generally about relationships based on their perceptions of others. Efforts were made to ensure that FGD participants in each group were recruited from different geographical areas and did not know each other personally.

PAGE 130

115 FGDs took place at comfortable, centrally loca ted community sites As part of the formal consent process, respondents w ere explained the purpose of the study, why it was important and what would be learned from the research, the risks and benefits of participating, the right to not answer any questions or stop the interview at any time, and who would see their resp onses. T hey were provided with contact information for the princip al investigator should any additional questions arise Participants were also as ked for their permission to audio record the interviews so that they could be translated from Chichewa into English. V erbal consent was obtained since signatures of the respondents would be the only piece of information linking the respondent to the study. A copy of the consent form (explaining all of the above) was translated into Chichewa and back translated into Englis h to ensure that the form was interpreted as I expect ed The final Chichewa version was left with respondents to read over at their leisure. Two witnesses (a Malawian research assistant and myself) were always present during the consent process. Following the completion of the FGD each participant was given a modest cash gift of 500 Malawi Kwatcha (MK) or the equivalent of $3.25 USD. T he amount of the gift was not be specified in the recruitment materials. Consent forms disclosed more details regarding th e gift. This amount of money given to respondents had been approved for the larger TLT project by Penn State University and the National Health Sciences Research Committee (NHSRC) in Malawi and the research team had not encountered any problems with this i ncentive over the first year of the project. The gift also addressed potential transportation and food costs participants might face by participating in the study.

PAGE 131

116 I anticipated that this study would pose minimal risks to research subjec ts. One possible risk of the FGDs was the breach of confidentiality in which private information could have been shared with others outside the group. To protect against this risk I ensured that the audio equipment containing project files and hardcopies of transcripts we re locked up and stored on a password protected computer that only I had access to. Although I collected demographic information such as age, education level, resident village/town, marital status, etc. on study participants, I did not collect any identify ing information such as names or identification numbers. Any names mentioned in the transcripts were anonymized (i.e., names were changed to a name other than the one mentioned) after translation/transcription.

PAGE 132

117 CHAPTER IV CONSTRUCTIONS OF GENDER AND POWE R: TRADITION, UNITY, AND RIGHTS In Chapter 3, I presented a new measure of relationship power for the Malawi context. The statistics revealed that two constructs were important domains of power in young couples: unity and discordance. While I used qualita tive interviews to develop the initial set of power scale items, this formative work was carried out in a rapid fashion during a short field trip to Malawi in 2009. Formative qualitative work conducted using a deduct ive framework and a priori theory may no t fully capture the breadth or depth of the construct of interest in the same way that more inductive qualitative methods allow for. This chapter presents the results for specific aim 1b: to evaluate the cultural relevance of the measure of power using qua litative data To accomplish this aim I use d the Theory of Gender and Power (TGP) to frame how rural Malawians construct meaning around gender and power, while at the same time allowing any additional theoretical constructs to emerge inductively from the data. To briefly recap from Chapter 3, Connell's TGP (1987) proposes that three social structures characterize the gendered relationships between men and women: 1) the sexual division of labor, or the economic inequalities that favor males; 2) the sexual division of power or abuses of authority and control in relationships; and 3) cathexis or social norms and affective attachments around femin in ity and masculinity The three social structures are overlapping, but distinct and explain the gender roles th at men and women act out in their relationships and society (Connell, 1987). In addition, these three social structures are thought to operate

PAGE 133

118 at two different levels: the societal (the higher level) and the institutional (which includes the family and rel ationship level) levels I focus mostly on how power is manifested at the relationship leve l. A t the end of this chapter, I describe how structural level interventions aimed to empower women filter down to influence relationship power dynamics. I focused the analysis on the focus group discussion data rather than the original set of semi structured couple interviews used to develop the power measure because I found them to be much richer and to offer a fresh perspective on gender and power relations. One e xplanation for this may relate to the willingness of respondents to disclose private information. In the focus group discussions, respondents were not asked to talk about their own personal experiences but rather their perceptions of people they know in t heir villages and therefore they may have been more forthcoming than the semi structured interview respondents. In the future, it may be worth spending more time with individual respondents to build rapport perhaps by conduct ing several consecutive interv iews and help respondents feel more comfortable sharing their private experiences. Tradition, Unity, and Rights Using the grounded theory approach outlined in Chapter 3, I found that respondents voiced their perceptions of gender and relationships using t hree interrelated narratives (or discourses), which I named tradition, unity, and rights. The thematic code of tradition was assigned to passages that referenced patriarchal gender roles, norms, and ideals (including those reflected in conservative religio us teachings), such as statements like "husbands are heads of households" and "wives must obey their husbands ." The unity code was assigned to

PAGE 134

119 passages around shared power, egalitarianism, communication, mutual respect, and collaboration. Finally, the righ ts code was assigned to passages that showed support for individual freedoms and privileges especially with regard to women's rights. In some cases, the narratives were not easily distinguishable as respondents invoked multiple codes in a single train of t hought suggesting that the narratives are not always mutually exclusive but rather intertwined. How do these three narratives tie to Connell's three social structures of labor, power, and social norms? I believe that e ach of these three narratives offers a higher order framework for categorizing the division of labor, the division of power, and social norms around gender in Malawian sexual relationships The TGP inherently assumes that relationships operate under a tradition orientation characterized by m ale d ominance and female submissiveness However, r ural M alawians' challenge d this notion ; while tradition was certain ly a prevalent narrative used to describe gender relations, it was not the only one. The o ther narratives of u nity and rights offer ed alte rnative explanations for power relations beyond notions of tradition While rights narratives tended to emphasize female autonomy and freedoms, unity narratives sugge sted the presence of egalitarian relationship dynamics I believe that these additions of unity and rights serve as extensions to the three TGP social structures of labor, power, and cathexis. Figure 4.1 illustrat e s how the TGP could be re conceptualized to include all three narratives related to tradition unity and rights. The short descript ions under each box are examples of what follows in this chapter.

PAGE 135

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igure 4.1: Re Conceptualizing the Connell's (1987) Theory of Gender and Power

PAGE 136

121 Cathexis: Gender and Relationship Ideals According to the TGP, cathexis refers to the attachment of women's sexuality to other social concerns around impurity, i mmorality, and the primary goal of bearing children (Wingood and DiClemente 2002) The same conventional notions of gender apply to men and constructions of masculinity. But h ow do rural Malawians define the structure of cat h ex is, that is, the social norms and ideals around gender? I st art by cataloging the short phrases used to describe ideal relationships as they related to each of the three narratives of tradition unity, a nd rights (as illustrated in Table 4.1). When asked to describe the characteristics of a perfect relationship, a large majority of focus group respondents used a unity discourse to describe their relationship ideals. Slightly nuanced patterns emerged by gender and marital status. Both married and single men mentioned similar characteristics of ideal relationships in cluding love, cooperation, mutual respect, trust, sex, and getting tested together (refer to Chapter 6 for more detail on testing). Married men also mentioned the importance of "understanding each other", referring to the need for good communication, peace in the household, and respect for relatives. Married and single women described ideal relationships using similar constructs as men (with the exception of sex) including peace, love, cooperation, understanding, good communication, trust, and getting teste d together. When asked to characterize an ideal spouse, respondents often drew upon multiple discourses ( unity rights, and tradition) In reference to tradition, the men noted that they wanted a wife who was virtuous and faithful to their husbands, beaut iful in terms of her physical

PAGE 137

122 appearance, and respectful in terms of clothing and behavior towards husbands, in laws, and neighbors. Married men, in particular, mentioned their desires for a wife who was god fearing, humble, religious, loving and did not gossip or share family secrets with others. When men were asked about what makes a good husband, all focus groups drew upon notions of tradition and stated that he must be a good provider for the family. Men also mentioned how he should be faithful and lov ing towards his wife. Women were also questioned about what makes an ideal husband. Common traits mentioned by both married and single women included being a good provider, trustworthy, loving, and an open communicator. Single women noted that he should be healthy and free of HIV, able to bear chil dren, good looking, and refrained from alcohol and smoking which reflected their status as unmarried in search of a man who demonstrated strong marriage potential. In addition, they often used a tradition narrat ive to describe how an ideal wife shou ld be obedient to their husband faithful, virtuous respectful, loving, and avoided gossiping about others. Mutual respect was noted as a relationship ideal in virtually all of the focus group discussions. Intuitivel y, mutual respect would be coded as part of a unity discourse. However, sometimes respondents really meant rights In one quote, a single man referred to respect as "they get along and agree on what to do without taking someone's human rights." (FGD #3, si ngle men). He went on to explain that men and women could violate eac h other's rights when they did not agree on an iss ue and when one person dominated all the decisions. He added, "People will not admire this type of couple ." His extrapolation to larger c ommunity

PAGE 138

123 perceptions of a good marriage suggests that mutual respect is an important relationship ideal that many couples desire in their marriages. Love, Materiality, and Power In about half of the focus groups, love was noted to be the most importa nt r elatio nship ideal I characterized love as an aspect of unity under the social structure of cathexis, although respondents' descriptions suggested that it intersected with the divisions of labor and power In nearby South Africa, Mark Hunter (2010) describes two types of love: the first type is called "provider love" and is intimately tied to the division of labor through cooperation and mutual ass istance. The second, a more modern type is called "romantic love" and could be explained by couple interactions such as the exchange of love letters. Hunter describes this type as more of an individualistic love reflecting a greater choice in marriage par tners whereby men and women can leave relationships at any time. We asked respondents to explain what love meant to them. Unsurprisingly, love carried different meanings to men and women. The meaning of love also diverged based on marital status. Female focus group respondents associated love with materiality and exchange. For married women in particular, love was strongly tied to tradition and the sexual division of labor that is, whether their husbands lived up to the provider role. In one example, a m arried woman replied, "it shows he has love when he leaves food in the house" (FGD #1, married women). Another group of married women talked about love as a collaborative act of unity i.e., "working together" (FGD #7, married women). While single women al so talked about materiality, their descriptions of love included the exchange of luxury items beyond

PAGE 139

124 subsistence items like food and other necessities required for a family. Other women described love as a means to control a partner as enacted through the sexual division of power They argued that when you love your partner, you do what they want, whether it be dressing a cer tain way, providing sex, or giving a girl what she asks for. One single woman said, "The love I am talking about here is when you tel l your partner that you want something, he does it immediately. When you tell your parents [to do something for you], they take a long time to give it to you. But your boyfriend does it fast and gives you want you are looking for." (FGD #4, single women). A few respondents conceptualized love as faithfulness and dedication to the relationship. According to these women, people who love each other do not search for love elsewhere and cheat on their partners. F or the men, love was also understood as a product of the sexual division of labor, specifically with regard to cooperation, working together, caring and loving each other, and helping each other out when it came to sex and decision making. In the following passage, a married man from one of the focus gro ups discussed the collaborative aspects of love using a unity narrative and pointed out how love makes a man put his wife first: Love is very difficult to understand. Love means that you and your wife are doing things together and doing everything in time [quickly/immediately], like work for the house. Maybe there is a time when she wants you to be with her, but you go somewhere else instead. Maybe you say I am going to watch the video [there are places in the villages where a person can pay to watch a vid eo in a group setting]. It means the woman has a feeling that when her husband goes somewhere, there is someone he loves more than me [suspects he could be cheating]. Love means that you should consider your wife before considering other people [women]. Bu t if you love your wife and you don't do what she wants, the woman takes herself as not good enough [she thinks she's not a priority]. (FGD #6, married men) Similar to the women's accounts, men also mentioned how love leads to obedience or power : when yo u love her, you do what she wants. According to this narrative, love provides a pathway to power such that the partner who receives more love has more control over the

PAGE 140

125 circumstances of the relationship. Much of the literature on relationship power focuses on decision making (refer to Blanc 2001 for a summary) with the assumption that men have the final say on many financial, sexual, and household decisions. But here, women whose material needs and wants were succe ssfully met by a loving partner indirectly controlled how family finances were spent through the enactment of love. Men (both single and married) emphasized the distinction between true love and false love For example, one single man lamented, "Some wo men are attracted to different things. It can be that she is in a relationship with you but she is attracted to a drummer [musician who plays African drums in ceremonies and other events] or what not. She is not satisfied [with you]. A g irl with real love cannot be found these days ." (FGD #3, single men) In this example, these men alluded to a second type of love reflected in Hunter's definition of romantic love something that is fleeting and rooted in an individual's ability to selectively choose a partner In the next passage, a group of married men discussed how real love is not superficially based on a woman's physical appearance or driven by sexual desires: Man #1: But I can also say that it can happen that one doesn't move about [cheat] and the other doesn't move about and you say there is love. But also when you say love, you don't regard the appearance of your partner or what she has. Because there are some who have love when the person is rich or well to do and say, "I love her ." But also there is a nother love where you say this one is good looking that is why I love her. So without looking good, there is no love. So when we say love, it doesn't regard the appearance [crosstalk]; it's like real love that has no doubts (group laughs) [crosstalk] Man #3: It can make us have another woman because we don't trust the woman. We don't love her. The main thing is that if there is love, you shouldn't have another relationship and the woman shouldn't have another relationship. Man #4: Because some say that I p roposed that woman, I love her. While others say that I don't love that woman, I just use her with the aim of just sleeping with her. So there is no love since he just wants to sleep with her. (FGD #8, married men)

PAGE 141

126 Though briefly mentioned in the above p assage, married men elaborated in other conversations that real love was not based on riches. The level of detail in their stories suggests that these are not just perceptions but rather reflections of personal experiences with women who claimed to love th em but later only wanted them for their money. According to these men, true love meant that you accept ed a person as they were during the good times and bad times. As one married man put it, "Love is not tied to money. In some families, they love each othe r because one of them is working, simply because of the money. And when one loses their job, love decreases because that person is no longer able to provide." (FGD #4, married men) Tremendous social pressures are placed upon young men to provide women wit h what they want and not just during marriage. This idea resonated in a follow up conversation that I had with James, a 23 year old, English speaking man from one of the focus groups of single men. When I asked James why he does not have a girlfriend and i s living at home with his parents, he replied that he couldn't afford a girlfriend right now since he was unemployed. The following is an excerpt from my field notes: I asked him to tell me about any current girlfriends or memorable relationships of the pa st. He does not currently have a girlfriend. His last girlfriend was from 2 months ago but they broke up because he wanted to focus on his education. He claims that they broke up because she was demanding too much and since he had no job, he couldn't affor d to give her what the wanted. He says that Balaka girls want money. When I asked him what he was giving her, he said money for clothes, units, biscuits (not school fees as I had presumed ). He says she wasn't materialistic, but came from a poor family and needed the money for these things. When I asked him if he wants a girlfriend, he says no, but qualifies his answer with having no money and she will need money (so its not a matter of not wanting one, its just not an option for him right now). He said tha t he wants to have a good job before getting married. (Field notes dated October 18, 2011)

PAGE 142

127 Materiality and the exchange of resources from men to women are important factors that bind relationships in rural Malawi; this exchange may empower both gender s. Women feel powerful when they receive, while men feel powerful when they give. As noted by Wamoyi and colleagues (20 11) in Tanzania, young women felt powerful when they used their sexuality to exploit men for money and gifts. These women considered themselves as "lucky", not disadvantaged, to have female bodies that could be cashed in to meet their material desires. In their seminal qualitative study in Malawi, Swidler and Watkins (2007) argue that transactional sex may be best understood as one of the many ties of unequal exchange between patrons and clients. The authors write, "just as women need patrons to provide them with material benefits, men need clients who provide them with an outward display of power, prestige, and social dominan ce and an inward sense of behaving morally." The consequences of being a poor man and lacking the very thing women desire limits men's ability to carry out their life aspirations a round marriage and childbearing: social states that bring tremendous status to men. In such systems of dependence, women rely on men for material gain just as much as men depend on women to perform their masculinity and for upward social mobility.

PAGE 143

Table 4.1: Focus group respondents' perceptions of ideal relationships Married Si ngle Male Perfect relationship: Love Cooperation Understanding each other Mutual respect Trust Sex Peace Relatives are treated right HIV testing together Perfect wife: Understanding Loving Virtuous Respectful Trustworthy God fearing/religious Beautiful Do es not gossip/share secrets Perfect husband: Understanding Loving Calm Trustworthy/faithful Good provider Cooperates with wife Uses his power wisely Perfect relationship: Love Cooperation Mutual respect HIV testing together Trust Sex Perfect wife: Beautifu l Respectful Virtuous Hard worker Trustworthy (does what she says) Perfect husband: Loves his wife Good provider Faithful Religious/spiritual Respectful Does not drink or smoke Female Perfect relationship: Peace Love Cooperation Understanding each other H IV testing together Trust Perfect husband: Good provider Understanding man Trustworthy Loves your relatives Open (communication)/discusses HIV with wife Loves his wife Obedient (gives wife what she wants) Perfect wife (not asked) Perfect relationship: Resp ect Love Cooperation Good communication HIV testing together Trust Perfect husband: Loves his wife Should be able to bear children Good provider Trustworthy (faithful) Open communication Respectful Religious/spiritual Does not drink or smoke Physical appea rance Healthy (no HIV) Perfect wife: Obedient Faithful Virtuous Respectful Does not gossip Loving

PAGE 144

129 The Sexual Division of Labor According to the TGP 's sexual division of labor men and women are allocated to certain professions and occupations It assumes that women are assigned to more unequa l positions as compared to men, which creates economic imbalance s that force women to rely on men financially a premise that centers on what I call tradition In order t o determine the extent to which this theory held in the Malawian context, respondents were asked to debate the meaning of the phrase "head of the household" and to verbalize whether men, women, or both genders are allowed to take on this role. After reading the transcripts, I was struck by how critically respondents contemplated their answers to this question in such a pensive and rational manner Both men and women used tradition, unity, and rights narratives to describe the division of labor Given the interconnectedness of the TGP social structures, re spondents also invoked t he social structure of cathexis to j ustify the appropriate division of labor between men and women. In what follows, I separated out examples of the different narratives used and then presented how respondents contested multiple dis courses together during a single conversation or debate. Tradition Narratives The tradition discourse was by far the most widely used narrative to explain the role of head of the household. Women described the head of the household as the leader and prov ider for the family, the one who works, the breadwinner, the one who takes care of the household needs and his wife and children, the one who sets t he rules and gives instructions, the owner of the house, the problem solver, the decision maker, and the one who plans for

PAGE 145

130 the family's future. With regard to tradition the men's narratives echoed many of the women's responses. In one of the groups of married men, a respondent used a tradition narrative to describe how husbands make the important decisions and wives obey their husbands' orders. Most women depend on their husbands for everything, saying "if my husband tells me that we won't go to the garden, then we won't." Three quarters of women in the village I can't say for women in the town because I haven 't lived there before just wait for the man. Even if the house is leaking and if there is no plastic paper and the woman tells the husband you should buy me a paper. If the man refuses, the woman can't get the money. Even if she sees the money, she can't b uy the plastic papers. She just waits for the rule of the man to happen. When it happens, she is just obedient and she just waits for that rule from the man. (FGD #6, married men) Male respondents provided additional justifications for who is the house hold head: some cited Christian teachings from the Bible, while others defined the head of the household by the one who proposes marriage which was largely perceived to be the man. Respondents also noted how the head of the household was supposed to be the older partner (presumably the man) given cultural rules that dictate more respect for elders. In general, many resp ondents believed that the man was always considered the head of the household as dictated by customs and long lasting ideals around male res pect regardless of what the circumstances may be. Why might male respect surpass the respect deserved by women? In South Africa, Mark Hunter (20 10) talks about male respect in relation to the head of the household role as something that came out of men's migration to urban areas in search of jobs that later brought status, wealth, and respect. Respect and its historical ties to labor is just one explanation. Yet women who are employed may not be privy to the same social benefits When focus group respondents were presented with the sc enario of a wife works or who

PAGE 146

131 was the sole breadwinner, some respondents argued that she could never be the househo ld head. A wife can rule a husband, "but not like the way that women are ruled" by their husbands (FGD #5, single women). In the following conversation, the facilitator presented a hypothetical scenario of a husband who did nothing to contribute to the hou sehold but the wife worked He then q uestioned the group about who was the head in this scenario, which elicited conflicting responses among them: Man ? She looks after children, buys food in the house, and does everything else that needs to be done? In such a case, the woman is the head of the household. Man #1: A train does not turn as a car does, when it wants to return, it just moves backwards. But though it moves backwards, we are still able to tell which side is the front and which one side is t he back. This is the same in the family. In a relationship wher e by the man does not work but the woman does, more people will say the head of the household is still the man. (group laughs) Man #3: What you are saying is not true because in this case, the woman is doing everything. (FGD #4, married men) According to the second man's perspective, larger societal definitions of gender trump any local derivatives. Although this may indeed be the majority's opinion, the presence of differin g opinions among t he focus group men suggests that gender roles related to the division of labor are not straightforward or easily definable. As further exemplified in the next section, not everyone believed that tradition justifies men's unquestioned authority over househo ld members. Unity Narratives Though unity was less frequently invoked as compared to the tradition narrative when explaining the division of labor some respondents believed that if both spouses contributed to the household, they were each entitled to t he role of "head of the household" and not just symbolically. For example, a group of married women reported that men and women

PAGE 147

132 could lead the household together since both partners were required to make important family decisions. In one group of married women, a respondent stated: They need to sit and make decisions together as one. Looking at the fact that these two are working together, the man nor the woman cannot be called head of the household without the next person who in this case is the woman or the man. They both need each other's help to be called the head of the household. (FGD #1, married women) In a different focus group, single women also shared the perspective that women could be household heads with their husbands through their contribu tion to the domestic sphere (i.e., washing clothes, cooking, and raising children). These beliefs resonated amongst the men as well. For example, one focus group of single men contended that a marriage is about two people and both husbands and wives should respect each other; thus, both of them could be considered household heads. In contrast to the tradition narratives and the emphasis on male respect, unity narratives encompassed ideas of collaboration and mutual respect for one another. Rights N arrative s I coded passages as rights narratives when respondents talked about the circumstances in which women could lead the household alone. I also cited the rights code when people spoke of women receiving an education and gaining jobs. I suspect that responde nts were most likely reciting language they learned from their exposure to government and community gender empowerment programs notably founded upon a human rights framework. Respondents, including men, identified several situations in which women could in dependently take on the role of head of the household. For some respondents, the title of

PAGE 148

133 head of the household was perceived as fluid, subject to change with a family's economic circumstances, and a position that was up for grabs. First, a woman could d eclare herself as the household head if she had a husband who was quiet and submissive. However these women who took control over the household were perceived as usurping their husband's legitimate power; something that was not rightfully theirs. One marr ied woman said, "Sometimes it is because their husbands treat them very well and they take advantage and take up the role of being the head of the household. Because of this, some women challenge their husbands." (FGD #1, married women). Second, women coul d be considered the head of the household when their husbands did not work and the women were forced to provide for the family. In the following group of single men, respondents carried on a conversation about how a woman could take over as the household h ead if her husband was lazy and failed to find work. Woman #2: A woman can be the head of the household because of what the man is doing. Because if you, the man, are docile [lazy], a woman can have more say in the household even though you are there. In terviewer: Can you explain that you cannot be docile, in what way? Woman #4: Meaning a docile person who just stays (cross talk), just waiting [not working, staying at home]. A woman is running a business but he just stays from morning to dusk. Woman #2: Just waiting for what the wife will bring [to eat]. Woman #4: When she brings things (cross talk), he eats (cross talk) and then he just goes and plays games with friends. Woman #1: Even if the woman says "we will eat not today", he will still stay [fa il to find work to buy food]. (FGD #3, single men) While these extenuating circumstances certainly increased women's responsibility and decision making power in the household, it was probably a situation that many women would have preferred to defer to their husbands. This is unlike in the West, where young

PAGE 149

134 people perceive notions of egalitarianism to be the ideal relationship arrangement (Gerson 20 10) Contested N arratives While the unity, tradition, and rights narratives above appear well defined and neatly bounded, men and women often used all three narratives sometimes in a single conversation thread or individual response as they made sense o f the division of labor in Malawian relationships In the first debate below, four single men debated back and forth about whether women can be household heads. The first man (man #5) argued with a unity narrative that women could help to provide for the h ousehold. A second man (man #4) chimed in with a tradition narrative contending that social norms dictate d that it was the man who provided for the family. He questioned where a woman would get her money from, suggesting that she may be receiving money fr om her extramarital partner. The first man replied to this man's remarks using a rights narrative, saying that these days women receive d higher education and could therefore contribute to the household through employment. They said, Man #5: If the man is failing to find money [through work], the woman can also help since the marriage is two people. She can help, by saying "here is the money, let's buy relish [food, usually vegetables or meat] so that we can eat in the household." (UNITY) Man #4: But man will still question the woman's money, sa ying "where has it come from?" Because at the household, it is I who provide everything. So you have to ask. You cannot just be receiving because in the end you will eat things from a fellow man [boyfriend of wife] (group laughs). (TRADITION) Man #5: On the point about how the household head is the man, I for one disagree because nowadays the world is changing. Women are getting educated and finding good jobs. Let's say I am working and the woman also working. We wi ll be doing things together. There is no relish, maybe the man has no money and the woman will give the money. So we can't say the household head is the man. There we should accept that the household is supposed to be run by two people. That's a household. (FGD #3, single men) (RIGHTS/UNITY)

PAGE 150

135 As the conversation continued, the interviewer questioned the group about whether there were other characteristics of the "household head" beyond those related to economics. He asked specifically about decision making dynamics in the family Interviewer: Ok, as you have all said, you are taking the head of the household as the one who provides the household needs, if I got you correctly. Does it mean that the household head [role] ends at providing? Because there are other households where the man is not working but the woman is the one working. But for something to be done, the man is needed [for example, to make an important decision]. So can we say in households like those, the man is not the household head? Man #4 : He's the head, as he is needed [to make decisions]. Man #2: Because it is the man who makes the big decisions. (TRADITION) Man #4: But also what we need to say is that a woman can't help provide the whole budget for the month in the household without the man doing something, even if you don't work. You can try doing other things with the aim that you should solve other problems. You should lighten the burden. You can't just say that because I am not working then I will be eating what the woman provides no. (TRADITION) (FGD #3, single men) In this scenario, two other men argued that even if the wife was the family br eadwinner, the husband still had the final authority on important family matters. In this case, the social structure of cathexis, specif ically men's legitimate right to power, had a stronger influence on power relations than the actual division of labor in the household In a subsequent conversation about the special circumstances of when a woman could be considered the family leader, t he same group of single men debated about changing gender roles and how people are taught about freedom and rights in school, but nevertheless, a woman cannot have more freedom than a man poin ting to the structure of cathex is. Interviewer: Ok, is there a ny other ways that a woman can be the head of the household? Because it seems we have shifted. At first we were saying that a woman can't be the household head, but it seems there are some times where a woman can be the head of the household. What other ti mes are there? Man #8: There are times where you as a man, you know about freedoms. Maybe you went to school and you know that personal freedoms are supposed to like this and that when in marriage. But the woman can have greater freedom such that she can direct the household. And you do not even know that she is directing you. (RIGHTS)

PAGE 151

136 Man #4: But her freedom can't be greater than you, the man, when you are present. It's not true, the woman giving you money to go and buy relish [vegetables and meat] and you rush out going to buy the relish (group laughs) Are you man enough or not? Or you have become a woman? (TRADITION) Man #2: A woman is just a helper. (TRADITION) Man #4: A woman is like a worker. You marry her and leave her at your house. Everything y ou buy, like relish, clothes, food at the home. So if you are buying those things, a woman cannot tell you to go and buy this, or do that no. Or even on dressing [clothing], you can tell her that what she is wearing is not proper dressing but someone else can't tell her. (TRADITION) (FGD #3, single men) In this group, the men worried about losing their masculinity to an overpowering woman who ordered them around. Women's r ights to power conflicted with what they perceived as a man's legitimate right t o authority in the family. In a different group of single men, respondents participated in a very lengthy, 7 single spaced page debate on the appropriate division of labor Eight men shared their opinions during the debate with many of them supporting a tradition justification for male entitlement to the head of the household role. Tradition narratives were contrasted with a unity viewpoint that both husbands and the wives could rule the house together. What follows is a condensed version of their convers ation. The first man started off th e debate by explaining how it was true that in some households women were the main providers but the Bible ultimately designated the man as the authority figure. He said, Man #8: I can explain that it all starts from th e Bible, which says that [crosstalk] women should respect their husbands and that the man is a very important person in the house. But even though the man is the household head, we should not say that in all families they depend on the man. Some households depend on the woman [crosstalk]. She can plan such that the household is run by the woman. But we just say that the household head is the man because of the Bible. (TRADITION) Interviewer: There, you have said it perfectly that some households, it is the woman who plans. And can we say that in that situation the household head is the woman? (Pause) Similar to the other group of single men presented earlier, what emerged from this question was a debate about the extent to which tradition dictates male auth ority. The men continued,

PAGE 152

137 Man #2: In that household, the household head can be the woman [crosstalk]. Man #4: But she can't be a household head. Man #7: But let us not say that because the woman does that [provides] then the household head is the woman. The head will still be the man. [crosstalk] (TRADITION) Man #4: She just works hard. Man #3: She does all that but she still has to respect the man as the husband. The household head is the man. (TRADITION) Man #6: We are saying that the household head i s the man, but I disagree. The household head is both of them because when they say marriage it is both of them. Marriage can't be one person only. Everything concerns both of them, like sex. Two people have sex. In the Bible, you said that the woman shoul d respect their husbands but how I heard it is that the woman should respect their husbands but also husbands should respect their wives. (UNITY) Man #8: Yes they said that, but brother, let us take the Bible, Jesus Christ, when he fed people with fish and bread, between men and women, who were counted? [meaning only men were counted] (TRADITION) (FGD #8, single men) In recognition of the strong passions people have for their religion s and the possibility of it of derailing the conversation, the facilita tor politely redirected the group's attention to the question at hand. This time, the group used a second metaphor around the military to make the argument that both spouses could not lead a household together. Interviewer: Ok, let us leave the Bible ther e, but in this area, the household head means what? Man ?: It is the man. Man #6: When they say a marriage, it is how many people? (UNITY) Man #3: A marriage is two people but there is a leader [crosstalk]. (TRADITION) Man #8: For example, let's take soldi ers. We say there are soldiers but there is also a leader, a commander. Even though they are all soldiers, there is a need to find one who will lead the group. Any group has a leader even here as a committee [referring to the focus group], there will be so meone. It can't work on its own [the committee] but there will be a need for a leader who will run the committee. That one is the head. Similarly in a family, whether five children or seven but the one who is there, the man, that is the one you all look up to. Maybe there is hunger and you have no food, you go there [to the leader, i.e., man]. (TRADITION) (FGD #8, single men) The men then transitioned to a discussion of house hold decision making and who had the final authority in that particular domain. In the next passage, several respondents discus sed how a double standard existed around how men could make decisions on their own, but women need ed to consult with their partners beforehand.

PAGE 153

138 Man #7: It may be that the woman is well to do, is rich. Perhaps a child says he doesn't have a pen, you'll hear that child tell the father first. The father should buy it for the child but the woman has the money. She still says tell [ask?] your father'. (TRADITION) In response to his statement, the next man ar gued t hat it was both spouses who share the household power. However, this man's opinion was counterpointed by other men who believed that husbands should have the ultimate authority. Man #4: Let me explain. We are saying that the household head is the man, whi ch means that we [the men] should all be household heads. Because now with money, the woman doesn't always depend on the man and the man doesn't always depend on the woman that she'll give me money, no. But together you'll find that the woman works and the man also works. Both of you receive the money you have made. Meaning that both of you are household heads and that you depend on each other. (UNITY) Man #8: When they go and work in someone's farm, they get money. Or the woman will farm in someone's field and the man in another field. It can't happen that the woman takes the money and makes plans on how to spend the money. What will happen is that the woman will get the money and give the husband. The man will get his and put it here. The man will do his p lans and he will tell his wife that what is here is for the children's notebooks. Get the money and buy them, meaning that the man is the household head. (TRADITION) Man #3: Even the woman, when she finds a contract to farm she can't just go and farm witho ut telling the man. It can't happen. She will tell the man that I have found this, I want to work in the farm. If the man has found another contract, he will tell her that don't go there, let's go and farm here. (TRADITION) Man #6: The man can just go on h is own without telling the woman. (TRADITION) Man #3: Meaning she can't complain. (FGD #8, single men) As they end ed the conversation, two men arrived at the conclusion that the man had all the power in the marriage but his power was passed down to him th rough unspoken traditional rules and norms around gender relations (cathexis). Man #2: I can say the man has all the power. The man has 100% power. (TRADITION) Man #1: Not having it by force, but by how the rules are here. (TRADITION) (FGD #8, single m en) To summarize this debate, r e spondents narratives allude to the interconnectedness of the three social structures of labor, power, and social norms on gender. T he division of labor could be based on husbands and wives respective economic contributions to the family but perhaps more from deeply embedded soc ial norms related to cathexis The division of power

PAGE 154

139 was perceived to result, in part, from the sexual division of labor who had more economic power as well as from social norms around male respect a nd legitimate authority. Even though the majority of respondents appeared to hold conservative gender ideals around tradition this did not necessarily mean that t he balance of power always favor ed men In the next section, I show how respondents talked mo re about unity when asked about sources of power. It is plausible that tradition narratives around social norms such as a man s legitimate and cultural right to power masks the actual distribution of power in relationships. The Sexual Division of Power In this section, I present what relationship power means to young Malawians. The most common narrative used by all focus groups to describe power was unity which included references to love, sex, respect, helping each other, and open communication as indica ted in Table 4.2. While love as a source of power was mentioned by almost all of the groups, the ways that men and women talked about love differed between the two groups. Women talked of feeling loved when their partners gave them gifts or things they ask ed for, which in turn made them feel powerful. Men associated love with obedience i.e., women did what they wanted and when they wanted it, such as cooking for them, drawing water from the well so that they could bathe, and having sex. In one conversation, the interviewer asked a group of single me n to explain exactly how love was associated with power. One man summarized the main point of the discussion with his response. H e argued that when a woman loved her man, she obeyed his orders which in turn brough t him power. The major point is that we propose to the woman. So when you propose you say to her, "I see you with another man, you should not move with him again ." The woman will listen if she loves you with all her heart and her trustworthiness. You give her rules and she follows the rules. So the power also

PAGE 155

140 comes from when she following rules. That means the power is working [cross talk]. For example, if the government declares that the people should not steal and they are not stealing, that means they a re following the orders from government. That's where the powers are. (FGD #8, single men) There was some common ground among men and women in the way that love as it related to power was tied to open communication. Both married and single groups of women mentioned how they felt powerful when their partners listened to their advice when they disagreed with his approach to important family decisions. Some people stated that when you love d each other, you could talk freely about things. In the following pass age, a single man provide d a vivid example of how love was inextricably linked to aspects of unity: understanding, mutual respect, and the ability to speak openly with a partner, which yield ed power: A person knows that he has power in the relationship wh en there is love. When he speaks to his partner, he speaks respectfully and the other partner responds respectfully as well. And when he brings up a story and the story is discussed properly amongst them, he knows that he has power and he is free and open to bring in different issues to be talked about with his wife. He knows there will not be a problem with coming up with better solutions. In such a way, one knows that he has power to speak. (FGD #4, married men) Similarly, in another example, a si ngle wo man believed that she had power when her partner listen ed to her thoughts and opinions: When you tell your friend something and they listen to you, you know you have power because when you tell a person and they don't listen, you know you don't have power. He belittles me. So when you tell him something and he listens, you know that you have power. (FGD #2, single women) Sex and sexual satisfaction were also noted as important sources of power among male and female respondents. Some men provided details about how "being on top" during sexual intercourse made them feel powerful while women spoke about sexual satisfaction as a source of power. The following group of married women suggested that feeling sexually desired and having one's own sexual desires me t made them feel powerful:

PAGE 156

141 Woma n #4: All that matters in a relationship is sex and that is what makes a woman feel powerful. Interviewer: In short, we can say for a woman to feel powerful it is all about sex? All: Yes. Woman #2: When the two satisfy each other's needs. For example, if the woman has strong sexual desires and the man doesn't have them, can we say that anything will work out there? All: No (group answers). Woman #2: For sure it cannot work because women are like fishing ho oks where b y men get stuck on them and women always want their hook to have something hooked to it. So if both parties have strong sexual desires and have sex, that's something that makes us feel powerful. (FGD #1, married women) In this narrative, the example of se x was used to typify how broader relationship characteristics such as reciprocity and consideration for the other partner's needs, thoughts, and circumstances brought individuals power in their relationships. In addition to unity male respondents also br ought in tradition narratives to describe additional sources of power for themselves specifically, regarding the division of labor. First, men spoke about their social roles as heads of the household, family providers, leaders (and the respect that came wi th leadership), and decision makers as sources of power. Second, men noted that they felt powerful through proposing women (a man can propose a woman to "date" him as well as to be his wife). Finally, one group of married men mentioned how their fertility and ability to produce children through the social structure of cathexis m ade them feel powerful. Surprisingly, none of the women mentioned children as a source of power.

PAGE 157

142 Table 4.2: What makes people feel powerful in their relationships? Married Sin gle Male Proposing a woman; being the head of the family/main provider/leadership; the respect that comes with leadership; love and speaking freely with your wife (open communication); bearing children; farming When women do what he wants (cooks for him, calls her and she comes); making decisions; sex; being the head of the household; proposing a woman; giving the wife rules; when the wife obeys your rules and listens to you; being on top during sex; love Female Sex and sexual satisfaction; when your husb and loves you; when he does what you ask; when he listens to your advice; being able to say no' to a husband; ruling the family with your husband Love; when a boyfriend gives a girl what she asks for; when he takes your ideas into consideration The Bala nce of Power in Relationships We asked respondents to tell us how much power (in the descriptive sense) men and women should have in their relationships. The social structure of cathexis was used to describe how gender norms determine the balance of powe r in relationship. Both men and women provided similar narratives centering on the idea that women should be treated with respect and given "a say" in the relationship. Most focus group respondents arrived at the conclusion that men should have more power in the relationship overall (with a minority of people leaning towards equal power) and virtually no one concluded that women should have more power than men. In 2010, when the wave 3 TLT couple sample was asked "who is generally in char ge of the relation ship", approximately 88% of men stated that they were the ones in control of the relationship. Approximately 80% of women confirmed their male partner's reports. Less than 1.5% of men and women reported that the women have more

PAGE 158

143 control than men in their re lationships. Approximately 18% of women and 10% of men said that they shared power equally with their partners. Chapter 5 will present more details on these statistics. Regardless, statistics such as these do not capture the nuances of power and the fact t hat respondents may be more likely to justify their responses to an absolute measure of power given the social norm that men are supposed to rule the household. Several male focus groups had respondents that said power in relationships should be more b alanced yet not equal and used a unity narrative to describe the ideal distribution of power. According to respondents, a man shouldn't abuse his power by beating up his wife or by treating them like "slaves ." Instead, the couple should collaborate on fami ly issues and decision making. In the following passage, a group of married men shared their opinions on when men's power had exceeded its limits and how this could affect the relationship quality. They explained that when a man was using his power wisely, he was respectful of his wife and her different opinions. The first man said, Man #4: There are some that have power in their relationships and they take their wives as slaves because they have power over everything. They say what should happen, whether the woman wants it to happen or not, she still does it because he husband has commanded and that he has power. Because she has less power, she fails to meet what the husband wants. The husband then beats her up, shouts at her, and sometimes disagreements occur. He becomes violent to her, and later people say he is abusing her. Mostly abuse comes with someone that has power because someone that does not have power cannot abuse the next person. In this case, it is good that the man should have enough power b ut he should use it accordingly. When we say accordingly, he should be able to respect others for example, his wife, children, and everybody else that lives there with him. A second man emphasized that a husband should take his wife's opinions into consid eration not just out of respect, but because it could ultimately benefit him to do so: Man #3: There are some men that say that they have more power in the relationship and do not want their wife to oppose what they have said. When the wife dares to oppo se the man tends to beat her up and even chasing her to go back to her parent's house. In this case, the man should have power to also welcome other people's ideas and suggesti ons.

PAGE 159

144 Because to the point where by he abuses his wife, that means his power has e xceeded its limits. But even though he has too much power, he should not use these powers by bringing in disagreements or by being violent to his wife. When the woman is opposing what her husband is saying it means that there is something that is wrong bec ause there are some women that can help you make better decisions that can even make you rich. We now see different women in towns getting involved in different businesses. When you ask them, they tend to tell you that they got their ideas from their husba nds. We cannot know who in the first place came up with the idea of starting a business; sometimes it is possible that the woman is the one that brought up such an idea. For example, she may say "To avo id over spending, you should open for me a shop that wh en you are away, me and the children should not have any problems and the business will provide everything for us ." Looking at the idea that the woman has brought, the man will take his money and give it to the woman to start the business but when the man is not using his powers accordingly, he will not agree with his wife and this also may cause disagreements. (FGD #4, married men) According to these two men, overly control ling male behavior that included disrespect and abuse of a wife were not acceptable uses of power even when a wife intentionally opposed her husband's authority. As a result, abuses of power and the failure to consider a wife's opinion could end up backfiring on men and limiting their opportunities to build a better life for themselves. Another group of married men used a rights narrative to explain how men's power shouldn't exceed that of the wife. For example, if the wife is sick, he shouldn't make her complete her farming work because such an act would violate her rights. In the next passage, one man summarized a previous point made in the group by stating that, "power and rights need to go together ." Another man added that the husband should also respect the number of children she wanted and give her a pause in between pregnancies. Man #4: Powers of a man should not exceed the point that it disappoints the wife. The powers should have limits not exceeding the ability of the wife. Because there can be the use of power such that the woman does not have peace in the marriage. For exampl e, maybe she is sick, but because of your power, you tell her "you'll not sleep until you water the nursery of tobacco seedings Or maybe it is a garden and you are not ploughing, but because the woman is the one who goes to the garden and you go to town, you won't listen that she is sick and you tell her to still go and farm. For those powers, the man

PAGE 160

145 should look at the limits. His powers should end there so that the woman has a role in terms of her rights. Power and rights need to go together. Man #3: On the same issue he has explained, you need to see how the woman is, especially how you are staying. Especially on the part of child bearing, you need to know that I should give my wife a break although I have the power to have a number of children. I shou ld have a small number considering how things are. (FGD #6, married men) According to the focus group respondents, a woman's rights could also be violated when a man did not consider her opinions in important family matters that could potentially impro ve the household well being. The following group of single men engaged in a debate about women's level of decision making power by contrasting rights with tradition. At the en d, they all agreed that women did not have to listen to everything their husbands said, especially when they felt his decisions were not wise for the family. Interviewer: You said you know you have power when you tell a woman to do something. So should the woman listen to anything because you have more say? Man #1: Its not that, it a ll depends on what things. When she is not happy about it she can't accept [agree to] everything. When you force her, its like you are infringing on her right. (RIGHTS) Man #4: Because its different, when she is coming from her mother's house and you are dating, she can refuse something. You can say "let's go somewhere" and she can say "I don't want to ." She stays with her parents then but with your wife, you can just tell her "let's go ." (TRADITION) Man #2: You have taken beer and you tell her to drink i t. So she should just accept because she is a woman (cross talk) and you are a man? (TRADITION) Interviewer: But here, let us say how we see things in our villages, are women expected to accept everything that their husbands say? Man #1: No, they are not expected to accept everything that the husband says. It depends on what he says. Man #2: If the things will be beneficial for the household, she will accept but if not [beneficial for the household], then she can't accept. Interviewer: Ok, as we have said that as people, you have powers. I believe women also have some power at the household. Since you have said that she cannot just accept everything, it shows that she does have some power right? All: Yes. (FGD #3, single men) Some male responden ts attempted to quantify exactly how much power men and women should be allocated using percentages. A group of single men argued back and forth about the figures and the transcripts revealed that there was a lot of cross talk going on during the

PAGE 161

146 discussio n suggesting that respondents were actively engaged and interested in coming to a group consensus on the ideal balance of power. Their conversation continued: Man #3: The man should not have a lot of power but should have a little more than the woman. Ma n #4: I can say that the man should have more power but the woman should have a little less power, but it should not be a lot [less]. Because if it is a lot [less], the woman will be low. Man #5: Maybe the woman should have 50 per cent and the man should h ave 100 per cent? Interviewer: You have said 50 per cent and 100. That can be a problem. [alluding to the incorrect math] Man #3: Yes, that is too much power. Interviewer: My point is that you are saying 100%, you mean the man should have all the power in the household? All: Yes. Man #3: Even to move the house into a tree, it can be done? Man #4: The woman can't have a say. Man #5: Maybe there we can say I am wrong. Man #6: The man should have 60% while the woman 40%. [crosstalk] Interviewer: So mainly we are agreeing that the man should have powers more than the woman? Man #5: The difference [between them] should be 10%. [crosstalk] (FGD #8, single men) Regardless of the exact number, the group generally seemed to conclude that men should have more power overall than women. It was not just the male respondents who believed men should have more power than women. When asked about how much power women should have in their relationships, women often drew upon social norms related to the tradition narrative to support men's higher power. One group of women discussed how women should not be more "intelligent" than men, meaning that it was unacceptable for a wife to rule her husband by demanding things from him or withholding sex until he met her requests. In a nother group of married women, one respondent stated that "A woman should have power in marriage but she should not have powers that will exceed the man's powers. Sometimes when you have more powers than your husband you can also ruin your marriage." (FG D #7, married women) This woman suspected that a wife with more power could become rude and disrespectful

PAGE 162

147 towards her husband thus causing marital conflict. Her account suggests that for some women, relationship aspirations and goals are a higher priorit y than the quest for more power in the relationship. In the next section, respondents expanded upon this idea that "too much female power can lead to problems" as they described the consequences of gender role transgressions related to female empowerment The Influence of Gender Policy on Relationship Power In 2000, the G overnment of Malawi passed its National Gender Policy in order to "enhance the participation of women, men, boys and girls for sustainable and equitable development for poverty eradicatio n" (White 2007) Following the developme nt of the National Gender Policy, a National Gender Program was launched in 2004 to bring about gender equality and the empowerment of women. Eight priority areas were identified and include but are not limited to education, health, HIV/AIDS, economic empo werment, and human rights. Human rights are often at the forefront of gender empowerment policies and programs worldwide as referenced in the United Nations 2010 Millennium Development Goals (United Nations 2010) Gender empowerment programs attempt to shift power imb alances by directly addressing the sexual division of labor as well as social norms around gender (i.e., cathexis). In sub Saharan Africa structural level interventions aimed at changing the sexual division of labor have ranged from microfinance solutions to land rights, food security, and education for women. Social norms around gender have been addressed through the diffusion of messages on gender equality, for example, as incorporated into secondary

PAGE 163

148 education curriculums These structural level factors operate at a different level of the Social Ecological Model than the interpersonal level studied in this dissertation. However changing gender roles at the societal level have real consequences at the relationship level by challenging the ways in which m en and women relate to each other In Brazil, for example, Hautzinger (2007) suggests that women's gains in the labor market could, in part, be interpreted as men's respective losses. She argues that the shifting gender organization of labor, household aut hority, and pe rceptions of entitlement creates new patterns of conflict within relation ships, particularly, the rise of gender based violence as men attempt to defend themselves against more powerful women (Hautzinger, 2007). In South Africa, Mark Hunter ( 2010) notes similar increases in gender conflict that ca me as a consequence of modernity, women's rights, and men's decreasing control over work and housing. These examples, among many others around the world, highlight the need to consider ho w gender empo werment programs and policies at the structural level change the way gender is talked a bout and ultima tely understood at the ground level. Focus group respondents were asked to share their opinions and perceptions of government sponsored policies and prog rams on gender. Overall, men and women shared mixed opinions on the benefits and the drawbacks of gender empowerment but appeared to place stronger emphasis on their negative consequences. Despite the widespread accolades of gender empowerment in the West rural Malawians carried very different perceptions.

PAGE 164

149 Gender Equality Fosters Unity in Marriage When respondents spoke of the benefits of gender empowerment, they used a unity narrative. From men's perspectives, the advantages included women's ability to help with family decision making and the opportunity to contribute to family income via their labor outside the household thus taking some of the pressure off men to provide everything for the family. In addition, men viewed additional income gained throu gh women's employment as a chance to improve their family's economic situation and plan for the future. Men also perceived women's employment as a form of insurance to prevent the family from starving should the husband die, become ill, or fail to find emp loyment. Some men even said that when both spouses are working, it builds unity love and openness b ecause both parties feel like they can contribute to family discussions. Respondents generally agreed that recent gender empowerment policies have increased women's power in society through the sexual divi sion of labor At the relationship level, increased earning potential was thought to buy women a level of decision making power on how to spend their contribution of the family income. But not everyone agree d with this perception For example, in one group of married men, a respondent disagreed saying that gender empowerment has given women rights, but not more power. This man juxtaposed rights with tradition to explain how women have the right to make decisi ons but they still need to gain approval from their husbands. In this account, new social norms around rights were secondary to norms around tradition that require d women to seek their partner's permission.

PAGE 165

150 Well, according to me, when a woman agrees or d isagrees does not mean that she has power but I can say that she has human rights than allow her to choose what she wants. But when she makes her choice, if she is married, there's a need for someone to analyze if it is good or not. Because even when a wom an wants to be a builder, she has to let her husband know first before she takes a step in doing so. Her husband should agree and see if she can manage or not. This does not mean that the woman has power because if she did, she would just go ahead without letting her husband know. (FGD #4, married men) As the above conversation conti nued, men talked about how it was not only women who need ed to review their decisions with their spouses in contradiction to the above man's perception of a double standard. In the following passage, the facilitator asked the question about whether a man could sell a bicycle (an expensive possession by rural Malawian standards) without informing his wife. Several respondents pointed out that a man could get himself into serious trouble with his wife by doing this. Interviewer: There were human rights before not power. Thirty years ago a man would just sell one of his cows without informing his wife. The wife would just get to know about it in the end when the cow was already sol d. So tell me, can you do the same today? Man ? No, that is impossible. Man #1: You cannot do that. Your wife will probably divorce you. Man #3: Even just a bicycle, you cannot sell it without your wife knowing about it. (FGD #4, married men) In an other exchange among a different group of single men, the facilitator brought up the same scenario around a man's ability to sell a bicycle without the wife's approval. One respondent replied that a man could not sell the bicycle without telling his wife. Another person in the group added that the couple would argue if he did this, unless he bought the bicycle on his own perhaps before the marriage started. A third person stated that even i f he got it on his own, he needed to tell his wife about his decisio n to sell it. A fourth person said that nowadays, it is important to consult with each other emphasizing the importance of unity From th is example, the group concur red that women's decision making power has

PAGE 166

151 increased as compared to days of the past. The f ollowing excerpt documents the remainder of the conversation: Man #8: Culture before and now are different. Things are changing with culture. If culture changes, everyone changes his behavior. Because before when you get married, they were saying a lot o f things, like you should listen to your husband, that he is the household head, that's your husband you should respect him, whatever he says you should listen [crosstalk]. So even if someone gets a bicycle and sells it, and the husband says, "I have sold the bike and used the money on beer", the woman will listen since it's what she was told [crosstalk], because it's how culture was. But now with how things are changing, everyone's actions are changing. Because you can't take an ox cart and sell it without telling the wife. It can't happen now, so she will sue you even at the police. Interviewer: So in that example, it is like the difference in power between a man and a woman is less. Man #2: It's like the powers of a woman are increasing. Interviewer: I t's increasing? All: Yes. (FGD #8, single men) The first man above seems to argue that not informing a wife would violate her rights rather than the unity in the relationship. This appears to be unrelated to the division of labor and more about changing social norms around women's property rights. As suggested above, some men attributed gains in female power to the enforcement of gender policies. Whether or not this is true is debatable; yet still, the perception that there are l egal ramifications to vio lating women's rights may be enough to curb overly dominant behavior among men. Most people seemed to agree that unlike the past, household decisions are now shared through women's increasing participation in family matters. However in almost every focus group discussion, a common theme emerged around the belief that changing gender norms were overall a bad thing for men. Competing discourses of rights and tradition were used to explain how as women gain rights, men lose their traditional authority in soc iety and in the household. Both women and men talked about how men felt threatened and emasculated by gender empowerment in several important ways. In their narratives, men expressed deep

PAGE 167

152 fears about the negative consequences of changing gender roles for t he family and the institution of marriage in Malawi. Loss of Respect in Society Respondents believed that if a wife was working, it usually meant that the man was a failur e in society or at least this was what other people would think of him 6 This percep tion would imply that as women's education and participation in the workforce shifts the division of labor, social concerns around masculinity (cathexis) threaten men's power at the societal level. It appear ed that many of the rural men were unfamiliar wit h the western idea of dual income earner families, especially ones where the wife was the primary breadwinner ; m en assumed that the only reason the wife was working was to keep the family from starving as a result of her husband's so called laz iness or fai lure to find employment. According to the men in this study, they found it embarrassing to have an income earning wife because it sen t the signal of their failure to live up to the quintessen tial responsibility of family provider. The women's n arratives pr ovided parallel accounts Respondents believed that people would gossip if they saw a man doing a woman's job and might even say that he was g iven a "love potion This finding parallels some of the earlier findings around the use of love to shift the divi sion of power. 6 Though notably, in the sample of 466 mostly married women from TLT's wave 3, approximately 40% had earned some level of personal income in the past month either through piece wo rk, temporary employment, or a steady job. The focus group men were likely referring to female breadwinners who take away men's status rather than women who earn a small amount of income alongside their husbands, e.g., selling produce in the markets. Futur e research may follow up on this finding in more depth, for example, by investigating how much income a woman needs to earn relative to her partner to decrease his social status.

PAGE 168

153 I initially presumed that the "love potion" analogy was just a symbolic use of language to depict extraordinary cases of female power and male submissiveness as opposed to real experiences. To the contrary, i n her article entitled "Of love potions and witch baskets", Wilson (2012) documents how Malawian women use these love poti ons to curtail undesirable male behavior, particularly, men's mobility and infidelity. As she writes, "some of these medicines increase the measure of love the recipient has for the giver and simultaneously decrease the recipient's interest in having other sexual partners" (Wilson 2012) However, as Wilson discovered, love potions could sometimes backfire causi ng men to become economically unproductive and bound to the domestic sphere due "loving a woman too much ." These men no longer "move around" as men are expected to do and paradoxically cause women to lose out on resources gained through their husband's emp loyment. In the focus groups, performing female jobs what thought to make men appear submissive in the eyes of other s as if they had been given a love potion. As one married woman stated, "with the way that I see it here in the village with gender [meanin g gender empowerment policies], for one to see somebody's man is working [washing dishes] stories are going to be everywhere in the village that the man has been given love potion. That will make the man appear stupid [submissive] when each time you ask hi m to do something, he does exactly that." This love potion analogy came up again during a male focus group (FGD #6, married men). In this case, men discussed how they don't want to be seen by their peers doing women's jobs like fetching water at the boreho le. People believed that men were

PAGE 169

154 embarrassed by performing women's work and so they refuse d to do these jobs and resist ed these changing gender structures. One of the men said, Some do it [female jobs] while others are shy. Maybe she says "you should was h a cloth [ chitenje ] for me at the water tap ." He tells her, "my friends could see me washing a cloth for you ." So, on the issue of gender, some men don't do it. (FGD #6, married men) It appears that love has limits and there are instances when too much l ove or perhaps socially inappropriate expressions of love are problematic. Respondents deemed material transfers, caring for each other, and listening to each other as normal signs of love presented in a positive light. Yet when men loved their partners so much that they performed female related tasks and deviated from traditional male gender roles, love was perceived as abnormal and could even carry a mystical property that only love potions" could explain. In a group of single men, one man pointed out t hat while it was acceptable for women to help men out at home with male dominated tasks like building a house, it was socially frowned upon for a man to do a woman's job. He said, They [men] are happy if the woman is doing work. (group laughs) A man is on the roof and is maybe roofing the house. For him to get down to get a stack of grass, it is difficult. The woman gets the stack and gives it to him and he opens his teeth [smiles] (group laughs) Maybe the woman is busy. Maybe she is washing dishes and doin g everything but is able to leave that job. She thinks that my husband should not get down, let me pass him the stack. Sometimes the man even tells her to give him the stack. She leaves her job and she helps him. But for the man to do a woman's job at the time the woman is busy, that can't happen. That's when you go to sleep, and only be woken up when food is ready. (group laughs) [meaning he won't help with cooking, he'll just eat it] After you eat, you say "W here is the water, I should bathe, I am tired ." So gender is there [things are changing with regard to gender], but its only with a few people. When it is known that you do those things, then you are [seen as] stupid [submissive] in the whole village. (group laughs). (FGD #3, single men) But even if m en hide these embarrassing jobs from public scrutiny, they cannot prevent their wives from telling their friends. Some men noted that when women gain ed too much power they become boastful and might brag to others about how they could coax their husbands to

PAGE 170

155 do whatever they want ed Thus, a husband could still be subjected to shame through the route of gossip. To summarize, shifts in the division of labor were believed to create a source of power for women while at the same time reducing men's value at the so cietal level This appeared to operate through the pathway of cathexis, that is, women's increased economic power caused men to lose power when they were perceived as failing to live up to quintessential ideals around masculinity and the provider role. Me n's power deficit at the societal level filtered down to the relationship level as they lost their authority in the household. Similar findings have been noted elsewhere, such as in Bahia, Brazil (Hautzinger 2007) Men's Loss of Authority in the House Women believed that, in general, men were not happy with gender policies that emphasize d an equitable division of household labor. In one focus group, a single wo man affirmed, "It cannot be something that will make them happy because men cannot wash dishes. Because they feel that they are male and they cannot wash dishes. Ladies are there to wash dishes." Likewise, some men lamented that they do not want women to h elp them with jobs reserved for men. Similar to losing f ace in the community if a man was seen doing women's work, there was also the risk of a tainted image in the eyes of a spouse. In one focus group, the facilitator summarized the general sentiment of t he group by stating, "Men do not like it [gender empowerment] because women are doing the same things that the men are doing. Then it is like the man is losing power because the woman is doing the work that

PAGE 171

156 was supposed to be done by men. This will make th em feel like they are small." (FGD #5, single women) Everyone in the group agreed with her assessment. Power changes may also be related to men's diminished sense of importance in the relationship. A group of married men argued that men's power was decrea sing because women could now do men's jobs such as plowing the fields ; w hen she put on her o veralls to help, the man appeared weak in her eyes since she was doing the same work. One man stated, "Mainly, it can be that a man is working and the woman is work ing [in the fields]. She is using a spanner [tool for farming] and I am also using a spanner, then at home you still remain light [weak] in the face of the woman." Another respondent chimed in to the conversation and said, "She regards you as a woman." (FG D #6, married men ). Here, i t was not the shift in the divis i on of labor itself that caused men to lose their power but rather the emergence of new social norms tied to the act of female labor (cathexis) According to a group of single men, gender policies were thought to cause men to lose some of their authority with women out of fear of the legal ramifications. The se new legal structures existed at the societal level, but created new social norms around women's rights in the relationship specifically, the right to a marriage free of violence This created fear among the m en because they believed that they could go to jail if they exerted their power in the household in previously tolerated ways. In the following passage, single men discussed how gender rel ations were changing due to new laws on domestic violence that punish ed men for abusive behavior: Man #8: So it can be that I have disagreed on a very small thing with my wife in the house. Discussing or bringing in the marriage counselors might end it [th e disagreement]. But w hen she just goes to the police. (group laughs)

PAGE 172

157 Man #4: You'll be locked up there. Man #3: They will say you are being violent. (group laughs) Man #6: Their powers have started to get low [decrease] with how women are able to express their ideas. Interviewer: It is like powers of men are decreasing? All: Yes. Man #4: Men are also bothered by how in the beginning [of the relationship], when the man says this or that, the woman was listening but now when the man says the same thing, sh e is not respecting the man. She knows that if he beats me I will go to the organizations (laughs) so now it's like the man is afraid of the woman. Man #3: If I beat her, I will be locked up [crosstalk]. So the power of men is declining because if I do thi s and we don't agree, then for sure I am going to be arrested. (FGD #8, single men) Men are Losing Out on Jobs Based on what they saw in their communities, respondents perceived that men were having trouble obtaining jobs because women were replacing them in the workforce. In one focus group, a female respondent said, "This hurts most men. For example, at the hospital where the construction is taking place, there are a lot of female builders and because of this men are failing to get jobs because the women have taken over what was supposed to be for men." (FGD #7, married women) A group of single men pointed out that a woman would be selected for a job or university spot when a man and woman were equally qualified because of gender empowerment policies. Man #4: Nowadays, we can say that a woman can have 6 points at MSCE [Malawi school certificate of education, equivalent to O level Cambridge]. And the man also has 6 points but what happens is that between the man and a woman, they pick the woman. [either for college or a job]. Man #3: They say the man has nothing in his head. (group laughs) Man #7: While he has passed [his exams]. Man #3: No, because he has the same [qualifications] as a woman. (group laughs) [crosstalk] Man #4: He is intelligent and the woma n is intelligent but they want the woman to rise up so that is how they are being empowered. (FGD #8, single men) Men's ability to provide for their families through the division of labor was previously noted as a source of power, respect, and characteriz ed as an ideal spousal trait. As a result, m en felt

PAGE 173

158 that gender policies were violating their right and entitlement to steady employment as the main family providers. Therefore, these changes to the division of labor could directly de power men through the loss of employment opportunities. The Destruction of Marriages Both men and women frequently mentioned the belief that gender empowerment destroys otherwise healthy marriages. Respondents noted that divorce happens through sev eral pathways, one of which was when a husband did not feel respected in the household that is, respected above what a woman should receive. In following passage, a married man presented an example of a farmer who was married to a wife with an upper class job. While such dramatic cla ss differences within a married couple are probably highly unusual in this population, his account nevertheless painted the scenario of a situation where a husband felt dishonored and less important by his wife's social status. This man believed that the h usband would leave the marriage because his wife had a better job and he would not feel respected as the main family provider. He said, Maybe the job of the woman is of higher class, but the man's is the same on of farming. It is usually that the man leave s the marriage because his wife has a better job. He says he is not respected. (FGD #6, married men) As previously mentioned, respondents believed that a woman could become "rude" towards her husband when she had too much power and her husband might leave her for her disrespectful attitudes. According to this man, it was not the division of labor and women's employment that caused a m a n to lose his power bu t rather her misuse of this power and how it violated social norms around tradition

PAGE 174

159 Respondents als o believed that marriages dissolve d when external gender equality messages that attempt ed to change social norms b eca me a source of dispute between spouses. Overall, respondents believed that it was very difficult for a man to do the job of a woman and if she tried to persuade him to adhere to new ideals on gender, this could cause conflict in the marriage. In the following powerful quote using a tradition narrative, a married man believed that western ideals about gender shouldn't come to the r ural village s because it created problems. It is becoming very difficult for the man to accept that the job of a woman can be done by a man. For example, even washing, even cooking nsima [the staple food of Malawi], all that a woman does, it is difficult for many men here to follow considering that they take old cultures. There are many questions about why, up until now, these customs were not around. Town life should not come to the villages. These are villages, those things are done in town. So in the villages, wome n do their chores even if they are sick. A woman is sick but she drags herself, saying "I should cook for my husband ." It's only a few men who are able to get water and heat it, maybe even heating it for your wife to bathe. So in the villages, it is very f ew who are following gender [gender equality ideals] and the man doesn't want to hear a woman saying the word gender. For example, saying "I thought there is gender, prepare relish ." This is a problem for that marriage. It will be difficult to stay togethe r, maybe the matter will even go to the counselors [marriage advisors]. The counselors are the ones who start to talk about things like this and that, and gender. When you follow gender, you will dissolve your marriage. To satisfy a man, satisfy the stomac h. But also you should take care of him. You should do everything even if you are sick. Even our parents encourage us that a man should not do a woman's job. This is for white people. We should leave these to white people. They will mislead us. That's how it is in the villages. (FGD #6, married men) While this quote was not inclusive of everyone's opinion, it suggested the presence of conspiracy beliefs around how "white people" might intentionally destroy families and relationships by promoting western id eas that did not fit with traditional village life. And he may be right. To conclude, rural Malawians emphasized more of the negative consequences of rights based gender policies than the perceived benefits. They struggled with how make sense of how rights based gender policies fit with notions of tradition and unity in their everyday lives. Gender policies that center on individual human rights and freedoms socially

PAGE 175

160 celebrated in the West may conflict with local values that many people desire in their rela tionships perhaps causing more harm than good. Discussion Theoretical Significance What can be learned about gender and power from the narratives of rural Malawians? A vast body of literature has almost exclusively centered on what I call tradition patr iarchal gender norms and roles that favor men and has shaped much of what is known about the gender and power While beliefs about tradition still played a role in the way gender and power were conceptualized and understood, rural Malawians invoked two add itional discourses that have received significantly less attention in the literature, namely, unity and rights At times there was contention between the three narratives of tradition unity, and rights. In other instances, respondents used multiple discou rses simultaneously to express their views on gender indicating that there are points of overlap and congruency between the narratives. Overall, respondents expressed difficulties defining gender roles and coming to a consensus on the ideal balance of pow er between husbands and wives. There were signs of resistance to the predominant gender hierarchy as rural Malawians actively challenged the tradition paradigm with alternative unity and rights perspectives. This suggests that gender constructions may be m ore elusive than well defined and likely undergoing serious change in Malawi as people interact with larger social structures that tend to promote a western ideology centered on human rights and equality and with each other as they try to make sense of gen der and redefine gender relations for themselves.

PAGE 176

161 I am drawn to believe that rights narratives may be rooted in western backed structural changes and foreign influences during the AIDS epidemic in Malawi. The concept of individual rights and freedoms is a n inherently western perspective, one that has been widely and deeply diffused throughout global AIDS interventions and policies notably crafted in Western corridors of power and often with little engagement of local communities. Unity narratives, on the o ther hand, may arise more organically through the difficult times that Malawians live in and their need for collectivism that ensures family functioning and survival. By using "the past" as a reference point, male respondents saw tremendous value in their wives' contribution to the household and beli e ved that their opinions on family matters could ultimately improve the family's circumstances. At the same time, men expressed fears about the legal consequences of rights violations and the failure of men to i nclude their partners in important family dilemmas. There were also concerns among both men and women that western ideals that seek to elevate women's power at men's expense may ultimately backfire and destroy Malawian relationships. Perhaps unity ideals a nd practices may be best understood as a by product of both structural changes that emphasize respect and rights for women and "bottom up" adaptations to everyday experiences with poverty and AIDS. During this time of gender evolution, unity may provide a sort of "win win" or more balanced solution for everyone it promotes male respect but also facilitates women's increased say in their relationships. How do these findings fit with the TGP ? In support for the TGP, the data illustrate the interconnectedness of the three social structures of labor, power, and social norms. This

PAGE 177

162 framework accurately ca ptures the complexity of gender relations and reaffirms the idea that one social structure cannot exist independent of the others. However, the T GP is historical ly rooted in a tradition orientation and this narrow focus misses the wider range and complexity of gender relation s For example, important aspects of unity such as love, communication, reciprocity, and respect are not accounted for in the current TGP. Li kewise, the TGP does not consider righ ts based approaches that strive to increase women's participation in the labor force and generate new social norms related to women's freedoms and choices Factors related to unity and rights have been largely ignored despite their relevance and importance in the Malawi context. I argue that the TGP should be extended to include additional aspects of unity and rights within the three social structures of labor, power, and cathexis. But beyon d t his recommendation I wou ld also like to point out a few areas of mismatch between the current TGP and rural Malawians' accounts Through the sexual division of labor, the family member with the greatest command of resources to meet the other's needs and goals has the greater powe r (Blood and Wolfe 1960) Rural Malawians indicated that most men are indeed the head of the household and carry the main responsibility for economic provisions. According to the TG P this would imply that the distribution of resources favors men and therefore places women in a disadvantaged state. But t o the contrary, the focus group data suggested that the receipt of more economic support from a partner may mean the very opposite t hat women who receive support actually have more power in their relationships through their use love to command economic support. For

PAGE 178

163 women, materiality symbolized that a male partner listened to her requests. Here, women may indirectly influence financia l decisions (i.e., how their partners spent money) if their partners love d them and by default, shower ed them with signs of affection such as money, food, and gifts. These findings suggest that innovative measures are needed to take into consideration bo th women's economic dependence on men and economic transfers that result from love and therefore signify a woman's power not disadvantage Other nuances of the sexual division of labor are noteworthy. Even though men generally have more access to wealth t han women in rural Malawi, male focus group respondents point out that unemployed men or those who cannot live up to the provider role risk losing their wives who may desert them in search of better opportunities. In Malawi, men are placed under tremendous pressure to provide for the family and know very well what the consequences of failing to do so might be. Family power structures (meaning men have more power and women have less) may fall apart if women find support from other sexual partners when their husbands fail to meet their needs. As such, the stories in this chapter illustrate that both couple members are mutually dependent upon each other and affected by resource exchange just in different ways depending on whether they are on the giving or recei ving side of the transaction. In addition, a discussion is warranted around the social structure of cathexis Legitimate power, also referred to as "authority", is based on an individual's legitimate, normally prescribed right to change another person's b ehavior (Cromwell and Olson 1975) Though not always, focus group respondents used tradition narratives to justify the husband's title as

PAGE 179

164 the head of the household regardless of any special circ umstances, i.e., the wife may be the sole provider. They admittedly blamed cultural norms for allocating more power to men. Similarly, the quantitative TLT data reflect that most individuals (both men and women) consider ed men to be overall in charge which is notably, also the most socially desirable response. However, as the narratives in this chapter unfolded, the social norm of legitimacy did not always determine the balance of power ; women were often consulted for their options on family decisions and s ome men even reported that they could not even sell a bicycle without their wife's approval. Therefore, while men were reported to possess more power than women, detailed examples suggest that power may actually be more balanced and democratic at least for some couples. Cromwell and Olson (1975) also point out the difference between "potential power" (or perceived power) and "actual power" (the ability to change the behavior of ot hers). Norms around male authority may serve as potential sources of power but whether or not this power is exerted over a partner is another question. Rural Malawians expressed deep worries about how an imported division of power from the West one that p rivileges women c ould destroy their own relationships, i.e., "the women become rude and disrespectful towards their husbands." Men, in particular, worried about their loss of authority and respect in the household for "doing ladies' jobs" and appearing wea k in the eyes of their partner or others in the community. Women also pointed out that resisting male dominance may bring larger troubles to the relationship, particularly divorce a risk that many did not appear to be willing to take. In studies on power, there

PAGE 180

165 appears to be an underlying, western oriented assumption that disadvantaged women desire more power in the first place but they are unable to rise to the occasion due to structural limitations around gender. The narratives in this chapter suggest tha t this underlying premise may not hold true. Gwendolyn Mikell (1997) in her book on African feminism states that western feminist debates about essentialism, the female body, and radical feminism are not characteristic of African feminism. Rather, emerging African feminism is "distinctly heterosexual, pro natal, and concerned with many bread, butter, culture, and power issues" (Mikell 1997) According to Mikell, African women view their responsibilities as dual: the bearing of children is a primary responsibility and their status as women depends on thi s, but their responsibility for maintaining the family, village, and community is also critical. No self respecting African woman fails to bear children or to be an autonomous economic contributor as compared to their western counterpart who actively choos es these circumstances for herself. Thus, women's choice to ostensibly maintain the "status quo" may be mistaken for female submissiveness or economic dependence when it is really a sign of agency a deliberate act rather than the inability to act. Importan t aspirations around marriage and childbearing may trump the desire to resist male authority. In this respect, women act strategically in order to maintain the very thing that is so important to their livelihoods. The TGP's emphasis on women's economic dep endence on men fails to consider how social benefits and status associated with marriage and childbearing may reinforce seemingly traditiona l power dynamics.

PAGE 181

166 Implications for Power Measures The complexity in which rural Malawians conceptualized gender a nd power relations suggests that quantitative measures may oversimplify the meaning of power and its relevance to the AIDS epidemic in Africa. Limitations aside, I address the extent to which the power measures are reflected in the focus group data. To bri efly recap on the power subscales, three items loaded on a first factor and 4 items loaded on a second factor, which were named unity ("My partner shows they care about me"; "When I need my partner's assistance, he/she is there to help me"; and "My partner and I discuss important matters together") and discordance ("If my partner was really angry with me, he/she might beat me"; "My partner punishes me when he/she is angry with me"; "When I disagree with my partner's relatives, my partner chooses their side over mine"; "My partner is probably having sex with someone else"). The construct of discordance was later deconstructed in to three single item measures of physical violence, sexual violence, and mistrust. Given the above discussion, I conclude that the tw o dimensions of power captured in the scale named unity and discordance do reflect what Malawians say power means to them in their relationships. The unity subscale specifically captures some but not all aspects of unity under the social structure of cathe xis (as shown in Figure 4.1). For instance, respondents felt powerful when their partners were understanding, listened to them, and took their opinions into account. The measure of unity went beyond couple communication and included aspects of love and car e giving that respondents believed were so important to their relationships. Feeling loved by a partner was thought to be a source of power. Even

PAGE 182

167 though social norms around unity were also noted by men as sources of power, masculine gender roles rooted in tradition were not included in the power measure developed in Chapter 3. For men, attachment to traditional gender roles such as providing for the family was also closely tied to feelings of power in their relationships. This could be an important missing dimension of male pow er. The di scordance subscale closely maps to aspects of tradition and appear s to center on factors at the TGP's th e sexual division of power. Wingood and DiC lemente (2002) point out that the sexual division of power is maintained by s ocial mechanisms such as the abuse of authority and control in relationships. They argue that physical violence, sexual violence, and having a unfaithful sexual partner are manifes tations of this power imbalance. One particular scale item that emerged in t he discordance subscale was the belief that one's partner was having an affair. As the TGP argues, women who have a cheating husband may have less power in the relationship. For men, an unfaithful wife may also reflect lower power. Although women engage in transactional sex with men for reasons unrelated to economic necessity, the men in this study were very concerned that their wives might cheat on them if they failed to provide for the family. For the theoretical justifications provided earlier, having a wife who is perceived to be cheating shows that husbands may have less power in the relationship especially if the infidelity is a result of their fai lure to provide. Two other items of the discordance subscale get at punitive aspects of power, particularl y, physical violence and punishment. In the focus group data, women and men reported that wi fe beating occurs when a man had too much power and abused his wife for opposing his

PAGE 183

168 views. Some r espondents from the 2009 semi structured couple interviews discuss ed how their partners had beaten them in the past for transgressing from traditional gender roles i.e., acting "rude" and "disrespectful" towards a husband or for coming home late. In these examples, the act of physical violence against women may be relate d to women challenging male authority or transgressing from ideals around female submissiveness. Certainly, physical violence against women may be less related to gender transgressions and more to adherence to patriarchal gender roles that celebrate male d ominance and female submissiveness. Given these scenarios, it is difficult to ascertain using a single item measure whether the experience of violence is a sign of intolerance and power, or lack thereof. Focus group respondents provided very few cases of women who abuse d men (when male IPV victimization was discussed, it was usually as mutual violence i.e., "they beat each other up"). The TLT data reflects this as well. TLT respondents were asked a separate question on their experiences of physical abuse a nd approximately 2% of men reported that their female partners had beat en them, indicating it could be a relatively rare occurrence among men. Yet interestingly, these two discordance items loaded more strongly for men than for women. More investigation is needed to explain how violence and punishment against men may be linked to power in this population of rural Malawians. One possibility is that women punish men in non physical ways, such as by withholding sex or refusing to cook food for them. Focus grou p participants attributed this to women having more power than their husbands to the point where they become disrespectful towards them. Future research

PAGE 184

169 should seek to uncover if and how women discipline husbands who m they perceive to demonstrate unaccepta ble behavior.

PAGE 185

170 CHAPTER V THE ASSOCIATION BETWEEN RELATIONSHIP POWER AND HIV TESTING BEHAVIOR Chapters one, two, and three provided the rationale for why and how the relationship context specifically power shapes HIV testing behavior. Chapter 4 examined h ow rural Malawians themselves conceptualize d gender and power relations and whether the measure of power developed in Chapter 3 is justified which I believe it is. This process also produced a modified conceptual framework for relationship power. In this c hapter, I present the results of the specific aim 2: to test whether relationship power influences two ty pes of HIV testing behavior: uptake of HIV testing services (a) and disclosure of test results to primary partners (b) Specifically, I utilized variab les related to power in Chapter 3 and the revised conceptual model on relationship power from Chapter 4 to test hypotheses related to HIV testing uptake and disclosure. Conceptual Framework and Hypotheses I begin with the modified versi on of the TGP fr om Chapter 4 to hypothesize that an association will exist between relationship power and two forms of HIV testing behavior: uptake and disclosure. Under each construct in Figure 5.1, I list the variables used to operationalize the TGP constructs related t o the sexual division of labor, sexual division of power, and cathexis. The red boxes refer to constructs that I will test to determine if associations exist with HIV testing behavior. Many of these variables are categorical, allowing me to simultaneously test the constructs of tradition and unity using a single variable

PAGE 186

171 For example, the relationship dominance variable under the division of power has a category for a male dominated relationship ( tradition ) and an egalitarian relationship ( unity ) In Figure 5.2, I show a condensed version of Figure 5.1 showing only the main predictor variables and their associations with HIV testing uptake and disclosure. My revised hypotheses from Chapter 3 relating each variable to the two testing behaviors are as follows. Hypothesis #1 I hypothesize that each power variable will be associated with uptake of future HIV testing over a 16 month period. In addition, I hypothesize that the perceived risk construct from the HBM will be associated with HIV testing uptake. A. Socio e conomic inequalities (division of labor) : Individuals in a lower socio economic position relative to their partners will be less likely to get tested since they will have stronger fear s around divorce or abandonment and loss of financial support that may come with HIV testing. With less economic power, these individuals may also be in a disadvantaged position to negotiate testing with their partners. Conversely, individuals were greater socio economic power relative to their partners will be more likely to test for HIV. B. Relationship dominance (division of power) : Being in a male dominated relationship as compared to an egalitarian relationship will be negatively associated with HIV testing among men due to male control over testing decision making. For men, male dominance will also be negatively associated with testing since these men may be more likely to adhere to traditional beliefs about masculinity and therefore be disinclined to test.

PAGE 187

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igure 5.1: Conceptual Model of Relationship Power based on the Modified TGP

PAGE 188

173 Figure 5.2: Conceptual Frame work for Main Predictor Variables and HIV Testing Behavior !"#$%&'()'*+,'-&./01 2&34&56&7'85.%'()'*+,' 9.&:)'$07'"$3#0&3; *&$:#<'=&:5&)'>(7&:' ?5.4:(.@3&'()'*+,'-&.#'8&.@:#.' #('235A$3B'2$3#0&3 !"#$%!&'(')*) !"#$%!&'(')+) 2$3#0&3'*+,'.#$#@.' 4(04(37$04&' 8&:$/(0.<5"'' 2(C&3' 8&:$/(0.<5"' ,5(:&04& 8&:$/(0.<5"' ?(A50$04& 2$3#0&3' 50D7&:5#B E4(0(A54' 50&F@$:5#B' !05#B ?5.4(37$04& ?5.4(37$04&' .@G.4$:&'3&H )(3A@:$#&7'$.' 65(:&04&'$07' 50D7&:5#B'

PAGE 189

174 C. Relationship v iolence (divisio n of power): Having a history of relationship violence (physical and sexual) is a proxy for fear of abuse, which has been shown to be a barrier to testing. Individuals in violent relationship s will therefore be less likely to test for HIV. Having a history of violence (sexual or physical) could also operate through the pathway of risk, thereby decreasing the likelihood of testing (assuming people are higher risk are less likely to test). D. Unity (cathexis/social norms). Unity will be positively associated wi th HIV testing uptake among men and women. Aspects of unity such as communication, reciprocity, and love will foster a more supportive environment for couples to discuss testing and get tested Therefore, i ndividuals in relationship s with higher levels of unity will be more likely to test for HIV. E. Mistrust/partner infidelity. This variable emerged as an underlying con struct of the discordance subscale. In Chapter 4, trust was perceived as an important social norm related to unity (cathexis) ; thus more trus t, higher power As such, mistrust (lack of t rust) will be negatively associated with t he act of testing. F. Perceived risk: Perceived risk of self and partner will be negatively associated with getting tested for HIV. For both men and women those who belie ve they are at higher ris k for HIV will fear being blamed for infidelit y and therefore less likely to test.

PAGE 190

175 Hypothesis #2 I hypothesize that each power variable will also be associated with HIV testing disclosure to a primary sexual partner. In additi on, I hypothesize that perceived HIV status concordance will be associated with HIV testing disclosure A. Socio e conomic inequalities (division of labor) : Individuals in a lower socio economic position relative to their partners will be less likely to discl ose their test results since they will more likely to fear divorce or abandonment and loss of financial support as a conseq uence of disclosure Conversely, individuals in a higher socio economic position relative to their partner will be more likely to dis close their test results. B. Relationship dominance (division of power) : Being i n a male dominated relationship as compared to an egalitarian relationship will make women less likely to disclose their test results out of fear of the violence that is associat ed with dominance. I also hypothesize that men in male dominated relationships as compared to egalitarian relationship s will be less likely to disclose their test results. C. Relationship v iolence (division of power) : Having a history of relationship (physic al and sexual) violence will reinforce fears of abuse, thereby decreasing the likelihood of disclosing HIV test results. Even if they tested negative, individuals in violent relationships may still avoid disclosure if they did not inform their partners of their plans to test. D. Unity (cathexis). Unity will be positively associated with disclosure of HIV status to a sexual partner, that is, couples with more unity will be more likely to disclose.

PAGE 191

176 E. Perceived HIV status concordance: The decision to disclose one 's HIV status depends upon the test result (positive or negative) and whether or not the result is perceived to be similar or different to a partner's HIV status. Individuals who test positive but perceive their partners to be negative will be less likely to disclose than those who test negative and also perceive their partners to be negative. Summary of Analysis Approach For the HIV testing uptake hypotheses, I used self reported HIV testing histories collected over a 16 month period to test whether rela tionship power w as associated with future uptake of HIV testing. Decisions to test for HIV do not necessarily occur immediately in response to certain social stimuli and thus I wanted to track people over time to give these relationship characteristics a c hance to play out in everyday life and affect HIV testing behavior. Random effects models were used to estimate predictors (i.e., socio economic inequality betwe en partners male dominance, relationship violence, unity, perceived risk, and perceived infide lity) of having a new HIV test over this period. For the disclosure hypotheses, I took advantage of the fact that a group of TLT respondents were randomized to receive testing as part of the TLT study. I investigated whether relationship power measured bef ore testing influence s whether respondents disclosed the ir test results to their main sexual partner In my analysis, I used multivariate logistic regression to model predictors of disclosure (i.e., so cio economic inequality, male dominance, relationship v iolence, unity, and perceived partner HIV concordance ).

PAGE 192

177 Before proceeding with the results, I wo uld like to reintroduce the table from Chapter 3 containing the variables used in both the HIV testing uptake and disclosure models (duplicated as Table 5.1).

PAGE 193

178 Variable Included in model? Type Wave(s) Included in model? Type Wave(s) Independent variables Marital status X Categorical 4-7 X Binary 3, 5 Age X Continuous 3 X Continuous 3 Years of education X Continuous 3 X Continuous 3 Household goods X Continuous 3 X Continuous 3 Age inequality X Binary 3 X Binary 3 Education inequality X Categorical 3 X Categorical 3 Income inequality X Categorical 3 X Categorical 3 Unity X Continuous 3 X Continuous 3 Relationship dominance X Binary 3 X Binary 3 Physical violence X Binary 3 Sexual violence X Binary 3 Perceived risk (self) X Categorical 4-7 Perceived risk (partner) X Binary 3 HIV status concordance X Categorical 4 Perceived partner infidelity X Binary 3 New antenatal care HIV test X Binary 4-7 HIV testing uptake models HIV test disclosure models Table 5.1: Independent and dependent variables for the two statistical models of uptake and disclosure

PAGE 194

179 Table 5.1 continued Variable Included in models? Type Wave(s) Included in models? Type Wave(s) Independent variables Previous number of TLT tests X Continuous 4-7 Previous number of external tests X Continuous 4-7 Tested at TLT's wave 4 X Binary 4 Dependent variables New HIV test (since previous wave) X Binary 4-7 Disclosure to sexual partner X Binary 5 HIV testing uptake models HIV testing disclosure models

PAGE 195

180 Characteristics of the Couple Sample A total of 932 men and women (466 couples) were asked the relationship power questions at TLT's wave 3. Table 5. 2 presents the descriptive statistics for the analytic sample. ANOVA was used to test for gender d ifferences in continuous variables including age, years of education, and household wealth. The majority of the sample reported being married (91.4%). Close to three quarters ( 73.5% ) of all couples reported having at least one child together (the mean numb er of shared children was 1.2). On average, couples had been together for 5.2 years. The mean age for the study population was 24.8 years. Men were on average 5.5 years older than women and differences by gender were significant ( p =0.000). In 41.4% of co uples, the man was over 5 years older than his female partner. Note that 1.8% of women were older than their male partners but the age difference was never greater than 2 years (results not shown). The mean years of education was 7.3 years, reflecting a p rimary school education. Men had approximately 1.3 more years of education than women and differences by gender were significant (p =0.000). In approximately 14.0% of couples, both members had the same level of education, in 62.2% of couples the man had mor e education, and in 23.8% of couples the woman had more education. The mean number of owned household items, a measure of household wealth ranging from 0 to 9, for the total sample was 3.0. Men also reported slightly higher household goods (3.1) as compare d to women (2.8) and the difference by gender was significant ( p =0.000). In 8.4% of couples, both members were unemployed. In 2.4% of couples, the wife worked but the husband did not. In 54.7% of

PAGE 196

181 couples, the husband worked but the wife did not. Finally, i n 34.6% of couples, both couple members were employed. The cross sectional TLT wave 3 sample would have allowed me to make stronger conclusions about the larger population of rural Malawians at least for the women, who were randomly selected from a house hold listing. The couple sample is inherently different from the cross sectional wave 3 sample because of self selection but to what extent? To determine this, I compar ed the wave 3 couple sample (466 women) to the wave 3 individual level sample (1,371 wom en) using an ANOVA on key demographic variables: age, marital status, and household economic status (results not shown). Significant differences were noted on marital status ( p =0.000). The individual level sample was significantly less likely to be married (46.3%) as compared to the couple sample (90.8%). In order to be considered part of the couple sample, both partners were required to answer the power questions and thus the sample was biased towards more serious partnerships (i.e., with serious male part ners more likely to participate than causal partners). The individual level sample of women was slightly younger with a mean age of 20.6 years as compared to 22. 1 for the couple sample They were also slightly wealthier with a mean number of household good s of 3.2 as compared to 2. 8 for the couple sample Differences in age ( p =0.000) and household economic status ( p =0.000 ) were statistically significan t

PAGE 197

182 Table 5.2 : Selected characteristics of the baseline sample of couples, TLT wave 3 Variable % Mean (SD) % Mean (SD) % Mean (SD) Individual characteristics Age (16-57) 24.8 (4.8)* 22.1 (2.7) 27.6 (4.9) Years of education (0-13) 7.3 (3.0)* 6.6 (2.6) 7.9 (3.1) Total (N=932) Women (N=466) Men (N=466) Household goods (0-7) 3.0 (1.5)* 2.8 (1.5) 3.1 (1.5) Couple characteristics Married 91.4 90.8 92.1 Relationship duration (1-14) 5.2 (2.9) 5.2 (2.9) 5.2 (3.0) At least one living child with partner 73.5 72.3 74.7 Age difference (age inequality) 0-5 years 58.6 6+ years 41.4 Education inequality Same education 14.0 Male higher education 62.2 Female higher education 23.8 Employment inequality Both unemployed 8.4 Female employed, male unemployed 2.4 Male employed, female unemployed 54.7 Both employed 34.6 Unity (total score) 3.77 (0.41) 3.77 (0.45) 3.77 (0.37) My partner shows that they care about me 3.87 (0.40) 3.86 (0.44) 3.88 (0.35) When I need my partner's assistance, he/she is there to help me 3.63 (0.69) 3.65 (0.67) 3.60 (0.70) My partner and I discuss important matters together 3.80 (0.47) 3.78 (0.51) 3.82 (0.43)

PAGE 198

183 Variable % Mean (SD) % Mean (SD) % Mean (SD) Relationship dominance Male-dominated 85.0* 81.5 88.4 Female-dominated/egalitarian 15.0 18.5 11.6 Relationship violence Ever experienced forced sex by partner 16.5* 21.5 11.6 Ever been physically abused by partner 4.0* 6.0 1.9 Perceived risk of self for HIV (1-10) 1.73 (2.43) 1.63 (2.48) 1.82 (2.39) Perceived risk of partner for HIV No or low likelihood of infection 93.9* 91.3 96.4 Medium likelihood of infection 2.5 3.8 1.4 High or certain likelihood of infection 3.6 4.9 2.3 Perceptions that partner is having affair Strongly disagree/disagree 85.7* 81.8 89.7 Strongly agree/agree 14.3 18.2 10.3 Unity: 1=Strongly Disagree; 2=Disagree; 3=Agree; 4=Strongly Agree. Higher scores indicate more unity. Unity scores were created by taking the mean across all three unity items. *Chi-square and ANOVA differences for gender were significant at p <.05. Total (N=932) Women (N=466) Men (N=466) Table 5.2, continued

PAGE 199

184 Gender Differences in Relationship Factors A multivariate MANOVA was used to test for gender differences in the unity variable using all three unity items as dependent variables. As illustrated in Table 5. 2 the mean score for unity was almost exactly the same for men and women (3.77), indicating that most people either agreed or strongly agreed w ith the unity statements (range: 1 4). The results showed that men and women's responses were not significantly different from each other ( p =0.137). Chi square tests were used to test for gender differences in categorical variables including relationship dominance and the two IPV measures. ANOVA was used for continuous variables. The majority of respondents indicated that their relationships were male dominated (85.0%). Men were more likely to state that their relationship was male dominated (88.4%) as com pared to women (81.5%) and the difference was significant ( p =0.003). Approximately three quarters ( 74.9% ) of couples were concordant in their responses to the relationship dominance question, e.g., if a wife said that her partner was the dominant one then the husband said that he was the dominant one. Couple members never agreed that a wife was dominant. Less than one fifth ( 16.5% ) of respondents reported a history of sexual IPV. Statistically significant gender differences ( p =0.000) in sexual IPV were fou nd with women indicating higher levels of sexual coercion (21.5%) than men (11.6%). It is noteworthy that almost 12% of men reported feeling pressured to have sex when they did not want to. Overall, 4.0% of respondents reported being physically abused by t heir partners. Statistically significant

PAGE 200

185 gender differences were also found for physical IPV ( p =0.001), with women reporting higher levels of abuse (6.0%) than men (1.9%). With regard to reciprocal violence, around 3.0% and 1.0% of couples reported that bo th couple members had experienced either sexual or physical IPV, respectively. Reciprocal violence was also calculated as a proportion of total violence (either sexual or physical IPV). In 18.2% of all sexually violent couples, both couple members reported sexual coercion. In 5.4% of all physically violent couples, both couple members reported physical abuse. Physical IPV only occurred within married couples. Sexual and physical IPV were moderately correlated in the overall sample ( r= 0.22) indicating that f or some couples, physical abuse and sexual abuse may occur simultaneously or in serial. Overall, respondents reported a low likelihood of being infected with HIV, with a mean of 1.73 (range: 0 10). Women reported lower levels of perceived risk for HIV (me an=1.63) than men (mean=1.82) but the difference was non significant ( p =0.211). Among the total sample, 93.9% reported that their partner had no or low likelihood of HIV infection, 2.5% believed their partners had a medium likelihood of HIV infection, and 3.6% believed their partners had a high or certain likelihood of HIV infection. The distribution for women was slightly different than for men. The majority ( 91.3% ) of women reported that their partners had no or low likelihood of HIV infection, 3.8% repor ted a medium likelihood, and 4.9% reported a high or certain likelihood of a partner's HIV infection. The corresponding figures for men were 96.4%, 1.4%, and 2.3%, respectively. Gender differences in perceived risk of a partner were statistically significa nt when the variable was trichotomized as no risk/low risk, medium risk, and high/certain risk ( p= 0.007). When the perceived infidelity measure was

PAGE 201

186 dichotomized as strongly agree/agree and strongly disagree/disagree, most respondents (85.7%) believed that their partners were faithful. Women were more likely to believe that their partner was having an affair than men (18.2% vs. 10.3%) and the gender difference was significant ( p= 0.001). Table 5. 3 shows couple HIV status for those tested as part of TLT at wa ve 4 (at the individual level). For 444 respondents (222 couples), HIV serostatus w as on file for both partners. As shown in the first column of the table, 92 .0% of couples were sero concordant negative, meaning both couple members tested negative. Approxim ately 2% of couples were sero concordant positive, meaning both couple members tested positive. In 2.5% of couples, the man was positive and the woman was negative. In the remaining 3.5% of couples, the man was negative and the woman was positive. For the e ntire sample, I used the combined mea sure of HIV status that included perception of HIV status for those not tested as part of TLT at wave 4 Therefore, r espondents were considered to be HIV positive if they had tested positive via TLT at any po int up to a nd including wave 4 or if they reported that they were 100% likely to have HIV at wave 5 (i.e., 10 beans on a scale of 0 to 10). Th e figures for estimated HIV status were very similar to those for actual serostatus. Table 5. 3 : Couple HIV status, TLT wave 4 Actual HIV serostatus data (%) Estimated HIV serostatus data (%) Both are negative 92.0 91.4 Both are positive 2.0 2.5 Man positive, woman negative 2.5 2.3 Man negative, woman positive 3.5 3.9 N (individuals, not couples) 400 880

PAGE 202

187 HIV Testing Hi stories Figure 5.3 shows the baseline HIV testing history for all 932 respondents when they first enrolled in the TLT study. Respondents were asked the question, "I don't want to know the results, but have you ever tested for HIV and received the results? Almost three quarters ( 71.6% ) of the total sample had previously been tested for HIV. More men than women had previously tested for HIV (63.2% compared to 79.8%), which reflects national statistics from the Malawi Demographic Health Survey (MDHS, 2011). Figure 5.3 : Percent of TLT Respondents Who Had Ever Tested for HIV at Baseline Figure 5.4 illustrates the percent of respondents who had a new HIV test (outside of the TLT study) since the previous wave. Since respondents were not asked to provide spe cific details on their last test, external testing could include voluntary counseling and testing 71.5 79.8 63.2 0 10 20 30 40 50 60 70 80 90 Total sample (N=932) Women (N=466) Men (N=466) Ever tested for HIV at baseline (%)

PAGE 203

188 (VCT), antenatal care testing, home based testing, or other possible testing modalities. The sample size (N) listed under the wave number refers to the total number of respondents participating in that particular wave. At wave 4, approximately 24% (or 217 respondents) of the 902 respondents who participated at that wave had a new HIV test since wave 3. The percentage of participating respondents who received a new HIV test since the previous wave appeared to slightly increase from wave 5 to 7. Approximately 20% (or 175 respondents) had a new tes t during the period from wave 4 to 5. Approximately 22% (or 189 respondents) had a new tes t during the period from wave 5 to 6. Finally, approximately 25% (or 210 respondents) had a new tes t during the period from wave 6 to 7. Table 5. 4 provides the breakdown by gender for the overall numbers presented in Figure 5.4 Higher rates of external testing among women across all four waves are attributed, in part, to antenatal care testing. Between waves 3 and 4, 30 pregnant women (13.8% of all who tested) had tested for HIV through antenatal care. Between waves 4 and 5, 33 pregnant women (18.9% of all who tested) had tested thro ugh antenatal care. Between waves 5 and 6, 25 pregnant women (13.2% of all who tested) had tested through antenatal care. Finally, between waves 6 and 7, 21 pregnant women (10.0% of all who tested) had tested through antenatal care.

PAGE 204

189 Figure 5.4 : TLT Res pondents Who Had a New HIV Test Since the Previous Wave Table 5. 4 : HIV testing history by gender, TLT waves 3 7 Total Sample Women Men Wave Total Sample N New HIV test since previous wave N (%) New HIV test since previous wave N (%) New HIV test si nce previous wave N (%) 3 932 N/A N/A N/A 4 902 217 (24.1) 138 (30.1) 79 (17.8) 5 867 175 (20.2) 111 (24.9) 64 (15.2) 6 861 189 (22.0) 112 (25.3) 77 (18.4) 7 836 210 (25.1) 123 (28.5) 87 (21.5) Figure 5.5 illustrates the distribution of cumulative new HIV tests (outside of the TLT study) recorded at wave 7 over the course of 16 months. Note that these data only include respondents who had complete data for all of the waves (N=799). The number of cumulative new HIV tests ranged from 0 (no new tests) to a maximum of 4 Over half 24.1 20.2 22.0 25.1 0 5 10 15 20 25 30 Wave 4 (N=902) Wave 5 (N=867) Wave 6 (N=861) Wave 7 (N=836) Percent of respondents who had a new HIV test since the previous wave (%)

PAGE 205

190 ( 52.7% ) of participating respondents had no new tests over the 16 month period, 21.8% had 1 new test, 12.6% had 2 new tests, 8.1% had 3 new tests, and 4.8% had 4 new tests. Figure 5.5 : Cumulative Number of New HIV Tests from Waves 4 to 7 Predictors of HIV Testing Uptake Table 5. 5 presents the crude and adjusted odds ratios predicting a new HIV test from wave 4 to 7. The adjusted models included all predictor variables and statistical controls for marital status, age, years of education, the household goods index, previous number of TLT HIV tests, previous number of external HIV tests, and whether or not pregnant women tested for HIV through antenatal care at each wave. I present exact p values for associations that were appr oaching statistical significance (with a p less than 0.10). 52.7 21.8 12.6 8.1 4.8 0 10 20 30 40 50 60 No new test 1 new test 2 new tests 3 new tests 4 new tests Cumulative new HIV tests (%)

PAGE 206

191 Hypothesis 1A: Socio E conomic Power and Testing Uptake I hypothesized that individuals in a lower SES position as compared to their partner would be less likely to test ; similarly, those who were in a higher SES position as compared to their partners would be more likely to test The results for women are discussed as follows. In the unadjusted models, women whose partners were at least 6 years older were 34% less likely to test than women paired with men of a similar age (the reference group). After controlling for covariates, the results showed that women with partners who were at least 6 years older were 30% less likely to test for HIV as compared to the reference group ( p =0.092). Regarding educ ation inequality being in a more disadvantaged state as compared to a male partner did not appear to play a significant role in women's uptake of testing ; both the crude and adjusted odds ratios were non significant Nor did having more education than a p artner influence women's uptake of testing. Regarding employment inequality, women who were unemployed when their male partner was employed were 55% less likely to test than when both women and their partners were unemployed (reference group). This associa tion, however, was attenuated and became non significant when other variables were added to the model. Conversely, being employed when a partner was unemployed did not appear to give women the advantage to test Finally, dual employment ( both men and women were employed) did not significantly influence women's uptake of HIV testing. The results for men are as follows. In the bivariate models, men who were at least 6 years older than their partners were 50% less likely to test for HIV than men of a similar age as their partners (reference group) Yet this association was attenuated and became non

PAGE 207

192 significant after controlling for other cov ariates In the multivariate models, men who had higher education levels than their fe male partners were 2 .1 times more l ikely to test than men in couples with the same level of education ( p= 0.059). However, m en who had lower education levels as compared to their partners were not less likely to test after controlling for the other covariates Regarding employment inequaliti es, n o significant associations were found in the bivariate or multivariate models for men. This means that, for men, being employed or unemployed as compared to one's par tner did not affect HIV testing uptake; similarly, dual employment did not play a sig nificant role in men's decisions to test. Hypothesis 1B: Relationship Dominance and T esting Uptake It was hypothesized that if the division of power was imbalanced such that men dominated the relationship, women would be less likely to get tested for HIV as compared to women in egalitarian relationships. To the contrary, the results for women show that being in a male dominated relationship significantly increased the odds of testing by 70% as compared to women in egalitarian relationships. For men, it was hypothesized that male dominance would be negatively associated with men's uptake of testing since these men may be more likely to adhere to traditional beliefs about masculinity and therefore feel disinclined to test. However, b eing in a male dominat ed relationship was not significantly associated with HIV testing uptake for men i n the crude or adjusted models. Hypothesis 1C: Violence and Testing Uptake For women, the unadjusted results show a trend towards significance for the association between ha ving a history of sexual coercion and receiving a new HIV test. Women who

PAGE 208

193 reported being coerced into having sex when they did not want to had a 53% higher odds of getting tested for HIV than women wh o did not report a history of sexual violence This resu lt conflicts with the hypothesis that women in sexually violent relationships may be less likely to test out of fear of violence. It is possible that these women may be more likely to test due to heightened risk for HIV. In the multivariate models, this as sociation was attenu ated after controlling for perceived risk and was no longer statistically significant. In accordance with the hypothesis, w omen with a history of physical violence w ere less likely to test for HIV both in the bivariate and multivariate models. However, the associations failed to achieve statistical significance in both cases. Interestingly, men who reported being pressured to have sex with their partners when they did not want to were 66% less likely to receive a new HIV test net of pe rceived risk. The same association did not hold for physical abuse, which may be an artifact of low rates of physical abuse among men in the sample. Hypothesis 1D: Unity and Testing Uptake In contradiction with the hypothesis that individuals with higher levels of unity in their relationships would be more likely to test due to the supportive environment it promotes, the results demonstrated that unity was negatively associated with having a new HIV test in the multivariate models for women. For each one unit increase in unity, the odds of having a new HIV test decreased by 41% after controlling for other factors. T he same hypothesis was thought to apply to me n Yet similar to the women each one unit increas e in unity decreased the odds of men's testing b y 47% after controlling for other factors ( p =0.092)

PAGE 209

194 Hypothesis 1E: Mistrust/Infidelity and Testing Uptake In contrad iction to the hypothesis that perceived partner infidelity was an indicat or of lower power, and thus a reduced likelihood of testing, w om en were slightly more likely to test if they suspected that their partner could be cheating H owever, this relationship was not significant in the bivariate or multivariate models Similarly f or men, mistrust was not significantly associated with testing u ptake in either the bivariate or multivariate models. Hypothesis 1F : Perceived Risk and Testing Uptake I hypothesized that individuals who perceived themselves to be at a higher risk for HIV would be less likely to test for HIV For women, the bivariate results for perceived risk of self were statistically significant and demonstrated that women who reported a low likelihood of HIV infection were 35% less likely to test than those who were sure they were HIV negative. Similarly, women who reported a mediu m likelihood of HIV infection were 50% less likely to test than women who were sure they were HIV negative ( p =0.069). However, these associations failed to retain their statistical significance in the multivariate models. In addition women who perceived t hat their partners were more likely to be HIV infected were 49% less likely to test for HIV as compared to women who p erceived that their partners had no or low likelihood of HIV infection net of perceived risk of self. In the bivariate models, men who pe rceived themselves to be a medium risk for HIV were 47% less likely to test for HIV as compared to men who considered themselves at no risk for HIV. This relationship was attenuated and became non significant in the multivariate

PAGE 210

195 models. The bivariate and m ultivariate results also showed that there was no association be tween perceived risk of partner and uptake of HIV testing among men.

PAGE 211

196 Variable Crude OR (95% CI) Adjusted OR (95% CI) 1 Crude OR (95% CI) Adjusted OR (95% CI) 1 Women Men Socio-economic inequality Age difference 0-5 years difference ref ref ref ref 6+ years difference 0.66 (0.45, 0.96)* 0.70 (0.47, 1.06)   0.50 (0.27, 0.91)* 0.91 (0.46, 1.81) Education inequality Same education (ref) ref ref ref ref Male higher education 0.75 (0.44, 1.28) 0.84 (0.46, 1.53) 1.71 (0.71, 4.13) 2.13 (0.97, 4.65)   Female higher education 0.77 (0.42, 1.42) 0.77 (0.40, 1.48) 0.82 (0.30, 2.28) 1.25 (0.47, 3.31) Employment inequality Both members unemployed (ref) ref ref ref ref Female employed, husband unemployed 0.89 (0.25, 3.18) 0.73 (0.18, 2.92) 1.86 (0.27, 12.55) 2.90 (0.50, 17.0) Male employed, female unemployed 0.45 (0.24, 0.87)* 0.83 (0.38, 1.79) 0.46 (0.16, 1.28) 1.10 (0.39, 3.07) Both members employed 0.59 (0.30, 1.17) 1.10 (0.49, 2.48) 0.46 (0.16, 1.33) 1.38 (0.46, 4.10) Relationship unity 0.58 (0.39, 0.87)** 0.59 (0.37, 0.95)* 0.57 (0.27, 1.21) 0.53 (0.25, 1.11)   Relationship dominance Female-dominated/egalitarian (ref) ref ref ref ref Male-dominated 1.36 (0.84, 2.21) 1.70 (1.00, 2.88)* 1.08 (0.44, 2.67) 1.18 (0.53, 2.64) Relationship violence Ever been sexually coerced by partner 1.53 (0.99, 2.38)   1.10 (0.67, 1.83) 0.33 (0.12, 0.89)* 0.33 (0.13, 0.84)* Ever been physically abused by partner 0.85 (0.39, 1.86) 0.60 (0.24, 1.53) 0.31 (0.02, 4.73) 0.69 (0.07, 7.18) Perceived risk for HIV (self) No likelihood (ref) ref ref ref ref Low likelihood 0.65 (0.45, 0.92)* 0.74 (0.50, 1.11) 0.82 (0.51, 1.30) 0.87 (0.55, 1.37) Medium likelihood 0.50 (0.32, 0.78)** 0.72 (0.44, 1.18) 0.53 (0.29, 0.98)* 0.66 (0.36, 1.23) High likelihood 0.65 (0.29, 1.47) 0.73 (0.30, 1.77) 0.77 (0.26, 2.27) 0.91 (0.30, 2.79) Certain likelihood 0.87 (0.41, 1.87) 0.99 (0.44, 2.23) 0.35 (0.09, 1.37) 0.36 (0.07, 1.83) Perceived risk for HIV (partner) No or low likelihood (ref) ref ref ref ref Medium, high, or certain likelihood 0.62 (0.30, 1.29) 0.51 (0.24, 1.06)   0.31 (0.05, 1.90) 0.79 (0.16, 3.93) Perception that partner is having an affair 0.96 (0.59, 1.54) 1.06 (0.63, 1.78) 1.00 (0.38, 2.60) 0.76 (0.32, 1.84) Unity scores ranged from 1 to 4, with higher values indicating more unity.   trend towards significance ( p <0.10) ; p <.05; ** p <.01; *** p <.001 1 Adjusted models control for all other predictor variables and marital status, age, years of education, household goods index, previous testing through the TLT study, previous testing outside of the TLT study, and testing for HIV through antenatal care (women only). Time-varying predictors include perceived risk of self, marital status, and all three testing controls. Table 5.5: Odds ratios predicting a new HIV test among women and men, TLT waves 4 7

PAGE 212

197 Disclosure of HIV Test Results Unlike a typical "repeated event" such as HIV testing disclosure of test results is conceptually different and methodologically more complicated. For example, would only disclosure of a new HIV test result be considered an "event"? Or should every time a respondent states that they disclosed be considered a repeated event? Unlike common medical or social events studied by sociologists or epidemiologists, disclosure is differen t the event is dependent upon whether a person decides to test and thus the number of people eligible at each wave depends upon who was tested. This would have impli cations for sample size, which is already limited to those who have previously been tested for HIV. Therefore, I chose to conduct a more straightforward analysis using a cross sectional sample of respondents starting with the baseline set of couples at wav e 3 who had ever tested at any time prior to wave 5. This allowed me to make use of a larger sample size than would otherwise be possible if I limited the sample to only those who had a new HIV test at each wave. While I realize the inherent limitations of cross sectional analyses such as when predictors are collected at the same time as outcome variables, I did take advantage of the fact that power and many of the relationship variables were collected prior to when disclosure occurred. This provides a stro nger argument that relationship factors may influence decisions to disclose instead of the other way around that the act of disclosure changes relationship dynamics. As such, I used relationship variables from wave 3 to predict disclosure as reported at wa ve 5.

PAGE 213

198 Of the 799 respondents eligible for the disclosure analysis, 85.7% reported that they had told either their spouse or sexual partner the results of their last HIV test. Among women, 87.5% reported disclosing their test results. Men disclosed less of ten than women (83.7%), but gender differences were not significant ( p =0.121). Table 5. 6 shows rates of disclosure broken down by martial status, HIV status, and whether or not respondents had their last HIV test at TLT. Chi square tests were used to test for differences in rates of disclosure by these three variables. For the total sample, married respondents were significantly more likely to disclose their HIV test results than unmarried respondents (87.1% vs. 59 .0 %; p= 0.000). The same pattern of disclosu re by marital status persisted for both women ( p =0.002) and men ( p= 0.000). Table 5. 6 : Disclosure of last HIV test to main partner, TLT wave 5 Variable Total Sample (N=799) Women (N=432) Men (N=367) Marital status (%) Married 87.1* 88.6* 85.3* Unm arried 59.0 65.0 52.6 HIV status (%) HIV positive 80.0 80.0 80.0 HIV negative 86.1 88.1 83.9 TLT test at wave 4 (%) N=469 N=258 N=211 Yes 89.1* 87.6 91.0*** No 80.9 87.4 73.7 Chi square differences were significant at p <.05; ** p <.01; *** p <.001 Differences in disclosures rates were also examined by HIV status. Of those with serostatus data (N=469), 6.2% were HIV positive. Of those who had never tested at TLT (N=330), 6.4% were estimated to be HIV positive. The combined measure of HIV status

PAGE 214

199 sh owed that 6.3% of all respondents (N=799) were HIV positive. Rates of disclosure among HIV positive and negative individuals were surprisingly similar (refer to Table 5. 6 ) Four fifths ( 80% ) of all respondents who were HIV positive reported disclosing to e ither their spouse or primary sexual partner. Rates of disclosure among HIV negative respondents were slightly higher, at 86.1%. Chi square tests revealed that HIV status was not significantly associated with decisions to disclose ( p =0.231). For two third s of the wave 5 couple sample, their last HIV test was administered at TLT 's wave 4 When disclosure rates were stratified by whether or not respondents were tested via TLT at wave 4, the overall figures showed that TLT testers were slightly more likely to disclose to a sexual partner (refer to Table 5. 6 ). At wave 4, 469 respondents had tested through TLT and approximately 89.1% reported that they disclosed their test results to their spouse or sexual partner. Of those not tested at TLT's wave 4, approximat ely 80.9% reported that they had disclosed. For the overall sample, chi square tests showed that rates of disclosure varied significantly by whether respondents had their last HIV test at TLT ( p =0.001). Men who tested at TLT's wave 4 were significantly mor e likely to disclose than men who tested elsewhere (91% compared to 73.7%) ( p =0.000). However, there were no differences in rates of disclosure between women who tested at TLT's wave 4 and women who tested outside the study ( p= 0.941).

PAGE 215

200 Reliability of Di sclosure Reports The disclosure data were self reported. Naturally, this prompts the questions, do respondents really disclose if they say they do and do their partners believe them? And do respondents tell their partners their actual HIV test result? Con sider the example of an HIV positive man who reported that he told his wife his status. We can examine her perception of his status to see if what she thinks about his status matches his actual HIV serostatus. If she reports that she is unsure about whethe r he is positive or negative, this would indicate that he did not really disclose (or it could reflect that she doesn't believe what he told her). If she believes that he is positive (and he is), then it is more likely that he told her the correct test res ult. Yet another explanation could explain these discrepancies, particularly with regard to gaps between the time of disclosure and the time perceptions of a partner's risk were ascertained. For instance, if a woman told her partner that she tested negativ e for HIV in January and was shortly thereafter found to be cheating on her husband, his perception of her risk for HIV in April may no longer be "no likelihood" given that he knows she has put herself at risk for HIV through an extramarital affair. In or der to assess whether respondents really disclose and if their partners believe what they tell them, I examined disclosure of HIV test result at wave 5 (partner's report) and perceived risk of partner at wave 5 (respondent's report of their partner). Respo ndents from the disclosure dataset were included in the analysis if they had complete data for both measures (N=723). To briefly recap, perceived risk of a partner was captured with the statement: What is the likelihood that your partner is currently infe cted with HIV?"

PAGE 216

201 Response options included no likelihood, low, medium, high, and "I know she/he is" (infected with HIV). In theory, respondents who were truly informed of a partner's negative or positive result would have either responded "no likelihood" (t old negative result) or "I know he/she is" (told a positive result). From this measure, I first created a three level categorical variable, where 1=HIV negative (no likelihood), 2=HIV positive ("I know she/he is"), and 3=uncertainty of a partner's status ( low, medium, or high likelihood). I then checked whether uncertainty about a partner's status was less common in the group of respondents whose partners reported that they disclosed by using chi square tests (refer to Table 5. 7 ). Note that for this test, I combined categories 0 and 1 to measure "certainty" (i.e., HIV positive or HIV negative) about a partner's status (so that 0 equaled "certainty" and 1 equaled "uncertainty"). Overall, uncertainty about a partner's status was higher among respondents whose partners did not disclose than among those whose partners did disclose (19.6% vs. 16.2%). For women, uncertainty about a partner's status was more common among women whose partners did not disclose than among women whose partners did disclose (28.9% vs. 22 .9%). However, chi square tests demonstrated that the relationship was not significant ( p= 0.359). For men, uncertainty about a partner's HIV status was more common among men whose partners had disclosed rather than not disclosed (10.9% vs. 7.5%). However chi square tests did not reveal a significant relationship ( p =0.512) 7 7 Note that the level of certainty of a partner's status may not be accurate if partners are reporting disclosure based on a non recent HIV test. Therefore, I also ran the analysis for respondents whose partners had a recent external HIV test (wave 4 or 5) or a TLT test at wave 4. Similar results were found as compared to the numb ers presented above.

PAGE 217

202 Table 5. 7 : Do respondents really disclose an d do their partners believe their reports ? For respondents whose partners reported disclosing their test results, I next che cked the level of concordance between perceived risk of partner at wave 5 (respondent's report of their partner) and actual HIV serostatus at wave 4 (partner's data) (refer to Table 5. 8 ). My tabulations for perceived risk of partner only included those who reported with certainty on their partner's status (HIV positive or negative). Overall, for partners who tested negative, 99.3% of the respondents believed that their partners were indeed negative. For partners who tested positive, only 55.3% of respondent s believed that they were indeed positive. Similar trends were found by gender. Almost all women ( 98.4% ) whose partners tested negative believed that they were HIV negative; while only 50% of women whose partners tested positive believed them to be HIV pos itive. All men ( 100% ) whose partners tested negative believed that they were indeed HIV negative. Only half ( 55.6% ) of men whose partners tested positive believed that the se women were HIV positive. All chi square tests showed significant differences at p = 0.000. Uncertainty about a partner's status was lower among those whose partners had not disclosed but the difference was non significant. Yes (N=629) No (N=92) Yes (N=279) No (N=52) Yes (N=350) No (N=40) Perception of partner's HIV status (%) Sure of partner's HIV status 83.8 80.4 77.1 71.2 89.1 92.5 Unsure of partner's HIV status 16.2 19.6 22.9 28.9 10.9 7.5 Total Sample (N=723) Women (N=331) Men (N=390) Partner disclosed? Partner disclosed? Partner disclosed?

PAGE 218

203 Table 5. 8 : Are respondents told the correct test result by their partners? Predictors of HIV Status Disclosure Table 5. 9 presents the crude and adjusted odds ratios predicting disclosure at wave 5. The adjusted models included all other predictor variables and statistical controls for marital status, age, years of education, the household goods index, and having a TLT test at wave 4. Exact p values are presented for associations approaching statistical significance. Hypothesis 2A: Soci o Economic Power and Disclosure It was hypothesized that individuals in a lower socio economic position relative to their partners would be less likely to disclose their test results since they would be more likely to fear divorce or abandonment and loss of financial support as a conseq uence of disclosing Regarding age inequality, having a partner who was greater than 5 years older was not associated with disclosure for women in the multivariate models. Regarding education inequality, women who were less educated than their partners were not any more likely to disclose than women who had a similar education level as their partners. In the bivariate models, unemployed women paired with employed men were approximately 3.8 times more likely to disclose than t he same women paired with unemployed men; however, this association was attenuated and became non significant after adding other variables to the Perception of partner's HIV status (%) HIV positive (N=15) HIV negative (N=310) HIV positive (N=6) HIV negative (N=127) HIV positive (N=9) HIV negative (N=183) HIV positive 55.3 0.7 50.0 1.6 55.6 0.0 HIV negative 46.7 99.3 50.0 98.4 44.4 100.0 Women (N=133) Partner serostatus Men (N=192) Partner serostatus Total Sample (N=325) Partner serostatus

PAGE 219

204 model. W omen who were a part of a dual income earning couple were not any more likely to disclose than those i n an unemployed couple. The results for men are as follows. Men who were greater than 5 years older than their par tners were not any more likely to disclose than men of a similar age as their partners. For education inequality, men who were more (or less ) educated than their partners were not any more likely to disclose than men paired with women of equal education. Finally, regarding employment inequality, differences between partners were not significantly associated with disclosure with one exception. Men in relationships where both couple members were employed were approximately 5.4 times more likely to disclose than men in relationships where both coup le members were unemployed, after adjusting for other covariates. Hypothesis 2B: Relationship Domina nce and Disclosure In the multivariate models, women in male dominated relationships were not any less likely to disclose as compared to women in egalitarian or female dominated relationships. This conflicts with the hypothesis that male dominance deters women from telling their partners their HIV test results. In fact, the odds ratios w ere in the opposite direction of what was hypothesized. For men the direction of the association was similar to that found for women such that being in a male dominated re lationship increased the odds of disclosure as compared to men in egalitarian or female dominated relationships ; however, like for women, it was also non significant.

PAGE 220

205 Hypothesis 2C: Violence and Disclosure It was hypothesized that h aving a history of re lationship violence (physical and sexual) would reinforce fears of abuse, thereby decreasing the likelihood of disclosing HIV test results. In support of this prediction, the results showed that women who had been physically abused by their partners were 6 8% less likely to disclose their test results than women who had never experience d physical violence even after controlling for other covariates ( p =0.077). Yet in contradiction to the violence hypothesis, women who had experienced sexual violence were 2.25 times more likely to disclose than women who had not ( p =0.097). Regarding physical abuse, men who had experienced physical abuse in their relationship were 93% less likely to disclose after controlling for other factors ( p =0.056). I question the conclus iveness of the finding for physical abuse given the low occurrence of abuse reported by men in the sample (only 2%). For sexual violence the bivariate models showed that men who had experienced sexual violence were 54% less likely to disclose than men who had not; however, this association was attenuated and became non significant after controlling for other factors in the model. Hypothesis 2D: Unity and Disclosure It was hypothesized that unity would be positively associated with disclosure of HIV statu s to a sexual partner. A s expected the multivariate models demonstrated significant associations between unity and disclosure for women and men The models for women showed that f or each one unit increase in unity, the odds of disclosure increased by 216%

PAGE 221

206 after controlling for other covariates. For men the odds of disclosure increased by 243% for each o ne unit increase in unity ( p= 0.084). Hypothesis 2E: Couple HIV Status and Disclosure I predicted that an individual's decision to disclose HIV status woul d depend upon the perception of his or her partner's status. After controlling for other covariates, women who tested positive and perceived their partners to be positive were 80% less likely to disclose than women who were found HIV negative and perceived their partners to be negative (reference group). In addition, women who tested positive but perceived their partners to be HIV negative were 69% less likely to disclose than women in the reference group. HIV negative women were 22% less likely to disclose if they perceived their partners to be HIV positive ; however, the association failed to reach statistical significanc e in the multivariate models After controlling for other factors, men who tested positive and perceived that their partners were HIV nega tive were 84% less likely to disclose as compared to men who tested negative and perceived their partners to be HIV negative (reference group) Men paired with suspected HIV positive partners were not any more or less likely to disclose than the reference group.

PAGE 222

207 Variable Crude OR (95% CI) Adjusted OR (95% CI) 1 Crude OR (95% CI) Adjusted OR (95% CI) 1 Women Men Economic inequality Age difference 0-5 years ref ref ref ref 6+ years 1.07 (0.60, 1.91) 1.09 (0.54, 2.18) 1.08 (0.61, 1.90) 0.52 (0.18, 1.53) Education difference Same education (ref) ref ref ref ref Male higher education 1.12 (0.48, 2.54) 1.41 (0.49, 4.09) 1.70 (0.79, 3.64) 2.30 (0.86, 6.92) Female higher education 0.97 (0.38, 2.47) 0.80 (0.26, 2.49) 1.49 (0.61, 3.60) 1.86 (0.48, 7.22) Employment difference Both members unemployed (ref) ref ref ref ref Female employed, male unemployed note 1 note 1 0.24 (0.04, 1.33) 0.66 (0.08, 5.42) Male employed, female unemployed 3.38 (1.35, 8.47)** 2.10 (0.65, 6.86) 1.41 (0.56, 3.55) 2.66 (0.67, 10.54) Both members employed 1.36 (0.56, 3.32) 0.71 (0.22, 2.29) 3.13 (1.11, 8.83)* 5.39 (1.17, 24.77)* Relationship unity 2.33 (1.41, 3.84)** 2.16 (1.12, 4.15)* 2.86 (1.49, 5.52)** 2.43 (0.89, 6.69)   Relationship dominance Female-dominated/egalitarian (ref) ref ref ref ref Male-dominated 1.56 (0.81, 3.04) 1.65 (0.74, 3.68) 1.52 (0.68, 3.37) 1.57 (0.54, 4.57) Relationship violence Ever been sexually coerced by partner 1.10 (0.55, 2.23) 2.25 (0.86, 5.85)   0.46 (0.22, 0.93)* 1.01 (0.29, 3.53) Ever been physically abused by partner 0.30 (0.10, 0.94)* 0.32 (0.09, 1.13)   0.17 (0.02, 0.79)* 0.07 (0.00, 1.07)   Perceived HIV status concordance R is HIV negative, P is perceived HIV negative (ref) ref ref ref ref R is HIV positive, P is perceived HIV positive 0.20 (0.07, 0.63)** 0.20 (0.05, 0.75)* 1.31 (0.16, 10.72) 1.71 (0.15, 19.40) R is HIV positive, P is perceived HIV negative 0.34 (0.15, 0.78)** 0.31 (0.12, 0.78)* 0.14 (0.04, 0.47)* 0.16 (0.02, 1.03)* R is HIV negative, P is perceived HIV positive 1.42 (0.18, 11.17) 0.78 (0.09, 6.52) 0.44 (0.11, 1.71) 0.28 (0.04, 1.85) Unity scores ranged from 1 to 4, with higher values indicating more unity. R refers to respondent; P refers to their partner   trend towards significance; p <.05; ** p <.01; *** p <.001 note 1: omitted due to multicollinearity. 1 Adjusted models control for all predictor variables and marital status (married/unmarried), age, years of education, household goods index, and having a TLT test at wave 4. Table 5.9: Odds ratios from logistic regression models predicting disclosure among women and men, TLT wave 5

PAGE 223

208 Discussion HIV Testing Uptake The hypothesis testing phase of this dissertation focused on whether relationship power predicted HIV testing uptake and disclosure among young couples approaching peak ages of HIV infection in Malawi. The res ults suggest that people are indeed getting tested at relatively high rates, repeatedly, and not just during pregnancy. In stark contrast to a host of studies from sub Saharan Africa that show low levels of lifetime testing among nationally representative samples (e.g., Sambisa, Curtis, and Mishra 2012; MacPhail et al. 2007; Peltzer et al. 2009) most respondents in this study had received at least one HIV test prior to enrolling in the TLT study. Rates of testing e xceeded population based estimates from the Malawi Demographic Health Survey, which found that 64% of women and 43% of men aged 15 to 24 had ever tested for HIV in 2010 (National Statistical Office & ORC Macro 2011) This suggests that young people in southern Malawi currently have good access to HIV testing and counseling services, which is likely to improve in upcoming years as testing be comes more integrated into routine health care. The use of longitudinal, HIV testing history data as opposed to a one time event of "ever tested" was a unique feature of this study that allowed for the investigation of predictors of HIV testing over a 16 month period. This is an important contribution to the literature, above and beyond the ability to circumvent methodological problems associated with temporality and cross sectional data (Singleton & Straits, 2005). With the recent scale up of ART and HIV testing services in sub Saharan Africa, more people are returning to

PAGE 224

209 health care centers for another HIV test (refer to Fiorillo et al. 2012 for an example in Tanzania) and thus it becomes critical to study how relati onship factors influence repeated testing for HIV. Regular HIV testing and counseling is also important from an HIV prevention perspective; a single HIV test is not enough information to make safe sex decisions throughout the reproductive years. R epeated H IV testing and counseling is also crucial for "treatment as prevention" interventions that rely on high rates of regular HIV testing in order to identify new seroconversions (Isingo et al. 2012) Still, few studies from sub Saharan Africa have measured the number of times people have ever been tested for HIV at the population level with a few exceptions 8 In South Africa, Kalichman and Simbayi (2003) reported that 29% of their cross sectional sample of men and women had tested twi ce and 19% had tested three times for HIV. Venkatesh and colleagues (2011) reported that approximately 58% of their Tanzanian respondents who had ever tested for HIV received multiple HIV tests in the past. Another study from Uganda found that 39% of respo ndents who agreed to be tested as part of the study were repeat testers (Matovu et al. 2007) While the current study captured HIV testing histories in a relatively short time period (as compared to lifetime history of testing), high rates of repeat testing were reported: almost 13% had reported two new HIV tests and another 13% had tested more than twice. This indicates that lifetime history of repeat testing is likely to be high in this population. 8 Note that while the sample in the current study is not a "true" population based sample due to inhere nt biases present in a couple sample, it is likely to be more representative of the larger population than clinic patients or other convenience samples used in studies elsewhere.

PAGE 225

210 Hypothesis 1A: Socio Economic Power and Testing Uptake Power and its effect on HIV testing was hypothesized to operate through the sexual division of labor, that is, socio economic inequalities between partners could either enable or inhibit uptake of testing depending on a couple member's relative position T here was some suggestion tha t women paired with significantly older partners were less likely to test for HIV. Similarly, older men were less likely to test for HIV than men closer in age to their partners. From the broader literature from Africa, we know that age confers social stat us and respect. Brown and Levinson (1987) argue that the social distance b etween two people influences an individual's commun ication strategy with a conve rsation partner. In this case, women may elect not to discuss HIV testing with older partners in order to avoid offending or disrespecting their partner. Older men, on the other hand, may be less pressured or persuaded by their younger female partners to t est for HIV. Economic inequalities, on the other hand, can be more closely mapped to the TGP's sexual division of labor. There was some evidence that being an unemployed women paired with an income generating man creates a situation of economic dependence that could limit women's use of HIV testing services. In Figure 5.1, this finding maps to a tradition orientation of labor. No support was found to indicate that equal economic power promotes HIV testing in the relationship (the unity orientation in Figur e 5.1).

PAGE 226

211 Hypothesis 1B: Relationship Dominance and Testing Uptake Through the sexual division of power being in a male dominated relationship (versus an egalitarian or female dominated relationship) was hypothesized to act as a barrier to testing thr ough the mechanism laid out by Maman et al. (2001) In contradiction to this hypothesis, the results showed that being in a male dominated relationship was associated with higher rates of HIV testing for both genders. Particularly, the results appear to be more conclusive for women The association between male dominance and higher testing conflicts with other research that suggests that male control over men's use of HIV testing services prevents women from getting tested (Perez et al. 2006; Dahl et al. 2008; Kranzer et al. 2009; Baiden et al. 2005) T hese studies did not measure male dominance directly but rather it is implied to be the underlying mechanism th at affects women's ability to s ay yes when offered an HIV test. It is possible that t raditional marital power structures may facilitate testing, particularly, if men take ownership over their responsibility for the family's health and well being. Hence, men in relationships characterize d as male dominant may be more likely to accompany their wives for testing and test with them Hypothesis 1C: Violence and Testing Uptake Relationship violence is one of the most widely studied manifestations of the TGP's division of power. Interestingly for women, a history of physical abuse was not predictive of HIV testing. This differs from studies that cite fear of abuse as a possible reason for why women may refuse testing (Irungu et al. 2008; Maman, Hogan, and Kilonza 2001) However, others scholars point out that the perc eived negative consequences of HIV testing and

PAGE 227

212 disclosure of results are not always the negative consequences that occur in reality (King et al. 2008) In Malawi, the same relationship may be true; be ing in an abusive relationship does not necessarily prevent women from testing or conversely, motivate them to get tested because of heightened risk for HIV. Interestingly, this study also found that men who had ever been pressured to have sex by their pa rtners were significantly less likely to get tested n et of perceived risk. It is difficult to attempt to understand this finding without supporting data from other research. Men's experiences of IPV are not well studied in Malawi or sub Saharan Africa desp ite some accounts of men being coerced into sex by more powerful women (Simpson 2009; Dunkle et al. 2007; Sikweyiya and Jewkes 2009) There may be other unmeasured characteristics of men who are unable to refuse se x making them less likely to get tested for HIV. Regardless, women's power over men in the sexual realm appears to be harmful for men's health when it comes to testing for HIV. Future studies should explore men's experi ences of sexual coercion in more dep th and how this may relate to their use of HIV services. Hypothesis 1D: Unity and Testing Uptake Although unity could exist at each of the TGP's three social structure s, it was specifically examined at the structure of cathexis or social norms around gen der and relationships. Unexpectedly unity was associated with lower rates of HIV testing for both genders. It is possible that individuals who report more unity with their partners (i.e., caring for partner, working together, and good communication) find less reason to get tested for HIV in the first place because the act of testing violates the very thing their relationships are

PAGE 228

213 built upon. Parallels can be drawn from the condom literature in Malawi. Chimbiri (2007) concludes that con doms are understood largely as "an intruder in the domestic sphere ," which may be attributed to public health advocacy for condom use in risky sexual partnerships not with in marriage. In another study using a set of qualitative, conversational journals fro m rural Malawi, condoms appeared to take on a symbolic meaning such that "if a relationship is defined as love, the connection between condom use and trust is inverted, as condom use signifies the absence of love, trust, and intimacy" (Tavory and Swidler 2009) Though we did not see a strong association between mistrust and testing uptake, the inverse relationship found between unity and testing reflect s what has been discovered for condom use. Hypothesis 1E: Mistrust/Infidelity and Testing Uptake Women were slightly more likely to test if they suspected that their partner could be cheating, however, this relationship was also not significant. Showing a relationship between mistrust and testing is challenging because this variable is inextricably linked with perceived risk, which sh owed a negative relationship with HIV testing. Thus, the predictive ability of the "mistrust" component of the perceived infidelity measure may have been cancelled out by risk. Hypothesis 1F : Perceived Risk and Testing Uptake Overall, p erceived risk of s elf and partner were negatively associated with HIV testing uptake among both genders, but failed to reach statistical significance in the multivariate models. The direction of this association for women and men is consistent with what others

PAGE 229

214 have found el sewhere in Africa (MacPherson, Corbett et al. 2012; Pool, Nyanzi, and Whitworth 2001) HIV Testing Disclosure This study is one of a few to report on relationship factors that influence both women and men's experiences of HIV status disclosure in sub Saharan Africa. Overall, respondents were very forthcoming about their HIV test results with their sexual partners. Over 80% of women and men reported disclosing their test results to their partner. This is in contrast to a large body of research that suggests that rates of disclosure to sexual partners are generally low (Obermeyer and Osborn 2007) However, these results coincide with findings from another study in rural Malawi that found similarly high rates of disclosure among ever married respondents In this study, 85% of women and 92% of men reported that they had disclosed th eir HIV status to their spouse (Anglewicz and Chintsanya 2011) It has been noted by others that self reported disclosur e data are sometimes unreliable (Anglewicz and Chintsany a 2011) The use of couples data in this study allowed me to assess the reliability of disclosure reports. Respondents were slightly more certain about a partner's status when their partners reportedly disclosed, however, the rates were not substantially different from partners who did not disclose. Thus, I cannot draw strong conclusions about whether people really disclose d their test results when they said they do. The findin gs are, however, more consistent when it becomes a question of who is more likel y to disclose. When respondents tested positive, half of the time they told their partners they were HIV positive; the other half of the time they told their partners they were HIV negative. But

PAGE 230

215 respondents who tested negative almost always told their part ners that they were HIV negative. Hypothesis 2A: Socio Economic Power and Disclosure As with the HIV testing uptake results, socio economic inequ alities between partners appear to play a role in decisions to disclose for women. The fear of abandonment i s widely cited in the literature as a barrier to women's disclosure of HIV status. It is not unreasonable to assume that these fears are inextricably tied to one's financi al situation. I suspected that women who had more education or were the family breadw inners would be less concerned if their relationships dissolved after disclosing test results. The results did not support this hypothesis. However, o ne exception was found for men. Men in dual i ncome earner relationships were significantly more likely to disclose their HIV status. Recall that in Chapter 4 some of the male respondents reported that they were relieved when their wives could help contribute to the family income; it reduced the pressure upon them to be the sole breadwinner for the family. It i s possible that men who have wives that help to support the household may have more time to take care of their own health, such as by testing for HIV. Other male respondents noted how when both spouses are working, it builds love and openness in the relati onship and allows both spouses to contribute to family decision making. When economic equality is present, it can help to build unity that fosters a more supportive environment for disclosure of HIV test results.

PAGE 231

216 Hypothesis 2B: Relationship Dominance a nd Disclosure Power imbalances that favor men were thought to create barriers for women to disclose their HIV status. However, no support was found for the idea that being in a male dominated relationship compared to an egalitarian one would limit disclo sure. Hypothesis 2C: Violence and Disclosure As hypothesized, women with a history of physical violence in their relationships tended to be less likely to disclose test results to their primary partners. The combined findings of this chapter suggest tha t relationship violence may not necessarily constrain women's ability to test, but rather their decisions to disclose the test results. The negative association with disclosure is consistent with numerous studies from sub Saharan Africa that cite fear of v iolence is a major barrier to disclosure among women (Nebie, Meda, and Leroy 2001; Maman et al. 2003; Kilewo et al. 2001; Farquhar et al. 2000) Men who had been victims of physical abuse were also less likely to d isclose their test results, p resumably for the same reasons. For women, t he opposite relationship was found for sexual violence; women who reported sexual coercion in their relationship tended to be more likely to disclose their HIV test results. Clearly, sexual IPV appears to operate very differently from physical IPV with regard to its influence on disclosure. Hypothesis 2D: Unity and Disclosure The unity factor appears to play an important part in both women's and men's decisions to disclose to each oth er. While unity was found to be a barrier to uptake of testing, it served as an enabling factor for disclosure of HIV test results. Individuals in relationships with

PAGE 232

217 higher levels of unity were more likely to discuss their HIV test results with their prima ry sexual partners. Aspects of unity may facilitate communication and provide the necessary social support needed for couple members to share their status with partners, particularly for HIV positive individuals who may be more susceptible to the negative consequences. Hypothesis 2E: Couple HIV Status and Disclosure HIV status did not play a significant role in decisions to disclose for men or women in bivariate analyses. The results for HIV status conflict with other studies that find HIV positive indivi duals are less likely to disclose than their HIV negative counterparts (Maman et al. 2003; Anglewicz and Chintsanya 2011) It is possible that some HIV positive respondents kne w about their HIV infection for quite some time, such as those who might have tested positive at an earlier TLT wave or agreed to re test via TLT when they already knew they were HIV positive. Thus, respondents may have been re disclosing what their partne rs already kne w. Others have found that knowledge of a partner's status plays a role in decision to disclose (King et al. 2008) In this study, the choice to disclose appears to depend on whether a r espondent's status is similar to or different from the perceived status of a partner. Women who tested positive but perceived their partners to be HIV negative were significantly less likely to disclose. Similar resul ts were found for men. HIV positive wom en were also less likely to disclose if they perceived their partners to be negative as compared to when both partners were perceived to be HIV negative. In this scenario, these women may be concerned with transmission to their partner if they failed to di sclose.

PAGE 233

218 One additional finding that was not part of the primary analysis but has important implications is worth a discussion. Men who tested through TLT as opposed to on their own were significantly more likely to disclose their test results to spouse s and partners. For women, it did not matter whether they were tested at TLT or elsewhere rates of disclosure remained constant between the two groups. A few explanations are possible. First, men tested at TLT may have been pressure d by their wives to disc lose if their wives also tested and expected to hear their partners' results. This is important because it suggests that men who test on their own perhaps without their partners' knowledge may be less likely to share their results. Women, on the other hand may be equally likely to disclose their results regardless of whether they tested at TLT or on their own because their partners are typically more aware of their plans to test and expect to hear the results. There may also be an underlying obligation to inform male partners out of respect and obligation for their role as the family leader. Couples testing may be an important intervention to increase rates of di sclosure specifically among men who may be more sensitive to their partner's particip ation in th e testing experience, if it can be done in a way that minimizes the negative consequences. Limitations As with any study, the hypothesis testing phase is subject to a number of limitations many of which have already been mentioned throughout this disser tation. The first set of limitations centers on the relationship power scale As already discussed, there are likely to be other uncovered forms of power not covered by the unity and discordance subscales

PAGE 234

219 Though inductive forms of power could emerge using the semi structured couple interviews collected in 2009, the dominant theory guiding the line of questioning was the TGP. I have already described the limitations of using this theory. The reliability of the unity scale could also be improved upon. This w ould likely occur as a natural consequence of more measure developm ent work that refines the unity construct. T he scale development research was conducted in the southern region of Malawi with a matrilineal/ local orientation (as opposed to the patrilineal/ local north), which may limit the ge neralizability of the findings T he main power differen ce between the two regions may be women's ability to leave their relationships (i.e., divorce) w ith more control over land. Although thi s was accounted for in the pr eliminary scale, these items did not make it into the final scale; they dropped out during the factor analysis It is possible that if the same factor analysis was run using a sample from the patrilineal areas, these underlying constructs may have been mor e salient. At the same time, it could be argued that l ineage structure plays less of a role given the colonial influences of patriarchy on power relations across Malawi. A second limitation relates to the data used in the hypothesis testing stage for HI V testing behavior. The HIV testing uptake measure included HIV testing through multiple possible venues, including home based and workplace testing, client initiated testing (or VCT), provider initiated testing thro ugh other health services, and antenatal care testing (although this was controlled for) Thus, since it is difficult to disentangle the testing modality in this measure, it is unclear how relationship power influence s decisions to test through these multiple venues. This limitation is not neces sarily limited to this dissertation; a

PAGE 235

220 vast number of studies that examine the outcome "ever tested" do not inquire further about where the test occurred and who initiated it. Finally, it is possible that the disclosure results may be sli ghtly biased by selection given that only respondents who had previously tested were included in the sample This is in contrast to other st udies where the entire sample was tested for HIV (as part of the study) and then asked about whether they disclosed the results of t he test to their sexual partners (Anglewicz and Chintsan ya 2011) The bias in the current study is mitigated to some extent given that over half of the 799 respondents were tested at TLT's wave 4.

PAGE 236

221 CHAPTER VI "WHEN THERE IS DOUBT IN THE HOUSE, THAT'S WHEN YOU GO FOR TESTING": HIV TESTING PERCEPTIONS AND EXPER IENCES In the previous chapter, I used quantitative data to show that an important form of relationship power called unity a composite measure that embodied social norms of communication, collaboration, intimacy, love, and trust was associated with HIV te sting behavior. I discovered that unity served as a barrier to testing uptake, but facilitated disclosure What might explain these findings? In this chapter, I present the resul ts of specific aim 3 to understand the me aning of HIV testing within a sexual relationship. What emerged from the results was the importance of the theoretical construct of unity and its subcompone n t of trust, which exist at the TGP structure of cathexis. This chapter offer s more depth and explanation around the meaning of unity and how it shape s decisions to test and disclose. Through an inductive approach Malawians' narratives also chart ed how couples negotiate testing within the context of relationship formation, marriage, childbearing, and intimacy. Summary of Research Methods For the analysis, I drew upon two sets of qualitative data: 8 focus group discussions conducted in 2011 and 34 semi structured interviews conducted with couples in 2009. In both sets of data, respondents were young, a mix of married and single individuals and from the same villages in southern Malawi. Focus group respondents spoke about their perceptions of testing while the semi structured interview respondents explained their actual

PAGE 237

222 experiences with HIV testing in their current relationships. A series o f vignettes were used to elicit responses among the focus group respondents, using a hypothetical young couple named "Lucy" and "Promise ." In four different scenarios, Lucy and Promise faced difficulties negotiating the terms of testing and the group was a sked for their recommendation s on what the couple should do and why. For e xample, Lucy thinks her husband Promise has another sexual partner and she is worried about HIV A chain of questions followed such as, should she go for testing on her own? Or does she need her husband's permission to test? What if the couple was not married but dating, is she still obligated to tell her partner of her plans? Subsequent scenarios asked other hypothetical questions about disclosure of test results and relationship dis solution or divorce. HIV testing also came up organically during early questions regarding relationship ideals and norms. In the semi structured interviews, respondents were asked about their HIV testing history and to describe the events leading up to and following the act of testing (these respondents will be referred to as "semi structu red interview respondents" here after). The Symbolic Meaning of an HIV Test Through their social interactions, people learn the meanings and symbols that ultimately shape their thoughts, actions, and interactions with others (Ritzer and Goodman 2004) In this study my analysis focused on the micro level and followed how rural Malawians came to understand HIV testing as they interacted with each other and the larger social structures surrounding HIV testing. I conceptualized "social structures" to refer to physical structures such as VCT testing sites located in rural communities as well as intangible structures related

PAGE 238

223 to national testing policies, the dominant gender hierarchy, and the institution of marriage. I specifically examine d how these social structures were manifested at the relationship level. Using a social interactionism perspective Kathryn Rhine (2009) argues that an HIV test is not simply a tool employed to measure immunological malfunction. A d iagnosis also illuminates a set of social facts. She suggests that the virtues and fears embedded in the act of taking a test are related to larger questions of how families and relationships might change in light of a positive result. Other qualitative re search that interviews people who have been tested indicates that test results have meanings that are tied to relationships, faithfulness, and trust. Individuals often take HIV tests when they are ending a relationship or starting a new one and thus testin g serves to mark these important life transitions (Lupton, McCarthy, and Chapman 1995) In the current study, respondents revealed that HIV testing decisions were intimately connected to aspects of un ity and trust. They described the relationship between HIV testing and trust as a reciprocal one. Testing could promote or destroy trust depending upon the results of the test, but trust (or lack thereof) could be the main motivating factor for seeking tes ting in the first place. Overall, study respondents perceived HIV testin g as problematic if the topic was brought up at the wrong time in the relationship and in such a way that raise d concerns about cheating. For some people, learning one's HIV status pro vided solid answers to their questions of fidelity. But for others, it created more confusion and uncertainty about the future. Despite their trepidation around HIV testing, rural Malawians did see the value of learning their HIV status and invoked creativ e strategies to maintain important relationship

PAGE 239

224 ideals such as love, trust, and intimacy while attempting to minimize thei r risk to HIV at the same time. The Relationship Ideal of Testing before Marriage' Finding an HIV negative partner is of critical im portance to young people contemplating marriage (Clark, Poulin, and Kohler 2009) In today's age of expanded access to HIV testing services, young people are likely to take advantage of more accurate ways to assess a partner's HIV status. In the current study, respondents from six of the eight focus groups mentioned almost immediately after being asked to describe a perfect relationship about the importance of going for "VCT" together while dating. This relati onship ideal transcended gender lines. Focus group respondents provided a variety of different reasons for why couples do and should get tested during the courtship phase of their relationship. Young people may go for testing while they are dating to deter mine if their chibwenzi 9 is marriage material. In several focus groups, respondents recommended that couples get tested when they start to fall in love because they are probably going to have sex soon and early testing can help prevent HIV transmission if one partner is positive. There was evidence among the semi structured interview respondents that these ideals and recommendations are carried out in everyday life. Pre marital testing was perceived as an important and often symbolic step towards advancing a relationship to the next level. Several single men informed us that they got tested during courtship as a way to demonstrate their trustworthiness and dedication to the relationship, and in some cases, to convince a partner 9 Chibwenzi is the Chichewa word (gender neutral) for a sexual partner outsid e of marriage, usually the equivalent to a boyfriend or girlfriend.

PAGE 240

225 to have sex. One young d ating man named Andrew shared a conversation he had with his girlfriend about getting tested for HIV before having sex for the first time. He later explained how he eventually agreed to her suggestion to test because of the dangerousness of AIDS. Interviewer: Who started this issue of VCT? Andrew: The girl told me and we agreed not to have sex until we get tested. I asked her to sleep with me. She said first we were to get tested then continue with using chishango (a popular brand of condoms in Malawi). Inter viewer: What was she afraid of? Andrew: I can say two things. She was afraid of these venereal diseases, STIs and pregnancy. This time she was just a girl, still in school. Remembering the bad times we are in [referring to HIV/AIDS], I knew she was sayin g the truth and I agreed to go with her for VCT. (semi structured interview, male #1) It is notable that his initial reason for testing was to sleep with his girlfriend, not to prevent HIV. For other people, future aspirations around marriage and childbea ring weighed heavily into their decisions to test. In the following passage, a married woman named Anna recalled about how she pressured her partner go for testing before she would officially commit to marrying him: Anna : Before we got married, we went fo r HIV testing. After we went for t esting, we got engaged. Interviewer: What made you go for testing? Anna : We wanted to know my and his status on HIV. I told him that if you want to marry me, we should go for HIV testing first. Interviewer: What did h e say? Anna : He agreed and he said that he will go. So he went and before that time we didn't do anything. Interviewer: What are you trying to say when you say you didn't do anything? Anna : Um, before we started having sex. Yes. So we went for testing first so that we could know how our bodies are. (semi structured interview, female #3) Beyond fears of AIDS and perceptions of risk, important life projects like marriage and desires for intimacy, trust, and closeness with a premarital chibwenzi were dee ply embedded in the way rural Malawians made decisions to test.

PAGE 241

226 In hig h fertility settings it is customary for young couples to begin childbea ring soon after the marriage begins There is evidenc e from Malawi that knowledge of one's HIV status plays an s ignificant role in desires to have children (Yeatman 2009) Thus, HIV testing before marriage help s couples plan out their future families. In the next passage, a married woman named Edith believed that learning her and her fiancÂŽe's HIV status would help them to bear healthy children after they got married. Interviewer: May you explain to me the conversation you had about it [HIV testing]? Edith: We said that our wedding day is near, and before it is reached, I would love if we could go for testing so that we can know how our blood is so that we can have a happy [meaning healthy] family. Then we agreed to go. Interviewer: Who was saying that? Edith: It was him but still we both had the same thoughts [idea to test] because these days you can't just get married before going for blood testing. (semi structured interview, female #15) At first, Edith talked about how her husband insisted on testing so that they could have healthy children but as she continued h e r reason for testing refocused on the marriage itself. One is left to wonder if Edith and her partner are reall y testing so that they can safel y continue with their marriage plans. By using concerns about the children as a way to justify testing, the coupl e can effectively circumvent the negative association between trust and testing and avoid falsely accusing each other of sexual promiscuity. But t o the contrary, focus group respondents pointed out that not everyone gets tested during courtship and this i s how a person can become entangled in a complicated web of love, sex, and HIV. Other research from rural Malawi shows that ideal relationship sequences, e.g., getting tested for HIV then having sex, do not always translate into real life events (Frye, Trinitapoli, and Nam adingo 2011) This disconnect between ideals and practice is understandable for young people who may be worried about testing positive, social

PAGE 242

227 rejection, and engaging in awkward conversations about sex and HIV testing. Other couples may simply get caught in the moment and end up having sex before they initially intended, as suggested by the man below. In this conversation, a group of four single men illustrated the problems of not going for testing in the beginning of the relationship: Man #3: A good rel ationship is the one that when you have proposed each other and you go for testing. If you are really in love, you can have sex when you get married but also after you go for testing. That's a good courtship. Man #5: My idea is the same that in a relation ship you should start sleeping with each other after you have gone for testing. But if you find that you are okay [without the virus], you should still use condoms. It can help. Man #3: But perhaps we want to have sex with the woman before we know our sta tus. Man #8: Maybe the woman had sex with someone that is HIV positive and she did not go for testing. But maybe you went for testing and you are okay [HIV negative]. So you get caught in the cobweb and you are found in the group of people being infected. (FGD #8, single men) In the passage above, the first man took on a relationship centered as opposed to risk centered narrative when discussing the benefits of early testing evidenced by his references to "love" and the quality of the relationship ("a good courtship"). The other men perceived HIV testing as a way to avoid risk and determine who is considered a safe sexual partner. Regardless of their orientation towards risk or the relationship, all men in the group seem to concur that testing together at t he start of the relationship is of critical value. The Incompatibility of Testing with Marriage While HIV testing was portrayed as a necessary precaution for dating couples to take before considering marriage, testing during marriage outside of the a ntenatal care setting was considered inherently problematic. From the respondents' accounts, I conclude that HIV testing is largely incompatible with marriage because it goes against social norms of trust, love, and intimacy that couples strive for in thei r daily lives. Married couples who love

PAGE 243

228 each other and have no reason to suspect cheating would have little reason to seek out HIV testing on their own accord. This conclusion was consistent across marital status and gender. In the following passage, a sin gle man summarized the general belief of the group about how it is critical to go for testing at the beginning of the relationship because bringing up testing during marriage raises questions of faithfulness: Testing has to start while in a relationship [d ating]. You have to tell each other to go for testing while in a relationship not just when you are married because if you do [wait until marriage to test], you might start having doubts about the partner, for example, asking why all of a sudden she wants to go for testing? Has she been sleeping around with men? (FGD #3, single men) When asked about why married people typically go for HIV testing, a female focus group respondent answered, "Mostly, it is because of the lack of trust in the relationship and in doing so, they are able to know the truth so that they live happily after knowing their status." (FGD #1, married women) Simply bringing up conversations around HIV testing with a spouse symbolizes mistrust and deception and can lead to major disagreeme nts in the marriage. In the first HIV testing vignette, a young girl named Lucy ponders whether she should tell her husband she wants to go for testing (she thinks he is putting her at risk for HIV by his extramarital affairs). A married man shared his opi nions with the group about the implications of bringing up testing unexpectedly with a spouse: Because by just saying that we should go for testing, it means there is doubt in the house. You can go for testing, but it doesn't just come up like "let's go f or testing", no. But when there is doubt [mistrust] in the house, like how is my partner moving [behaving sexually]? If I think he is cheating, that's when you go for testing. But if you are not doubting [trusting] each other, you'll see that even if peopl e say that others are getting tested over there [at the VCT clinics], you don't regard [consider] it. (FGD #6, married men)

PAGE 244

229 In this man's account, he contrasted his opinion about testing with what he saw going on in his community, i.e., "others are gettin g tested over there" (referring to VCT clinics). Worldwide, there have been tremendous attempts to normalize the experienc e of HIV testing in order to increase the number of people who know their HIV status. In Malawi, this conflict ed with how rural villag ers defined appropriate use of HIV testing services for themselves. As the young man in the passage above insinuated, testing during marriage was labeled as an unusual event, one reserved for special or problematic circumstances rather than for everyday r outine screening of disease. The semi structured interviews illustrated how negative perceptions of testing during marriage correspond ed with people's lived experiences. In the following account, a married woman named Ruth lamented how her husband though t HIV testing was for people who suspect ed they were p ositive and therefore he refused to go. As a consequence of his response, she started to question his faithfulness as well as his HIV status: Interviewer: Ok! So when you went for testing did you tell y our husband or you just went? Ruth: I told him that there was no reason to be afraid of knowing how your blood is but he denied [refused to go], so I went by myself. So he was saying I was doubting myself. That was what he was saying. Interviewer: Alright, so did you encourage him to go one day or did you just give up? Ruth: No, I didn't give up, I always tell him. The first time I went I told him that you should also go so that we can have proof that we are alright, but if you don't then I will be having d oubts that maybe my husband is positive or negative. I will not be sure. The second time I went I told him the same thing, but to no avail. Interviewer: So what does he really say, I'll go later or I can't go? Ruth: He says that he can't go, that those who go for testing are doubting themselves [think they could be HIV positive because of past sexual behavior]. They don't trust themselves. Interviewer: Oh! How do you feel when he answers you that way? Ruth: It pains me, I see that he is not helping me and t he children. (semi structured interview, female #9)

PAGE 245

230 Ruth's interview illustrates the paradoxical connection between risk, trust, and the perceived need to go for testing. Rural Malawians understand very well that risky sexual behavi or warran ts an HIV test. Yet it is the adm ission or more likely suspicion, of guilt associated with the act of testing that may trump concerns with HIV infection. Not everyone views HIV testing from a western, public health lens that emphasizes disease pr evention and early detection or in the case of HIV testing, as a means to protect a spouse f rom HIV or to initiate ART In rural Malawi, conceptions of "risk" carry deep social meanings tied to infidelity and promiscuity which do not necessarily align wit h a purely biomedical or public health understanding Risk, Infidelity, and HIV Testing In general, respondents talked about HIV status as if it was an absolute indicator of marital fidelity. Smith and Watkins (2005) found that Ma lawians vastly overestimate d their chances of infection through a single sexual encounter wi th an infected person. Men worried about their extramarital partners as a source of infe ction while women worried about their husbands' affairs. In another study fr om Malawi, marital infidelity was found to be the strongest correlate of overestimating one's own and a spouse's risk for HIV (Anglewicz et al. 2008) According to this logic, mar ried people who were cheating and found to be HIV positive would presumably have become infected through these extramarital affairs. On the other hand s ome participants perceived HIV testing as a strategy to substantiate their own or their partner's fait hfulness. In one conversation about jealousy, single women shared their beliefs about how rural villagers could try to destroy a healthy marriage by spreading rumors

PAGE 246

231 about a couple's HIV status and sexual history. One woman pointed out that HIV testing was one way to prove to your spouse (and your accusers) that you weren't c heating and by extension, that you did not have HIV: Interviewer: So people can go and say your woman has done this and that [referring to cheating]? Woman #1: They can lie that the women has diseases when she doesn't have diseases. Woman #3: But, if you love each other you can go for testing. They can spread lies, but you can say [to your husband] that if you love me, let's go for testing so that you believe that I don't have [HIV] (FGD #2, single women) The use of HIV testing for "fidelity confirmation" resonated in the men's narrati ves as well. In Malawi, men are cognizant of the dominant HIV prevention discourse that blames unfaithful married men for transmitting HIV to their w ives. This topic also came up in a conversation I had with Joseph, a VCT counselor working at a private clinic in the Balaka town (see Chapter 2 for more details). His general impression for why men refuse to test was because he thought men believe d and fe ar ed that they would be blamed for bringing HIV into the family. While this belief may deter some men from getting tested, it is likely to motivate others. In the following interview excerpt, a married man named George described how he used HIV testing as a way to prove his faithfulness to his wife: Interviewer: Have you ever talked about being tested for HIV/AIDS? George : Yes, we have tested for the virus. Interviewer: Who started talking about this issue? George : I am the one. Interviewer: What made you to do this? George : We wanted to promote trusting each other. We wanted to remove [stop] doubting each other. Just because many times the blame goes to the men, with women saying you are the one who brought the virus to me'. Then I decided that we should just go and be tested. Interviewer: What was her reaction to this issue? George : She was happy and agreed. She said she will follow what I was saying because I am the head of the family. (semi structured interview, male #8)

PAGE 247

232 Sometimes, HIV testing raised more questions than provided answers. A few focus group respondents pointed out that just because a person tests positive, it doesn't mean they necessarily cheated; a spouse could have been positive before marriage and may have contracted HI V through non sexual routes, such as by caring for a sick loved one, exposure to contaminated needles, or from their HIV positive mother during childbirth. One strategy to retrospectively assess when a spouse became infected with HIV was by going for testi ng together during the courtship phase. The following group of married men discussed how failing to go for testing at the start of the relationship can lead to false accusations of infidelity later on during marriage. Man #2: The problem is when you are proposing a girl, she does not tell you that she has the virus. Man ? But together, you can make the decision of going for testing at that particular time. Man #4: It just goes wrong in the beginning because you did not take each other for testing. Ma n #7: Because we can think of the woman as being a bitch [meaning cheater] when she is not. The problem is really based in the beginning when you did not go for HIV testing. (FGD #4, married men) As the above passage suggests, early testing provides a ba seline measurement to use in future judgment around cheating. For example, if a woman tested negative before marriage but positive afterwards, her faithfulness could be questioned. Without this additional piece of knowledge, it becomes difficult for partn ers to assess their faithfulness to each other using current serostatus alone. They come to rely strongly on subjective assessments of a partner's infidelity, which are often less reliable. When making life changing decisions using this information, it bec omes very important for partners to accurately evaluate the relationship integrity before moving forward. Indeed, in rural Malawi, Anglewicz and colleagues (2008)

PAGE 248

233 found that couples' perceptions of a partner's risk and their actual risk are often mismatched, possibly leading to premature divorce or breakup. Now that I have outlined the broader framework for how rural Malawians understand HIV testing as it intersec ts with their relationships, I move onto the role of gender and power in couple decisions to test for HIV. In what follows, I paid special attention to the broader set of beliefs respondents used to make sense of the relationship between power and getting tested for HIV. Tradition, Rights, and Unity In Chapter 4, I outlined three narratives respondents used to construct meaning around gender and power relations: unity tradition and rights. While I hypothesized that these three perspectives could interse ct with the TGP's divisions of labor, power, and cathexis, in this chapter, I specifically focus on the structure of cathexis. When presented with the hypothetical HIV testing scenarios, respondents used the same three narratives to verbalize their perspec tives on relationship power and HIV testing. Broadly, I coded passages with rights when respondents talked about human rights and "the right to test ." The t radition code was applied to passages that referenced traditional gender roles (e.g., "she has to te ll him, he is the head of the household"). I invoked the unity code when respondents talked about love, intimacy, couple communication, the need to test together, and the notion of a married couple as "one body ." I next discuss how these three narratives w ere invoked in discussions of testing autonomy, the refusal to test, and alternative testing strategies.

PAGE 249

2 34 Testing Autonomy In the first hypothetical vignette, focus group respondents were asked to comment on Lucy and Promise's level of autonomy and decis ion making power to test. Most focus group respondents drew upon a unity narrative to argue that both husbands and wives were supposed to tell each other about their plans for HIV testing, especially if they were married and love d each other. Other respond ents contrasted tradition with rights in their justifications for whether women should inform a spouse of their plans to test and how women should respond if husbands impede d them from going. In their tradition narratives, respondents invoked language aro und traditional gender roles to argue that a woman could be in big trouble if she goes secretly for HIV testing because women were not supposed to be "moving around" without their husbands' knowledge. Covert testing i mplied that one was hiding something fr om a partner presumably a secret love affair. Other tradition arguments suggested the importance of the division of labor. In the following passage, a married man argued that the man (husband) needs to know of family health matters and ultimately, a woman 's decision to test falls in his domain: Because first of all, the man needs to know that, in our family we went to the clinic, we have tested and our status is like this or that. She doesn't have the power to go on her own for testing. It will be like I do not have trust how she is regarding herself [perhaps as promiscuous or unfaithful]. But it will also depend whether they tell each other and with how things are, saying "What if we go to the hospital for testing?" So it should be a man's decision afte r you have thought very well [deeply about the decision to test], and saying [to his wife] "let's go." [for testing] (FGD #6, married men) Similar beliefs resonated among a group of single women who argued that the husband should know of a wife's intentio ns of going for testing since he is considered the head of the household:

PAGE 250

235 Interviewer: So she [Lucy] should tell the husband that she wants to go to the hospital for blood testing and then she should hear what the husband will say. All: Yes. Interviewer: Why should she ask the husband? Woman #1: Because he is the household head. (FGD #2, single women) Another group of single women reported that if Lucy goes for testing without telling her husband, her husband may challenge her intentions by saying "w hat made you go for testing without telling me?" and the couple could quarrel something that is highly undesirable and thought to cause the breakdown of marriage. However, not everyone share d this belief. Several groups of female participants shared their opinion that many husbands do respect their wives decisions to test. Other respondents, notably men, contested the role of tradition wit h the belief that HIV testing was legal in Malawi, a human right and therefore a husband could not prevent his wife f rom learning her status. In one particular conversation about Lucy, two married men argued with each other about whether a woman could rightfully go for testing without informing her husband: Man #1: HIV testing is legal and I don't think he can refuse her to go and do so. Saying, "do not go for testing because I am cheating on you ." Because HIV testing is a human right. He can say "do not go for testing" but because she really wants to go, one day or the other, she will make her own plans and go for tes ting. Man #4: To be clear enough, if the woman went for HIV testing without telling him, her husband, and she has been found HIV positive, she is supposed to tell him but it is at the same time difficult on how to tell him because maybe the man can ques tion her saying: "why you were going for testing?" It is true that testing is her right but as the head of the household, she is supposed to let him know. He can ask that "when you were going for HIV testing, did you tell me?" And she can say "no, I did no t." (FGD #4, married men) These two men exemplify the difficult dilemma that couples face when making decisions to test. Human rights advocacy in Malawi has provided women with the legal freedom to learn their HIV status, however, this structural change c ollides with the everyday realities of

PAGE 251

236 couples on the ground. As the men in the focus groups suggest, exercising global human rights to test has local implications particularly between husbands, who have been granted cultural rights to the family leadershi p role, and the ir wives who should respect them. Research from Malawi indicates husbands may prevent their wives being tested (Kranzer et al. 2009) even when press ured by antenatal care staff to test during pregnancy (Angotti, Dionne, and Gaydosh 2010) Interestingly in the couple interviews, I did not find a single instance of a man prohibiting his wife or girlfriend fro m getting tested even if he refused testing himself. The following semi structured interview excerpts provide candid examples of women's ability learn their HIV status when they so desired. In the first passage, a single woman replied how she was free to g et tested even if her boyfriend refused to go. When she showed him her results, she said she could use it as a teachable moment to convince him to test. Interviewer: Let's just go back on the issues of HIV, if you tell your boyfriend that you should both g o for testing and he refuses, can you still go on your own? Agnes: If it can be that I have told him to go for testing and he refused, I can go on my own for testing. Maybe we agreed earlier on to go and he has just changed his mind, so I can go. And when I come back I can show him my test results so that I can encourage him to go for testing as well. (semi structured interview, female #10) In the second passage, a married woman named Ellen to ld the interviewer how she faced no resistance from her husband when she wanted to go for testing: Interviewer: So when you are going for testing, when you tell him [that you are going], does he agree that you should go or does he not want you to go? Ellen: He lets me go, he doesn't care. (semi structured interview, female #11) Finally, a third woman explained how she could go for te sting as many times as she wanted especially if she suspected that he was cheating: Interviewer: So you trust your husband, you don't think there is anything he can do [referring to

PAGE 252

237 che ating]? Catherine: He can have an affair behind my back, but if there is any doubt I can go and get tested again. (semi structured interview, female #0) In contrast to some of the earlier narratives in this chapter on the negative associations of testing with mistrust, these women spoke of testing as if it were any other normal activity and encountered little opposition from their male partners. Yet none of these women went covertly for testing (as least that we know of) and partner reactions could be ver y different if he found out she went for testing without his knowledge. Informing a male partner out of respect and to maintain transparency may be a wise strategy for women to exercise their rights while maintaining norms around tradition at the same time Refusing to Test Just as the act of bringing up testing with a partner conv eyed underlying meaning regarding the status of the relationship, so did refusing to go with a partner for testing. Focus group respondents provided a variety of different reaso ns for why young people might refuse to go for testing with their partners. A few respondents mentioned that f ear of testing positive prevented people from testing. In addition, the belief that spouses share the same serostatus given they have had unprotec ted sex might lead some people to the conclusion that they can infer their own status from their partner's test result and thus why go yourself? For example, one single man said: "Mostly when in a relationship, one partner always refuses to go and when the other one has gone for testing and has come with the results, the other partner assumes he/she is also the same." (FGD #3, single men) Respondent talked about how if "Lucy ," the hypothetical wife, brought up testing with her husband and he wo uldn't go with her, she should interpret his reluctance as a sign of

PAGE 253

238 unfaithfulness or HIV infection. Similarly, refusing to go for testing while dating signified that there was a problem and this could be a deal breaker. In the following excerpts, single men from t wo separate focus groups commented on how a par tner's refusal to test indicated a promiscuous past or a sign of infidelity: Man #3: In the village where I am living, when people are dating they talk to each other about HIV testing and g o for testing. In th is case where by one is afraid of going for testing, the relationship ends there. Because it is taken as though he/she does not have enough trust in his/her partner. (FGD #3, single men) In the second focus group, a man responded: Man #4: Even if she hears that the man is cheating, she should tell him that let's go for testing. When he refuses, that's when she has a point. Why has my husband refused? Then it must be true. (FGD #8, single men) When a partner refused to test, it provided an important opportun ity for people to re evaluate the quality of their relationship and changing risk for HIV infection. Testing Together as a Workaround Strategy Overall, study respondents perceived HIV testing as problematic if the topic was brought up at the wrong time i n the relationship and in such a way that raise d concerns about cheating. For some people, learning one's HIV status provided solid answers to their questions of fidelity. But for others, it created more confusion and uncertainty about their relationship. This is not to say that young Malawi ans did not see the value in getting tested for HIV. In fact, people offered various strategies to overcome perceived barriers to testing and to circumvent the negative associations between HIV testing and trust. In the ir proposed solutions, focus group respondents primarily drew upon unity narratives. Seven out of eight focus groups suggested that a couple (i.e., Lucy and Promise) should ideally get tested together and sometimes the entire group came to this conclusion

PAGE 254

239 after debating back and forth about what Lucy should do. In one conversation, married women chatted about whether women should tell their partners about going for testing. The women stressed that it was important for couples to go together and learn their status: Interviewer: She should tell her husband that she is going for testing? All: Yes (group answer). Woman #3: She should tell her husband that they should go together for HIV testing. If she doesn't want to let the husband know, she should just go for the HIV test on her own and when she comes back, she should tell her husband that she went for HIV, stating and that she has been found negative. It's now up to her husband to also take the same route [get tested] and be happy about his wife's statu s. Woman #8: The most important thing is to agree on one thing. Going for testing together and know their status together. (FGD #1, married women) The same belief resonated among single women. In another conversation about whether Lucy should go for tes ting on her own, single women in a focus group emphasized the need to go together for testing even if he initially refuse d : Interviewer: So in the first place she told her husband, let's go for testing and he refused, but she went on her own. She should still force or tell the husband to go to the hospital to be tested? Woman #3: You go together for testing. Interviewer: When you are tested you should go together? All: Yes. Interviewer: Ok, what if Lucy is found without the virus, should she tell the husband? Woman #1: She should tell him so that they go together. (FGD #2, single women) It is important to highlight the last woman's response the idea that Lucy should not tell him her status but rather suggest that they go together. This strategy emerged in other interviews as well, as I will discuss in the paragraphs to follow. Numerous benefits of couples testing were mentioned mostly as they pertain to relationships, not risk avoidance. Male respondents, in particular, discussed how the act of testing together symbolize d love, unity, and a strong marriage. For example, one focus group respondent described testing together as a way to demonstrate caring for a partner during a

PAGE 255

240 time of intense worry. Others saw HIV testing as inseparable from the c ouple context. In the next passage, a male focus group respondent shared his belief that when you are married you are one and therefore should learn your status together: With the way I see it [regarding] the scenario that says, this woman would like to go for testing so should she seek for permission from her husband or not? According to me, I think that she should tell him because once people are married they are considered as one. They should also go together and kno w their status. In a case where by th e two have not gone together, still, the woman should let her husband know and when she comes back from the hospital she should also let her husband know about the results. (FGD #3, single men) Once married, HIV status becomes a couple level issue that re quired collaborat ion between spouses This understanding was repeated in another conversation among a different group of married men, who pointed out that spouses need to tell each other about their plans to test since "they are considered one ." The last m an perceived testing together as the proper use of a man's power. Interviewer: What about man #2, you said she should seek permission from her husband [when going for testing]? Man #2: Yes. Interviewer: Why? Man #2: She is supposed to seek permissi on because they are one and they are suppose d to help each other when there is something that needs to be done. So she needs to ask because maybe the man can also go with her for HIV testing. Maybe this can help. Man #1: There's a need of asking for perm ission from her husband because if the man is using his powers properly, he can also join the woman and go for testing together. (FGD #4, married men) In general, respondents perceived "permission" not as male control over female autonomy but as an oppor tunity for couples to communicate with each other about HIV testing and assist each other out by testing together. While a unity narrative was used to emphasize the importance of couples testing irrespective of marital status, focus group respondents were more likely to emphasize a rights narrative when it came to dating couples. Once married, HIV transition ed from a strictly

PAGE 256

241 individual level concern to a family health matter that require d the involvement of both couple members to make health decisions A group of married men concluded that Lucy should ultimately tell her husband about her plans to test, but they changed their mi nd when Lucy's marital status was presented as single. Both narratives are contested in the following conversation: Interviewer: What if Lucy is not married, should she ask for permission from her boyfriend before going? Man #3: They should discuss before going. Man #2: That is just a relationship [dating] and that is what she wants. If the man wants [to test], he will also go o n his own because they do not know how their partner was behaving before they met each other. Man #7: Because it is just a relationship, she has the right to go for testing and know her status. At first the men talked about how each of the partners has t he right to go for testing on their own as unmarried individuals. But if an individual just exercises their right to test without considering their partner, there could be problems. The first m an ( #3 ) took issue with each of them going alone, particularly, regarding the inability to really know each other's HIV status. Therefore, he argued they need ed to go together for testing. Man #3: You are saying she should just go because it is just a relationship, now tell me how will she know her partner's status if she just goes without letting him know? Man #4: Everyone has the right to go for testing when they want and where they want to go. And when she goes for testing, whether she has a boyfriend, she will not just go to her boyfriend and say "I went for HIV testing and they found that I am HIV negative ." Or "I went for testing and they found that I am HIV positive ." If there is a need of telling her boyfriend she needs to tell him that "we should go for HIV testing" not telling him that "I went for testing ." If she went for testing and she was found HIV negative or she was found HIV positive, she should just keep quiet because she did not inform him when she was going. But if she wants to tell her partner, she should say that they should go together. "What ab out we go for testing?" (FGD #4, married men) The fourth man's opinion came up in an earlier focus group. He explained in a very matter of factly manner that anyone could learn their HIV status for personal benefit. In order to avoid conflict, he stressed that there was a need to communicate with a chibwenzi about

PAGE 257

242 testing only if the goal was to learn each other's status together. With this strategy, unmarried young people could maintain their relationships and protect their own health and reputation by in volving a partner only if necessary. Another benefit of testing together was that it ensured that both partners learn ed each other's HIV status given that the couple was told their test results in front of each other by an HIV counselor thus avoiding a s ituation of deception or failure to disclose. In one of the married men 's focus groups, a man noted how a person doesn't really know their partner's HIV status unless they tested together. Respondents pointed out that if a woman went for testing on her own and tested positive, she risks being labeled as the cheater despite the fact that th e status of her husband remains unknown. When asked how Lucy's husband would respond to her positive test result, one man said: "The way I see it, the man will not accept what he has heard. Because of this, it shall be a burden to the man where by the man will call her a bitch. You said you were HIV positive? Yes, that means that you are a bitch." (FGD #4, single men). Finally, testing together help ed to avoid the uncomfort able situation of having to inform a partner of a positive test result without knowing the other's status. When asked about whether a young man named Promise should tell his wife about testing, a young woman noted how if they were married, he should go wit h his wife for testing because it could be too difficult to tell her the results later on if he tested positive: If he loves her, he needs to tell her. Maybe his wife was not told that Promise was going for testing because of whatever reasons. So if they a re married, it's better they go together. Maybe the wife is also like that [sick] since they have stayed together [had sex] for a long time during marriage. So it's difficult. If he goes alone, maybe he will be found positive. Now it's going to be difficul t to tell his wife that he is positive. (FGD #2, single women)

PAGE 258

243 Again, there exists a widespread assumption that if one spouse is positive, the other person must be as well. In a setting where cell phones are becoming increasing ly popular, o ne focus group respondent metaphorically used the phrase "t he virus is in the voicemail" to mean that it was just a matter of time before the other couple member test ed positive (FGD #8, single men). The young man, "Promise", is presumed to d read telling h is wife that he was positive because he believed he would be indir ectly informing her that she had AIDS. No one wants to inform a loved one that they have an incurable, deadly disease. In this respect, testing together takes this burden off of his shoulder s and shifts responsibility to the counselor's domain. Respondents offered many clever suggestions for Lucy to avoid confrontation with her husband over testing. One particularly ingenious strategy to use when worried about a partner's infidelity was to go for testing secretly and then suggest to the spouse or chibwenzi that they go together wh ile acting as if one never tested on their own. This strategy is likely commonplace, as it came up during four different focus groups of different demographic chara cteristics In the following quote, a single women suggested that Lucy could keep her status a secret while she tried to implicate her husband in giving her the disease by saying to him: "My husband, let's go for testing, while already knowing that she has the virus so that he finds out that they both have the virus." (FGD #2, single women). For dating couples, one partner might go secretly for testing on their own first and then if they test ed negative, try to convince their partner to go with them for tes ting as if they never tested before. In this scenario, they had nothing to lose and could only gain by learning their partner's status.

PAGE 259

244 Love could also be used as a strategy to convince a reluctant partner to get tested. In a setting where men emphasize d the importance of having a loving, respec tful wife who took care of them, love could be a powerful tool for women to use to get what they want. One man pointed out that in an attempt to get a cheating man to go for testing, the wife could kindly make the a rgument that they haven't been tested in a w hile and it was good to go for regular testing to know their status. In another example, a married woman suggested sitting down with the husband after eating and explaining to him in a loving manner that it would be a good idea for them to get tested again: He does not know the status of the girlfriend that he has is sleeping around with. So he thinks that if you go for testing or maybe you went for testing together with him before he started cheating and you all tested negative then. Now, you ask him in a good [nice] way just to see what he will say then if he asks why, you just say that "I would like to see if it is going to be the same as last time ." If he knows [he is positive], he will not let you go because h e knows that when you will be found HIV positive you will know that you got the virus from him. But he will refuse [deny] you. Yes, [ask him] in a good [nice] way. For example, after you have eaten, you say, "why don't we go for testing ." Not in a harsh wa y, but in a loving way so that you can see whether he is going to allow [agree] or refuse to go. (FGD #7, married women) If love failed sex could be used as a bargaining platform. The Malawi Demographic Health Survey data from 2011 (National Statistical Office & ORC Macro 2011) demonstrates that almost 90% of women believe d that a wife was justified in taking action to protect herself from HIV, for example, by refusing to have sex. A male focus group respondent (as if speaking from his personal experiences) suggested that a woman could withhold sex from her husband until he agree d to go for testing: "Here when she goes to the hospital and she is negative, she needs to tell him that I went to the hospital got tested. These are my results, I am negative. When he refuses, she should tell him that they will only have sex after he goes for testing." (FGD #8, single men). In the following semi st ructured interview passage, a

PAGE 260

245 young man named Francis told the story of how his girlfriend withheld sex from him until they learned their HIV status together: Interviewer: You said you got tested for HIV. Who started this issue and what happened? Franci s: I am the one who started the issue when I was angry. She was refusing to have sex with me, saying she was afraid of the virus and getting pregnant. I started scrutinizing my background [sexual history], what I have done before this lady? Of course this is not my first partner but the previous one was not as serious as this. You know how children are. We were doing zawana [children games where they decide one to be a father and one a mother, games like hide and seek] and we agreed and got tested. I was h appy. I was more confident. Lucky enough, I was negative. (semi structured interview, male #10) Antenatal Testing as an Alternative to Couples VCT While couples testing was considered the ideal by many respondents, alternative strategies were offered specifically, antenatal testing. Any visit to a rural health center in Malawi reveals the overwhelming presence of pregnant women and sick children rather than men. Health facilities may be unreceptive to men, with little space for male partners to wait, l imited access to patient rooms, and unfair treatment by health service workers. Research is currently trying to build a strong evidence base for the inclusion of men in women's based health services (Baiden et al. 2 005; Mlay, Lugina, and Becker 2008; Desgrees du Lou et al. 2009) In rural Mala wi, modern day health care in Malawi is still considered a feminized space: a place where men generally feel unwelcome and overwhelmed by the unfamiliar and daunting task of na vigating a disjointed health care system that privileg es women and children. Despite these challenges, couples testing via antenatal care may provide a promising alternative to traditional couples VCT (or CVCT'). In Malawi, women of childbearing age are more likely to be tested for HIV through antenatal care than though any other venue

PAGE 261

246 (Angotti, Dionne, and Gaydosh 2010; National Statistical Office & ORC Macro 2011) This may provide an important opportunity for C VCT by recruiting men to participate in an already established system. Indeed, research has shown that couple counseling facilitated through couple friendly antenatal car e services should be explored as a potential alternative to VCT (Baiden et al. 2005; Mlay, Lugina, and Becker 2008) There have also been claims th at antenatal care provides the chance for women to sensitize their male partners to the sexual risks around HIV/AIDS (Des grees du Lou et al. 2009) While antenatal care testing was not a central topic of discussion in the focus groups, some respondents suggested that when the authority to test comes from health care providers instead of their wives, men ma y be more inclin ed to test. In a lengthy conversation about antenatal care, a group of married men explained how men were recruited from the antenatal clinics when their pregnant wives were found with HIV. The first man stated: Man #7: It is possible that Lucy has a chi ld so she goes to antenatal clinic. At the clinic, the women are tested. They are not just weighed [on a scale] but they check their HIV status. So if she was at the antenatal clinic, it is a good reason to tell the husband that she was at the clinic and t ested. So if in the first visit she was tested and found negative, she needs to tell her husband that. Then they are supposed to do another blood test at the clinic. So [let's say] they test her and she doesn't have the disease. The next time she tests neg ative and then after that, the next time she tests positive. And so then they call the man and tell him that he should come in on such such a day [to test]. So you the man, you follow not of your wife, but the one who sent Lucy to tell the husband to come. So it is up to that man to accept or refuse. So Lucy has no rule [authority], but the decision has come from the person who has sent Lucy [VCT counselor or medical professional]. The other men in the group agreed with this man's asses sment and offered u p other strategies to increase men 's rates of testing via antenatal care: Man #5: Indeed, maybe the person can feel uncomfortable to go but if the doctor finds a way of calling him, [he might agree to test]. Man #7: He can't be uncomfortable.

PAGE 262

247 Man #6: But for her to just say that she has been found positive, now he already finds [knows] the answer [his status]. So why would the person go? The doctor should write something that he is required [to come]. "There is something that we should discuss. But com e with your wife." Man #5: But now when they go to the clinic they want the man. So if he doesn't go, they send the woman back [without receiving the antenatal care she is supposed to receive at that appointment]. When they send her back, he knows he nee ds to go so that she can be weighed. They should test you together with the woman. Man #1: Now, when the woman is expectant [pregnant], they can tell her when she comes back to please come with your husband so that we should also test him. (FGD#4, marrie d men) These men's perceptions of antenatal care testing were corroborated by real experiences as documented in the semi structured interviews. Several men were tested after doctors recommended to their pregnant wives that they bring in their partners fo r testing. In three different semi structured interview exc e r pt s with married women, Mary, Chisomo, and Rose recalled how their partners agreed to get tested during their pregnancies after receiving advice from health care providers. Perhaps this acted as a motivating factor for men who might have otherwise declined Mary: We talked about it before we had our baby, but he was refusing to go. He was afraid and not only him, I was also afraid. When I was pregnant, I was advised to get tested but I was afraid so I didn't until the baby was born. Then we were living but we knew that we were supposed to go for HIV testing. That time I had a long term cough and flu. So I went to the hospital to get tested for TB but it wasn't. Then we just decided to go for HIV te sting and when we went, they found me HIV positive and so was my husband. (semi structured interview, female #4) Chisomo: Yes, we went for testing. We talked. At that time we went because I was pregnant with our first child so I explained to him that we s hould go for testing. I went and I am HIV negative and so they told me to tell my husband to go for testing as well. So when I told him that he has to go for testing, he agreed. Lucky enough at his workplace there were people who test blood, so he just got tested there and came home with the results and he was found negative. (semi structured interview, female #5) Rose: Yes, when I was pregnant I was told to go for HIV testing so I went and I was found negative. When I came home I told my husband to go and get tested as well, at first he denied, but after encouraging him he went and was found negative too. (semi structured interview, female #6) In another semi structured interview, a married man named Steven explained how he perceived HIV testing to be com pulsory during pre gnancy and per the doctor's authority, he

PAGE 263

248 consented to testing: Interviewer: On HIV/AIDS, have you ever discussed about it? Steven: Yeah, this is a difficult issue and not a difficult issue. When we felt that we were trusting each othe r we discussed it. I bore two children, the issue now came during the pregnancy with this child, the third one. The doctors told her that whether you like it or not you have to be tes ted for HIV. We discussed it, "O k if the doctors say that you need to do it, we have to" since they gave us motivation. She was tested and found negative. We have a VCT centre but we haven't gone there for testing. So when my partner came with the results [from antenatal testing], I also went and got tested. I was negative. (se mi structured interview, male #5) Succumbing to the advice of health care professionals who make testing difficult to refuse instead of their wives allows men to maintain their masculinity and household power. As noted in Chapter 4, men who obey ed every c ommand of their wives could be seen as "been given a love potion" and looked down upon by their peers. Of course, both antenatal testing and couples testing carry their own set of barriers and costs. In an effort to avoid HIV infection, sometimes more dra matic strategies are required including the option of leaving a dangerous partner altogether (refer to Schatz 2005 for other examples; Reniers 2005) Relationship Dissolution as a Fallback Strategy Divorce, separation, and breakup are not ideal alternatives. But i f all else fails, it may be an effective strategy to avoid HIV infection in the absence of testing. In the southern region of Malawi, divorce rates are among the highest in the country with around 33% of all marriages ending before their 5 th year anniversa ry (Reniers 2003) In her article entitled, "Take your mat and go!", Schatz (2005) discusses how women wi ll leave partners who refuse to reform their sexual behavior and threaten to bring HIV into the household. The exchange of meaning through interactions with a spouse or partner around HIV testing may provide another piece of information to use in decisions to leave a troublesome or risky partnership.

PAGE 264

249 According to both men and women from five f ocus groups, refusing to test was grounds for terminating the chibwenzi and possibly the marriage because of its association with promiscuity. Many people believed th at a person with nothing to hide would openly agree to go for testing. As previously mentioned, the refusal to test signifies se xual immorality: a trait that was highly undesirable in a potential spouse Married people were not as worried about a partner's sexual past as they were about their partner's current sexual indiscretions and the danger these actions present ed to the family's wellbeing. Overall, these beliefs about dissolution emerged in every type of focus group (women, men, married, dating). In t he following passage, a single man commented on how relationships dissolve in his village when a partner refuse s to test: In the village where I am living, when people are dating they talk to each other about HIV testing and go for testing. In this case we re by one is afraid of going for testing, the relationship ends there. Because it is taken as though she does not have enough trust on his/her partner. (FGD #3, single men) In another conversation among single women, everyone agreed that if a man refuse d to test with his girlfriend, the relationship would end immediately since he was most likely concealing something about his sexual past. Interviewer: That woman [another focus group respondent] has said that the two should go for testing. They have just started dating, the two should take each other and go for HIV/AIDS testing to see or know about their status. That is what she is saying. What if the man refuses? All: It is going to end there. Man #2: That just shows that there is a reason and that i s why the man is refusing. (FGD #5, single women) One married man told the story of how men set "traps" to catch women in cheating by asking her to test and t hen divorcing her if she refused because of her presumed unfaithfulness. He said, "When the man s ays we should go for testing, maybe the woman

PAGE 265

250 refuses. But the refusal of the woman means the man has set a trap that if she refuses, I should [will] tell her that the marriage has ended." (FGD #6, married men) It is unclear from the data, however, about w hether the same strategy is common among women with their unfaithful partners. Furthermore, w hile breaking up with a chibwenzi who refuses to test may be feasible, divorcing over a partner's unknown HIV status and presumed infidelity may be significantly m ore complicated especially when children, land, and other marital assets are involved. For those who do, it is plausible that the dec ision to divorce over testing was just one piece of evidence in a couple's larger story around infidelity and HIV infection Discussion Rural Malawi is a setting where HIV risk is at constant odds with other important life and reproductive aspirations. What I presented in this chapter reinforced the notion that social responses to HIV testing programs go beyond perceptions o f risk and are strongly influenced by factors at the couple level such as communication, love and trust, and what an HIV test means for the future of the relationship. I argued that for these young couples who are just beginning their sexual and reproducti ve lives, an HIV test is more than a marker of HIV serostatus; it also symbolizes the relationship status. Similar to what others have observed for condoms, HIV testing may violate core ideals around love and trust that couples strive for in their relation ships (Chimbiri 2007; Tavory and Swidler 2009) Bringing up testing. Refusing to test. Testing together as a couple. These testing behaviors provide a few examples of the symbolic acts that rural Malawians use to evaluate their current circumstances and react accordingly. As the data in this chapter illustrate,

PAGE 266

251 Ma lawians do not dismiss the benefits of HIV testing altogether nor do they embrace it wholeheartedly. Instead, they actively respond ed to the HIV testing enterprise and create d meaning around the act of testing that fits with their local realities. As they interact ed with each other, rural Malawians utilize d the local symbols and signs associated with the act of HIV testing as tools to simultaneously navigate the AIDS epidemic and their futures. These conclusions are consistent with the association found in Chapter 5 using the quantitative TLT couples dataset. Couples with higher levels of unity were less likely to test for HIV as compared to others with lower levels of unity in their relationships because they see less reason to do so. The narratives in the current chapter also suggest the presence of another possible association between mist rust, specifically the perception t hat a partner might be cheating and increased rates of testing. Surprisingly, the quantitative analysis failed to find a strong assoc iation. Mistrust may still be a contributing factor but could be less important as compared to the other factors in the model. The social construction of gender is continually evolving in Malawi as notions of rights unity, and tradition are contested a t the local level. As social norms around gender relationships, and HIV testing take on new shapes, rural Malawians are forced to grapple with how to make sense of this for their everyday lives. Therefore, a decision to test may be best understood as a pr ocess that involves constant evaluation of one's circumstances as the world changes rather than a n isolated event that occurs at one point in time. The process of creating new understandings was not straightforward either. Often respondents invoked these three narratives in a contradictory fashion; for example, respondents believed that

PAGE 267

252 people have the legal right learn their status at individuals, but should still follow cultural traditions and embrace unity with their loved ones. Men and women had to the refore walk a fine line as they attempt ed to balance imported western values around human rights with the preservation of local values and beliefs. One particular ly interesti ng finding was the overemphasis on the role of unity in testing decisions for mar ried couples as opposed to single people. Why were respondents so adam ant that married couples test together? In her study of dual income earning families in Los Angeles, Linda Garro (2011) arg ues that studying hea lth as a part of family life reveals the centrality of a socially embedded and relational view of health as "family well being" that coexists with health construed as an individual level concern. Regardless of which partner is the brea dwinner and homemaker in Malawian marriages, both spouses depend on each other just in different ways in order to meet their life aspirations and daily needs This family or unity orientation carried over into their perceptions about how HIV testing should be undertaken Among respondents, there was an overwhelming need to consider the broader couple context rather than just the individual when it came to HIV testing yet this was not without its challenges. Indeed, there were times when unity could not win over a reluctant partner who was unwilling to test. Here, rural Malawians suggested the use of alternative strategies to navigate their risk for HIV while maintaining the very relationships so important to people's livelihoods.

PAGE 268

253 Public Health Implication s The narratives covered in this chapter suggest that there are important implications for public health related to the disconnect ion between the meaning of "risk" at the policy level and what it means to rural Malawians. Epidemiologists define risk as "t he probability of a disease occurring in people exposed compare d to the probability of the disease in non exposed people." (Gordis 2009) Rural Malawians are well aware that the exposed condition referred to here is relate d to "immoral sexual behavior" but this is something that many people are not willing to own up to by having an HIV test. Early on in the epidemic when testing programs were rolled out in rural Malawi, prevention messages and mass media campaigns may have inadvertently undermined their own initiatives by emphasizing the need for testing among people deemed to be a "high risk" for HIV, notably, commercial sex workers. Among rural Malawians, there are still strong attachments to the idea that HIV testing is o nly for people who exhibit these behaviors. Today, however, Malawians are targeted with public VCT advertisements containing vaguely worded phrases like "know your status" (I emphasize the "your" here) in an attempt to de stigmatize the act of testing. Ind irectly, these messages are appealing to the human right to know one's HIV status. But t his is problematic for several reasons. First, as respondents in this study emphasized through their perceptions and experiences, HIV testing is not always considered an indiv idual level decision, especially during marriage. For some people, it was difficult to disentangle the individual from the married couple There is an overwhelming need to

PAGE 269

254 consider the couple context when targeting the average individual and their personal liberties and freedoms. Second, western ideas of routine screening for disease are likely to conflict with local health beliefs shaped by the current health care system in Malawi. In the US, for example, there is strong public health advocacy aro und disease prevention and health promotion which is based on the argument that it is more clinically and cost effective to treat infection earlier in the disease course In settings with good access to screening and treatment services people are more li kely to see the value in prevention and early detection of disease In Malawi, however, people require d strong reason s to initia te HIV testing when they did not perceive that there were problems HIV testing can be expensive, time consuming, and requires m uch planning and calculation to ensure that the benefits outweigh the costs (Beardsell and Coyle 1996; Fylkesnes et al. 1999) Furthermore, AIDS related concerns may simply not make the list of top priorities (Dionne, Gerland, and Watkins 2013) During the time when VCT was th e primary diagnostic approach Malawians often delayed learning their HIV status until they were very sick (Hatchett et al. 2004) Globally, there have been increasing calls for universal access to effective HIV prevention, treatment, care, and support (UNAIDS, 2010 ) Specifically in Africa, Bunnell and Cheru tich (2008) argue for the urgent scale up of HIV testing with an emphasis on couples and a goal of universal coverage. In theory, t he provision of universal testing c ould circumvent an otherwise exceptio nally complex negotiation process fo r both women and men who desire to know their status Advances towards universal testing are currently

PAGE 270

255 underway in Malawi starting with the mandatory screening for HIV of pregnant women seeking antenatal car e (Office of the President and Cabinet and National AIDS Commissi on [Malawi] 2003) In addition a new policy in Malawi referred to as "Option B+ allows all pregnant women who test HIV positive to be placed on ART for life regardless of their CD4 counts or clinical stage (Schouten et al. 2011) Despite this progress wi despread access to testing and treatment in the general population especially among men is far from a reality Economic, logistical, and infrastructural limitations continue to imped e progress towards universal coverage of HIV and AIDS services in Malawi. Other public health implications of this chapter are noteworthy. Although rural Malawians generally possess high levels of knowledge of HIV/AIDS transmission (Watkins 2004) there are local nuances to these beliefs and understandings. Three additional insights gained from this chapter are relevant for public health and could be incorporated i nto new HIV testing media campaigns and programs, particularly: Perceptions of a shared serostatus among couples The belief that HIV status is an absolute indicator of marital fidelity Acceptance of antenatal care testing among men, when suggested by a hea lth professional The first two relate to risk. As described in Chapter 2, it is well documented that individuals use perceptions of their partner's risk of HIV to inform perceptions of themselves. Although, their assessments of each other are not always co rrect (Anglewicz and Kohler 2009) Additionally, many people believe a person can become infected through a single

PAGE 271

256 sexual encounter with an HIV positive individual; however, the likelihood of infection from one sex act is actually q uite low (Gray et al. 2001) In the current study, respondents also believed that married couples share a single corporal body at least in the symbolic sense. Giv en these beliefs together, it was not surprising to hear that people subscribe d to the idea of a shared serostatus among married couples who presumably have had unprotected sex with each other and regularly. Indeed, i n most couples who are s eroconcordant negative, they share a common HIV status. Ho wever, when one couple member tests HIV positive, it would be mistaken to assume that the other is HIV positive. The data presented in Chapter 5 showed that of those with actual HIV status data, 88.8% of respondents were sero concordant (HIV positive or HIV negative), but only 12.2% were serodiscordant. These findings are of course biased in the fact that most people tested HIV negative and thus most couples would be sero concordant by default. Yet in instances where at least one partner tested positive, couples were more likely to have different statuses than to both be positive (12.2% com pared to 5.5%). Similar to how people overestimate their chances of HIV infection, they were more likely to assume that if a spouse tested positive, it was because she or he became infected from an extramarital partner. In line with these beliefs, it woul d only make sense that HIV testing could be employed as a sure way to confirm a partner's infidelity and serostatus. Men knew they could win back a suspecting wife if they showed her a negative test result, as if saying, "I was faithful to you ." Along the same vein, husbands could refuse to go for

PAGE 272

257 testing if they were indeed cheating and assumed that a positive test result would seal their fate with their wives. Overall, these findings provide additional evidence for the broader understanding that local be liefs about risk and how risk is manifested at the couple level are important considerations for why people may accept or refuse to test. As others have argued (Anglewicz and Kohler 2009) it may be necessary to consider risk from t he couple rather than individual level when improving HIV testing regimens in sub Saharan Africa. In addition, counseling messages should work to clear up some of the more common misbeliefs about a shared serostatus and high transmission probability t hroug h extramarital sex HIV interventions should proceed carefully with more effective strategies to increase H IV status knowledge. There is currently an ongoing shift in Malawi and within the region more broadly from patient initiated testing to routine, pr ovider initiated testing in all health care clinics. Tremendous advances (at least from a public health standpoint) have been made in the antenatal care sector to increase rates of testing among pregnant women, although not always in accordance with women s and their partner's desires to test (Angotti, Dionne, and Gaydosh 2010) On one hand, antenatal care testing allows couples to circumvent power issues around rights and tradition in their own relationships by shifting authority to the health care system level. Yet significant pressures are placed upon women to test during pregnancy to the point where they often forgo their rights to test in order to protect the healt h of their unborn children While this is cer tainly not ideal, and in fact conflicts with the very notion of human rights, growing pressure for men to test through the

PAGE 273

258 antenatal care sector may help to curb growing gender disparities in diagnosis, care, and treatment.

PAGE 274

259 CHAPTER VII IMPLICATIONS AND F UTURE DIRECTIONS Theoretical and Policy Implications Bertrand Russell observed that "the fundamental concept in social science is power, in the same sense in which energy is the fundamental concept in physics. Like energy, power has many forms" (1938, as cited in Murphy and Meyer 1991) While it h as been widely acknowledged among scholars and public health practi ti oners alike that power is multidimensional and varies by locale, the leading discourse on gender inequality and HIV/AIDS in sub Saharan Africa continues to perpetuate a portrayal of overl y dominant men paired with subordinate women who have presumably little control over their lives. Beyond the influence of colonialism in the region, p olitics and a culture of altruism among western donors and policymakers may be partially to blame. Watkins and Swidler (2012) write that, "Western donors imagine women as poor and weak, victims who urgently need empowerment so that they can "just say no" to unsafe sex and thereby turn the tide of the epidemic in sub Sahar an Africa ." The data from this dissertation indicate that these depictions do not fully capture the nuances and complexity of power or instances of female agency and may have unintended consequences for men who fear being blamed as transmitters of HIV to their seemingly innocent wives. Conceptions of power in this dissertation move our understanding of gender relations in an entirely new direction away from relying solely on problematic stereotypes of gender towards the idea that powe r may be better unde rstood as an dynamic factor related to the

PAGE 275

260 couple itself. Using a unique combination of numbers and narratives, I found that the notion of unity embodied by aspects of reciprocity, communication, trust, and love emerged as a key construct in the measure of power for Malawian couples. Indeed, the statistics were supported by rural Malawians' own voices about what makes them feel powerful in their sexual relationships. I believe that unity as a source of power goes beyond the receipt of social support to inc lude intimacy a broad term that consists of fertility, love, marriage, and sexual pleasure (Hunter, 2010). Though the relationship unity measure does not capture all aspects of intimacy as defined here, future studies can take what was learned in this diss ertation to expand the unity construct more fully. The reconstruction of power as something that is inextricably tied to the union largely diverges with how others have typically conceptualized and measured power in this region using constructs such as ma le dominance, control, or decision making dominance. Perhaps research has been blinded by a western lens that fails to consider the nuances of gender and power relations in Africa. In an exception, Mbweza and colleagues (2008) found that men and women used a combination of gender based cultural scripts (e.g., "husbands are the head of the household") and non gender based scripts that emphasized harmony and open communication to explain how decisions are made in everyday life. In the current study, respo ndents invoked similar cultural scripts to emphasize men's legitimate authority to the household leader role (which I called tradition narratives) but when interviewers probed deeper about whether men could sell basic household possessions without their w ife's permission, they responded with "she will divorce me" thus, emphasizing the need for

PAGE 276

261 unity. This suggests that there are important lessons to be learned from the differences between how people talk about gender in terms of ideals and how these ideals may be actualized in everyday life. Truly grasping the complexity of gender relations may require an ethnographer or a persistent interviewer to be able to tear away the layers of ideals and socially desirable conceptions of gender to find out what happen s in real life. Of course, everyday reminders of the female vulnerability/male dominance paradigm still reverberate across the Malawian gender landscape. Respondents told stories of men beating up their wives or secretly engaging in an extramarital love af fair. Respondents also suggested the presence of many double standards around sex, infidelity, and autonomy that granted men privilege, but punished women for the very same behaviors. I do not intend to dismiss the importance of these findings. Malawian wo men continue to face many difficulties negotiating the terms of their relationships and navigating societal boundaries around gender. But at the same time, respondents provided examples of "docile" men who had been given love potions by their wives or resp ectful men who considered their wives opinions in family matters. In contrast to other studies that illustrate a more absolute or clear cut gender hierarchy, the qualitative data from this dissertation paint a very different picture; one marked by shades o f gray, uncertainty, and a state of flux rather than fixation. Through the use of different narratives around tradition, unity, and rights, rural Malawians actively debated with each other in a public forum about how men and women should behave. Through th is process, there was much contestation and contradiction around the meaning of gender. Thus, I believe that a "one size fits all" paradigm rooted in either

PAGE 277

262 extreme position that women have no agency or that women have unconstrained agency is problematic g iven that gender relations are more likely to fall somewhere along the continuu m at any given time. Final Conceptual Model for Relationship Power and HIV Testing I would like t o briefly summarize the model building activities in this dissertation. Ground ed theory was used not only as an approach for the qualitative phases, but as an overall theory building method for this dissertation. First, I started with a preliminary model based on three social structures of the TGP: the sexual division of labor, the sexual division of power, and social norms or cathexis. Three constructs comprised the preliminary model of power: s ocioeconomic inequalities (division of labor), relationship violence (division of power), and relationship dominance (division of power). Af ter running factor analyse s two additional aspects of power emerged : unity and discordance. The discordance factor was later deconstructed into separate measures of relationship violence (which I had already included) and mistrust/infidelity (new construc t). In Chapter 4, I used grounded th eory to inductively study aspects of relationship power not accounted for by the TGP. Th rough this process rural Malawians illustrated that unity was an important component o f relationship power T he measure of power no w consisted of socio economic inequalities (division of labor), relationship violence (division of power), relationship dominance (division of power), unity (cathexis), and infidelity (division of power). In Chapter 6, I again used grounded theory with qua litative data to inductively study how relationship factors influence HIV testing behavior. This process confirmed the importance of the constructs of unity and

PAGE 278

263 infidelity (trust) in shaping HIV testing behavior. Future model testing activities could colle ct new measures related to unity to further unpack its underlying characteristics, which I believe includes aspects of love, intimacy, reciprocity /social support and communication. The final conceptual model that could be tested is presented in Figure 7.1 Figure 7.1: Final Conceptual Model on Relationship Power and HIV Testing Behavior M ixed support was found for the Theory of Gender and Power (TGP) in explaining HIV testing behavior (as indicated in Table 7.1). An important limitation of t he TGP is that it takes on a western feminist perspective that emphasizes women's vulnerability to HIV/AIDS as a result of patriarchal, male dominant social structures. There was some !"#$%&'()'*+,'-&./01 2&34&56&7'85.%'()'*+,'' 9.&:)'$07'"$3#0&3; <5.4:(.=3&'()'*+,'-&.#'8&.=:#.' #('235>$3?'2$3#0&3 !"#$%!&'(')*) !"#$%!&'(')+) 2$3#0&3'*+,'.#$#=.' 4(04(37$04&' 8&:$/(0.@5"'' 2(A&3' 8&:$/(0.@5"' ,5(:&04& 8&:$/(0.@5"' <(>50$04& B5.#3=.#C2$3#0&3' 50D7&:5#? E4(0(>54' 50&F=$:5#?' !05#? <5.4(37$04& <5.4(37$04&' .=G.4$:&'3&H )(3>=:$#&7'$.' 65(:&04&'$07' 50D7&:5#?' I(6&C+0/>$4? J(>>=054$/(0 8&45"3(45#?C .(45$:'.=""(3#

PAGE 279

264 support found for the e conomic dependence theory, that is, women with le ss economic power in the relationship would be less likely to test. For men, more education as compared to a partner resulted in a higher likelihood of testing. Economic factors related to division of labor were less imp ortant for disclosure. Although for men, being in an economic ally balanced relationship (a dual income earning relationship) increased the likelihood of men's disclosure perhaps by creating more unity in the relationship around labor. T he findings do strongly support the role of relationsh ip unity in HIV testing behavior. Couples based interventions that target everyday relationship factors may provide more immediate results as compared to longer term, economic ally based approaches, especi ally for couples who just need to learn the right ne gotiation and communication skills. Findings from this dissertation also highlight the continued need for gender based violence mitigation among both men and women, which should be incorporated i nto any unity building exercise Couples with high levels of discord who are not already protected by unity may need customized attention coupled with othe r structural approaches that address the root problem. It is also interesting that male dominance actually facilitated HIV testing among women. If this is true, then we need to be cautious about assuming that traditional gender based power structures are inherently problematic for women and men. There may be aspects of male authority and cultural legitimacy that are advantageous when it comes to health seeking beh avior. Future programs should harness the positive aspects of masculinity to encourage

PAGE 280

265 the adoption of health behaviors (also noted by Mankowski and Maton 2010) and not necessarily attempt to change them to fit with western notions of rights and equality Rural Malawians may be more likely to voluntarily test for HIV when they perceive problems in the relationship than when there is harmony. This is a good thing; we want people who are deemed to be at higher risk to be more likely to test. At the same time, the fact that decisions to test are based on subjective perceptions which are highly imperfect means that people are likely to fall through the cracks and perhaps present themselves at health clinics only after the y have started to exhibit the physical signs and symptoms of AIDS. Provider initiated testing offers one promising solution to circumvent unreliable risk assessments if it can be done in a way that seriously considers the couple context. Additional traini ng of clinicians and HTC counselors may be warranted in order to increase their sensitivity to issues around trust, intimacy, and the aftermath of status disclosure and to encourage couples' HIV testing In instances when couples' te sting is not easily ac cessible, strategies that continue to increase rates of testing among women antenatal care or health services without applying equal efforts to men c ould create new gender disparities and ethical issues for women. Thus, partner recruitment will continue to be an essential approach to target men th rough the pathway of their wives.

PAGE 281

266 Table 7.1: Theoretical assertions and supporting evidence found in quantitative data Number Prediction and TGP construct Finding Hypothesis #1: HIV testing uptake 1A Socio ec onomic disadvantage deters individuals from testing (division of labor) Some support Women: being significantly younger than a partner limit s uptake of testing ; being unemployed with an employed partner limits uptake of testing. Men: Being the older partn er limits uptake of testing; being the more educated partner facilitates uptake of testing 1B Male dominance prevents individuals from testing (division of power) Support found among men and women, but in the opposite direction of hypothesis (male domina nce, more testing) 1C Physical and sexual violence prevents individuals from testing out of fear (division of power) Physical abuse: no support found for men or women. Sexual abuse: Men who were sexually coerced were less likely to test. Women were more likely to test. 1D Relationship unity create s a supportive environment for testing (cathexis) Support found among men and women but in opposite direction of hypothesis (more unity, less testing) 1E Mistrust/perceived infidelity limits uptake of testing (division of power) No support found for men or women. 1F Perceived risk for HIV (self and partner) may be a barrier to testing Some support for men and women; higher risk, less testing. Hypothesis #2: HIV testing disclosure 2A Socio economic disadv antage deters individuals from disclosing test results (division of labor) Some support for men. Men in dual income earner relationships were more likely to disclose. 2B Male dominance prevents individuals from disclosing (division of power) No support f ound for men or women.

PAGE 282

267 Table 7.1, continued Number Prediction and TGP construct Finding 2C Physical and sexual violence prevents individuals from disclosing out of fear of the consequences (division of power) Physical abuse: Barrier to disclosure amo ng men and women. Sexual abuse: Men who were sexually coerced were less likely to disclose. 2D Relationship unity creates a supportive environment for disclosure (cathexis) Support found for both men and women; more unity, more disclosure. 2E Positive i ndividuals will be less likely to disclose if they perceive their partners to be HIV negative Support found for men and women.

PAGE 283

268 While HIV testing during marriage was deemed problematic, testing during the courtship phase of the relationship was mentioned as an important relationship ideal. On one hand, this finding opens up possibilities for new HIV testing and counseling programs. Youth and couple friendly services should continue to target young people before they get married as they are negotiating the terms of their relationships. While at the same time, it is important to cautiously emphasize to young people in premarital partnerships that a one time test is not enough to prevent HIV. Future Directions The findings of this dissertation provide a sol id foundation for future studies that seek to understand how the relationship context shapes HIV testing behavior. The next phase would be to build and test a more inclusive theoretical model using the unity construct as the foundation. Given that relation ship unity was somewhat of an unexpected finding from both the quantitative and qualitative data analysis, a logical next step would be to develop and collect new data related to unity, particularly, trust, intimacy, love, couple collaboration and reciproc ity, sexual pleasure, respect and other constructs that emerged as salient sources of power in the qualitative sources. Taking a strong gender centered approach would help to ensure that the model held for both men and women. Similarly, there are likely to be better ways to capture the idea of marital discord and its connection to infidelity, punishment, and violence using multi item scales instead of single item measures. Validated scales exist for many of these constructs and their use would first require adaptation to the African context. A small pilot study among young couples using these existing scales for unity and discord

PAGE 284

269 could yield a fully tested and validated model that could later be applied to larger study populations. This model could then be used to develop a couple based intervention that harnesses the unity in primary partnerships in order to increase rates of individual or couples testing for HIV. Several other areas of research could arise from this dissertation. First, more exploration is needed of repeat testers and how the relationship context influences regular use of HIV testing services. Second, it would be useful to examine how changing relationship patterns or characteristics over the course of the relationship motivates or constr ains decisions to test for HIV. Individuals change over time, but so do their relationships. In this study, I used relationship factors mostly at one point in time to predict future testing, however, many of these factors are likely to shift over time as w ell Second, this study did not explore the depth of people's HIV testing hi story. For example, in the semi structured couple interviews, respondents were mostly asked about testing in their current relationship. Qualitative research using life history met hodologies could yield additional individual, relationship, and structural patterns of influence on HIV testing behavior as they evolve over the life cycle Third, it would be novel to examine how relationship factors in secondary relationships, e.g., with an extramarital partner, influence decisions to test. Multiple concurrent partnerships was a topic of great concern, however, respondents were not asked to provide details on these relationships despite the fact that they are likely to influence HIV testi ng behavior within the primary relationship.

PAGE 285

270 Conclusions In western settings such as in the United States, the public health approach to HIV/AIDS emphasizes prevention and early detection of disease. Health promoting and health seeking behaviors only oc cur if individuals are motivated to do so. As such, public health interventions for HIV/AIDS rely heavily on individual responsibility and comm unity engagement in order for them to be effective. With client initiated HIV testing in Malawi, the underlying a ssumptions are largely similar. But what if Malawians simply do not see the value in HIV prevention and treatment services? In a recent study in rural Malawi, Dionne and colleagues (2013) note that although AIDS has been treated as "exceptional" and a priority issue over other d evelopment problems, Malawians perceived other problems in their communities as being more pressing matters. Even people who were HIV positive prioritized resource allocation to clean water, agricultural development, and general health services over HIV/AI DS services. As part of the same study, the authors also used a set of ethnographic journals that captured people's everyday conversations and found that economic survival was the most frequently discussed topic. Conversations about "domestic matters" such as marriage, family, and children also surpassed the number of conversations about HIV/AIDS. These data combined with the findings of this dissertation suggest that people are forced to make decisions about HIV testing in light of other competing priorit ies related to marriage, childbearing, and basic economic survival. HIV testing policy and programs cannot divorce individuals from their broader social context simply for the sake of increasing the

PAGE 286

271 numbers of testers and decreasing the rates of transmissi on. Pushes towards the human "right" to know one's status will only be meaningful if the social environment supports it. Otherwise, it exists only in the imaginations of western policy makers with grandiose visions of how HIV/AIDS services should function in sub Saharan Africa. Approaches that emphasize unity over rights may elicit a better response at the local level given the importance of this relationship ideal and source of power for Malawians themselves.

PAGE 287

272 REFERENCES Adock, A.A. 2006. A gentle introduction to Stata College Station, TX: Stata Press. Allison, P. 2004. Event History Analysis. In Handbook of data analysis edited by M. Hardy and A. Bryman. London: Sage, 369 385. Anglewicz, P. 2008. Disclosure of HIV status among rura l Malawians. Paper presented at the Annual Meeting of the Population Association of America, 17 19 April 2008, New Orleans, LA Anglewicz, P., and J. Chintsanya. 2011. Disclosure of HIV status between spouses in rural Malawi. AIDS Care 23 (8):998 1005. Ang lewicz, P., and H.P. Kohler. 2009. Overestimating HIV infection: The construction and accuracy of subjective probabilities of HIV infection in rural Malawi. Demographic Research 20 (6):65 96. Anglewicz, P.A., S. Bignami Van Assche, S. Clark, and J. Mkandaw ire. 2008. HIV risk among currently married couples in rural Malawi: What do spouses know about each other? AIDS and Behavior 14 (1):103 112. Angotti, N., A. Bula, L. Gaydosh, E.Z. Kimchi, R.L. Thornton, and S.E. Yeatman. 2009. Increasing the acceptability of HIV counseling and testing with three C's: Convenience, condentiality and credibility. Social Science & Medicine 68 (12):2263 2270. Angotti, N., K.Y. Dionne, and L. Gaydosh. 2010. An offer you can't refuse? Provider initiated HIV testing in antenatal clinics in rural Malawi. Health Policy and Planning 26 (4):307 315. Antelman, G., M.C.S. Fawzi, and S. Kaaya. 2001. Predictors of HIV 1 serostatus disclosure: A prospective study among HIV infected pregnant women in Dar es Salaam, Tanzania. AIDS 15:1865 1874. Ante lman, G., M.C.S. Fawzi, S. Kaaya, J. Mbwambo, G.I. Msamanga, D.J. Hunter, and W.W. Fawzi. 2001. Predictors of HIV 1 serostatus disclosure: a prospective study among HIV infected pregnant women in Dar es Salaam, Tanzania. AIDS 15:1865 1874. Baiden, F., P. R emes, R. Baiden, J. Williams, A. Hodgson, M. Boelaert, and A. Buve. 2005. Voluntary counseling and HIV testing for pregnant women in the Kassena Nankana district of northern Ghana: Is couple counseling the way forward? AIDS Care 17 (5):648 657.

PAGE 288

273 Beardsell, S., and A. Coyle. 1996. A review of research on the nature and quality of HIV testing services: a proposal for process based studies. Social Science & Medicine 42 (5):733 743. Becker, M.H. 1974. The health belief model and personal health behavior. Health Education Monographs 2:324 473. Becker, S., R. Mlay, H.M. Schwandt, and E. Lyamuya. 2010. Comparing Couples' and Individual Voluntary Counseling and Testing for HIV at Antenatal Clinics in Tanzania: A Randomized Trial. AIDS and Behavior 14 (558 566). Bened ict, R. 1959. Patterns of culture New York, NY: Houghton Mifflin [originally published in 1934]. Berendes, S., and R.N. Rimal. 2011. Addressing the slow uptake of HIV Testing in Malawi: The role of stigma, self efficacy, and knowledge in the Malawi BRIDGE Project. Journal of the Association of Nurses in AIDS Care 22 (3):215 228. Blanc, A. 2001. The effect of power in sexual relationships on sexual and reproductive health: An examination of the evidence Studies in Family Planning 32 (3):189 213. Blanc, A., B. Wolff, A.J. Gage, A.C. Ezeh, S. Neema, and J. Ssekamatte Ssebuliba. 1996. Negotiating reproductive outcomes in Uganda. Calverton, MD: Institute of Statistics and Applied Economics (Uganda) and Macro International. Blood, R.O., and D.M. Wolfe. 1960. Husb ands and wives: The dynamics of daily living Glencoe, IL: Free Press. Blumer, H. 1969. Symbolic Interactionism: Perspective and Method Englewood Cliffs, NJ: Prentice Hall. Bradley, H., A. Tsui, A. Kidanu, and D. Gillespie. 2011. Client Characteristics an d HIV Risk Associated with Repeat HIV Testing Among Women in Ethiopia. AIDS and Behavior 15:725 733. Brown, P., and S. Levinson. 1987. Politeness: Some universals in language use Cambridge, UK: Cambridge University Press. Bunnell, R., and P. Cherutich. 20 08. Universal HIV testing and counselling in Africa. The Lancet 371 (9631):2148 2150. Butler, J. 2003. Performance Acts and Gender Constitution: An Essay in Phenomenology and Feminist Standpoint. In Feminist Theory Reader: Local and Global Perspectives ed ited by C. R. McCann, and Kim Seung Kyung. New York: Routledge, 415 427.

PAGE 289

274 Bwambale, F.M., S.N. Ssali, S. Byaruhanga, J.N. Kalyango, and C.A.S. Karamagi. 2008. Voluntary HIV counselling and testing among men in rural western Uganda: Implications for HIV prev ention. BMC Public Health 8 (263). Caldwell, J., P. Caldwell, and I. Orubuloye. 1992. The family and sexual networking in sub Saharan Africa: Historical regional differences and present day implications. Population Studies 46:385 410. Campbell, C. 2003. 'L etting them die' Why HIV/AIDS prevention programmes fail Oxford: The International African Institute. Carpenter, L.M., A. Kamali, A. Ruberantwari, S.S. Malamba, and J.A.G. Whitworth. 1999. Rates of HIV 1 transmission within marriage in rural Uganda in rel ation to the HIV sero status of the partners. AIDS 13 (9):1083 1089. CDC. 2006. Revised recommendations for HIV testing of adults, adolescents, and pregnant women in health care settings. MMWR 55:1 17. Champion, V.L., and C.S. Skinner. 2008. The Health Bel ief Model Edited by K. Glanz, B. K. Rimer and K. Viswanath. 4th ed, Health Behavior and Health Education San Francisco: Jossey Bass. Chimbiri, A.M. 2007. The condom is an 'intruder' in marriage: Evidence from rural Malawi. Social Science and Medicine 64: 1102 1115. Clark, S., M. Poulin, and H.P. Kohler. 2009. Marital Aspirations, Sexual Behaviors, and HIV/AIDS in Rural Malawi. Journal of Marriage and the Family 71 (2):396 416. Cohen, C.R. 1998. Statistical power analysis for the behavioral sciences (Rev. e d.) Hillsdale, NJ: Erlbaum. Connell, R.W. 1987. Gender and power Stanford, CA: Stanford University Press. Conroy, A. 2010. Adapting the sexual relationship power scale for HIV/AIDS research in rural Malawi. Paper presented at the Annual Meeting of the Po pulation Association of America, 15 17 April 2010, Dallas, TX. Creel, A.H., and R.N. Rimal. 2012. Factors related to HIV testing behavior and interest in testing in Namibia. AIDS Care 23 (7):901 907. Cromwell, R.E., and D.H. Olson. 1975. Power in families New York: Sage Publications. Cronbach, L., and P. Meehl. 1955. Construct validity in psychological tests. Psychological Bulletin 52:281 302.

PAGE 290

275 Dagne, G.A., G.W. Howe, C.H. Brown, and B.O. Muthen. 2002. Hiercharical modeling of sequential behavioral data: An empirical Bayesian approach. Psychological Methods 7:262 280. Dahl, V., L. Mellhammar, F. Bajunirwe, and P. Bjorkman. 2008. Acceptance of HIV testing among women attending antenatal care in south western Uganda: Risk factors and reasons for test refusal. AIDS Care 20:746 752. deBeauvoir, S. 1989. Introduction. In The Second Sex edited by H. M. Parshley. New York: Vintage Books, xix xxxv. deGraft Johnson, J., V. Paz Soldan, A. Kasote, and A. Tsui. 2005. HIV Voluntary Counseling and Testing Service Preferen ces in a Rural Malawi Population. AIDS and Behavior 9 (4):475 484. Denison, J.A., K.R. O'Reilly, G.P. Schmid, C.E. Kennedy, and M.D. Sweat. 2008. HIV voluntary counseling and testing and behavioral risk reduction in developing countries: A meta analysis, 1 990 2005. AIDS Behavior 12:363 373. Desgrees Du Lou, A. 2005. The couple and HIV/AIDS in sub Saharan Africa: Telling the partner, sexual activity, and childbearing. Population 60 (3):179 198. Desgrees du Lou, A., H. Brou, A.T. Traore, G. Djohan, R. Becquet and L. Valeriane. 2009. From prenatal HIV testing of the mother to prevention of sexual HIV transmission within the couple. Social Science & Medicine 69:892 899. deZoysa, I., M.D. Sweat, and J.A. Denison. 1996. Faithful but fearful: Reducing HIV transmis sion in stable relationships. AIDS S10A:S197 S203. Dionne, K.Y., P. Gerland, and S. Watkins. 2013. AIDS Exceptionalism: Another Constituency Heard From. AIDS and Behavior 17:825 831. Dunkle, K.L., R. Jewkes, M. Nduna, N. Jama, J. Levin, Y. Sikweyiya, and M .P. Koss. 2007. Transactional sex with causal and main partners among young South African men in the rural Eastern Cape: Prevalence, predictors, and associations with gender based violence. Social Science & Medicine 65:1235 1248. Dunkle, K.L., R.K. Jewkes, H.C. Brown, G.E. Gray, J.A. McIntyre, and S.D. Harlow. 2004. Gender based violence, relationship power, and risk of HIV infection in women attending antenatal clinics in South Africa. The Lancet 363:1415 21. 2004. Transactional sex among women in Sow eto, South Africa: Prevalence, risk factors and association with HIV infection. Social Science & Medicine 59:1581 1592.

PAGE 291

276 Dunkle, K.L., R. Stephenson, E. Karita, E. Chomba, K. Kayitenkore, C. Vwalika, L. Greenberg, and S. Allen. 2008. New heterosexually tran smitted HIV infections in married or cohabiting couples in urban Zambia and Rwanda: an analysis of survey and clinical data. Lancet 371:2183 2191. Emerson, R. 1981. Social exchange theory. In Social psychology: Sociological perspectives edited by M. Rosen berg and R. H. Turner. New York: Basic Books, 30 65. Epstein, H. 2007. The invisible cure: Africa, the west, and the fight against AIDS New York: Farrar, Straus and Giroux. Esacove, A.W. 2013. Good sex/bad sex: the individualised focus of US HIV prevention policy in sub Saharan Africa, 1995 2005. Sociology of Health and Illness 35 (1):33 48. Farquhar, C., J.N. Kiarie, B.A. Richardson, M.N. Kabura, F.N. John, R.W. Nduati, D.A. Mbori Ngacha, and G.C. John Stewart. 2004. Antenatal Couple Counseling Increases U ptake of Interventions to Prevent HIV 1 Transmission. JAIDS Journal of Acquired Immune Deficiency Syndromes 37:1620 1626. Farquhar, C., D. Ngacha, R. Bosire, R. Nduati, J. Kreiss, and G. John. 2000. Prevalence and correlates of partner notification regardi ng HIV 1 in an antenatal setting in Nairobi, Kenya. Paper presented at the XIII International AIDS Conference, July 9 14 2000, Durban, South Africa. Fiorillo, S.P., K.Z. Landman, A.C. Tribble, A. Mtalo, D.K. Itemba, J. Ostermann, N.M. Thielman, and J.A. Cr ump. 2012. Changes in HIV risk behavior and seroincidence among clients presenting for repeat HIV counseling and testing in Moshi, Tanzania. AIDS Care 24 (10):1264 1271. Foa, E., and U. Foa. 1980. Resource theory: Interpersonal behavior as exchange. In Soc ial Exchange: Advances in Theory and Research edited by K. Gergen, M. Greenberg and R. Willis. New York: Plenum Press, 77 94. Frye, M., J. Trinitapoli, and H. Namadingo. 2011. Sexual Relationship Scripts: Adapting a Card Sort Technique for Use with a Semi literate Population in a Developing Country. Paper presented at the Annual Meeting of the Population Association of America, March 31 April 2 2011, Washington DC. Fylkesnes, K., A. Haworth, C. Rosenvard, and P.M. Kwapa. 1999. HIV counselling and testing: o veremphasizing high acceptance rates, a threat to confidentiality and the right not to know. AIDS 13:2469 2474.

PAGE 292

277 Gage, A.J., and D. Ali. 2005. Factors associated with self reported HIV testing among men in Uganda. AIDS Care 17 (2):153 165. Garro, L.C. 2011. Enacting Ethos, Enacting Health: Realizing Health in the Everyday Life of a California Family of Mexican Descent. Journal of the Society for Psychological Anthropology 39 (3):300 330. Gerson, K. 2010. The unfinished revolution: How a new generation is res haping family, work, and gender in America Oxford: Oxford University Press. Gipson, J.D., C.J. Muntifering, F.K. Chauwa, F. Taulo, A.O. Tsui, and M.J. Hindin. 2010. Assessing the Importance of Gender Roles in Couples' Home Based Sexual Health Services in Malawi. African Journal of Reproductive Health December 14 (4):63. Glaser, B.G. 1978. Theoretical sensitivity: Advances in the methodology of grounded theory Mill Valley, CA: Sociology Press. Glaser, B.G., and A.L. Strauss. 1967. The Discovery of Grounded Theory: Strategies for Qualitative Research Chicago: Aldine de Gruyter. Goffman, E. 1959. The presentation of self in everyday life New York: Doubleday. Gordis, L. 2009. Epidemiology (4th edition) Philadelphia: Saunders Elsevier. Gray, R., M. Wawer, R. Brookmeyer, N. Sewankambo, D. Serwadda, F. Wabwire Mangen, T. Lutalo, X. Li, T. vanCott, and T. Quinn. 2001. Probability of HIV 1 Transmission Per Coital Act in Monogamous, Heterosexual, HIV 1 Discordant Couples in Rakai, Uganda. The Lancet 357 (9263):11 49 53. Gupta, G.R. 2000. Gender, sexuality, and HIV/AIDS: The what, the why, and the how. Plenary address at the XII International AIDS Conference, July 9 14 2000, Durban, South Africa. Harries, A.D., S.D. Makombe, E. Libamba, and E.J. Schouten. 2011. Why did the scale up of HIV treatment work?: A case example from Malawi. JAIDS Journal of Acquired Immune Deficiency Syndromes 57:S64 S67. Harrison, A., L.F. O'Sullivan, S. Hoffman, C. Dolezal, and R. Morrell. 2006. Gender role and relationship norms among you ng adults in South Africa: Measuring the context of masculinity and HIV risk. Journal of Urban Health: Bulletin of the New York Academy of Medicine 83 (4):709 723. Hatcher, L. 1994. A step by step approach to using the SAS system for factor analysis and st ructural equation modeling Cary, NC: SAS Institute.

PAGE 293

278 Hatchett, L.A., C.P.N. Kaponda, C.N. Chihana, E. Chilemba, M. Nyando, A. Simwaka, and J. Levy. 2004. Health seeking patterns for AIDS in Malawi. AIDS Care 16:827 833. Hautzinger, S.J. 2007. Violence in t he city of women: Police and batterers in Bahia, Brazil Berkeley: University of California Press. Heise, L.L., and C. Elias. 1995. Transforming AIDS prevention to meet women's needs: A focus on developing countries. Social Science & Medicine 40 (7):931 94 3. Helleringer, S., H.P. Kohler, J.A. Frimpong, and J. Mkandawire. 2009. Increasing Uptake of HIV Testing and Counseling Among the Poorest in Sub Saharan Countries Through Home based service provision. JAIDS Journal of Acquired Immune Deficiency Syndromes 51 (2):185 193. Higgins, J.A., and J.S. Hirsch. 2008. Pleasure, power, and inequality: Incorporating sexuality into research on contraceptive use. American Journal of Public Health 98 (10):1803 1813. Higgins, J.A., S. Hoffman, and S.L. Dworkin. 2010. Rethi nking gender, heterosexual men, and women's vulnerability to HIV/AIDS. American Journal of Public Health 100 (3):435 442. Hirsch, J.S., H. Wardlow, D.J. Smith, H.M. Phinney, S. Parikh, and C.A. Nathanson. 2009. The secret: Love, marriage, and HIV Nashvill e, TN: Vanderbilt University Press. Hunter, M. 2002. The materiality of everyday sex: thinking beyond 'prostitution'. African Studies 61 (1). 2010. Love in the time of AIDS: Inequality, gender, and rights in South Africa Bloomington, Indiana: Indiana University Press. Hutchinson, P.L., and X. Mahlalela. 2006. Utilization of voluntary counseling and testing services in the Eastern Cape, South Africa. AIDS Care 18 (5):446 455. Irungu, T.K., P. Varkey, S. Cha, and J.M. Patterson. 2008. HIV voluntary coun selling and testing in Nakuru, Kenya: findings from a community survey. HIV Medicine 9:111 117. Isingo, R., A. Wringe, J. Todd, M. Urassa, D. Mbata, G. Maiseli, R. Manyalla, J. Changalucha, J. Mngara, E. Mwinuka, and B. Zaba. 2012. Trends in the uptake of voluntary counselling and testing for HIV in rural Tanzania in the context of the scale up of antiretroviral therapy. Tropical Medicine and International Health 17 (8):E15 E25.

PAGE 294

279 Izugbara, C.O., C.C. Undie, N.N. Mudege, and A.C. Ezeh. 2009. Male youth and Vo luntary Counseling and HIV Testing: the case of Malawi and Uganda. Sex Education: Sexuality, Society, and Learning 9:243 259. Jean, K., X. Anglaret, R. Moh, F. Lert, and R. Dray Spira. 2012. Barriers to HIV Testing in Cote d'Ivoire: The Role of Individual Characteristics and Testing Modalities. PLOS One 7 (7):e41353. Jeon, Y. H. 2004. The application of grounded theory and symbolic interactionism. Scand J Caring Sci 18:249 256. Jewkes, R. 2002. Intimate partner violence: Causes and prevention. The Lancet 35 9:1423 1429. Jewkes, R., K. Dunkle, M. Nduna, J. Levin, N. Jama, N. Khuzwayao, M. Koss, A. Puren, and N. Duvvury. 2006. Factors associated with HIV sero status in young rural South African women: Connections between intimate partner violence and HIV. Inter national Journal of Epidemiology 35:1461 68. Jewkes, R., J. Levin, and L. Penn Kekana. 2002. Risk factors for domestic violence: Findings from a South African cross sectional study. Social Science and Medicine 55:1603 1617. Jewkes, R., M. Nduna, P. Jama, a nd J. Levin. 2002. Measuring relationship power: adaptation of the SRPS for South Africa. Abstract presented at the XIV International Conference on AIDS, July 7 12 2002, Barcelona, Spain. Jewkes, R., Y. Sikweyiya, R. Morrell, and K. Dunkle. 2009. Underst anding men's health and use of violence: Interface of rape and HIV in South Africa. Technical report. Pretoria: Medical Research Council, 2009. Jewkes, R.K., K. Dunkle, M. Nduna, and N. Shai. 2010. Intimate partner violence, relationship power inequity, an d incidence of HIV infection in young women in South Africa: A cohort study. The Lancet 376:41 48. Jewkes, R.K., J.B. Levin, and L.A. Penn Kekana. 2003. Gender inequities, intimate partner violence, and HIV preventive practices: Findings of a South African cross sectional study. Social Science & Medicine 56:125 134. Kaler, A. 2001. "Many Divorces and Many Spinsters": Marriage as an Invented Tradition in Southern Malawi, 1946 1999. Journal of Family History 26:529. Kaler, A., and S.C. Watkins. 2010. Asking G od about the date you will die: HIV testing as a zone of uncertainty. Demographic Research 23 (32):905 932.

PAGE 295

280 Kalichman, S.C., and L.C. Simbayi. 2003. HIV testing attitudes, AIDS stigma, and voluntary HIV counselling and testing in a black township in Cape T own, South Africa. Sex Transm Infect 79 (442 447). Katz, D.A., J.N. Kiarie, G.C. John Stewart, B.A. Richardson, F.N. John, and C. Farquhar. 2009. HIV testing men in the antenatal setting: understanding male non disclosure. International Journal of STD and AIDS 20 (11):765 767. Kelley, H.H., and J.W. Thibalt. 1978. Interpersonal relations: A theory of interdependence New York: Wiley. Kelly, R.J., R.H. Gray, N.K. Sewankambo, D. Serwadda, F. Wabwire Mangen, and T. Lutalo. 2001. Age differences in sexual partn ers and risk of HIV 1 infection in rural Uganda. JAIDS Journal of Acquired Immune Deficiency Syndromes 32 (4):446 451. Kenny, D.A., D.A. Kashy, and W.L. Cook. 2006. Dyadic data analysis New York: The Guilford Press. Keogh, P., S. Allen, C. Almedal, and B. Temahagili. 1994. The Social Impact of HIV Infection on Women in Kigali, Rwanda: A Prospective Study. Social Science & Medicine 38 (8):1047 1053. Kilewo, C., A. Massawe, E. Lyamuya, I. Semali, F. Kalokola, E. Urassa, M. Giattas, F. Temu, K. Karlsson, F. M halu, and G. Biberfeld. 2001. HIV Counseling and Testing of Pregnant Women in Sub Saharan Africa: Experiences From a Study on Prevention of Mother to Child HIV 1 Transmission in Dar Es Salaam, Tanzania. JAIDS Journal of Acquired Immune Deficiency Syndromes 28:458 462. King, R., S. Allen, A. Serufilira, E. Karita, and P. Vandeperre. 1993. Voluntary confidential HIV testing for couples in Kigali, Rwanda. AIDS 7 (10):1393 1394. King, R., D. Katuntu, J. Lifshay, L. Packel, R. Batamwita, S. Nakayiwa, B. Abang, F Babirye, P. Lindkvist, E. Johansson, J. Mermin, and R. Bunnell. 2008. Processes and Outcomes of HIV Serostatus Disclosure to Sexual Partners among People Living with HIV in Uganda. AIDS and Behavior 12:232 243. Kishamawe, C., D.C. Vissers, M. Urassa, R. Isingo, G. Mwaluko, G. Borsboom, H.A. Voeten, B. Zaba, J.D.F. Habbema, and S. de Vlas. 2006. Mobility and HIV in Tanzanian couples: both mobile persons and their partners show increased risk. AIDS 20 (4):601 608. Kline, R. 2005. Principles and practices of structural equation modeling New York: Guilford Press.

PAGE 296

281 Kranzer, K., N. McGrath, J. Saul, A.C. Crampin, A. Jahn, S. Malema, D. Mulawa, P.E.M. Fine, B. Zaba, and J.R. Glynn. 2009. Individual, household, and community factors associated with HIV test refusa l in rural Malawi. Tropical Medicine and International Health 13 (11):1341 1350. Leclerc Madlala, S. 2003. Transactional sex and the pursuit of Modernity. Social Dynamics 29 (2):213 233. Lewis, M.A., C.M. McBride, K.I. Pollak, E. Puleo, R.M. Butterfield, a nd K.M. Emmons. 2006. Understanding health behavior change among couples: An interdependence and communal coping approach. Social Science & Medicine 62:1369 1380. Lindgren, T., S. Rankin, and W. Rankin. 2005. Malawi women and HIV: Socio cultural factors an d barriers to prevention. Women and Health 41 (1):69 86. Lowrance, D., S. Filler, S. Makombe, A. Harries, J. Aberle Grasse, M. Hochgesang, and E. Libamba. 2007. Assessment of a national monitoring and evaluation system for rapid expansion of antiretroviral treatment in Malawi Tropical Medicine and International Health 12 (3):377 381. Lugalla, J.L.P., C.M. Madihi, H.L. Sigalla, N.E. Mrutu, and P.S. Yoder. 2008. Social context of disclosing HIV test results: HIV testing in Tanzania. Dar es Salaam, Tanzania: U SAID. Luke, N. 2003. Age and economic asymmetries in the sexual relationships of adolescent girls in sub Saharan Africa. Studies in Family Planning 34 (2):67 86. 2005. Confronting the 'Sugar Daddy' stereotype: Age and economic asymmetries and risky se xual behavior in urban Kenya. International Family Planning Perspectives 31 (1):6 14. 2006. Exchange and condom use in informal sexual relationships in urban Kenya. Economic Development and Cultural Change 56 (2):375 396. Luke, N., R. Goldberg, B.U. M beru, and E.M. Zulu. 2011. Social exchange and sexual behavior in young women's premarital relationships in Kenya. Journal of Marriage and the Family 73 (5):1048 1064. Luke, N., and K.M. Kurz. 2002. Cross generational and transactional sexual relations in sub Saharan Africa: Prevalence of behavior and implications for negotiating safer sexual practices. In Research Paper Series Washington, D.C.: International Center for Research on Women, Population Services International.

PAGE 297

282 Lupton, D., S. McCarthy, and S. C hapman. 1995. 'Doing the Right Thing': The Symbolic Meanings and Experiences of having an HIV Antibody Test. Social Science & Medicine 41 (2):173 180. Lurie, M.N., B.G. Williams, K. Zuma, D. Mkaya Mwaburi, G.P. Garnett, M.D. Sweat, J. Gittlesohn, and S.S.A Karim. 2003. Who infects whom? HIV 1 concordance and discordance among migrant and non migrant couples in South Africa. AIDS 17 (15):2245 2252. Lurie, M.N., B.G. Williams, K. Zuma, D. Mkaya Mwaburi, G.P. Garrett, and A.W. Strum. 2003. The impact of migra tion on HIV 1 transmission in South Africa: A study of migrant and non migrant men and their partners. Sexually Transmitted Diseases 30:149 156. Luseno, W., and W.M. Wechsberg. 2009. Correlates of HIV testing among South African women with high sexual and substance use risk behaviors. AIDS Care 21 (2):178 184. MacPhail, C.L., A. Pettifor, T. Coates, and H. Rees. 2007. "You must do the test to know your status": Attitudes to HIV voluntary counseling and testing for adolescents among South African youth and p arents. Health Education and Behavior 35 (1):87 104. MacPherson, P., E.L. Corbett, S.D. Makombe, J.J. van Oosterhout, E. Manda, A.T. Choko, D. Thindwa, S.B. Squire, G.H. Mann, and D.G. Lalloo. 2012. Determinants and Consequences of Failure of Linkage to An tiretroviral Therapy at Primary Care Level in Blantyre, Malawi: A Prospective Cohort Study. PLoS One 7 (9):e44794. MacPherson, P., D.G. Lalloo, A.T. Choko, G.H. Mann, S.B. Squire, D. Mwale, E. Manda, S.D. Makombe, N. Desmond, R. Heyderman, and E.L. Corbett 2012. Suboptimal patterns of provider initiated HIV testing and counselling, antiretroviral therapy eligibility assessment and referral in primary health clinic attendees in Blantyre, Malawi. Tropical Medicine and International Health 17 (4):507 517. Mak wiza, I., L. Nyirenda, G. Bongololo, T. Banda, R. Chimzizi, and S. Theobald. 2009. Who has access to counseling and testing and anti retroviral therapy in Malawi an equity analysis. International Journal for Equity in Health 8:13. Maleta, K., and C. Bowi e. 2010. Selecting HIV infection prevention interventions in the mature HIV epidemic in Malawi using the mode of transmission model. Bio medical central health services research 10:243 249.

PAGE 298

283 Maman, S., N.M. Hogan, and G.P. Kilonza. 2001. Women's barriers to HIV 1 testing and disclosure: challenges for HIV 1 voluntary counselling and testing. AIDS Care 13 (5):595 603. Maman, S., J. Mbwambo, M. Hogan, G. Kilonzo, M. Sweat, and E. Weiss. 2001. HIV and partner violence: Implications for HIV voluntary counseling and testing programs. New York: The Population Council. Maman, S., J. Mbwambo, N.M. Hogan, E. Weiss, G.P. Kilonzo, and M.D. Sweat. 2003. High rates and positive outcomes of HIV serostatus disclosure to sexual partners: Reasons for cautious optimism from a voluntary counseling and testing clinic in Dar es Salaam, Tanzania. AIDS and Behavior 7:373 382. Maman, S., J.K. Mbwambo, N.M. Hogan, G.P. Kilonza, J.C. Campbell, E. Weiss, and M.D. Sweat. 2002. HIV positive women report more life time partner violence: Fi ndings from a voluntary counselling and testing clinic in Dar es Salaam, Tanzania. Am J Public Health 92:1331 1337. Mankowski, E.S., and K.I. Maton. 2010. A Community Psychology of Men and Masculinity: Historical and Conceptual Review. American Journal of Community Psychology 45:73 86. Marin, B.V., J.M. Tschann, C.A. Gomez, and S.M. Kegeles. 1993. Acculturation and gender differences in sexual attitudes and behaviors: Hispanic vs. non Hispanic white unmarried adults. American Journal of Public Health 83:175 9 1761. Matovu, J.K.B., R.H. Gray, N. Kiwanuka, G. Kigozi, F. Wabwire Mangen, F. Nalugoda, D. Serwadda, N.K. Sewankambo, and M.J. Wawer. 2007. Repeat Voluntary HIV Counseling and Testing (VCT), Sexual Risk Behavior and HIV Incidence in Rakai, Uganda. AIDS and Behavior 11:71 78. Mbweza, E., K.F. Norr, and B. McElmurry. 2008. Couple decision making and use of cultural scripts in Malawi. Journal of Nursing Scholarship 40 (1):12 19. McGrath, N., J.R. Glynn, J. Saul, K. Kranzer, A. Jahn, F. Mwaungulu, M.H. Ngwir a, H. Mvula, F. Munthali, V. Mwinuka, L. Mwaungulu, P.E. Fine, and A.C. Crampin. 2010. What happens to ART eligible patients who do not start ART? Dropout between screening and ART initiation: a cohort study in Karonga, Malawi. BMC Public Health 10 (1):601 McKenna, S.L., G.K. Muyinda, D. Roth, M. Mwali, N. Ngandu, A. Myrick, C. Luo, F.H. Priddy, V.M. Hall, A.A. vonLieven, J.R. Sabatino, K. Mark, and S.A. Allen. 1997. Rapid HIV testing and counseling for voluntary testing centers in Africa. AIDS 11:S103 S11 0.

PAGE 299

284 McLeroy, K.R., D. Bibeau, A. Steckler, and K. Glanz. 1988. An Ecological Perspective on Health Promotion Programs. Health Education Quarterly 15:351 377. Mead, M. 1935. Sex and temperament: in three primitive societies New York, NY: William Morrow. Med ley, A., C. Garcia Moreno, S. McGill, and S. Maman. 2004. Rates, barriers and outcomes of HIV serostatus disclosure among women in developing countries: implications for prevention of mother to child transmission programmes. Bulletin of the World Health Or ganization 82:299 307. Mikell, G. 1997. African feminism: The politics of survival in sub Saharan Africa Philadelphia, PA: University of Pennsylvania Press. Miles, B.M., and A.M. Huberman. 1994. Qualitative data analysis Thousand Oaks, CA: Sage. Ministry of Health [Malawi]. 2003. Treatment of AIDS. Guidelines for the use of antiretroviral therapy in Malawi, 1st edn. Lilongwe, Malawi: Ministry of Health. 2005. Report of a country wide survey of HIV/AIDS services in Malawi for the year 2005 Lilongwe. 2006. Report of a country wide survey of HIV/AIDS services in Malawi for the year 2006 Lilongwe. 2009. Malawi Antiretroviral Treatment Programme Report for Quarter 4 2009 Lilongwe, Malawi: Ministry of Health. Mkandawire Valhmu, L., C. Wendland, P.E. Stevens, P.M. Kako, A. Dressel, and J. Kibicho. 2013. Marriage as a risk factor for HIV: Learning from the experiences of HIV infected women in Malawi. Global Public Health 8 (2):187 201. Mlay, R., H. Lugina, and S. Becker. 2008. Couple counselling a nd testing for HIV at antenatal clinics: Views from men, women and counsellors. AIDS Care 20 (3):356 360. Morgan, D.L. 1998. Practical strategies for combining qualitative and quantitative methods: Applications to health research. Qualitative Health Resear ch 8 (3):362 376. Morin, S.F., G. Khumalo Sakutukwa, E.D. Charlebois, J. Routh, K. Fritz, T. Lane, T. Vaki, A. Fiamma, and T.J. Coates. 2006. Removing Barriers to Knowing HIV Status: Same Day Mobile HIV Testing in Zimbabwe. Journal of Acquired Immune Defic iency Syndromes 41 (2):218 224.

PAGE 300

285 Morris, M., and M. Kretzschmar. 1997. Concurrent Partnerships and the Spread of HIV. AIDS 11:641 648. 2000. A microsimulation study of the effect of concurrent partnerships on the spread of HIV in Uganda. Mathematical P opulation Studies 8 (2):109 133. Murphy, G.W., and S.L. Meyer. 1991. Gender, power, and violence in marriage. The Behavior Therapist 14:95 100. National Statistical Office & ORC Macro. 2005. Malawi demographic and health survey Calverton, Maryland: NSO an d ORC Macro. 2011. Malawi demographic and health survey 2010. Maryland: NSO and ORC Macro. Nebie, Y., N. Meda, and V. Leroy. 2001. Sexual and reproductive life of women informed of their HIV seropositivity: A prospective cohort study in Burkina Faso. J Acquir Immune Defic Syndr 28 (4):367 372. Njau, B., M.H. Watt, J. Ostermann, R. Manongi, and K.J. Sikkema. 2011. Perceived acceptability of home based couples voluntary HIV counseling and testing in Northern Tanzania. AIDS Care 24 (4):413 419. Nunnally, J., and I. Bernstein. 1994. Psychometric theory 3rd ed. New York: McGraw Hill. Nunnally, J.C. 1967. Psychometric theory New York: McGraw Hill. Obermeyer, C.M., and M. Osborn. 2007. The utilization of testing and counseling for HIV: a review of the social and behavioral evidence. American Journal of Public Health 97 (10):1762 1774. Office of the President and Cabinet and National AIDS Commission [Malawi]. 2003. National HIV/AIDS Policy, A Call for Renewed Action. Lilongwe, Malawi: National AIDS Commission. Painter, T. 2001. Voluntary counseling and testing for couples: a high leverage intervention for HIV/AIDS prevention in sub Saharan Africa. Social Science & Medicine 53 (11):1397 1411. Parikh, S. 2009. Modern wives, men's infidelity, and marriage. In The secret: Love, marriage, and HIV edited by J. S. Hirsch, H. Wardlow, D. J. Smith, H. M. Phinney, S. Parikh and C. A. Nathanson. Nashville, TN: Vanderbilt University Press, 168 196.

PAGE 301

286 Peltzer, K., G. Matseke, T. Mzolo, and M. Majaja. 2009. Determinants of kno wledge of HIV status in South Africa: results from a population based HIV survey. BMC Public Health 9 (174). Perez, F., C. Zvandaziva, B. Engelsmann, and F. Dabis. 2006. Acceptability of Routine HIV Testing ("Opt Out") in Antenatal Services in Two Rural Di stricts of Zimbabwe. Journal of Acquired Immune Deficiency Syndromes 41:514 520. Peters, P.E. 1997. Against the odds: matriliny, land and gender in the shire highlands of Malawi. Critique of Anthropology 17 (2):189 210. 2010. "Our daughters inherit ou r land, but our sons use their wives' fields": matrilineal matrilocal land tenure and the New Land Policy in Malawi. Journal of Eastern African Studies 4 (1):179 199. Petersen, T. 2004. Analyzing Panel Data: Fixed and Random Effects Models. In Handbook of data analysis edited by M. Hardy and A. Bryman. Thousand Oaks: Sage, 331 345. Pettifor, A.E., D.M. Measham, H.V. Rees, and N.S. Padiant. 2004. Sexual power and HIV risk, South Africa. Emerging Infectious Diseases 10 (11). Phiri, K. 1983. Some changes in the matrilineal family systems among the Chewa of Malawi since the nineteenth century. Journal of African History 24:254 274. Pool, R., S. Nyanzi, and J.A.G. Whitworth. 2001. Attitudes to voluntary counselling and testing for HIV among pregnant women in ru ral south west Uganda. AIDS Care 13 (5):605 615. Porter, L., L. Hao, D. Bishai, D. Serwadda, M.J. Wawer, T. Lutalo, and R. Gray. 2004. HIV status and union dissolution in sub Saharan Africa: The case of Rakai, Uganda. Demography 41 (3):465 482. Poulin, M. 2007. Sex, money, and premarital partnerships in southern Malawi. Social Science and Medicine 65:2383 2393. Pulerwitz, J., H. Amaro, W. DeJong, S.L. Gortmaker, and R. Rudd. 2002. Relationship power, condom use, and HIV risk among women in the USA. AIDS Car e 14 (6):789 800. Pulerwitz, J., S.L. Gortmaker, and W. DeJong. 2000. Measuring relationship power in HIV/STD research. Sex Roles 42 (7/8). Rankin, W. 2001. Malawi Report: Global AIDS Interfaith Alliance.

PAGE 302

287 Reniers, G. 2003. Divorce and remarriage in rural M alawi. Demographic Research SI(Article 6):173 206. 2005. Marital strategies for managing exposure to HIV in rural Malawi. Paper presented at the Annual Meeting of the Population Association of America, March 31 April 2 2005, Philadelphia, PA. Rhine, K 2009. Dilemmas of Diagnoses: HIV Testing in Northern Nigeria. Northeastern Anthropological Association Newsletter 31 (4):2 11. Riley, N.E. 1997. Gender, power, and population change. In Population Bulletin Washington, DC: Population Reference Bureau. Ri tzer, G., and D. Goodman. 2004. Symbolic interactionism Edited by G. Ritzer and D. Goodman, Modern sociological theory Boston: McGraw Hill. Sallis, J.F., N. Owen, and E.B. Fisher. 2008. Ecological models of health behavior. In Health behavior and health education edited by K. Glanz, B. K. Rimer and K. Viswanath. San Francisco: Jossey Bass, 465 485. Sambisa, W., S. Curtis, and V. Mishra. 2012. AIDS stigma as an obstacle to uptake of HIV testing: evidence from a Zimbabwean national population based survey. AIDS Care 22 (2):170 186. Santelli, J., M.A. Ott, M. Lyon, J. Rogers, D. Summers, and R. Schleifer. 2006. Abstinence and abstinence only education: A review of US policies and programs. Journal of Adolescent Health 38:72 81. Schatz, E. 2005. 'Take your ma t and go'! Rural Malawian women's strategies in the HIV/AIDS era. Culture Health & Sexuality 7 (5):479 492. Schensul, J.J. 1999. Focused group interviews. In Ethnographer's toolkit: Enhanced ethnographic methods edited by J. Schensul, M. LeCompte, B. Nast asi and S. Borgotti. Walnut Creek, CA: Altamira press, 51 114. Schouten, E.J., A. Jahn, D. Midiani, S.D. Makombe, A. Mnthambala, Z. Chirwa, A.D. Harries, J.J. van Oosterhout, T. Meguid, A. Ben Smith, R. Zachariah, L. Lynen, M. Zolfo, W. Van Damme, C.F. Gil ks, R. Atun, M. Shawa, and F. Chimbwandira. 2011. Prevention of mother to child transmission of HIV and the health related Millennium Development Goals: time for a public health approach. The Lancet 378:282 284.

PAGE 303

288 Semrau, K., L. Kuhn, C. Vwalika, and e. al. 2005. Women in couples antenatal HIV counseling and testing are not more likely to report adverse social events. AIDS 19 (6):603 609. Shannon, K., K. Leiter, N. Phaladze, Z. Hlanze, A.C. Tsai, M. Heisler, V. Lacopino, and S.D. Weiser. 2012. Gender inequity norms are associated with increased male perpetrated rape and sexual risks for HIV infection in Botswana and Swaziland. PLOS One 7 (1):1 8. Shelton, J. 2008. Counseling and testing for HIV prevention. The Lancet 372 (9635):273 275. Shettima, A.G. 1998. Ge ndered work patterns in the endangered Sahelian rural environment: Exploring three layers of exploitation. Africa Development 23:163 183. Sikweyiya, Y., and R. Jewkes. 2009. Force and temptation: contrasting South African men's accounts of coercion into se x by men and women. Culture Health & Sexuality 11 (5):529 541. Simbayi, L.C., S.C. Kalichman, A. Strebel, A. Cloete, N. Henda, and A. Mqeketo. 2007. Internalised stigma, discrimination, and depression among men and women living with HIV/AIDS in Cape Town, South Africa. Social Science & Medicine 64:1823 1831. Simpson, A. 2009. Boys to men in the shadow of AIDS: Masculinities and HIV risk in Zambia New York, NY: Palgrave Macmillan. Singleton, R.A., and B.C. Straits. 2005. Approaches to social research 4th e d. New York: Oxford University Press. Skovdal, M., C. Campbell, C. Madanhire, Z. Mupambireyi, C. Nyamukapa, and S. Gregson. 2011. Masculinity as a barrier to men's use of HIV services in Zimbabwe. Globalization and Health 7 (13). Smith, D.J. 2006. Love and the Risk of HIV: Courtship, Marriage and Infidelity in Southeastern Nigeria. In Modern Loves: The anthropology of romantic courtship and companionate marriage edited by J. S. Hirsch and H. Wardlow. Ann Arbor: University of Michigan Press, 135 156. 2 009. Gender inequality, infidelity, and the social risks of modern marriage in Nigeria. In The secret: Love, marriage, and HIV edited by J. S. Hirsch, H. Wardlow, D. J. Smith, H. M. Phinney, S. Parikh and C. A. Nathanson. Nashville, TN: Vanderbilt Univers ity Press., 84 107.

PAGE 304

289 Smith, D.J., and B.C. Mbakwem. 2007. Life projects and therapeutic itineraries: Marriage, fertility, and antiretroviral therapy in Nigeria. AIDS 21 (suppl 5):S37 S41. Smith, K.P., and S.C. Watkins. 2005. Perceptions of risk and strategi es for prevention: Responses to HIV/AIDS in rural Malawi. Social Science & Medicine 60 (3):649 660. Starks, H., and S.B. Trinidad. 2007. Choose Your Method: A Comparison of Phenomenology, Discourse Analysis, and Grounded Theory. Qualitative Health Research 17 (10):1372. Strauss, A., and J. Corbin. 1998. Basics of qualitative research: Techniques and procedures for developing grounded theory Thousand Oaks, CA: Sage. Swidler, A., and S.C. Watkins. 2007. Ties of dependence: AIDS and transactional sex in rural Malawi. Studies in Family Planning 38 (3):147 162. Tanur, J. 1992. Questions about questions: Inquiries into the cognitive bases of surveys New York: Russell Sage Foundation. Tavory, I., and A. Swidler. 2009. Condom semiotics: Meaning and condom use in r ural Malawi. American Sociological Review 74 (April):171 189. Tawfik, L., and S.C. Watkins. 2007. Sex in Geneva, sex in Lilongwe, and sex in Balaka. Social Science & Medicine 64:1090 1101. The World Bank. 2011. World Development Indicators. Washington, D.C Ulin, R.R., E.T. Robinson, E.E. Tolley, and E.T. McNeill. 2002. Qualitative methods: A field guide for applied research in sexual and reproductive health, edited by F. H. International. UNAIDS. 2006. HIV testing and counseling. Geneva: UNAIDS. 2010. Report on the global AIDS epidemic Geneva: UNAIDS. United Nations. 2010. The Millennium Development Goals Report 2010. Van Oosterhout, J.J.G., J.K. Kumwenda, T. Hartung, B. Mhango, and E.E. Zijlstra. 2007. Can the initial success of the Malawi ART scale up programme be sustained? The example of Queen Elizabeth Central Hospital, Blantyre. AIDS Care 19 (10):1241 1246.

PAGE 305

290 VCT Efficacy Group. 2000. Efficacy of voluntary HIV 1 counseling and testing in individuals and couples in Kenya, Tanzania, and Trinidad: A r andomized controlled trial. The Lancet 356:103 112. Venkatesh, K.K., P. Madiba, G. De Bruyn, M.N. Lurie, T.J. Coates, and G.E. Gray. 2011. Who Gets Tested for HIV in a South African Urban Township? Implications for Test and Treat and Gender Based Preventio n Interventions. JAIDS Journal of Acquired Immune Deficiency Syndromes 56 (2):151 165. Vissers, D.C., H.A. Voeten, M. Urassa, R. Isingo, M. Ndege, Y. Kumogola, G. Mwaluko, B. Zaba, S.J. de Vias, and J.D.F. Habbema. 2008. Separation of spouses due to travel and living apart raises HIV risk in Tanzanian couples. Sexually Transmitted Diseases 35 (8):714 720. Voeten, H.A., D.C. Vissers, and S. Gregson. 2009. Strong association between in migration and HIV prevalence in urban sub Saharan Africa. Sexually Transmi tted Diseases 36:1 4. Wamoyi, J., A. Fenwick, M. Urassa, B. Zaba, and W. Stones. 2011. Women's Bodies are Shops'': Beliefs About Transactional Sex and Implications for Understanding Gender Power and HIV Prevention in Tanzania. Archives of Sexual Behavior 40:5 15. Watkins, S.C. 2004. Navigating the AIDS epidemic in rural Malawi. Population and Development Review 30 (4):673 705. Watkins, S.C., N. Rutenberg, and D. Wilkinson. 1997. Orderly theories, disorderly women. In The continuing demographic transition edited by G. W. Jones, R. M. Douglas, J. C. Caldwell and R. M. D'Souza. Oxford: Oxford University Pres, 213 245. Watkins, S.C., and A. Swidler. 2012. Working Misunderstandings: Donors, Brokers, and Villagers in Africa's AIDS Industry. Population and Devel opment Review 38 (Supplement):197 218. Weir, S., I. Hoffman, and A. Muula. 2008. Final Report Malawi Prevalence Study. Lilongwe, Malawi UNC Project Malawi. Weiser, S.D., M. Heisler, K. Leiter, K.F. Percy de, S. Tlou, S. DeMonner, N. Phaladze, D. Bangsber g, and V. Iacopino. 2006. Routine HIV Testing in Botswana: A Population Based Study on Attitudes, Practices, and Human Rights Concerns. PLoS Medicine 3 (7):e261. White, S. 2007. Malawi: Country Gender Profile.

PAGE 306

291 WHO. 2003. Gender dimensions of HIV status dis closure to sexual partners: Rates, barriers, and outcomes; A review paper Geneva, Switzerland: World Health Organization. 2007. Towards universal access: Scaling up priority HIV/AIDS interventions in the health sector. Progress report Geneva, Switze rland: World Health Organization. 2010. Scaling up priority HIV/AIDS interventions in the health sector. Progress report 2010 Geneva, Switzerland: World Health Organization. Wilson, A. 2012. Of Love Potions and Witch Baskets. Western Folklore 71 (2): 149 173. Wingood, G.M., and R.J. DiClemente. 2000. Application of the Theory of Gender and Power to Examine HIV Related Exposures, Risk Factors, and Effective Interventions for Women. Health Education & Behavior 27:539 565. 2002. The theory of gender and power: A social structural theory for guiding public health interventions. In Emerging theories in health promotion practice and research edited by R. J. DiClemente, R. A. Crosby and M. C. Kegler. San Francisco: Jossey Bass, 313 346. Wojcicki, J.M. 20 02. "She drank his money": Survival sex and the problem of violence in taverns in Gauteng Province, South Africa. Medical Anthropology Quarterly 16 (3):267 293. Wong, L.H., H. Van Rooyen, P. Modiba, L. Richter, G. Gray, J.A. McIntyre, C.D. Schetter, and T. Coates. 2009. Test and Tell: Correlates and Consequences of Testing and Disclosure of HIV Status in South Africa (HPTN 043 Project Accept). JAIDS Journal of Acquired Immune Deficiency Syndromes 50 (2):215 222. Yeatman, S.E. 2007. Ethical and Public Health Considerations in HIV Counseling and Testing: What Do We Know and What Should That Mean for Policy? Studies in Family Planning 38 (4):271 278. 2009. The impact of HIV status and perceived status on fertility desires in rural Malawi. AIDS and Behavior 13 (Supplement):S12 S19. Zulu, E.M. 1996. Social and cultural factors affecting reproductive behavior in Malawi: University of Pennsylvania, Unpublished Ph.D. dissertation. Zulu, E.M., and G. Chepngeno. 2003. Spousal communication about the risk of contra cting HIV/AIDS in rural Malawi. Demographic Research S1 (8):247 278.

PAGE 307

292 APPENDIX A R elationship Power Scale Items Preliminary Relationship Power Scale Autonomy Subscale: 1. Under no circumstances would I ever leave my partner. 2. If my partner were to leave me, I would be in serious trouble. 3. If things were really bad with my partner, I w ould leave the relationship. 4. If my partner failed to meet my needs, I could easily find another partner. Communication Subscale: 5. My partner and I sit down and discuss impor tant matters together. 6. My partner sh ows that they care about me. 7. If I suspect my partner is having an affair, I would talk with my partner. 8. I would consult with my advisors (or friends) if my partner was behaving ba dly. Love and Trust Subscale: 9. When I need my partner's assistance, he/she is always there to help me. 10. I initiate sex with my part ner when I want to have sex. 11. I am able to buy expensive items wi thout my partner's approval. 12. I have my own mon ey to buy things I want. Relationship Dominance S ubscale: 13. My partner punishes me when he/she is really angry with me. 14. When I disagree with my partner's relatives, my partner chooses their side over mine. 15. My partner is probably having sex with someone else. 16. If my partner was really angry with me, he/she m ight beat me.

PAGE 308

293 Final Relationship Power Scale Unity Subscale: 1. My partner and I sit down and discuss important matters together. 2. My partner shows that they care about me. 3. When I need my partner's assistance, he/she is always there to help me. Disco rdance Subscale: 1. My partner punishes me when he/she is really angry with me. 2. When I disagree with my partner's relatives, my partner chooses their side over mine. 3. My partner is probably having sex with someone else. 4. If my partner was really angry with me, he/she might beat me.

PAGE 309

294 APPENDIX B Semi S tructured Interview Guide Relationship Power and HIV Testing in Rural Malawi Participant must have a primary partner to be eligible for this interview. Introduction: Introduce yourself, and state your name and affiliation (University of Colorado) Purpose of interview: We are trying to learn more about decision making, power, and sexual behavior within romantic relationships in Malawi. Relationship dynamics can affect the health of both men and women and so thi s new information will help us to understand what relationship factors affect health. Consent: Would you be willing to participate in our study? Also, would it be ok if we tape record our conversation? If at any point during this interview you wish to sto p the tape recorder, we can do so. (Obtain written consent and provide a copy of consent form to participant). 1. INTRODUCTION Thank you for taking the time to participate in our research on relationships and HIV/AIDS in Malawi. Are you from this village? How long have you lived here? Are these your children? How many children you do you have? 2. RELATIONSHIP CHARACTERISTICS Now, I would like to talk with you about your romantic relationship. How long have you been with your spouse or partner? How did you and your spouse or partner meet? What attracted you to him or her? For non marital relationships: How serious would you say your relationship is? Do you think this partner could be your husband/wife one day? Are you dating anyone else? 3. DEPENDENCE Befor e we talk specifically about your partner, can you tell me what an ideal partner or spouse would be?

PAGE 310

295 Does your partner meet these ideals? Why or why not? What does your partner do that makes you happy? What frustrates you? Can you give me examples? Pro bes: Tell me about the things your partner gives you. When does your partner give you these things? The last time you saw your partner, did he/she give you anything? What do you do when your partner does not give you want you want? Do you feel like you ha ve to stay in this marriage or relationship? If your partner were seeing someone else, what would you do? 4. RELATIONSHIP CONTROL I would like to talk with you more about your partnership or marriage. Do you feel "free" in your relationship? Why or why not ? Probes: Are you able to do the things you enjoy? What do you enjoy? Can you give me examples? Does your partner ever tell you where you can go, who you can be friends with, or what you should wear? Are there certain things your partner tells you to do o r not do? Do you feel that you can talk with your partner about anything? When you do something your partner does not like, how does he typically respond? When your partner does something that you do not like, how do you typically respond? 5. DECISION MAKIN G How do you and your partner decide on important things in your relationship? Probes: Are decisions made as couple or does one partner make most of the decisions? What types of decisions do you have more say in, what types of things does your partner h ave more say in? (finances, childbearing) Do you need to consult with your partner on certain types of decisions? Do you and your partner/spouse talk about sex? Probes: Who decides when you have sex, use condoms, or the types of sex positions? How do you feel about this? 6. CONDOMS Tell me about a conversation you and your partner had about condoms. Probes: Do you and your partner like condoms? Do you use them?

PAGE 311

296 Do your friends and neighbors use condoms? What do you know about condoms and HIV/AIDS? 7. ALCOHOL USE Do you and your partner drink alcohol? How does alcohol affect your relationship? Probes: What usually happens after one of you (or both of you) has been drinking? 8. RELIGION What religion are you? What church do you belong to? Is your husband religio us? What religion/church does he practice? What does your church tell you about sex, HIV/AIDS, and condoms? Do you agree with this? 9. HIV TESTING We are now going to talk about your thoughts and experiences with HIV testing. This information will help us to understand how relationships affect HIV testing and will remain completely confidential. Have you and your partner discussed getting tested for AIDS? Tell me about what you talked about. If YES, ask: Why did you decide to get tested? How did you feel about getting tested? Afraid? What did you think your test result would be before testing? If NO, ask: Have you ever considered being tested for HIV? Why or why not? Would you talk with your partner before being tested? Why or why not? Probes: If your p artner did not want you to be tested, would you get tested anyways? How would your partner react if he found out you were tested? ***CONTINUE with interview if respondent was previously tested for HIV*** 10. KNOWLEDGE OF HIV STATUS Did you inform your part ner of the test result? Why or why not? If informed partner of HIV test result, ask: How did your partner respond to the HIV test result?

PAGE 312

297 If did not inform partner of HIV test result, ask: Why did you not inform your partner of your HIV test result? I f tested together: What did you and your partner talk about after getting your test results? In either scenario, ask: Has anything changed in your life since you learned your HIV status? Probes: Have your beliefs or behaviors changed since testing? If ye s, what has changed? What caused these changes? How did you feel after receiving your test result? Surprised? Happy? Sad? 11. CONCLUSION OF INTERVIEW Is there anything else you would like to tell me? Are there any other questions that you have for me? Tha nk the respondent for their time. After turning off the recorder, ask the respondent about their impressions of the interview. What was it like to do this interview? Was it difficult or easy to talk about these things?

PAGE 313

298 APPENDIX C F ocus Group Discussion G uide Gender, Relationship Power, and HIV Testing in Rural Malawi Principal Investigator: Amy Conroy University of Colorado Denver HSRC No: 10 0861 Rev: 3.0, Rev Date: July 25, 2011 PART A: Questions on power within couples First, I will ask you some th ought provoking questions on relationships. 1. Can you describe the perfect relationship? Probes: What makes a good wife (girlfriend)? What makes a good husband (boyfriend)? What are the most important things that men/women get out of their relationships? W hat is most important when choosing a marriage partner (or a chibwenzi )? Note: if "love" comes up, ask: what does it mean to love your partner? 2. What happens when couples disagree on an issue? Are men/women free to bring up their issues with their partners ? What about if the issue is serious, for example, if a man/woman is having sex with another person? 3. What does it mean to be "the head of the household"? Probes: Do women expect men to be the head of the household? Can women ever be the head of the house hold? Under what circumstances? 4. What makes a man/woman feel powerful in their relationships? Probes : How does a person know if they have power in the relationship? How much power do you think men and women should have in their relationships? 5. In Malawi, gender empowerment programs help to make women and men equal. For example, women now work in jobs that were once only for men (give examples). Because of working and education, women are becoming more independent/gaining freedom. Probes: Has women's power changed because of gender empowerment? How does it affect relationships? How do men feel about this? 6. Does men's/women's power in the relationship change after marriage? How so? Probes: Do men take on a different role once they are the head of the househol d?

PAGE 314

299 What is now expected of a wife/husband once they are married? Does sex change after marriage? How so? PART B: Vignettes on HIV testing Thank you very much for your responses. Now, I will present a series of scenarios on HIV testing will be presented f or discussion. These are completely hypothetical situations, but we want to know what you think the individual should do in the given situation. Also, please let us know if you think this scenario would never happen in the first place so we can revise the scenarios to better capture Malawian life. Again, please do not discuss your own personal experiences with HTC, just what you think the outcome of the situation will be and why. Scenario #1: Lucy is a 20 year old married girl from the Balaka area. She wants to know her HIV status and goes to a VCT center to get tested. She hears rumors that her husband might be cheating on her. Does she need to get her husband's permission before going? Why or why not? Could he prevent her from going to get tested? Supp ose that Lucy never told her husband she was going for VCT and she ended up testing positive for HIV. Should she tell her husband? How do you think he will react? What if Lucy tested negative, should she still tell him? What will he think? What if Lucy is not married, should she obtain her boyfriend's permission before she goes? Should she tell him she is going for testing? Scenario #2: Promise is a young man from Balaka. He is worried about his HIV status. His uncle just died of HIV and Promise has no t been feeling well himself for the past 6 months. He has been married for 3 years now and loves his wife. Should Promise tell his wife he is thinking of going for testing? Scenario #3: Suppose that Lucy and Promise are engaged to be married. Before t hey get married, they decide to go to HTC so that they can find out their HIV status. Lucy tests positive and Promise tests negative they are told their results together in the same room. What could happen to their relationship after learning their statu s? Would it change, if so how? What if Lucy tests negative and Promise tests positive. Does this change their future together? What if Lucy and Promise were married, not dating. Suppose Lucy is positive, Promise is negative. Would they stay together?

PAGE 315

300 P ART C: General reasons for getting tested Do people go for VCT/HTC here in the Balaka district? What factors make people decide to get tested? What prevents people from getting tested?