Citation
Gender, power, & vertical HIV prevention in urban Zambia

Material Information

Title:
Gender, power, & vertical HIV prevention in urban Zambia
Creator:
Hampanda, Karen Marie ( author )
Place of Publication:
Denver, Colo.
Publisher:
University of Colorado Denver
Publication Date:
Language:
English
Physical Description:
1 electronic file (231 pages) : ;

Thesis/Dissertation Information

Degree:
Doctorate ( Doctor of philosophy)
Degree Grantor:
University of Colorado Denver
Degree Divisions:
Department of Health and Behavioral Sciences, CU Denver
Degree Disciplines:
Health and behavioral sciences

Subjects

Subjects / Keywords:
HIV infections -- Prevention -- Zambia ( lcsh )
AIDS (Disease) -- Prevention -- Zambia ( lcsh )
AIDS (Disease) -- Prevention ( fast )
HIV infections -- Prevention ( fast )
Zambia ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Review:
Biomedical and behavioral interventions can virtually eliminate the risk of vertical (i.e., mother-to-child) HIV transmission. Pregnant and breastfeeding women's adherence to prevention of mother-to-child transmission (PMTCT) interventions, however, remains a challenge across sub-Saharan Africa. Using a concurrent mixed methods approach, including a survey and semi-structured interviews, I test whether a relationship exists between women's low power within married couples (based on domains from the Theory of Gender and Power) and adherence across the PMTCT cascade of care, including drug adherence during and after pregnancy, safe infant feeding practices, and pediatric HIV testing. The results of this study indicate that intimate partner violence is particularly detrimental to PMTCT adherence. Certain PMTCT protocols are also affected by partner controlling behaviors, participation in household decisions, and economic dependence, but not to the same extent as violence. Women with low power cite a lack of partner support and an unwillingness to disclose their HIV status to the husband due to fear of violence or abandonment as reasons for low PMTCT adherence. Conversely, women with high power cite partner support and the ability to prioritize PMTCT, sometimes even over the marriage, as enabling adherence. Based on these results, augmented efforts to address gender power dynamics both in society and within the home are recommended to promote the health of HIV-positive women and their families.
Thesis:
Thesis Ph. D.)--University of Colorado Denver.
Bibliography:
Includes bibliographical references.
System Details:
System requirements: Adobe Reader.
Statement of Responsibility:
by Karen Marie Hampanda.

Record Information

Source Institution:
University of Colorado Denver Collections
Holding Location:
Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
982958761 ( OCLC )
ocn982958761
Classification:
LD1193.L566 2016d H46 ( lcc )

Downloads

This item has the following downloads:


Full Text
GENDER, POWER, & VERTICAL HIV PREVENTION IN URBAN ZAMBIA
By
KAREN MARIE HAMPANDA
M.P.H., Boston University, 2010 B.A., Beloit College, 2008
A thesis submitted to the faculty of the Graduate School of the University of Colorado Denver in partial fulfillment of the requirements for the degree of Doctor of Philosophy Health and Behavioral Sciences Program
2016


2016
KAREN MARIE HAMPANDA
ALL RIGHTS RESERVED


This thesis for the Doctor of Philosophy degree by Karen Marie Hampanda has been approved for the Health and Behavioral Sciences Program by
Sara Yeatman, Advisor Karen Spencer, Chair Sheana Bull
Patrick Krueger


Hampanda, Karen Marie (PhD, Health and Behavioral Sciences)
Gender, Power, & Vertical HIV Prevention in Urban Zambia Thesis directed by Associate Professor Sara Yeatman
ABSTRACT
Biomedical and behavioral interventions can virtually eliminate the risk of vertical (i.e., mother-to-child) HIV transmission. Pregnant and breastfeeding womens adherence to prevention of mother-to-child transmission (PMTCT) interventions, however, remains a challenge across sub-Saharan Africa. Using a concurrent mixed methods approach, including a survey and semi-structured interviews, I test whether a relationship exists between womens low power within married couples (based on domains from the Theory of Gender and Power) and adherence across the PMTCT cascade of care, including drug adherence during and after pregnancy, safe infant feeding practices, and pediatric HIV testing. The results of this study indicate that intimate partner violence is particularly detrimental to PMTCT adherence. Certain PMTCT protocols are also affected by partner controlling behaviors, participation in household decisions, and economic dependence, but not to the same extent as violence. Women with low power cite a lack of partner support and an unwillingness to disclose their HIV status to the husband due to fear of violence or abandonment as reasons for low PMTCT adherence. Conversely, women with high power cite partner support and the ability to prioritize PMTCT, sometimes even over the marriage, as enabling adherence. Based on these results, augmented efforts to address gender power dynamics both in society and within the home are recommended to promote the health of HIV-positive women and their families.
The form and content of this abstract are approved. I recommend its publication.
Approved: Sara Yeatman
IV


ACKNOWLEDGEMENTS
This dissertation would not have been possible without the intellectual, emotional, and logistical support from several key individuals. First and foremost, a tremendous thank you goes to my advisor, Dr. Sara Yeatman. Your guidance was absolutely crucial for this research and my academic progression. Thank you for always taking the time to work through ideas with me and provide detailed feedback. In addition, I was fortunate to have an amazingly dedicated committee: Drs. Sheana Bull, Karen Spencer, and Patrick Krueger. Thank you for your outstanding mentorship and consistent help with this study over the past several years. I deeply appreciate your time and assistance.
I would also like to extend my gratitude to my research team in Zambia. First, Yusuf Ahmed has been a dedicated mentor on this project since its conception. Your ongoing help and support has truly been essential to the success of this project. I cannot thank you enough for your guidance. In addition, my sincere appreciation goes to the Lusaka District Health Office, the University Teaching Hospital, and Kanyama Health Center for allowing me into your community. I would also like to thank my research assistants in Lusaka, Christine Chewe Sakala, Grace Lungeani Phiri, Franklin Munsanje, and Bibi Lambert Manda, for your dedication to this study. I could not have asked for a better research team and owe much of the success of this study to your dedicated efforts. Zikomo Kwambili!
In addition, I would like to thank my family for their unconditional love and support during this process. My parents, Johanna and Jack McManemin, are the most generous, kind, and loving people I have ever met. Thank you for your ongoing support of my education and for always taking the time to help in any capacity needed, including emotional reassurance, editing my work, and financial support I honestly do not know what I would have done without you!
v


Also, to my brother, John McManemin, thank you for your persistent encouragement and love.
You have helped me more than you know.
In addition, I am extremely grateful to my husband, Mwiza Hampanda, for his support, love, and patience during this research. Not only have you helped me understand the Zambian context and translated interviews for me, but you also are my best friend whose love and kindness during the project have kept me sane. Thank you so much for your assurance and comfort when things got tough and for your enthusiasm for my accomplishments. All of your kind words and gestures have made this challenge bearable because I had you by my side at every step.
Finally, this study was made possible through funding from 1) the National Institute of Mental Health (Award Number F31MH107348) and 2) the Center for Global Health at the University of Colorado Denver.
vi


TABLE OF CONTENTS
CHAPTER
I. INTRODUCTION..............................................................1
The Public Health Problem: Pediatric HIV/AIDS.............................2
The Solution: Prevention of Mother-to-Child Transmission..................3
The Neglected Factor: Gender Power Dynamics...............................4
Research Questions and Hypotheses.........................................5
Chapter Summary...........................................................7
II. BACKGROUND & THEORETICAL FRAMEWORK........................................8
Prevention of Mother-to-Child Transmission in Zambia......................8
Individual-Level Perspective.............................................10
Family Systems Perspective...............................................12
The Theory of Gender and Power...........................................13
Conceptual Model.........................................................16
Gender, Power, and Vertical HIV Prevention...............................18
Economic Inequities..................................................17
Household Decision-Making............................................18
Intimate Partner Violence and Controlling Behavior...................19
Disclosure of HIV Status.................................................22
Chapter Summary..........................................................23
III. METHODS..................................................................24
Study Setting............................................................24
Research Design..........................................................25
Ethical Considerations...................................................26
vii


Formative Research and Preparation
26
Sampling Technique...........................................................30
Sample Description...........................................................32
Survey Measures..............................................................34
PMTCT Adherence..........................................................33
Gender Power Dynamics....................................................36
Quantitative Analytic Strategy...............................................38
Data Entry and Cleaning..................................................38
Dependent Variables...............................................39
Independent Variables.............................................40
Covariates........................................................43
Missing Data and Multiple Imputation.....................................45
Semi-Structured Interviews...................................................48
Qualitative Analysis.........................................................50
Limitations..................................................................52
Chapter Summary..............................................................53
IV. THE SEXUAL DIVISION OF POWER.................................................54
Review of Methods............................................................54
Intimate Partner Violence and Controlling Behaviors..........................54
Experiences With Specfic Forms of Violence and Control.......................57
Predictors of Intimate Partner Violence and Control..........................60
Positive Deviants............................................................65
Gender Asymmetry in Intimate Partner Violence................................68
Chapter Summary..............................................................70
viii


V. DRUG ADHERENCE DURING AND AFTER PREGNANCY
70
Review of Methods..........................................................71
Adequate and Optimal Drug Adherence........................................72
Unprotected Children at Risk of Vertical HIV Infection.....................75
Intimate Partner Violence and Drug Adherence...............................75
The Severity and Frequency of Partner Violence and Drug Adherence..........78
Specific Types of Partner Violence and Drug Adherence......................82
Partner Control and Drug Adherence.........................................87
Participation in Household Decisions and Drug Adherence....................87
Womens Relative Income and Drug Adherence.................................87
The Combined Influence of Gender Power Dynamics on Drug Adherence..........91
HIV Status Disclosure and Drug Adherence...................................93
Chapter Summary............................................................94
VI. SAFE INFANT FEEDING........................................................95
Review of Methods..........................................................97
Infant Feeding Practices...................................................97
Intimate Partner Violence and Safe Infant Feeding.........................100
The Severity and Frequency of Partner Violence and Safe Infant Feeding.....102
Specific Types of Violence and Safe Infant Feeding.........................108
Other Gender Power Dyanmics and Safe Infant Feeding.......................109
The Combined Influence of Gender Power Dynamics on Safe Infant Feeding.....113
HIV Status Disclosure and Safe Infant Feeding.............................114
Chapter Summary...........................................................115
IX


VII. PEDIATRIC HIV TESTING
116
Review of Methods...........................................................117
Adherence to Pediatric HIV Testing..........................................117
Intimate Partner Violence and Pediatric HIV Testing.........................119
The Severity and Frequency of Partner Violence and Pediatric HIV Testing....120
Specific Types of Partner Violence and Pediatric HIV Tesitng................123
Other Gender Power Dynamics and Pediatric HIV Testing.......................126
The Combined Influence of Gender Power Dynamics on Pediatric HIV Testing....129
HIV Status Disclosure and Pediatric HIV Testing.............................131
Adherence to Other PMTCT Protocols and Pediatric HIV Testing................132
Chapter Summary.............................................................134
VIII. PMTCT IN CONTEXT............................................................135
Review of Methods...........................................................135
Qualitative Sample Characteristics..........................................136
High Power and High PMTCT Adherence.........................................137
Prioritizing Their and the Childs Health.............................137
HIV Status Disclosure to Husbands.....................................138
Partner Support.......................................................139
HIV Status Concordance................................................140
Low Power and Low PMTCT Adherence...........................................141
Prioritizing the Marriage.............................................141
Non-Disclosure of HIV Status..........................................142
Lack of Partner Support...............................................144
x


Discordance and Emotional Abuse...................................144
Controlling Behavior..............................................145
Following the Husbands Advice....................................145
High Power and Low PMTCT Adherence......................................146
Poor Mental Health................................................146
Maternal Illness..................................................147
Poor Understanding of Protocols and Erratic Supply of Medication..147
Family Influence..................................................148
Low Power and High PMTCT Adherence......................................148
Family Support....................................................148
Child Death.......................................................149
Chapter Summary.........................................................150
IX. CONCLUSIONS............................................................151
1. Does Womens Low Power Within Couples Effect PMTCT Adherence?........151
1 .a. Which Power Dynamics Are the Most Detrimental?....................151
1. b. Which PMTCT Protocols Are the Most Affected?......................155
2. What is the Role of HIV Status Disclosure?...........................156
3. Why Do Gender Power Dynamics Affect PMTCT Adherence?.................157
Final Conceptual Model..................................................159
Addressing Gender Power Imbalances and PMTCT Adherence..................160
Next Steps..............................................................163
REFERENCES....................................................................164
xi


APPENDIX
A. Consent Form...............................................................194
B. Survey Questionnaire.......................................................196
C. Semi-Structured Interview Guide............................................208
D. My Data compared to the Most Recent ZDHS...................................211


LIST OF TABLES
TABLE
1. WHO Options for PMTCT Programs in 2014........................................9
2. Constructs and Questions on Gender Power Dynamics............................37
3. Missing Data for Key Variables...............................................46
4. Participant Characteristics..................................................47
5. Sampling Domains for Semi-structured Interviews..............................48
6. Example of Qualitative Coding Scheme.........................................52
7. Multivariate Logistic Regression Results for the Odds of Experiencing IPV and Three
or More Partner Controlling Behaviors........................................61
8. Correlation Coefficients Between Gender Power Dynamic Measures...............64
9. Intimate Partner Violence and the Adjusted Odds of Drug Adherence............77
10. Adjusted Logistic Regression Results for the Odds of Drug Adherence by the Number
of Violent Events a Woman Experienced........................................78
11. Adjusted Logistic Regression Results for the Odds of Drug Adherence by the
Frequency of Violence in the Past Year.......................................79
12. Adjusted Logistic Regression Results for the Odds of Drug Adherence by Injuries
from Intimate Partner Violence...............................................82
13. Adjusted Logistic Regression Results for the Odds of Drug Adherence by Experiences
with Emotional Violence......................................................85
14. Adjusted Logistic Regression Results for the Odds of Drug Adherence by Experiences
with Sexual Violence.........................................................86
xiii


15. Adjusted Logistic Regression Results for the Odds of Drug Adherence by Experiences
with Physical Violence.....................................................87
16. Partner Controlling Behaviors and the Adjusted Odds of Drug Adherence......89
17. Womens Participation in Household Decision-Making and the Adjusted Odds of Drug
Adherence..................................................................90
18. Womens Relative Income and the Adjusted Odds of Drug Adherence............91
19. Adjusted Logistic Regression Results for the Odds of Drug Adherence by Gender
Power Dynamics in Combined Models..........................................92
20. HIV Status Disclosure and the Adjusted Odds of Drug Adherence..............94
21. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices
by Experiences with Intimate Partner Violence.............................101
22. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices
by the Number of Violent Events...........................................103
23. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices
by the Frequency of Violence in the Past Year.............................104
24. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices
by Injuries from Violence.................................................105
25. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices
by Emotional Violence.....................................................107
26. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices
by Sexual Violence........................................................108
27. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices
by Physical Violence......................................................109
xiv


28. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices
by Partner Controlling Behavior..........................................110
29. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices
by Womens Participation in Household Decision-Making....................111
30. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices
by Womens Relative Income...............................................112
31. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices
by Gender Power Dynamics.................................................113
32. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices
by Gender Power Dynamics.................................................115
33. Logistic Regression Results for the Odds of Pediatric HIV Testing by Experiences with
Intimate Partner Violence................................................119
34. Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by The
Number of IPV Events.....................................................120
35. Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by the
Frequency of IPV in the Past Year........................................121
36. Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by Injuries
Experienced as a Result of Violence......................................122
37. Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by
Emotional Violence.......................................................123
38. Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by Sexual
Violence.................................................................124
xv


39. Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by
Physical Violence..............................................................125
40. Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by Partner
Control........................................................................127
41. Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by
Womens Participation in Household Decisions...................................128
42. Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by
Womens Relative Income........................................................129
43. Adjusted Logistic Regression Results for the Odds of Pediatric HIV Testing by Gender
Power Dynamics ................................................................130
44. Adjusted Logistic Regression Results for the Odds of Pediatric HIV Testing by HIV
Status Disclosure..............................................................132
45. The Odds of Pediatric HIV Testing by Other PMTCT Protocols.....................133
46. Qualitative Sample Characteristics.............................................137
xvi


LIST OF FIGURES
FIGURE
1. Map of Zambia..............................................................2
2. The PMTCT Cascade of Care..................................................3
3. Conceptual Model..........................................................16
4. Map of Lusaka and Kanyama Compound........................................24
5. Study Timeline............................................................27
6. Children's Under-Five Card................................................31
7. Sample Description........................................................32
8. Example of Drug Adherence Question........................................34
9. Women's Experiences with IPV and Partner Controlling Behaviors............55
10. Womens Experiences With Specific Violent Events..........................57
11. Experiences with Specific Forms of Male Controlling Behavior..............58
12. Types of Intimate Partner Violence Reported...............................59
13. Household Decision-Making and Experiences with IPV/Controlling Behavior...66
14. IPV Victimization in Physically Violent Relationships by Gender...........68
15. Adequate and Optimal Drug Adherence Levels................................72
16. Number of Medication Protocols Where Women Achieved Adequate
Adherence................................................................73
17. Proportion of Children Not Fully Protected by PMTCT at Each Time Point....74
18. Adjusted Odds Ratio Estimates for Adherence to PMTCT Medication at Each Time
Period by the Number of Violence Events...................................80
xvii


19. Adjusted Odds Ratio Estimates for Adherence to PMTCT Medication at Each Time
Period by the Frequency of IPV in the Past Year Score....................81
20. Proportion of HIV-infected Women Practicing Three Types of Infant Feeding
Modalities...............................................................97
21. Reported Food/Liquids Given to Infants Prior to Six Months Among HIV-Infected
Women Practicing Mixed Feeding..........................................100
22. Adjusted Multinomial Odds Ratio Estimates for Adherence to Safe Infant Feeding by
the Number of IPV Events.................................................106
23. Adjusted Multinomial Odds Ratio Estimates for Adherence to Safe Infant Feeding by
the Frequency of IPV in the Past Year Score..............................106
24. Proportion of Children with Any Pediatric HIV Testing....................118
25. Proportion of Children Testing at 6 Weeks and 6 Months...................118
26. Adjusted Odds Ratio Estimates for Adherence to Both Recommended Pediatric HIV
Tests by the Frequency of IPV in the Past Year...........................122
27. Final Conceptual Model...................................................159
xvm


LIST OF ABBREVIATIONS
AIDS Acquired immune deficiency syndrome
ANC Antenatal care
ARV Antiretroviral
ART Antiretroviral therapy
CTS2 Conflict Tactics Scale
HIV Human immunodeficiency virus
IPV Intimate partner violence
IQR Interquartile range
MCH Maternal and child health
NVP Nevirapine
PCA Principal component analysis
PMTCT Prevention of mother-to-child transmission
PTSD Post-traumatic stress disorder
sdNVP Single-dose Nevirapine
STI Sexually transmitted infection
UTH University Teaching Hospital
VAS Visual analog scale
VCT Voluntary counseling and testing
WHO World Health Organization
ZDHS Zambian Demographic and Health Survey


CHAPTER I
INTRODUCTION
Throughout my public health career, one of the most emotionally challenging events I have witnessed was an HIV-positive mother being informed by a nurse that her child is also HIV-positive. The reason this is so heart wrenching is because vertical (i.e., mother-to-child) HIV transmission is almost entirely preventable through biomedical and behavioral interventions, such as antiretroviral (ARV) medication and safe infant feeding practices. Without intervention, however, approximately 40% of infants bom to HIV-positive women will become infected in utero, during childbirth, or while breastfeeding (WHO, 2010c). Zambia is one of the countries hardest struck by HIV one in five pregnant women are currently living with the vims (CIDRZ, 2012). Zambia also has one of the highest global fertility rates with an average of almost six children per woman (CIA, 2014). This translates into approximately 78,000 HIV-positive women giving birth annually with 15% currently transmitting the vims to their children (UNICEF, 2014) do to lack of access and poor adherence to PMTCT care.
My research took place in the capital city of Zambia, Lusaka (see Figure 1), which has a population of 1.8 million individuals (CIA, 2014). I chose to conduct my research here for two reasons. First, women of reproductive age living in Lusaka have disproportionately high HIV rates (CSO, 2009). The most recent estimates indicate that the overall HIV prevalence for adults1 in the country is 13%, yet women in Lusaka have an HIV prevalence of over 19% (CSO, 2014). Second, Lusaka has the best health care in the country, making it an ideal location to study non-adherence to prevention of mother-to-child transmission (PMTCT) because generally, women have access to health care.
1 Estimate is for adults age 15-49 years
1


Figure 1 Map of Zambia
Interpersonal relationships, including the family, are a critical influence over health behaviors (Glanz et al., 2008). Within families, however, sexual divisions exist leading to more or less relative power between heterosexual partners. One aspect that has been neglected in the PMTCT literature is the role that gender power dynamics within married couples may play on HIV-positive womens ability to adhere to various PMTCT protocols in high HIV prevalence settings, such as Zambia. In this chapter, I first describe the disease burden of mother-to-child transmission of HIV and why gender power dynamics are a likely contributing factor for womens non-adherence to PMTCT, followed by my a priori research questions and hypotheses.
2


The Public Health Problem: Pediatric HIV/AIDS
Although tremendous improvements have been made, hundreds of thousands of infants unfortunately continue to contract HIV each year. In 2013, 240 0002 children became newly infected with HIV that is approximately 660 children every day (UNIADS, 2014). The main route of new pediatric HIV infections is from mother-to-child, which accounts for 15% of all new HIV infections annually a substantial proportion of the overall global HIV incidence (Msellati, 2009; UNAIDS, 2012; WHO, 2010c). Sub-Saharan Africa carries the largest pediatric HIV burden with 94% of HIV-infected children living in this region (WHO, 2013). In Zambia alone, over 10,000 infants are newly infected with HIV annually (UNAIDS, 2014).
Moreover, pediatric HIV is a major contributor to infant and child mortality in sub-Saharan Africa (Khan, Michaels, & Eley, 2006). In southern Africa, where Zambia is located, 20% of child deaths are attributable to HIV infection (Torpey et al., 2012). HIV progresses more rapidly to AIDS in children, and without treatment, 50% of infected children die by their second birthday (Zambia Ministry of Health, 2008). Every year, there are over 260 000 pediatric deaths worldwide due to AIDS-related illnesses one in seven of all AIDS-related deaths (WHO, 2010d). In sub-Saharan Africa, AIDS remains one of the leading causes of death for children less than five years (Khan et al., 2006). If the epidemic of mother-to-child transmission of HIV is not halted, pediatric AIDS threatens to reverse years of steady progress in infant and child survival in the region (UNAIDS, 1999). Thus, eliminating the transmission of HIV from mother-to-child is currently one of the most pressing global health priorities.
2 Estimate is from 210,000 to 280,000
2


The Solution: Prevention of Mother-to-Child Transmission
Fortunately, vertical HIV transmission is almost entirely preventable through interventions commencing during pregnancy and continued throughout the breastfeeding period, known as the PMTCT cascade of care. The cascade of care for HIV-positive women and their children includes an HIV diagnosis during antenatal care (ANC); maternal ARV medication during and after pregnancy; ARV medication during childbirth (if applicable); infant ARV prophylaxis; specific infant feeding practices; and pediatric HIV testing (see Figure 2). If HIV-positive women adhere to these protocols, the risk of vertical HIV transmission is reduced to less than 5% (WHO, 2010a).
Source: Based on 2010 WHO PMTCT Guidelines for Zambia
Figure 2 The PMTCT Cascade of Care
Major strides have been made at the political and institutional levels to increase coverage and accessibility of PMTCT services throughout sub-Saharan Africa (WHO, UNAIDS, & UNICEF, 2011). In 2003, only 3% of HIV-positive pregnant women living in the region had access to ARV drugs for the purpose of PMTCT (Hardon et al., 2012). This percentage dramatically increased to 33% in 2007 and 59% in 2011 (WHO, 2010d).
3


Commendably, Zambia is one of six countries in sub-Saharan Africa currently reporting PMTCT drug coverage of more than 75% (UNAIDS, 2012). During my field research, every one of the 25 public health centers in Lusaka offered PMTCT care.3
Despite these admirable gains in access, adherence to PMTCT among HIV-positive women remains a challenge across sub-Saharan Africa (Nachega et al., 2012). To date, the global health communitys efforts to eliminate vertical HIV transmission have been primarily focused on scaling up biomedical services. While this is an essential component, it has left a gap in our understanding of the social and behavioral determinants of womens adherence to PMTCT. There is currently a poor understanding of why HIV-positive women are not adhering to PMTCT protocols despite their accessibility throughout much of urban sub-Saharan Africa. This knowledge is critical for our ability to create effective behavioral interventions aimed at curbing the HIV epidemic and reducing child mortality in the region. The Neglected Factor: Gender Power Dynamics
Gender is fundamentally a matter of social embodiment and is considered one of the social forces with the greatest constraint over human agency (Connell, 1987; Rylko-Bauer, Whiteford, & Farmer, 2009). For this research, I define gender as the culturally expected norms shared within a society about appropriate male and female behavior, characteristics, and roles (Blanc, 2001). Gender is an essential consideration in preventing vertical HIV because only women can pass HIV to offspring, and, for this reason, protocols to prevent transmission are aimed exclusively at women. Paradoxically, however, HIV-positive women may not hold the necessary power within the family to make independent decisions regarding adherence to the various PMTCT protocols in many settings, such as Zambia.
3 Although stock-outs of medication, particularly infant Nevirapine, were common
4


One major theoretical shortcoming in the current literature on PMTCT adherence is a lack of examination into dynamics within heterosexual couples, particularly gender power imbalances. Policies and counseling regarding PMTCT in Zambia do not adequately take into consideration dynamics within couples affected by HIV, which may exert a large influence over HIV-positive womens relevant health behaviors. HIV-positive women are assumed to have the necessary agency and power to make these health care decisions, even in cultures where large gender inequities persist in society and the home.
Research Questions and Hypotheses
The overall aim of this study is to examine the relationship between gender power dynamics within married couples, which stem from larger societal inequities, and adherence to PMTCT protocols across the cascade of care for HIV-positive women and their children. The theoretical framework for this study comes primarily from the Theory of Gender Power and Family Systems Theory. Using a concurrent mixed methods approach, including a survey and semi-structured interviews, I test whether a relationship exists between womens low power within married couples and adherence to PMTCT. I specifically analyze how intimate partner violence (IPV) and control, household decision-making, and economic inequity influence HIV-positive womens PMTCT-related behaviors. I additionally explore the role of HIV status disclosure to husbands and why gender power dynamics may have a negative influence on HIV-positive womens PMTCT adherence. The specific research questions and a priori hypotheses of my study are:
Question 1: Does womens low power within couples reduce the odds of PMTCT adherence?
5


Hypothesis 1: There is a relationship between womens low power within couples and non-adherence to PMTCT. The presence of violent or controlling behavior by a husband, low household decision-making power, and womens economic dependence on the husband will reduce the odds of PMTCT adherence.
Question 1A: Which power dynamics are the most likely to hinder PMTCT adherence?
Hypothesis 1A: HIV-positive women in relationships characterized by high levels of male partner controlling behavior (e.g., through violence/threat of violence, economic control, etc.) will have the lowest PMTCT adherence.
Question IB: Which PMTCT protocols are most affected by gender power imbalances within couples?
Hypothesis IB: Gender power imbalances will have the strongest effect on womens adherence to postpartum protocols because these are the most perceptible to a spouse (i.e., maternal medication, infant medication, and safe infant feeding practices). Gender power imbalances will have the weakest effect on childbirth protocols because this is a single-dose medication taken at the clinic without the need for a husbands awareness or permission.
Question 2: What is the role of HIV status disclosure in the relationship between gender power dynamics and PMTCT adherence?
Hypothesis 2: HIV status disclosure mediates the relationship between gender power dynamics and non-adherence to PMTCT.
Question 3: Why do gender power dynamics within couples affect HIV-positive womens ability to adhere to various PMTCT protocols?
6


Hypothesis 3: Relationships with large power inequities are characterized by less
spousal communication and support, making PMTCT more arduous on the woman. Additionally, fear of violence or abandonment prevents women from disclosing their HIV status, leading to difficulty adhering to PMTCT.
Chapter Summary
Eliminating mother-to-child transmission is one of the top global HIV/AIDS priorities, but poor understanding of the social barriers to PMTCT adherence among HIVpositive women is limiting progress towards this goal. Currently, there is limited knowledge regarding how gender power dyanmics within married couples affect HIV-positive womens adherence to PMTCT in sub-Saharan Africa. The present study addresses this gap by testing the hypothesis that womens low relative power within couples decreases adherence across the PMTCT cascade of care in Lusaka, Zambia. In the following chapter, I provide a review of the current PMTCT literature and describe the theoretical framework for this research. Chapter Three then describes my methodology, followed by my specific findings in Chapters Four through Eight.
7


CHAPTER II
BACKGROUND & THEORETICAL FRAMEWORK Prevention of Mother-to-Child Transmission in Zambia
Zambia has had a national PMTCT program in place since 1999; however, Zambia and other African countries PMTCT recommendations have significantly evolved over the past decade with growing scientific evidence. Prior to 2010, Zambia and many other African countries only offered single dose Nevirapine (sdNVP) intrapartum (i.e., at the time of labor/delivery) to HIV-positive women for PMTCT. This intervention occurred at the health facility in the delivery room out of sight of the husband, which enabled women to adhere to the medication without having to disclose their HIV status to the husband and/or obtain his permission. In 2010, PMTCT protocols in Zambia became more complicated because of improved medical knowledge, recommending an extended duration of taking combination ARVs throughout pregnancy and postpartum, including providing the infant with prophylaxis.
At this same time, the World Health Organization (WHO) began recommending, and still do recommend, that HIV-positive women in low/middle income settings like Zambia exclusively breastfeed their infants for six months, followed by the introduction of complementary foods and continue breastfeeding to at least one year of life (WHO, 2010b). In addition, infants who are exposed to HIV from their mothers should be tested for the virus at 6 weeks, 6 months, 12 months, and 18 months, and started on antiretroviral therapy (ART) immediately if infected (Zambia Ministry of Health, 2010).
Due to the extended duration of taking ARVs and the increased number of necessary PMTCT interventions that commenced in 2010, it is plausible that adherence to PMTCT
8


became more challenging for HIV-positive women. This may be particularly true for women who have low relative power within the family because the current recommendations require adherence to protocols that take place outside of the health care facility and in the home.
Table 1 WHO Options for PMTCT Programs in 2014
Woman receives Infant receives
Option Treatment (for CD4 count <350 cells/mm) Prophylaxis (for CD4 count >350 cells/mm)
A* Triple ARVs starting as soon as diagnosed, continued for life Antepartum: AZTi starting as early as 14 weeks gestation Intrapartum: at onset of labor, single dose NVP and first dose of AZT/3TC2 Postpartum: daily AZT/3TC through 7 days postpartum Daily NVP from birth until 1 week after cessation of all breastfeeding; or, if not breastfeeding or if mother is on treatment, through age 4-6 weeks
B Same initial ARVs for both: Daily NVP or AZT from birth through 4-6 weeks regardless of infant feeding method
Triple ARVs starting as soon as diagnosed, continued for life Triple ARVs starting as early as 14 weeks gestation and continued intrapartum and through childbirth if not breastfeeding or one week after cessation of all breastfeeding
B+ Same treatment and prophylaxis: Daily NVP or AZT from birth through 4-6 weeks regardless of infant feeding method
Regardless of CD4 count, triple ARVs starting as soon as diagnosed, continued for life
1 AZT = zidovudine (specific type of ARV)
2 AZT/3TC = zidovudine/lamivudine (specific type of combination ARV)
Source: WHO. 2012 Programmatic Update: Use of Antiretroviral Drugs for Treating Pregnant Women and Preventing HIV Infection in
Infants
At the time of the study (2014), WHO recommendations included three PMTCT options for countries to target vertical HIV transmission based on their individual health system capabilities (See Table 1). Zambia opted for Option A4 in 2010, which included two different medication regimens depending on womens level of HIV disease progression.
4 In 2014, towards the end of my field research, Zambia began transitioning to Option B+, which is what the country currently follows. Essentially this is the same regimen as the treatment option under Option A, but all HIV-positive women regardless of their CD4 count or clinical stage initiate lifelong ART.
9


First, pregnant or breastfeeding women with a CD4 cell count5 of less than 350 or WHO clinical stage three or four commenced lifelong ART. Second, HIV-positive women who did not meet the treatment criteria were given short-course ARV prophylaxis starting at 14 weeks gestation and continued until one week postpartum (WHO, 2012a). HIV-exposed infants received Nevirapine (NVP6) prophylaxis on both regimens, but for differing amounts of time.
While effective biomedical interventions and national policies exist, health behaviors and compliance to these interventions is a critical, underemphasized, component of PMTCT success. A thorough understanding of womens PMTCT-related behaviors is essential to achieve the goal of global eliminating vertical HIV transmission. In the next sections, I explain the theoretical perspectives typically applied to the analysis HIV-positive womens behaviors related to PMTCT, and why a focus on the familial-level, in particular gender power dynamics, is imperative.
Individual-Level Perspective
Individual-level theories of health behavior, such as the Health Belief Model (Glanz, Rimer, & Viswanath, 2008) and the Theory of Planned Behavior (Glanz et al., 2008) have been widely used to explain HIV-positive womens PMTCT-related behaviors. The construct of perceived barriers from the Health Belief Model is a somewhat helpful construct in understanding womens PMTCT behaviors. For example, established perceived barriers in the literature on PMTCT include fear of knowing ones own HIV status; stigma and discrimination of ones HIV status being disclosed to sexual partners, family, or the community; and opposition from the husband (Kebaabetswe, 2007; Nyasulu & Nyasulu, 2011).
5 Blood test for HIV-positive individuals to diagnosis HIV progression/stage of disease
6 A type of ARV drug given in syrup form to infants
10


Additionally, the Theory of Planned Behavior has been used to explain individual factors that may motivate an HIV-positive woman to use PMTCT. Constructs of attitude, perceived norms, and personal agency are frequently cited as ways to understand and address womens PMTCT adherence (Awiti Ujiji et al., 2011; Hardon et al., 2012; Kebaabetswe, 2007; Nyasulu & Nyasulu, 2011; Varga & Brookes, 2008). For example, a study in South Africa used the Theory of Planned Behavior to analyze a clinic-based health education intervention aimed to increase womens intention to use PMTCT in South Africa (Igumbor, Pengpid, & Obi, 2006). One important finding from the 2006 study by Igumbor et al., however, is that HIV-positive women consistently reported low control beliefs regarding PMTCT, and as a result, the authors recommended womens empowerment efforts. Another study by Igumbor and Obi (2006) recommended expanding and enhancing interventions that empower HIV-positive women in order to improve behavioral intention to use PMTCT. Womens empowerment appears to be an under-researched yet crucial component of increasing PMTCT uptake and adherence in sub-Saharan Africa.
PMTCT interventions focused only on the individual-level (i.e., HIV-positive woman) typically attempt to increase knowledge and behavioral intentions through education and counseling (Creek et al., 2009; Igumbor et al., 2006). For example, Msellati (2009) recommended a major information campaign focused on the advantages for pregnant women and their future children to increase PMTCT adherence (Msellati, 2009, p. 810). Currently in many sub-Saharan African countries, however, such as Zambia, widespread PMTCT and HIV education campaigns are already in place (Besser, 2010a; MOH, 2008), but PMTCT adherence is poor nonetheless. For instance, over 88% of women knew that HIV could be transmitted by breastfeeding in the last Zambia Demographic and Health Survey
11


(ZDHS), but only 21% of HIV positive mothers took the necessary ARVs while breastfeeding to prevent HIV transmission (UNICEF, 2012). Education regarding PMTCT is certainly a necessary component, but not sufficient by itself for behavior change.
The problem with using only individual-level constructs to understand PMTCT adherence is that the sole responsibility to prevent HIV transmission is placed on the HIV-infected mother, without taking into consideration the context in which her behaviors occur (Rylko-Bauer et al., 2009). This grossly overestimates the power HIV-positive women typically have, both in society and within intimate sexual relationships, especially in populations that are historically patrilineal7 and have large gender inequalities, such as Zambia. Perhaps it is no surprise, then, that individual-level interventions aimed at PMTCT adherence focused solely on education or behavioral intention have not been successful (see Igumbor et. al., 2006).
Family Systems Perspective
The family is arguably the most influential social unit in regard to health behaviors due to both the proximity to the individual and the longevity of these relationships (Simons-Morton, Hayne, & Noelcke, 2009). Moreover, marriage within African culture is one of the most significant social relationships (Smith & Mbakwem, 2007). It is also relatively well established that in many African societies, women are often not independent decisionmakers regarding the use of healthcare and need a husbands support/permission to use services (Auvinen, Suominen, & Valimaki, 2010). Given this background, it is perplexing that womens low relative power in couples has not been more thoroughly considered in the analysis of PMTCT adherence (For exceptions, see Hatcher et al., 2014; Kiarie et al., 2006; and Mepham et al., 2011).
7 Inheritance is passed from men to their children and women traditionally come to live in the mans village.
12


Gender and power are interrelated factors that influence health behaviors. Unfortunately, gender-based power in intimate relationships is often unbalanced, with women typically having less power (Blanc, 2001). In this study, I conceptualize power as the ability of one person to control another, occurring at the societal, institutional, and interpersonal levels of the social ecological model (Yoder & Kahn, 1992). Within the context of this study, I use the definition put forth by Blanc (2001), stating that power in married couples refers to the relative ability of one partner to act independently, to dominate decision-making, to engage in behavior against the partners wishes and to control a partners actions (p. 189).
The Theory of Gender and Power
The Theory of Gender and Power claims that relative power within heterosexual relationships arises from the global dominance of men over women (Connell, 1987). The patriarchal ordering of power and privilege is not only embedded in our social and political systems but also permeates into families, negatively affecting all members (Hare-Mussen, 1978; Libow, Raskin, & Caust, 1982). There are three relevant interrelated social structures related to this gender order: the sexual division of labor, the sexual division of power, and the structure of cathexis (Connell, 1987). These societal structures are exhibited both at the institutional level (e.g., family and relationships) and through social mechanisms (e.g., imbalances in control power and disparities in social norms and expectations; Connell, 1987).
Research on HIV/AIDS in sub-Saharan Africa using the Theory of Gender and Power has primarily focused on womens risk of contracting HIV through heterosexual intercourse. Aside from increased biological vulnerability to HIV compared to men, women are also more vulnerable to heterosexual HIV acquisition for numerous social reasons, including sexist
13


cultural norms, psychosocial and legal factors, economic vulnerability, and attitudes regarding feminine sexuality that are conducive to unsafe sexual practices (Travers & Bennett, 1996). All of these factors combined result in women generally having less relative sexual negotiating power within their intimate relationships compared to men. For example, women may be unable to enforce condom use with male partners or control their partners extramarital sexual practices because of gender power imbalances, increasing the likelihood that women will be exposed to, and eventually contract, HIV (Dunkle et al., 2004; Langen, 2005). Indeed, numerous studies indicate that gender power imbalances are associated with increased HIV prevalence rates for women across sub-Saharan Africa (Drain, Smith, Hughes, Halperin, & Holmes, 2004; Gupta, Parkhurst, Ogden, Aggleton, & Mahal, 2008).
I argue that the Theory of Gender and Power is also a useful framework when investigating HIV-positive womens PMTCT-related behaviors because many of the same social factors that put women at increased risk of sexual HIV acquisition likely also prevent them from adhering to PMTCT protocols. In this study, I specifically focus on inequities of power and control within married couples, which are a result of gendered social structures (Michalski, 2004) and may exert strong influence over womens PMTCT-related health behaviors.
The first structure/domain from the Theory of Gender and Power, the sexual division of labor, is essentially the allocation of particular types of work to men and women, which can become a constraint on additional aspects of life due to disparities in allocation of resources (Connell, 1987). This structure is related to the organization of household work and childcare, the division between paid and unpaid work, the segregation of labor markets through the creation of mens jobs and womens jobs, and unequal wages and unequal
14


exchanges (Connell, 1987). Inequity in skills and training make the sexual division of labor a powerful discriminatory mechanism creating a social constraint for women (Connell, 1987). The result is womens lower educational and economic status compared to men in many societies, often leaving women dependent on their male partners.
The second structure, the sexual division of power, describes how gender power inequities in society and within relationships isolate women from the ability to control their lives (Travers & Bennett, 1996). This may also include HIV-positive womens ability to control decisions and behaviors regarding PMTCT adherence. The sexual division of power is related to authority, control, and coercion, both of institutions over individuals and other individuals over individuals (e.g., husbands over wives). Masculinity in most societies is closely tied to authority, giving men increased power over women (Connell, 1987), which of course, permeates into familial life. Power within families can be used to win contested decisions and dictates who is charged with carrying out decisions, who monitors this process, and who is accountable to whom (Broderick, 1993).
The final structure, the structure of cathexis, describes affective attachments and social norms (Wingwood, Camp, Kristin, Cooper, & DiClemente, 2009). At the societal level, this structure describes gender norms and is characterized by the emotional and sexual attachments between women and men (Wingwood et al., 2009). The structure of cathexis creates behavioral constraints because cultural gender roles assigned to women typically involve subordination to men (Travers & Bennett, 1996). Indeed, many Zambian women report being counseled upon reaching puberty to be submissive and obey their future husbands (Human Rights Watch, 2007). Also related to the structure of cathexis are norms around mens power, including physical, sexual, or emotional violence. The use of violence
15


to maintain male dominance tends to be acceptable in society, especially where laws and customs combine to uphold the differential power between men and women (Michalski, 2004).
Conceptual Model
Based on the Theory of Gender and Power within family systems, Figure 3 presents my conceptual model, taking into consideration the potential relationship between PMTCT adherence and gender power imbalances within marriages related to the structure of cathexis and the sexual divisions of power and labor. I hypothesize that power imbalances within couples, conceptualized through economic inequity, household decision-making, and IPV/control, decreases adherence to protocols across the PMTCT cascade of care and that womens HIV status disclosure to husbands mediates this relationship. In the following sections, I describe the hypothesized pathways in the conceptual model, highlighting what is known in the current literature and the gaps that this study will fill. In the conclusion of the dissertation, I provide an updated conceptual model detailing the hypotheses that my data ultimately supported.
Economic
inequity
Low household decisionmaking ability
Intimate
partner
violence
Partner
controlling
behaviors
(Controlling for covariates of age, infant age, parity, education, knowledge of PMTCT, wealth, and type of PMTCT regimen)
Figure 3 Conceptual Model
16


Gender, Power, and Vertical HIV Prevention
Economic Inequities
The sexual division of labor in Zambia is readily apparent in the educational and employment statistics disaggregated by gender. Zambian Women generally receive less education and have fewer economic opportunities than men. For instance, in the most recent ZDHS, only 8% of Zambian women completed secondary education, compared to 13% of their male counterparts. In terms of employment status, only 49% of Zambian women were formally employed at the time of the survey, compared to 74% of men (CSO, 2014). There are also very few female politicians or women in power in the country to advocate for womens rights. As a result of the sexual division of labor at the societal level, socioeconomic disparities exist between men and women at the interpersonal level.
Due to income inequity within marriages, wives are often economically reliant on their husbands, creating power imbalances in the relationship. This notion has been termed Marital Dependency Theory (Kalmuss & Straus, 1990). Womens economic dependence on the husband constrains their ability to make independent decisions on, for example, the use of health care (Blanc, 2001). A woman may be at a disadvantage if she needs to ask her partner for money to pay for the transportation costs associated with getting to a health clinic for PMTCT services8 if she does not have an independent source of income of her own (Duff, Kipp, Wild, Rubaale, & Okech-Ojony, 2010; Tuller et al., 2010).
In addition, fear of abandonment has been cited by HIV-positive women in sub-Saharan Africa as a reason for why they have not disclosed their HIV status to their husband (Kebaabetswe, 2007; Nyasulu & Nyasulu, 2011) which may be related both to economic
8 PMTCT services at all public health clinics throughout Zambia are free through funding from the Presidents Emergency Plan for HIV/AIDS Relief (PEPFAR)/USAID and the Global Fund. Thus, there are no health care costs related to these services for HIV-positive women/mothers.
17


reliance and the social significance of being married. If a woman is economically reliant on her husband, abandonment is a legitimate concern since she may not possess the necessary education or skills to obtain her own income. For this reason, a woman may be highly motivated to hide the fact that she is HIV-positive from her husband, which could in turn, result in poor adherence to PMTCT.
Household Decision-Making
As a result of the aforementioned economic inequities, in addition to sexist cultural practices, women often have low participation in household decision-making. This is consistent across many countries, including Zambia. Kishor and Subaiya (2008) conducted a review of womens empowerment, including participation in household decisions, across 23 developing countries and found that there are no countries in which women make the majority of household decisions alone. In addition, they report that in most countries, there is strong agreement about mens right to control women (Kishor & Subaiya, 2008). In Zambia, for example, 24% of women on the last ZDHS reported that their husband alone makes decisions regarding visits to family and friends and 33% reported the husband alone makes decisions regarding major household purchases (CSO, 2014).
Most indicators and accounts would consider Zambia a highly patriarchal society. This is apparent at both the structural and interpersonal levels. Men have considerably more control over social, political, economic, cultural, and familial institutions. This partially stems from the fact that most tribes in Zambia are patrilineal, meaning inheritance passes from the husband to his children and women traditionally left their villages to live with the husbands family (Barfield, 1997). In addition, the cultural practice of bride price is common in Zambia, where a husband will pay the wifes family (usually the father) not the
18


wife a set amount of money or other commodity of value, such as cows, as compensation for the bride (Barfield, 1997). In some cases, this may result in mens perception of ownership over the woman.
In addition, it is well established that husbands often serve as gatekeepers to womens health care, such as a woman needing permission to seek HIV testing or treatment (Heckert & Fabic, 2013; Human Rights Watch, 2007; Ngom, Debpuur, Akweongo, Adongo, & Binka, 2003). Indeed, over one-fourth of Zambian women reported that their husband alone makes decisions about the womans use of healthcare on the last DHS (CSO, 2014). Generally speaking, if a husband does not want the woman to use PMTCT for whatever reason, if she has low relative power in the marriage regarding household decisions, she will be obliged to agree with his demands (Auvinen, Suominen, & Valimaki, 2010). Indeed, during my interviews with women in 2009 for my masters program, Zambian women repeatedly made statements, such as he [the husband] is the head of the household and whatever he says goes in reference to their use of pediatric HIV services.
Intimate Partner Violence and Controlling Behavior
IPV against women is one of the most compelling manifestations of unequal power in sexual relationships and the larger phenomenon of gender inequality (Blanc, 2001). For this study, IPV is conceptualized as actual or threatened physical or sexual violence or psychological/emotional abuse directed towards a female spouse that is part of a general strategy of power and control (Johnson, 2008; Population Council, 2008). IPV is thus an extreme expression of male domination and female subordination through the use of both violence and other control tactics (Michalski, 2004). It is important to note that male-directed
19


IPV most certainly exists; however, because this study focuses on womens adherence to PMTCT protocols, the concentration of this research is on female-directed IPV.
In societies where there is unequal access to economic or political resources by gender, the likelihood of IPV against women increases (Levinson, 1989). Additionally, a woman may be forced to remain in an abusive, controlling relationship because of her economic dependence on the husband as discussed above or because her family would have to repay the bride price. Based on traditional Zambian patrilineal customs, in the event of a divorce, the land and the children generally go to the husband and his family, which may prevent women from leaving abusive situations.
In Zambia, both IPV against women and male partner controlling behaviors are normative. Almost half of all women (47%) reported on the last ZDHS that a husband is justified in beating his wife for at least one specified reason, such as burning the food or arguing with him. In addition, almost half (47%) of all women also reported experiencing physical or sexual violence from their spouse and 24% reported experiencing emotional violence. Lastly, 35% of women reported that their husband displayed three or more controlling behaviors, such as insisting on knowing where she is at all times (CSO, 2014).
Although a quantitative relationship between IPV/controlling behavior and PMTCT has not been established, there is cause to believe an association exists. IPV is associated with numerous negative health outcomes among women and children, including reproductive outcomes, such as low birth weight, preterm delivery, and maternal and infant mortality (Boy & Salihu, 2004; Emenike, Lawoko, & Dalai, 2008). IPV also increases the risk of having gynecological problems, including sexually-transmitted infections (STIs) and HIV (Campbell, 2002; Harvey, 2007; Garcia-Moreno, et al., 2007). In addition, IPV increases
20


emotional distress, depression, and post-traumatic stress disorder (PTSD) among women, which are known barriers to medical adherence in general (Ellsberg, Jansen, Heise, Watts, & Garcia-Moreno, 2008; Fischbach & Herbert, 1997). Lastly, IPV has been shown to hinder HIV-related health behaviors, including HIV testing and ART adherence for womens own health (Maman, Campbell, Sweat, & Gielen, 2000; Maman, Mbwambo, Hogan, Kilonzo, & Sweat, 2001).
A qualitative study from 2007 conducted by Human Rights Watch found that Zambian women were physically, emotionally, and sexually abused by their husbands upon disscussing HIV testing or treatemnt or after disclosing their HIV positive status to the husband. In addition, fear of abandoment and divorce in an environment where women suffer insecure property rights were major impediments to HIV-positive womens ability to start and continue using ART for their own health (Human Rights Watch, 2007). Although this provides compelling evidence that IPV negatively impacts adhernce to ARV drugs for Zambian womens own health, the study did not specifically examine ARV adherence for the purpose of PMTCT. Based on this study, however, there is reason to believe that PMTCT would also be negatively affected by IPV, but this warrants further research to establish any conclusive associations.
More recently, two qualitative studies from South Africa reported that IPV or fear of IPV was a barrier for PMTCT (Hatcher et al., 2014; Mepham, Zondi,
Mbuyazi, Mkhwanazi, & Newell, 2011). However, these studies did not examine PMTCT protocols besides medication, such as infant feeding or pediatric HIV testing.
In addition, the findings of these studies cannot be generalized to the Zambian
21


context, although they do inform my hypotheses about IPV and other power dynamics and adherence cross the PMTCT cascade.
To my knowledge, the only study that has used quantitative data to examine PMTCT and IPV in sub-Saharan Africa, Kiarie et al. (2006), who found no association between IPV and uptake of PMTCT in Kenya. However, there are several explanations for why I believe more research is needed in this area beyond the Kiarie et al. (2006) study. First, Kiarie et al. examined physical, financial, and psychological abuse but did not include questions regarding partner control, household decision-making, or economic dependence key factors related to womens relative power within a couple. Second, the operational definition of PMTCT in the Kiarie et al. study was limited to medication during childbirth (i.e., sdNVP), which is the protocol most under medical personnels control9 and the least observable to the husband since it occurs within the delivery room at a clinic. At the time of the Kiarie et al. research, sdNVP was the only PMTCT protocol in place in Kenya. However, many changes have occurred since then in terms of PMTCT guidelines, and thus, a more recent, thorough investigation of the effect of IPV and other gender power dynamics on adherence across the PMTCT cascade is warranted.
Disclosure of HIV Status
A final factor relevant to the analysis of PMTCT adherence and dynamics within couples is a womans self-disclosure to her husband that she is living with HIV. Numerous studies from sub-Saharan Africa have established that PMTCT adherence improves when women have disclosed their HIV-positive status to the husband/partner (Auvinen et al., 2010; Doherty, Chopra, Nsibande, & Mngoma, 2009; Jasseron et al., 2011; Peltzer, Jones, Weiss, & Shikwane, 2011; Theuring et al., 2009). Unfortunately, HIV continues to be a highly
9 Compared to ARVs during and after pregnancy, which must be taken 1 -2 times a day in the home
22


stigmatized disease (USAID & Zambia Ministry of Health, 2009). According to Goffman
(1963), stigma is an attribute that is deeply discrediting, considered shameful, and can
therefore lead to a variety of animosities or discriminations directed at the individual. It is
therefore no surprise that HIV-positive women may want to hide their status, particularly if
the husbands HIV status is unknown at the time of her diagnosis.
HIV-positive women often go to great lengths to ensure that their status remains
hidden from the husband due to perceived and often very real repercussions. HIV-
positive women across sub-Saharan Africa report fear of a husbands reaction, including
violence or abandonment, as barriers to disclosure (Etiebet, Fransman, Forsyth, Coetzee, &
Hussey, 2004; Medley et al., 2004; Msellati, 2009). Indeed, women have reported IPV after
disclosure of an HIV-positive status across sub-Saharan Africa (Auvinen et al., 2010; Creek
et al., 2006; Ezechi et al., 2009; Gaillard et al., 2002; Gielen, O'Campo, Faden, & Eke, 1997;
Medley, Garcia-Moreno, McGill, & Maman, 2004; Ntaganira et al., 2008). The Zambian
Ministry of Health specifically describes some of the key barriers to disclosure as:
... fear of abandonment, rejection, violence, upsetting family members and accusations in infidelity. The risks of disclosure include loss of economic support, blame, abandonment, physical and emotional abuse, discrimination and disruption of relationships. (Zambian Ministry of Health, 2010, p.3)
Chapter Summary
Women often have low relative power within the family in comparison to the husband. It is within this framework of unequal power relations that HIV-positive women are expected to take preventative actions aimed at minimizing the risk of vertical HIV transmission. In order to understand how gender power dynamics influence HIV-positive womens adherence to various PMTCT protocols, I designed a mixed methods study in Lusaka, Zambia; the specific methods of this study are discussed in the following chapter.
23


CHAPTER III
METHODS
Study Setting
The study setting, Kanyama Health Center, is located next to a quarry in a busy area of Lusaka surrounded by makeshift shops selling a variety of items from fish to clothes. I purposefully selected Kanyama Health Center as my recruitment site because of the large, diverse population that it serves. Kanyama Health Center is the largest public health center in Lusaka providing primary care and serving a low socioeconomic population of 160,000 individuals who live in the neighborhoods just west of the city. This densely populated area (see figure 4) is one of the poorest areas of Lusaka and home to the largest compound (i.e., low income housing). There is no sewer system in Kanyama compound, and during the rainy season, bouts of cholera and other diseases are especially common. Although HIV rates are not available for Kanyama specifically, it is safe to assume they are likely above average given the compounds low socioeconomic urban setting.
Figure 4 Map of Lusaka and Kanyama Compound
24


I was thoroughly impressed with the health center where I worked in Kanyama.
Health care workers at Kanyama Health Center demonstrate a consistent commitment to improving the health of Kanyamas population. For example, they have an active community outreach program to raise awareness on issues, such as cholera and HIV/AIDS. In addition, community health workers regularly make home visits to conduct voluntary rapid HIV counseling and testing (VCT). Despite human resource constraints and occasional stock-outs of supplies, the clinic is highly functional in regards to PMTCT and maternal and child health (MCH) care.
All HIV-positive postpartum women for the study were recruited through the MCH Department in the clinic, which sees hundreds of women and children every day for routine health care. Certain days of the week are dedicated to pregnant women while others days are primarily for children. The MCH department provides numerous services, namely, ANC,
HIV testing and counseling, pediatric immunizations, vitamin supplements, deworming treatment, growth monitoring, and PMTCT for women and their children.
Research Design
I used a concurrent mixed methods research design (Leech & Onwuegbuzie, 2006), including a verbally administered cross-sectional survey (n=320) and semi-structured interviews (subsample of n=32). The participants for this study were HIV-positive women attending the MCH department of Kanyama Health Center for routine pediatric health care. By using both quantitative and qualitative methods within the same population, I identify statistical relationships between gender power dynamics and PMTCT while drawing on the richer interview data to shed light on contextual explanations. In addition, the semi-structured interviews facilitate opportunities to highlight cases that contradict my hypotheses (e.g.,
25


women with low power in their relationship but high PMTCT adherence and women with high power but low adherence) to uncover additional barriers and facilitators related to PMTCT adherence.
Ethical Considerations
The sensitive nature of my research topic necessitates attention to confidentiality and appropriate handling of situations where women report violence. I held a three-day training for my research assistants, which included an in-depth discussion of research ethics. In addition, the study was designed and implemented in accordance with the WHO Ethical and Safety Recommendations for Research on Domestic Violence Against Women (WHO,
2001). Lastly, women reporting IPV were offered referrals to the Young Womens Christian Association (YWCA) in Lusaka for counseling and victim support services. The study was approved by the Colorado Multiple Institutional Review Board (COMIRB; Protocol 13-1979) in September 2013 and approved by Excellence in Research Ethics (ERES) Converge in Lusaka (Protocol 2014-JAN-010) in March 2014.
Consent forms and information sheets (see Appendix A) were translated into the local languages, which the research assistants verbally read aloud to all participants. Consenting participants provided written informed consent or a thumbprint (in cases where women were illiterate). Participants were compensated $3 (15 Zambian Kwacha) for travel expenses if they completed at least half of the questionnaire, which was meant to be an incentive, but not overly coercive and based on local standards.
Formative Research and Preparation
Prior to implementing this study, I conducted significant formative research, including key informant interviews (See Figure 5). I travelled to Lusaka during the summers
26


of 2012 and 2013 to conduct interviews with both PMTCT clinicians and gender experts. I met with PMTCT physicians and nurses as well as with individuals working for nongovernmental organizations focused on gender-based violence and womens empowerment. The goal of these trips was to determine the relevance of my research questions, understand PMTCT care and gender power dynamics in Lusaka, and plan for the logistics of the study. I was able to navigate these meetings through the help of my in-country mentor, a physician within the Department of Obstetrics and Gynecology at the University Teaching Hospital (UTH) in Lusaka.
Key informant interviews
Survey (n=320) and semi-structured
Expert panel review interviews (n=32)
V
June July 2012
V
February 2014
Y
March August 2014
July 2013
A
March 2014
A
Key informant interviews
Pilot Study (n=35)
Figure 5 Study Timeline
After the formative key-informant interviews, I conducted an expert panel review of my survey questionnaire in order to establish its face validity in February of 2014 (Singleton & Straits, 2010). This was especially important for questions regarding PMTCT adherence because there was no previously validated instrument measuring this outcome. The expert
27


panel review included an obstetric and gynecology physician at UTH, a nursing officer with the Lusaka District Health Office MCH Division, three PMTCT nurses/counselors from UTH and Kanyama Health center, and the MCH coordinator at Kanyama Health Center. Each individual reviewed a draft of the questionnaire and provided feedback based on their expertise. I modified the survey questionnaire slightly based on their recommendations prior to the pilot study. For example, one clinician pointed out that it would be relevant to ask women if they were offered drugs for PMTCT since stock-outs of medication are a continual challenge in the country. At this time, I also had the questionnaire translated into both Nyanja and Bemba to have standardized instruments in the most widely spoken local languages.
Unfortunately, I am not fluent in the local languages, and thus, recruited and trained four local research assistants to administer the survey and semi-structured interviews. Kanayma Health Center would only approve the use of research assistants who were current employees at the clinic due to confidentiality concerns. I worked closely with the MCH Coordinator at Kanyama Health Center to find health care workers who were a good fit for this research. We ultimately selected three other nurses/counselors as research assistants for the data collection. In addition, the MCH Coordinator herself volunteered to be a research assistant. All research assistants had previous experience with public health research projects. The other health care workers at the clinic were also hired to help with recruitment and were paid a small amount based on the number of HIV-positive women they referred to the study during pediatric immunizations.
I conducted several trainings with the four selected research assistants on data collection and research ethics. I created the training modules based on information from the WHO, COMIRB, and other materials from colleagues in Lusaka. In addition to my
28


PowerPoint presentations, I had the research assistants practice going through the questionnaire with each other as well as discuss any questions or concerns they had. I was very pleased with the level of competency of my research team. They were highly motivated to participate in the research and also often served as my cultural translators regarding events that I observed at the clinic (or in Lusaka in general).
Following the training of research assistants, I conducted a pilot study (n=35) at Kanyama Health Center to test recruitment procedures, survey administration, and participant comprehension of the questionnaire. I made minor modifications to the survey instrument and recruitment procedures after the pilot survey. For example, we originally were using invitation cards that nurses gave to eligible mothers inviting them to participate in the research. However, during the pilot study, we realized that this was causing undo attention among women at the clinic. Non-eligible women (i.e., HIV-negative or with an infant not of appropriate age) were asking why they were not given an invitation card, drawing too much attention to the study. Thus, we decided to change procedures and have nurses verbally invite eligible mothers to participate in the study, discarding the use of invitation cards.
For data analysis of the pilot study, I conducted descriptive analysis of the data in Stata 12 to examine the variance in the distribution of my variables of interest. The questions measuring PMTCT adherence, my outcomes of interest, showed variation with a range from 0% to 100% in self-reported medication adherence during and after pregnancy. The main independent variables measuring gender power dynamics also displayed adequate variability. For example, women reported zero to six controlling behaviors from their husband and zero to five forms of physical/sexual violence from their husband. Household decision-making
29


also deviated with some women participating in no household decisions and others participating in the majority of decisions.
Sampling Technique
The population of interest for this study is HIV-positive women with a young infant living in Lusaka who attend Kanyama Health Center for health care. Based on a population size of 160,000 individuals in Kanyama Health Centers reported catchment area, if 50% of the population are female and 22% are HIV-positive (USAID & ZMOH, 2009), the overall sampling frame includes roughly 18,000 HIV-positive women of reproductive age.10
Recruitment for the study took place during routine pediatric immunization at Kanyama Health Center using a nonprobability purposive sampling design (Onwuegbuzie & Collins, 2007). I chose to recruit women during routine pediatric immunizations in order to capture HIV-positive women who were both adherent and non-adherent to PMTCT since immunization compliance is very high in Lusaka: only 1.7% of children have no immunizations by 12 months of age (CSO, 2014). In addition, other researchers have affirmed that sampling at pediatric immunization clinics offers a robust method for studies examining PMTCT (Horwood et al., 2012).
Women were eligible for the study if they were married (or living with a man as if married) with a known HIV-positive status, over 18 years of age (legal age to provide consent in Zambia), and their youngest infant was between three and nine months of age. Infant age was selected in order to capture all of the essential PMTCT protocols, match the immunization schedule in Zambia, and limit recall bias. As a safety measure, we excluded any women who were at the clinic with their husbands to avoid potential confrontations
10 The sampling frame was less because we did not recruit anyone under the age of 18, not married, women who had an infant under three months or over 9 months of age, and women who did not use Kanyama Health Center for routine child health care.
30


(and/or violence) over womens participation in the research. However, it is very rare for husbands to attend well-child healthcare, and only one woman was excluded for this reason.
From March to August of 2014, nurses recruited all eligible women they encountered (i.e., HIV-positive with a child of the appropriate age)11 during pediatric immunizations. Nurses determined eligibility using the childs Under-Five Card (i.e., a mothers copy of her childs health record that she is required to bring to all health care visits), which includes the childs birthdate, height/weight, immunizations, medications, and PMTCT (see Figure 6). If a woman is HIV-positive, her card indicates CE (i.e., child exposed to HIV). If women did not have an Under-Five Card, nurses used other available medical records to determine womens eligibility. 11
Figure 6 Children's Under-Five Card
11 Additional screening questions were asked to women by the research assistants immediately following informed consent.
31


Sample Description
My a priori sample size calculation indicated a need for data from at least 274 participants to conduct a two-tailed z-test Poisson regression to detect a rate ratio of at least 1.2 with a power of 0.80 and alpha of 0.05 (Faul, Erdfelder, Lang, & Buchner, 2007). Of the 517 women who were approached to participate during their childs immunization visit, 136 (24%) were ineligible because they were not married (see figure 7). Among the eligible women, 326 (86%) provided informed consent and 320 women ultimately completed the survey questionnaire.
Figure 7 Sample Description
The main reasons women cited for declining to participate in the research were lack of time and unfounded fears regarding biomarkers being taken or future required participation related to the monetary reimbursement participants would receive. My research assistants reported that some women were skeptical they would be given money just for answering some questions.
32


Survey Measures PMTCT Adherence
I measured the dependent variables of interest, PMTCT Adherence, by creating a set of questions based on the current PMTCT and ART adherence literature. My main references for these measures includes the 2010 WHO PMTCT Recommendations for Zambia (see Table 1 in Chapter 2 for specifics; Zambia Ministry of Health, 2010), the 2011 Malawi DHS PMTCT questions (National Statistical Office & ICF Macro, 2011), and a published review article by Simoni et al. (2006), detailing self-reported measures for ART with specific recommendations. I created a series of PMTCT questions on the survey (see Appendix B) capturing data regarding womens drug adherence during pregnancy, childbirth, and postpartum; infant feeding practices; and pediatric HIV testing at 6 weeks and 6 months.
Despite the limitations of recall and social desirability bias, self-reporting by HIVpositive women is one of the most common methodologies used to collect data on PMTCT (see Mepham et al., 2011; Futterman et al., 2010; Nassali et al., 2009). In addition, studies have shown that self-reported adherence to ARVs and plasma drug levels are significantly associated and a reliable measure (Fabbiani et al., 2015; Fletcher et al., 2005; Murri et al., 2000). Finally, the research assistants also used the womans Under-Five Card to validate responses. For example, if a woman reported 100% drug adherence but had numerous missed appointments indicated on her card, the research assistant would probe the woman to honestly report her adherence or explain where she was getting medication. For pediatric HIV testing, any tests the child had were written on the Under-Five Card and also were indicated in a space on the survey by the research assistants during the interview.
33


For ARV adherence, I was specifically interested in womens behavior and wanted to rule out health care system factors. Thus, the first question women were asked was if they were offered medication during pregnancy. If women reported being offered medication during pregnancy, they were then asked if they took the medication. Many studies on ARV adherence stop at this point; however, I felt it necessary to try and capture more variability in adherence levels rather than just any or none. Thus, women who reported taking medication during pregnancy were asked to indicate the percent of medication they took from 0 to 100% using a visual analog scale (VAS) that was verbally explained by the research assistants (see Figure 8; Simoni et al., 2006).12
Many women find it difficult to take their HIV medicines exactly as they are told during pregnancy. What is your best guess of how much of the HIV medication you took while you were pregnant? We would be surprised if this were 100% for most women.
I------1--------1-------1-------1-------1-------1-------1--------1-------1--------1
0 10 20 30 40 50 60 70 80 90 100
Figure 8 Example of Drug Adherence Question
At the time of the study, the Zambian Ministry of Health recommended two slightly different drug regimens depending on womens CD4 count or WHO clinical stage (see Table 1; ZMOH, 2010). Both groups of women were required to take medication throughout pregnancy, but only women on the short-course prophylaxis regimen were given medication during childbirth. Thus, for the childbirth period, women were again asked if they were offered medication. A possible response category for this item was no (mother was on
12 Although participants understanding of numeracy was high, research assistants also helped some women conceptualize proportions by rephrasing the question as given a typical 10 day time period, how many days would you take medication?
34


treatment). If women reported being on treatment, there was a skip pattern in the survey to a specific section for women on the treatment regimen (explained below). If women were not on treatment, they were then asked if they took the medication during childbirth (this is a single dose medication so no continuous measure of adherence was used).
For the postpartum time period, women on the short-course prophylaxis regimen were asked if they were offered medication, if they took the medication, and how much of the medication they took from 0 to 100% during the week after giving birth (women on this regimen only take medication for seven days postpartum). Women on the treatment regimen were also asked if they were given medication, if they took the medication, and what their adherence was from 0 to 100% since giving birth (women on the treatment regimen do not stop taking the medication after one week, but continue taking ARVs for life).
Lastly, both women on the treatment and the short-course prophylaxis regimens were required to give prophylaxis to the infant postpartum, which we also captured on the survey -asking women if they were offered infant prophylaxis and if they gave the medication to the child. However, women on the treatment regimen were asked about their adherence during the six-week period after giving birth, and women on the short-course prophylaxis regimen were asked about their adherence to the infant medication since giving birth (women on short-course prophylaxis continue giving the infant prophylaxis until one week after breastfeeding cessation).
In addition to medication adherence, safe infant feeding practices are also a critical component of PMTCT. For the measurement of infant feeding, we asked women if they had ever breastfed the infant, if they were currently breastfeeding the infant, and if they had ever given any other foods from a list of 11 possible items, such as cows milk, infant formula,
35


water, fruit juice, porridge, etc (see Appendix B for all food items). Women who reported giving other food besides breast milk were asked at what age they first introduced that particular food to the child. Using these questions, I was able to capture which women were following the PMTCT recommendation of exclusively breastfeeding the infant to six months of life. All questions regarding infant feeding came directly from the ZDHS (CSO, 2009).
I measured the final PMTCT protocol, pediatric HIV testing, through a combination of mothers self-reporting and by using the Under Five Card. I developed these questions based on the WHO PMTCT guidelines for Zambia, which recommend HIV-positive women have their children tested at six weeks, six months, 12 months, and 18 months. Since the oldest infants in this study were 9 months of age, I was only able to capture the first two tests. Women were first asked if they had ever taken the child for pediatric HIV testing. Second, using womens Under Five Card, my research assistants wrote in a designated space on the survey whether the child was in fact tested at six weeks and six months based on the childs card. Unfortunately, I was not able to capture reliable data regarding the test results (i.e., the childs HIV status) because the turn around time for pediatric HIV tests was several months, and the vast majority of mothers were awaiting their childs most recent results. Gender Power Dynamics
I measured gender power dynamics within couples using a version of the Revised Conflict Tactics Scale (CTS2). The CTS2 is one of the most widely used gender power measurement tools worldwide and has strong reported psychometric properties (Straus, Hamby, BoneyMcCoy, & Sugarman, 1996). The version of the CTS2 used in this study came directly from the ZDHS Domestic Violence Module, which has been used in Zambia since 2007 (CSO, 2009, CSO 2014).
36


Table 2 Constructs and Questions on Gender Power Dynamics
Construct Specific Questions
Sexual Division of Labor
Economic Inequity A. Would you say that the money you earn is more than what your husband earns, about the same, or less than what he earns?
Sexual Division of Power
1. Household Decision- A. In your current relationship, who usually makes decisions about health care
Making for yourself?
B. In your current relationship, who makes decisions about major household purchases? For example, a sofa or TV.
C. In your current relationship, who makes decisions about daily household purchases? For example, food.
D. In your current relationship, who has the final say on how to spend money?
2. Intimate Partner Violence Does your husband ever:
2a. Physical Violence A. Slap you?
B. Twist your arm or pull your hair?
C. Push you, shake you, or throw something at you?
D. Punch you with his fist or something that could hurt you?
E. Kick you, drag you or beat you up?
F. Try to choke you or bum you on purpose?
G. Threaten to attack you with a knife, gun, or other weapon?
2b. Sexual Violence A. Physically force you to have sexual intercourse with him even when you do not want to?
B. Force you to perform any sexual acts that you do not want to?
2c. Emotional A. Say or do something to humiliate you in front of others?
Violence B. Threaten to hurt you or someone close to you?
C. Insult or make you feel bad about yourself?
3. Controlling Behaviors Please tell me if these apply to your relationship with your husband:
A. He is always jealous or angry if you talk to other men?
B. He frequently accuses you of being unfaithful?
C. He does not permit you to meet your female friends?
D. He tries to limit your contact with family?
E. He insists on knowing where you are at all times?
F. He does not tmst you with money?
Structure of Cathexis
Perceptions of Wife Beating In your opinion, is a husband justified in hitting or beating his wife in the following situations:
A. If she goes out without telling him?
B. If she neglects the children?
C. If she argues with him?
D. Is she refuses to have sex with him?
If she bums the food? |
37


The DHS Domestic Violence Module was first developed and standardized in 2000 and has been used in more than 80 surveys, including throughout sub-Saharan Africa (MacQuarrie, Winter, & Kishor, 2014). I chose this instrument because it appropriately captures all of the domains from the Theory of Gender and Power, my overarching theoretical framework (see Table 2).
Additionally, the formatting of questions on this instrument has several advantages over a single question approach (e.g., have you ever experienced IPV?), particularly in the context of cross-cultural research (Kishor, 2005). By asking women about separate specific acts of violence, the violence measure is not affected by different understandings between women of what constitutes violence. For example, all women are likely to agree on what constitutes being slapped, but not all women may agree on what constitutes physical violence. Furthermore, the instrument has been praised for asking women about violence from many different angles, which encourages disclosure because it gives women more time to think about their experiences and multiple opportunities to disclose their experiences of violence (Kishor, 2005).
Quantitative Analytic Strategy Data Entry and Cleaning
I created a data entry tool in CSPro13 that was designed to capture and record responses to all of the questions on the survey. I double-entered each survey and addressed any inconsistencies in the two data entry files by going back and checking the original survey for the correct value. Data cleaning consisted of running descriptive statistics and checking skip patterns to find any abnormalities, such as out of range variables or responses within
13 The Census and Survey Processing System: a public domain software package for entering, editing, tabulating, and disseminating census and survey data.
38


skip patterns (see below for description of missing data and the imputation method). For example, if a question was asked when it was supposed to be skipped, I re-coded the data (in a new dataset) as missing. In addition, throughout data analysis, I was continually cleaning the data and checking the original surveys when any irregularities arose.
Dependent Variables
I dichotomized womens self-reported ARV adherence levels into yes (adherent) or no (non-adherent) variables for 1) ARV medication during pregnancy; 2) sdNVP during labor/delivery (if applicable); 3) ARV medication postpartum; and 4) giving the infant prophylaxis. Women who reported being offered the respective medication, but did not take it (or give it to the infant) were coded as non-adherent. Women who were not offered medication were excluded from the analysis. I defined adherence during and after pregnancy as the woman reporting she took (or gave to the infant) at least 80% of the prescribed medication doses.14 I chose the cut-off of 80% based on recent literature indicating that this is the adherence level needed to suppress HIV using the current ARV regimens that the women in my study were taking (Gordon et al., 2015; Kobin & Sheth, 2011).15
For infant feeding practices, I created three categories based on womens self-reports of the food they gave to the infant and at what age: 1) exclusive breastfeeding to six months; 2) exclusive replacement formula feeding; and 3) mixed infant feeding. Women whose infant was less than six months of age were characterized as mixed feeding if the mother reported giving any food in addition to breast milk to the child. For infants over six months, if women
14 Except for medication during childbirth, which is a single dose. I coded women who reported taking the tablet during childbirth as adherent for this protocol.
151 also explored the cut-off of 95% and found very similar associations, indicating that my findings were not particularly sensitive to the cut-off point used.
39


reported giving any food to the infant prior to six months of age in addition to breast milk, the mother was also classified as mixed feeding.
Lastly, in order to analyze pediatric HIV testing, I created two new variables based on a combination of womens self-reporting and what was indicated on the Under Five Card: 1) if the child had any HIV testing (for all infants); and 2) if the child had both HIV tests at six weeks and six months (for infants ages six to nine months, n=158). If there were any discrepancies between womens self-reporting and the Under Five Card, I opted to use data from the Under Five Card. One possible response to the self-reporting question regarding pediatric HIV testing was that the mother attempted to bring the child for HIV testing but did not receive a test due to lack of supplies at the clinic. The women who selected this response category were not included in the analysis (n=24). Again, this was meant to exclude participants who encountered health system barriers.
Independent Variables
The first gender power dynamic of interest, womens economic dependence on the husband (i.e., the sexual division of labor), was measured via one question asking women if their earnings were 1) more than the husband, 2) about the same as the husband, or 3) less than the husband. I collapsed the first and third categories (i.e., women who have the same or greater earnings) compared to women whose husband has greater earnings.
The second power dynamic of interest is womens participation in household decision-making, which is part of the sexual division of power. In order to capture this dynamic, I created four new decision-making variables regarding 1) womens use of health care, 2) daily purchases, 3) large household purchases, and 4) the final say over money. In the original questions, women were able to choose three responses regarding who primarily
40


makes the respective decisions: 1) the woman alone, 2) the husband alone, or 3) they make the decision together. Similar to my economic dependence variable, I collapsed the first and third categories for all four variables to capture womens participation in household decisionmaking (i.e., she makes the decision alone or with the husband) compared to when the husband alone makes the decision. It was additionally appropriate to collapse the two categories because of the rarity of women reporting they alone make decisions. For example, only 8% of women report they alone make the decision regarding major household purchases and only 10% of women report having the final say over how money is spent. Finally, I summed the four household decision variables to create a final count of the number of decisions a woman participates in from 0 to 4, which is a standard way of examining household decision making (Upadhyay & Karasek, 2012; CSO, 2014).
I analyzed the third power dynamic, IPV, by creating a series of new variables measuring different aspects of IPV. First, although the data collection instrument I used to measure IPV (i.e., the DHS Domestic Violence Module) refers to the items as scales, generally, researchers dichotomize the variables into those who experience IPV (or certain forms of IPV) and those who have not experienced IPV (Alio, Nana, & Salihu, 2009; Kishor & Johnson, 2004; Lawoko, Dalai, Jiayou, & Jansson, 2007). I followed in this trend, using the module as a way to capture different forms of violence by asking women various questions related to IPV but not necissarily analyzing the questions as scales.
I created several new binary variables to capture different aspects of IPV. First, I created dummy variables for emotional, physical, and sexual violence, as well as for experiencing any IPV. In addition, to better understand the nature of IPV, I used some unique approaches with the module that went beyond simply examining the presence of violence.
41


First, I created two ordinal variables to capture the mean number of violent events a woman experienced and the frequency of violence in the past year. I calculated the mean number of violent events by summing the total number of specific emotional, physical, or sexual events a woman reported out of a total of 12 possible events (3 possible emotionally violent events,
7 possible physically violent events, and 2 possible sexually violent events). I also gave women a frequency of IPV score by summing how frequently each violent event occurred in the past year (often=3, sometimes=2, rarely ever=l), yielding a scale from 0 (no violent events in the past year) to 36 (experiencing every violent event often in the past year). Lastly, I created a dummy variable for any injuries as a result of violence to capture the severity of physical/sexual IPV. If a woman reported experiencing any type of injury from a list of four possible injuries, such as cuts, bruises, or aches, I classified her as experiencing injury from IPV.
In addition to creating ordinal variables, I additionally created two sets of dummy variables to measure whether the number of violent events or the frequency of violence in the past year has a dose-response relationship with the various PMTCT adherence measures.
Five dummy variables were created for each measure. First, for the number of IPV events, I used the following cut-off points: 1) no IPV events; 2) 1 to 3 IPV events; 3) 4 to 6 IPV events; and 4) 7 to 9 (highest reported number) IPV events. Second, for the frequency of IPV in the past year, I used the following cut-off points: 1) no IPV in the past year; 2) a frequency of IPV score from 1 to 5; 3) a frequency of IPV score from 6 to 10; and 4) a frequency of IPV score greater than 10.
Despite not analyzing the these variables as scales per se, I did run internal consitency checks of the IPV scales to establish how well the questions were inter-related. Generally, an
42


alpha over 0.70 is considered an acceptable internal consistency (Tavakol & Dennick, 2011). In which case, only the physical violence and sexual violence scales had adequate scale reliablity coefficients in my study (alpha=0.73 for both). The scales for controlling behavior (alpha=0.66) and acceptance of wife beating (0.67) were very close to the 0.70 alpha threshold, but emotional violence (alpha=0.36) displayed very poor internal consistency.
In addition to IPV, I am also interested in the effect of a husbands controlling behavior, another component of the sexual division of power. For controlling behaviors, I created a new variable to capture if women if women experienced three or more partner controlling behaviors. First, I summed the number of male controlling behaviors (from 0 to 6) and then dichotomized the variable into 1) three or more controlling behaviors or 0) 2 or less controlling behaviors. This is a standard way of examining partner controlling behaviors using the DHS data (CSO, 2014). This method was also appropriate due to the extremely small number (n=23) of women who reported no controlling behaviors.
Finally, for the structure of cathexis, I created one variable measuring womens acceptance of wife beating. The questionnaire asked women whether a husband is justified in beating his wife for five specified reasons (e.g., neglecting the children, burning the food, etc.; see Appendix B). In order to capture the magnitude of womens acceptance, I generated a new count variable summing the total number of reasons women agreed a husband is justified in beating his wife (from 0-5).
Covariates
Covariates include the mothers age in years (continuous variable); the infants age in months (continuous); the mothers highest level of education attainment (ordinal); parity (continuous); knowledge of PMTCT (count, see below); relative wealth (standardized index,
43


see below); PMTCT regimen (binary); and disclosure of HIV status to the husband (binary). All covariate questions on the survey came directly from the ZDHS, with the exception of PMTCT regimen, which I determined during the data analysis stage based on the medication a woman reported taking on the questionnaire: either lifelong ART or short-course prophylaxis (CSO, 2009). Knowledge of PMTCT was measured by four questions on the survey (CSO, 2009) that were analyzed as a count variable based on correct answers, for example, Can HIV be transmitted from a mother to her baby during pregnancy? and Are there any special drugs that a doctor or nurse can give to a woman infected with HIV to reduce the risk of transmission to the baby?
In order to measure socioeconomic status, I generated a relative wealth index from a list of 21 possible household assets on the survey. An alternative method would have been to simply count the number of reported assets; however, this can lead to biased results because two individuals with very different economic resources can be assigned the same wealth score. Instead, I opted to use the method that Filmer and Pritchett (2001) recommend for use with the DHS household assets measure principal component analysis (PCA). The PCA procedure first standardizes the indicator variables, then calculates the factor coefficient scores, and lastly, for each household, multiplies the indicator values by the loadings. These numbers are summed to produce the households index value (Filmer & Pritchett, 2001). The estimation of relative wealth using PCA is based on the first principal component, which yields a wealth index that assigns a larger weight to assets that vary the most across households. Thus, luxury household assets are weighted more heavily and assets that everyone owns are weighted zero. The scale was standardized prior to inclusion in the regression models.
44


Missing Data and Multiple Imputation
Overall, I analyzed data from 320 surveys. Surveys with more than 50% missing data were not included in the analyses (n=4). Missing data (2.3%) were imputed using multivariate-chained equations in Stata 12 (Raghunathan, Lepkowski, Hoewyk, & Solenberger, 2001).16 Data converged, indicating that the multivariate-chained model was a good fit for the dataset (StataCorp, 2009). See Table 3 for the extent of missing data among my key variables. In order to determine possible auxiliary variables (those that explain missing values), I generated indicator variables coded as 1 if the value was missing and 0 if the value was present. Next, I ran a series of logistic regression models with the possible auxiliary variables (Acock, 2010). Variables that were significantly associated with missing data were included in the imputation model as auxiliary variables in addition to standard controls, such as age, education, and wealth.
I included the dependent variables in the imputation model to help predict missing values among the independent variables but did not use them in the final statistical models because skip patterns were used in the survey and the missing values were genuinely not applicable to the outcome questions. For example, if a woman was not offered medication during pregnancy, she should not have answered the question on adherence to medication during pregnancy. A total of 20 auxiliary variables were included in the final imputation models. Through chained imputation, 20 datasets were created and pooled for missing values among the variables. I used the imputed dataset in all of my subsequent statistical models. The statistical analyses I used were specific to each outcome, which is discussed in detail in
16 Imputed data were analyzed using maximum likelihood estimations, which is specifically discussed in the methods sections of the proceeding chapters
45


the following chapters. Table 4 displays the participant characteristics of the survey sample
after multiple imputation.
Table 3 Missing Data for Key Variables (n=320)
No. Percent
Variable Missing Missing
Dependent Variables
Pregnancy adherence* 49 15%
Childbirth adherence* 10 3%
Postpartum adherence* 35 11%
Infant prophylaxis adherence* 16 5%
Infant feeding 1 0.3%
HIV test at 6 weeks 4 1%
HIV test at 6 months* 16 5%
Independent Variables
Acceptance wife beating scale 11 3%
Emotional violence scale 1 0.3%
Physical violence scale 1 0.3%
Sexual violence scale 1 0.3%
Controlling behavior scale 0 0%
Health care decision 3 0.9%
Minor purchases decision 11 3%
Large purchases decision 4 1%
Final say over money decision 3 0.9%
Economic dependence 10 3%
* Missing due to skip patterns (e.g., women were not offered that medication or child was not over 6 months of age)
46


Table 4 Participant Characteristics
Demographic Characteristics: Total (n=320) Median (IQR) or %
Age (years) 28.0 (24,34)
Age of infant (months) 6.0 (3, 9)
Parity 3.0 (2, 4)
Completed primary education 72.0%
Completed secondary education 13.7%
Job past 12 months 63.4%
Household had electricity1 74.9%
PMTCT knowledge 95.6%
Relationship Characteristics
Length of relationship (years) 5.0 (2, 9)
Husband tested for HIV 80.3%
Discordant couple (n=254)2 35.1%
Disclosed to husband HIV-positive status 91.6%
Sexual Division of Labor
Greater or about the same earnings as the husband 23.7%
Sexual Division of Power
Any IPV 60.9%
Any controlling behaviors 92.80%
Ever hit partner 8.5%
Daily household items 73.8%
Major household items 55.6%
Health care for the woman 67.2%
Final Say over money 37.8%
Structure of Cathexis:
Believe a husband is justified in beating his wife for at least one reason 64.9%
Example of household asset possession. There were 21 possible assets that a participant could own, which were then converted into a standardized wealth index for the multivariate analyses.
2 Among women who reported knowing their husbands status.
47


Semi-Structured Interviews
Concurrently with the survey, my research assistants conducted semi-structured interviews with a subsample of 32 women in order to explore confirmatory and contradictory cases regarding gender power dynamics within couples and PMTCT adherence. I chose to use a subsample of participants for the qualitative interviews because in mixed methods research, it is appropriate to use the same individuals for both forms of data collection so the data may be more easily interpreted (Creswell & Plano Clark, 2011). During the survey questionnaire, my research assistants were asked to purposively select participants for the semi-structured interviews based on four previously established domains of interest (see Table 5).
Table 5 Sampling Domains for Semi-structured Interviews
1. Women with high power in their sexual relationship and high PMTCT adherence 3. Women with low power in their sexual relationship and high PMTCT adherence
2. Women with high power in their sexual relationship and low PMTCT adherence 4. Women with low power in their sexual relationship and low PMTCT adherence
There were no cut-off points based on survey responses for the research assistants to use at the time of qualitative sampling;17 however, I held a formal training session prior to data collection to explain the four domains of interest. For example, I explained that power was operationalized as the combination of partner violence and control, as well as household decision-making and womens relative earnings. I also explained that I was interested in high adherence across the PMTCT cascade. Although the criteria were relatively subjective and
17 Cut-off points were established post-hoc and described in further detail below
48


based on how the research assistant interpreted the four categories, this technique aimed to include individuals with different informative experiences (Shenton, 2004).
Domains 1 and 4 (i.e., women with high power/high adherence and women with low power/low adherence) are the groups that provide support for my hypotheses regarding gender power dynamics. These two groups help explain why gender power dynamics, such as IPV, decrease womens adherence to PMTCT. Domains 2 and 3 (i.e., women with high power/low adherence and with low power/high adherence), however, uncover additional barriers (Domain 2) and facilitators (Domain 3) related to PMTCT adherence unrelated to gender power dynamics within couples.
The qualitative interviews were conducted using a semi-structured interview guide (see Appendix C) that I developed based on my formative research in Lusaka and with the guidance of my local research team. On the interview guide, research assistants indicated womens level of adherence from the survey and asked questions, such as why would you say it was difficult for you to take the medication? or conversely, what helped you have such high adherence to your medication? I also included several questions specific to the relationship with the husband, such as how does your husband feel about PMTCT? or what would happen if you disobeyed your husband? The research assistants were also free to ask follow-up questions or continue interesting discussions that helped inform the research questions.
The interviews took place in the same location on the same day immediately following the survey with the selected subsample of participants. Prior to beginning the interview, the research assistants went through a second informed consent process with each woman, asking her to stay for additional time and for consent to voice record the interview.
49


The interviews were also conducted in the local languages. The interviews continued until the research team felt that we had a diverse sampling of women who fell into one of the four domains.181 later transcribed the interviews into Microsoft Word while one of my research assistants listened to the audio recordings and verbally translated them into English.
Interviewing women on the same day immediately following the survey had several notable strengths and limitations. First, since women were already recruited into the study and present, there was no hassle of tying to do any follow-up. Indeed, this was my rationale for the decision to conduct the interviews on the same day as the survey. However, by using this method, women were asked to stay for an additional 45 minutes to one hour on top of the 30-45 minutes they spent taking the survey. It is possible that selection bias may have occurred among women who had controlling partners and needed to account for their whereabouts without disclosing their status and/or use of PMTCT to the husband. These women may have opted not to participate in order to get home sooner and not raise suspicions from the husband. Unfortunately, we did not collect data on the response rates of women for the qualitative interviews or why women declined the interview (if any did). In hindsight, this would have been useful information.
Qualitative Analysis
For the semi-structured interviews, I used a content analysis method (Green & Thorogood, 2014) in Atlas.ti within a predetermined thematic framework. This framework was in the form of the four original power/adherence dimensions discussed above. In addition to this deductive coding, I also incorporated inductive approaches within the four previously established dimensions to determine significant themes. This technique has been
18 We also unfortunately had a time constraint with how long I was planning to stay in Zambia, which limited the number of women we could interview.
50


referred to as a hybrid approach and has the advantage of allowing both data-driven and theory-driven analysis (Fereday & Muir-Cochrane, 2006).
For the first step of my qualitative analysis, I placed each interview into one of the four previously established domains by examining womens responses on the survey questionnaire and determining appropriate cut-off points. I labeled women who reported less than 80% medication adherence for all protocols as having low adherence. Women were classified as having low power if they reported three or more partner controlling behaviors in addition to experiencing IPV and participating in less than two household decisions. This coding scheme provided the following distributions with the four domains: high adherence/high power (n=12), low adherence/low power (n=12), high adherence/low power (n=4), and low adherence/high power (n=4).
After indexing all if the interviews into one of the four-power/adherence domains, I then used a more iterative approach. For the first step, familiarization, I carefully read through each transcript (Green & Thorogood, 2014) of interviews in each particular domain. While reading through the transcripts, I assigned tentative open codes to chunks of data that were relevant to my understanding of adherence and gender power dynamics. I also simultaneously wrote memos for myself regarding my emerging hypotheses about the relationships between codes. After I established a comprehensive code list, I re-read through the transcripts assigning additional codes to quotes (Daly, Kellehear, & Gliksman, 1997). I continued this processes of re-reading the trasnscripts in Atlas.ti and assigning codes to quotes as well as creating new codes until I acheleived theoretical saturation (LeCompte & Schensul, 2010) for each of the four power/adherence domains.
51


Once all of the data had been coded in Atlas.ti, I arranged the relaevant quotes via a chart in Microsoft Word. I placed each quote with its affliated code into one of the four adherence/power domains for which that interview had been assigned (based on the womans survey responses). Finally, I created additional codes (i.e., axial codes) to represent the specific themes within that particular domain of interest (see Table 6). I used these themes to organize my presentation of the data within each domain (see Chapter Eight) and present illustrative quotes.
Table 6 Example of Qualitative Coding Scheme
Quote Domain Open Code Axial Code
He [the husband] has never reminded me about giving the child medication, not even one day. I am the only one that gives the child medication. Low power/ low adherence Relationship dynamic Lack of partner support
Limitations
The study has limitations to note. First, it is cross-sectional and cannot establish causality or the timing of events. There is no way to determine, for example, whether nonadherence or violence occurred first. Second, the results are primarily based on self-reporting, which is vulnerable to recall and social desirability biases. Third, questions regarding womens knowledge of PMTCT were also vulnerable to acquiescence bias (e.g., can HIV be passed from a mother to her baby during pregnancy?). Fourth, the sample is non-representative, limiting the generalizability of findings outside of low socioeconomic women in urban Zambia. Lastly, due to the relatively small sample size, there may have been inadequate statistical power to detect all significant quantitative relationships.
52


Chapter Summary
This chapter details the specific methods used in this study, including a survey and semi-structured interviews with over 300 HIV-positive postpartum women attending a large public health center in Lusaka, Zambia. In the following chapters, I report on the results of my quantitative and qualitative analyses, starting with an in-depth description of gender power dynamics within couples and moving into significant statistical relationships between power dynamics and adherence to PMTCT. Finally, I discuss important findings from the semi-structured interviews, which help inform the statistical associations.
53


CHAPTER IV
THE SEXUAL DIVISION OF POWER
Globally, large gender power inequities exist both within society and within intimate relationships with women generally having less power. The Theory of Gender and Power, which I discussed in-depth in Chapter Two, views inequities between men and women through three key domains: 1) The structure of cathexis (i.e., unequal gender norms), 2) the sexual division of power, and 3) the sexual division of labor (Connell, 1987). Although these domains often overlap and perpetuate one another, I specifically focus on the sexual division of power because womens relative power within couples is the main explanatory mechanism of interest in this study regarding adherence to PMTCT. Moreover, IPV/control is one of the most compelling manifestations of power differentials with couples, which is why I particularly focus on interpersonal power constructs in this chapter.
Review of Methods
I measured gender power dynamics within intimate relationships using the ZDHS Domestic Violence Module (CSO, 2009; see Chapter Three for Details). The statistical methods used in this chapter include descriptive statistics, simple and multivariate logisitic regression, and linear regression in Stata 12, using womens experiences with IPV and partner controlling behaviors as the dependent variables of interest.
Intimate Partner Violence and Controlling Behaviors
A large proportion of participants in this study experienced IPV (61%) and partner controlling behavior (93%). Only 5% of women report no experiences of violent or controlling events from their current husbands (see Figure 9).19 Not only is IPV and partner controlling behavior shockingly prevalent, but there is also a large overlap between IPV and
19 Characteristics of these positive deviants are discussed at the end of this chapter.
54


controlling behavior among the HIV-positive women in this study: 59% of women experienced both IPV and partner controlling behavior. Notably, almost all of the IPV women experienced occurred with the concomitance of controlling behaviors: only 2% of women report IPV without partner controlling behaviors.
2%
IPV only
Controlling behaviors only
IPV and controlling behaviors
No IPV or controlling behaviors
Figure 9 Women's Experiences with IPV & Partner Controlling Behaviors (n=320)
Violence that occurs in combination with controlling behavior deserves particular attention because of the nature and repercussions of these behaviors. IPV that includes partner control has been termed coercive controlling violence, as opposed to situational couple violence, which results from conflicts between partners that occasionally escalate into violence (Kelly & Johnson, 2008). Notably, coercive controlling violence tends to occur more frequently, results in more physical injuries, and results in greater psychological distress compared to situational couple violence (Johnson, 2006; Pico-Alfonso et al., 2006).
Importantly, the levels of IPV and partner control in my sample are much greater than what is reported in the most recent ZDHS (see Appendix D for a detailed comparison of my
55


data and the 2013-2014 ZDHS). The 2014 ZDHS reports that 38% of women in Lusaka and 47% of Zambian women experience IPV (CSO, 2014). In addition, 65% of women in Lusaka and 74% of Zambian women on the ZDHS report that their partner displays at least one type of controlling behavior. There are two factors that likely account for the greater prevalence of IPV/control found in my sample compared to the ZDHS: 1) socioeconomic status, and 2)
HIV infection.
First, the women in my sample are from one of the most impoverished areas of Lusaka, and low socioeconomic status is an established risk factor for IPV (Heise & Garcia Moreno, 2002; WHO/LSHTM, 2010). The WHO Multi-Country Study on Womens Health and Domestic Violence, which surveyed women in 10 countries from various regions, including sub-Saharan Africa, report that higher socioeconomic status is generally protective of physical and sexual partner violence (Abramsky et al., 2011). Indeed, in the most recent ZDHS, women in the lower wealth quintiles report greater levels of IPV and partner control than women in the wealthier quintiles (CSO, 2014).
Secondly, the women in my sample are all HIV-positive, which is an additional known risk factor for IPV (Campbell et al., 2008). Across sub-Saharan Africa, women report IPV following disclosure of an HIV-positive status to male partners (Kilewo et al., 2001). Moreover, a recent longitudinal study in South Africa also established that women who experience IPV are at greater risk of incident HIV infection (Jewkes, Dunkle, Nduna, & Shai, 2010). Another cross-sectional study reports that in addition to IPV, male control is associated with an HIV-positive diagnosis among women attending ANC in South Africa (Dunkle et al., 2004). Given these associations, it is not surprising that the low
56


socioeconomic HIV-positive women in my study report greater levels of IPV/control than women in the general Zambian population.
Experiences With Specfic Forms of Violence and Control
Differentiating between the causes, consequences, and nature of IPV and control is essential in order to develop an appropriate understanding and ability to design effective interventions. In terms of experiences with specific violent events, the most commonly reported event women experienced in this study is the husband/partner insulting her or making her feel bad about herself, followed by being slapped and physically forced to have unwanted sex (see Figure 10). Being choked or threatened with a weapon was extremely rare in this cohort of women. Lastly, only 8% of women report ever hitting their husband.
EMOTIONAL VIOLENCE: Insult you or make you feel bad Threaten to hurt you or someone you love Say /do something to humiliate you
PHYSICAL VIOLENCE: Slapped you
Twisted your arm or pulled your hair Pushed you, shook you, or threw something at you Punched you with his fist of something else Kicked you, dragged you, or beat you up Tried to choke you Threatened you with a weapon
SEXUAL VIOLENCE: Physically forced you to have sex Forced you to perfrom a sexual act
MALE-DIRECTED: Ever hit partner
Figure 10 Womens Experiences With Specific Violent Events (n=320)
57


The most commonly reported male controlling behaviors women were: the husband is always jealous if she talks to other men and he insists on knowing where she is at all times (see Figure 11). The high prevalence of male controlling behavior highlights the strong patriarchal cultural in Zambia something that I regularly observed while conducting this research. In addition, there are also strict gender norms (i.e., the structure of cathexis) that condemn women socializing with non-familial men, which may explain the high prevalence of women reporting their husband is always jealous if she talks to other men.
Always jealous if you talk to other men Frequently acceses you of being unfaithful Does not permit you to meet friends
Limits contact with family
Insists on knowing where you are at all
times
Does not trust you with money
0% 20% 40% 60% 80% 100%
Figure 11 Experiences with Specific Forms of Male Controlling
Behavior (n=320)
In addition to controlling behavior, classifying IPV into three different types (i.e., emotional, physical, and sexual) helps provide a better understanding of the power dynamics within these relationships. The most prevalent form of IPV in this study is emotional violence, with 40% of women reporting at least one emotionally violent event, such as the husband humiliating her in public or threatening to hurt her (see Figure 12). Often, emotional
58


violence is not evaluated in studies examining IPV (see Abramsky et al., 2011; Harling, Msisha, & Subramanian, 2010; Pallitto et al., 2013). My research highlights that this could be problematic because it overlooks a large group of women who are, for all intents and purposes, experiencing IPV. In addition, over one-third of women also report sexual violence, and over 30% report physical violence. The most frequently reported combinations of violence are experiencing only emotional violence or experiencing all three types of IPV. There are also a large proportion of women who either report experiencing emotional and physical violence together or only sexual violence.
Figure 12 Types of Intimate Partner Violence Reported (n=320)
59


Predictors of Intimate Partner Violence and Control
While treating IPV and partner controlling behavior as dependent variables, several significant associations emerge (See Table 7). First, younger women are more likely to experience male controlling behaviors: for every year of age, women have 8% lower adjusted odds of experiencing three or more partner controlling behaviors (p<0.05). Age, however, is not associated with experiencing IPV. To my knowledge, no other research from sub-Saharan Africa has established younger age as a risk factor for partner controlling behaviors. This may be potentially explained through womens decreased power as a result of the husband being much older among the younger women in my study.20 Indeed, some research has indicated that younger age is a risk factor for IPV, while others have found null results (Jewkes, 2002).
An additional risk factor for partner controlling behavior among women in this study is greater parity. For each additional child, women have 31% higher adjusted odds of experiencing three or more partner controlling behaviors (p<0.05). To my knowledge, this association has not been directly explored in the literature around IPV and partner control. However, there is evidence that parental stress, child health outcomes, and IPV are interrelated (Huth-Bocks & Hughes, 2007). It is plausible that families with more children have more stressors, which could lead to increased violent/controlling behaviors, although parity and IPV are not specifically associated in this study. In addition, there is a marginally significant relationship between the infants age and partner controlling behaviors: for each month of infant age, women have 13% higher adjusted odds of experiencing three or more controlling behaviors (p=0.08).
201 unfortunately did not capture data on the husbands age.
60


Table 7 Multivariate Logistic Regression Results for the Odds of Experiencing IPV and Three or More Partner Controlling Behaviors
Participant/Relationship Characteristics
Age (years)
Infant age (months)
Parity
Highest educational attainment Wealth index score Length of relationship Disclosed HIV status Dont know husbands HIV status* 1
Model 1
Experienced Intimate Partner Violence
aOR
(95% Cl)
1.0
(0.93 1.07)
0.98
(0.87- 1.06)
1.10
(0.69- 1.40)
0.97
(0.75- 1.23)
1.21
(0.89- 1.65) 0.94+
(0.88-1.01)
0.88
(0.24-3.28)
1.78
(0.81-3.94)
Model 2
Experienced Three of More Controlling Behaviors aOR
(95% Cl)
0.91*
(0.84 0.98)
1.13+
(0.98- 1.29) 1.31*
(1.00- 1.72) 1.35*
(1.00- 1.80) 0.65*
(0.46-0.91)
0.98
(0.91 1.06)
0.88
(0.20 3.78)
1.02
(0.42 2.48)
Discordant couple1
Number of acceptable reasons to beat a wife Experienced IPV
Experienced 3+ partner controlling behaviors
Number of household decisions the woman participates in
Equal or greater earnings than the husband
0.99 1.10 (0.55-2.23)
(0.53 1.85)
1.31** 1.80***
(1.12-1.53) (1.52-2.14)
N/A 2.88**
(1.56-5.31)
2.85** N/A
(1.55-5.21)
0.88 0.90
(0.71 -1.12) (0.67-1.17)
1.20 0.85
(0.60-2.40) (0.38- 1.94)
+ p<0.10 p<0.05 ** p<0.01 ***p<0.001
1 comparison group: Husbands HIV status is positive (i.e., HIV concordant couple)
61


Womens educational attainment is also positively associated with partner control.
For each additional increase in a womans educational attainment,21 she has 35% higher adjusted odds of experiencing three or more partner controlling behaviors (p<0.05). This is not only counterintuitive but is also contradictory to research on womens education and IPV. Womens education is generally accepted as a protective factor, at least for IPV (Abramsky et al., 2011; Garcia-Moreno, Jansen, Ellsberg, Heise, & Watts, 2006; R. Jewkes, 2002) -much less is known about the relationship with partner controlling behaviors. Indeed, education is negatively associated with IPV in my study, although the relationship was not significant (OR 0.98, p=0.871). A possible explanation for this finding is that womens education may make the husband feel emasculated, and as a result, he attempts to enforce his dominance in the relationship through controlling behaviors. There is evidence from low and middle-income countries, such as India, Bangladesh, and Nicaragua, that report a positive association between womens higher income (which is related to educational attainment) and IPV (Kishor & Johnson, 2004; Vyas & Watts, 2009b), but this association was not established in my models.
In contrast, wealth is negatively associated with partner controlling behaviors. For each standard deviation on the wealth index, women have 35% lower adjusted odds of experiencing three or more partner controlling behaviors (p<0.05). As mentioned earlier in this chapter, the most recent ZDHS also reports that women in the lower wealth quintiles have higher levels of partner control (CSO, 2014). It is unclear, however, why such an association exists in this study between wealth and partner control, but not with IPV, which has been reported elsewhere (Abramsky et al., 2011; CSO, 2014).
21 Ordinal scale: no education, some primary education, completed primary, some secondary, completed secondary, some college, completed college, and graduate education
62


Not surprisingly, greater endorsement of wife beating as an acceptable response for womens transgressions is positively associated with experiencing both IPV (p<0.01) and three or more partner controlling behaviors (p<0.001). Due to the cross-sectional nature of my research, it is not feasible to establish if IPV/control or acceptance of wife beating occurred first, but I suspect that there is a bidirectional relationship. Women who are accepting of IPV may be more likely to end up in violent/controlling relationships, while women who experience IPV/control may begin to justify their partners behaviors.
Overall, the HIV-positive women in my study endorse wife beating to an even greater extent than the women in the most recent ZDHS: 65% of my participants agree that a husband is justified in beating his wife for at least one reason compared to only 47% of Zambian women on the 2014 ZDHS (CSO, 2014). However, my participants also experience higher levels of IPV, which may account for their greater acceptance of wife beating.
Interestingly, there is evidence from sub-Saharan Africa that women tend to endorse wife beating to an even greater extent than men. Uthman, Lawako, & Moradi (2009) report that in sub-Saharan African countries where IPV is an accepted response to women's transgressions, women find more justification for the practice than men. Indeed, the most recent ZDHS found that 47% of Zambian women agree a husband is justified in beating his wife for at least one specified reason, compared to only 31% of Zambian men who endorsed this belief (CSO, 2014), highlighting the gendered interpretations of the structure of cathexis.
Lastly, experiencing IPV and the partner displaying three or more controlling behaviors are significantly associated with each other in the multivariate models (p<0.01). Surprisingly, however, household decision-making and womens greater/equal earnings are not associated with either IPV or partner control in the multivariate models. This may be
63


because there is a large amount of collinearity between the gender power measures. Indeed, at the bivariate level, each of the four gender power measures (i.e., IPV, partner control, household decision-making, and womans earnings) are significantly correlated with one another (see Table 8).
Table 8 Correlation Coefficients Between Gender Power Dynamic Measures
(n=320)
IPV 3 or More Controlling Behaviors Number of Decisions the Woman Participates in Womans Earnings Greater or Equal to Husbands
IPV 1.00
3 or More Controlling Behaviors q 40*** 1.00
Number of Decisions the Woman Participates in -0 24*** -0 32*** 1.00
Womans Earnings Greater or Equal to Husbands 0 16**** 0.18*** -0 11*** 1.00
***p <0,001
Despite null results in this study in the models from Table 7, it is important to note that numerous other studies from various global settings have reported a connection between womens low participation in household decisions and female-directed IPV (Gage & Hutchinson, 2006; Hindin & Adair, 2002; Murphy & Meyer, 1991; Murray A. Straus, Gelles, & Steinmetz, 1980). Research from the United States (Murphy & Meyer, 1991; Murray A. Straus et al., 1980), Asia (Hindin & Adair, 2002), and the Caribbean (Gage & Hutchinson, 2006) has reported that autonomy in decision-making generally has a negative association with IPV. Lastly, research from Zambia and Kenya found that women who lack autonomy in household decisions are more likely to justify IPV than their more empowered peers (Lawoko, 2006, 2008).
64


Positive Deviants
Only 5% of participants report no IPV or partner control (n=16; see figure 9). Upon comparison with women who experience IPV/controlling behavior, two significant differences emerge at the bivariate-level. First, women who have not experienced IPV/controlling behavior have been in their relationship longer than women who experience IPV/controlling behavior: 10.3 years compared to 6.4 years, respectively (p<0.01). Second, women who have not experienced IPV/controlling behavior universally disclosed they were HIV-infected to their husband (100%), compared to just over 90% of women who experienced IPV/controlling behavior (p<0.01).
Although not statistically significant, women who have not experienced IPV/controlling behavior also report a higher prevalence of discordant relationships (i.e., when one partner is HIV positive, in this case the woman, and the other is HIV negative). In contradiction, other studies have found an increased risk of IPV among HIV-positive women in discordant relationships (Emusu et al., 2009; Ezechi et al., 2009; Shuaib et al., 2012; Were et al., 2011). A possible explanation for my finding is that discordance in my study was based on womens self-reporting of the husbands status which may or may not in fact be his actual HIV status. As a result, my null finding may be based on Type II error. If I had biomarker information regarding the actual HIV status of the husband, I may have found quantifiably different results. Optimistically, it is also possible that this is an indication that HIV is becoming less stigmatized and that HIV-positive women in discordant relationships are not at greater risk of IPV/control from HIV-negative partners.
65


Lastly, the women who did not experience any IPV/controlling behaviors report more income or about the same as their husband (instead of the partner having a higher income), but the differences are not statistically significant. Likewise, women who have not experienced IPV/controlling behavior report greater participation in all household decisions (see Figure 13); but again, these differences are not statistically significant. However, this is possibly due to small sample size and inadequate statistical power. Given the consistent trend, I would like to see results using a larger sample size because I suspect with more statistical power, a significant relationship may be present.
80%
60%
40%
20%
0%
75%
67%
63%
55%
75%
74%
Own health Major Daily Final say over
care household household money
decisions decisions
No IPV or Controlling Behaviors (n=16)
IPV or Controlling Behaviors (n=304)
Figure 13 Household Decision-Making and Experiences with IPV/Controlling Behavior (N=320)
66


Gender Asymmetry in Intimate Partner Violence
There is an ongoing debate among sociological scholars regarding whether IPV is primarily female-directed and a result of patriarchal structures or conversely, gender symmetrical, with men and women perpetrating and experiencing IPV at similar levels (Straus, 2008). Studies from various settings although heavily focused in the United States report gender symmetry and that men and women commit violence equally in heterosexual relationships (Anderson, 2002; Capaldi & Owen, 2001; Moffitt, Robins, & Caspi, 2001; Straus, 2008; Straus & Ramirez, 2002). Feminist scholars, on the other hand, argue that IPV is a form of male domination over women in order to maintain power and control (Dobash & Dobash, 1979; Michalski, 2004; Stark & Flitcraft, 1996; Yllo, 1993).
The Theory of Gender and Power supports the feminist perspective and postulates that men perpetrate greater violence against women than vice versa (i.e., part of the sexual division of power). Indeed, studies from South Africa (Gass, Stein, Williams, & Seedat, 2011); Australia (de Vries Robbe, March, Vinen, Homer, & Roberts, 1996; Roberts, O'Toole, Raphael, Lawrence, & Ashby, 1996), the Netherlands (Romkens, 1997), the United Kingdom (Dobash & Dobash, 2004), and South Korea (Lee, Stefani, & Park, 2014) support the argument that men do in fact account for the majority of IPV perpetration.
Unfortunately, the only measure for male-directed IPV on the instrument I used is one question asking women if they have ever hit their partner. This limited my ability to create a comprehensive picture of what type of violence/control is occurring towards men in these relationships. However, using my physical violence against women measurement22 and the
22 This measurement includes seven items on the CTS2. If women reported yes to any of the items, they are considered to have experienced physical IPV.
67


one question regarding whether a woman ever hit her partner, I am able to make some inferences about the symmetry/asymmetry of physical violence in these marriages.
In the vast majority (77%) of relationships where physical violence is reported (n=l 18), it is only female-directed (see Figure 14). Surprisingly, more women report male-directed physical violence compared to bidirectional violence. If we are to examine only one item from the CTS2 that asks women has your partner ever hit or slapped you, which is the most comparable to have you ever hit your partner, the extent of difference between male and female-directed physical IPV decreases. However, even comparing just these two questions, women still report their husband hitting them at 3.5 times the rate they hit him (i.e., 28% compared to 8%, respectively; data not shown in figure). Based on these findings, my data finds overwhelming support for gender asymmetry in physical violence, which is by and large perpetrated by men against their HIV-positive female spouses.
Male-directed physical IPV only
Female-directed physical IPV only
Bidirectional physical IPV
Figure 14 IPV Victimization in Physically Violent Relationships by Gender (n=118)
68


Chapter Summary
This chapter explores HIV-positive womens experiences with various forms of IPV and partner controlling behaviors. I also establish factors associated with these experiences and the characteristics of the women who did not experience any IPV or male controlling behaviors. Lastly, this chapter indicates that physical violence is gender asymmetrical with women more likely to be victimized than their male partners. In the next several chapters, I examine how gender power dynamics within couples including different forms of IPV, male controlling behavior, and household decision-making influence womens health behaviors related to PMTCT, beginning with drug adherence during and after pregnancy in the following chapter.
69


CHAPTER V
DRUG ADHERENCE DURING AND AFTER PREGNANCY
PMTCT depends critically on HIV-positive womens adherence to ARVs during and after pregnancy. Suboptimal adherence by HIV-positive pregnant and breastfeeding women not only increases the risk of mother-to-child transmission but also increases the likelihood of maternal HIV-related disease progression and drug resistance for both the mother and the infant (Nachega et al., 2007; Ngoma et al., 2015; M. Onono et al., 2015). Recent studies suggest that HIV-positive individuals need to take at least 70-80% of prescribed ARV doses to adequately suppress the virus (Kobin & Sheth, 2011; Martin et al., 2008; Nachega et al., 2007). However, the optimal standard is generally considered to be at least 95% of ARV doses in order to maximize health outcomes, including reducing vertical HIV transmission (Alexander et al., 2003; Garcia de Olalla et al., 2002; Gross et al., 2006; Harrigan et al., 2005; Howard et al., 2002; Paterson et al., 2000; Press, Tyndall, Wood, Hogg, & Montaner, 2002; Wood et al., 2003; Wood et al., 2006).
WHO recently stated that in order to achieve the global goal of eliminating mother-to-child transmission, it is crucial to establish current trends in ARV adherence during and after pregnancy as well as a clear understanding of the barriers to adherence (WHO, 2010c). Nachega et al. (2012) confirms in their recent review that adherence to ARVs during and after pregnancy remains a challenge, including in sub-Saharan Africa. However, studies on adherence to ARVs for PMTCT vary vastly across the region with reports as high as 98% in urban Kenya (Imbaya, Odhiambo-Otieno, & Okello-Agina, 2008), to as low as 38% in rural Uganda (Barigye et al., 2010). Recent research from Zambia indicates that PMTCT medication adherence ranges from 63% to 79% (Conkling et al., 2010; Megazzini et al.,
70


2010; Stringer et al., 2010). These estimates, however, are based on sdNVP at the time of labor/delivery. Little is known about HIV-positive Zambian womens ability to adhere to the newer combination ARV regimens during and after pregnancy and what social barriers hinder adherence.
In this chapter, I answer an important research question: is there a relationship between gender power dynamics within couples and non-adherence to ARVs for the purpose of PMTCT? First, I describe HIV-positive womens adherence levels to ARVs during and after pregnancy. Second, I estimate the number of children at-risk for vertical HIV transmission among the mother-baby pairs in this study by combining the women who reported inadequate adherence with those who were not offered medication. The final sections of this chapter focus on how non-adherence to ARVs during and after pregnancy is affected by gender power dynamics, including IPV, male controlling behaviors, participation in household decision-making, and womens economic dependence on the husband.
Review of Methods
Women who were offered ARVs for PMTCT were subsequently asked questions regarding their drug adherence during and after pregnancy. In this chapter, I employ descriptive statistics to show levels of adequate (>80%) and optimal (>95%) drug adherence. I also use multivariate logistic regression models to determine the adjusted odds of adequate adherence to each ARV medication protocol by womens experiences with different gender power dynamics. Due to the collinear nature of my gender power dynamic measures (see Chapter Four), I first examine the effect on adherence of each power dynamic as a separate independent variable in models both with and without HIV status disclosure. I then estimate the models again combining all of the power dynamic measures.
71


Adequate and Optimal Drug Adherence
Of the women who were offered medication during pregnancy (n=271), 88% report adequate ARV adherence (i.e., took >80% of the medication doses), but only 62% report optimal ARV adherence (i.e., took >95% of doses; see Figure 15). Among the women on short-course prophylaxis, who were offered sdNVP during childbirth (n=131), 92% report taking the medication (i.e., were adherent). During postpartum, 91% of women who were offered medication (n=285) report adequate adherence, but only 70% report optimal adherence. Lastly, adherence to giving the infant prophylaxis during the postpartum period is adequate among 86% of women and optimal among 72% of women who were offered infant prophylaxis (n=303). My research thus supports the growing body of literature indicating that achieving high PMTCT adherence remains a challenge, including among low socioeconomic urban women Zambia.
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Adequate Optimal Took Adequate Optimal Adequate Optimal
Pregnancy (n=271) Childbirth* (n= 131) Postpartum (n=284) Infant Prophylaxis (n=303)
* Childbirth medication is a single dose that the woman either took or did not take.
Figure 15 Adequate and Optimal Drug Adherence Levels
72


Figure 16 additionally displays the number of medication protocols where women achieved adequate adherence out of a possible total of four protocols.23 Thirty-nine percent of women (n=125) report inadequate adherence to at least one of the medication protocols and 5% (n=17) report inadequate adherence to all of the medication protocols. However, the majority of women (61%) report being adherent to all four of the medication protocols -having their child completely protected across the PMTCT cascade of care.
* Includes medication during pregnancy, childbirth24, postpartum, and giving the infant prophylaxis_____________________________________________________________________________________
Figure 16 Number of Medication Protocols Where Women Achieved Adequate Adherence (n=320) 23 24
23 Pregnancy, childbirth, postpartum, and infant prophylaxis
24 Women on lifelong treatment (opposed to short-course prophylaxis) were considered adherent during childbirth if they were adherent during pregnancy because this would be protective for the infant during birth; Women on lifelong treatment were not offered any specific additional medication intrapartum to protect the infant unlike women on the short-course regimen.
73


Unprotected Children at Risk of Vertical HIV Infection
Figurel7 displays the percent of children who were unprotected through PMTCT during each time period as a result of either womens poor adherence or lack of access to drugs. It is not clear whether the women in my study who report not being offered ARVs were in fact not offered drugs by the health center or if they did not take the medication despite being offered it. I suspect this group constitutes a combination of women who 1) were genuinely not offered any medication, 2) were offered medication but did not understand what the medication was for, and 3) were offered medication but decided not to take the medication and did not want to admit this in the interview. Nonetheless, infants were unprotected through PMTCT in any of these scenarios.
* Includes women who were non-adherent to either their medication or giving the infant prophylaxis (or both)__________________________
Figure 17 Proportion of Children Not Fully Protected by PMTCT at Each Time Point (n=320)
74


During pregnancy, one-quarter of infants (n=81) were not fully protected through PMTCT. During childbirth, 21% of infants (n=67) were unprotected. During postpartum,
28% of infants (n=90) were not fully protected through maternal medication and infant prophylaxis. Of the 90 children during the postpartum period who were not fully protected by both medication protocols for the woman and the infant, over one-third (n=32) were not protected by either medication protocol, putting the child at high risk for vertical HIV transmission through breastfeeding.
By examining these gaps in coverage and adherence, it becomes clear why mother-to-child transmission is still a public health concern in Zambia. Moreover, my study only captures women coming to the clinic for well child-care, which may exclude women who are the least adherent those who do not come to the clinic for any medical care. Addressing both health system problems (e.g., stock-outs of drugs and heath care workers missing HIVpositive patients), in addition to HIV-positive womens drug adherence will be critical to achieving the PMTCT targets in Zambia. In order to adequately address HIV-positive womens adherence, however, a thorough understanding of the key social barriers is necessary, including the role of gender power dynamics within couples, such as IPV. Intimate Partner Violence and Drug Adherence
As I hypothesized, IPV is negatively associated with adherence to all PMTCT medication, except sdNVP during childbirth (see Table 9). After controlling for status disclosure and other covariates, women who experienced IPV have 75% lower adjusted odds of adherence during pregnancy (p<0.05); 89% lower adjusted odds of adherence postpartum (p<0.05); and 91% lower adjusted odds of adherence to giving the infant prophylaxis (p<0.001) compared to women who have not experienced IPV.
75


Table 9 Intimate Partner Violence and the Adjusted Odds of Drug Adherence
(n=320)
Power Dynamic Variable: >80% ARV Adherence during Pregnancy (n=27l) aOR (95% Cl) Took sdNVP during Childbirth (n=!31) aOR (95% Cl) >80% ARV Adherence Postpartum (n=285) aOR (95% Cl) >80% Infant Prophylaxis Adherence (n=303) aOR (95% Cl)
la! lb' 2a3 2b} 3a3 3b' 4a3 4b1
Any 1PV 0.23** 0.25* 0.24 0.36 0.12** 0.11* 0.09*** 0.09***
(0.07 0.70) (0.07 0.87) (0.03-2.16) (0.04-3.52) (0.03 0.52) (0.02 0.59) (0.03-0.31) (0.03 0.34)
Controls: Disclosed status 18.07*** (5.96 56.65) 11.03* (1.45 83.72) 19.88*** (5.86-67.37) 12.42*** (4.05 -38.12)
Age 1.03 1.11+ 1.04 1.12 1.06 1.16* 1.03 1.08
(0.93 1 1.14) (0.98 1.24) (0.83- 1.30) (0.86-1.42) (0.95 1.19) (1.01 1.33) (0.94 1.13) (0.98- 1.20)
Infant age 0.95 1.01 0.86 0.92 0.94 0.98 0.92 0.96
(0.81 1.12) (0.84- 1.22) (0.63-1.18) (0.66- 1.28) (0.77-1.13) (0.79- 1.21) (0.79- 1.07) (0.81 1.13)
Parity 0.87 0.72+ 0.90 0.73 0.82 0.66- 1.09 0.97
(0.63 1.20) (0.51 1.03) (0.43-1.89) (0.34- 1.56) (0.56-1.18) (0.44 1.01) (0.78 1.52) (0.68- 1.38)
Highest educational attainment 0.88 0.89 0.44* 0.42* 1.06 1.18 1.07 1.12
(0.61 1.27) (0.59- 1.36) (0.23 0.89) (0.20-0.87) (0.69- 1.63) (0.72 1.94) (0.77- 1.50) (0.78-1.61)
Knowledge of 1.62+ 1.61 1.43 1.23 1.27 1.24 1.09 1.16
PMTCT (0.99-2.64) (0.89-2.89) (0.68-2.98) (0.54-2.79) (0.77 2.08) (0.72-2.15) (0.71 1.67) (0.76- 1.78)
Wealth 1.27 1.03 1.28 0.92 1.82* 1.55 1.32 1.13
(0.85 1.91) (0.64-1.65) (0.64 2.57) (0.41 -2.08) (1.13-2.92) (0.90 2.65) (0.91 1.93) (0.75- 1.70)
On treatment11 3.28** (1.39-7.71) 2.90* (1.13-7.46) N.'A N/A 2.87* (1.12-7.35) 2.27 (0.80 6.46) 3.40** (1.59 -7.29) 3.02** (1.35-6.78)
+ {KO.IO p<0.05 p<0.01 ***p<0.001
1 Different sample sizes because women were offered medication differentially 3 Before adjusting for HIV status disclosure 1 After adjusting for HIV status disclosure
J Reference group was women on short-course prophylaxis_______________________
On


My findings regarding IPV and non-adherence to ARVs for PMTCT are supported by qualitative research from South Africa (Hatcher et al., 2014; Mepham et al., 2011). My findings, however, run contrary to what Kiarie et al. (2006) report in Kenya, who found no association between IPV and uptake of PMTCT. However, at the time of the Kiarie et al. study, PMTCT in Kenya was limited to sdNVP during childbirth. In my analysis, sdNVP intrapartum is also not significantly associated with IPV. Yet, experiencing IPV is associated with non-adherence to PMTCT for all other medication protocols in the PMTCT cascade of care.
There are several reasons why sdNVP during childbirth may not be as vulnerable to non-adherence as other PMTCT medication protocols. First, sdNVP during labor/delivery is easier to conceal from a male partner who does not know the womans HIV status because men are generally not present in the delivery room. Second, if HIV-positive women come to deliver in a health center, they should receive this dose of medication by the health care personnel, whereas the other protocols require constant refills of ARV prescriptions and daily consumption within the home, which may be challenging if women have limited power within their household or are concealing their HIV status.
Although the current PMTCT regimens are more effective than only sdNVP at childbirth (WHO, 2010a), the extended duration of taking ARVs is likely more difficult for HIV-positive mothers to follow, especially those in violent relationships and with less autonomy. Alternatively, there may be a relationship between IPV and sdNVP, but this study did not have enough power to detect such an association given the small sample size of women on the short-course prophylaxis regimen who were offered the medication intrapartum (n=131).
77


The Severity and Frequency of Partner Violence and Drug Adherence
Not only does IPV reduce the odds of adherence to drugs across the PMTCT cascade, but the number of violent events and the frequency of violence in the past year also affect adherence. The women in this study who experienced more violent events in their current relationship have reduced odds of adherence to all PMTCT medication protocols except sdNVP during childbirth (see Table 10). For each additional violent event, women have 19% lower adjusted odds of adherence during pregnancy (p<0.05); 30% lower adjusted odds of adherence postpartum (p<0.01); and 32% lower adjusted odds of adherence to giving the infant prophylaxis (p<0.001).
Table 10 Adjusted Logistic Regression Results for the Odds of Drug Adherence by the Number of Violent Events a Woman Experienced
Model 1: Model 2: Model 3: Model 4:
>80% ARV Took sdNVP >80% ARV >80% Infant
Adherence during during Adherence Prophylaxis
Pregnancy Childbirth Postpartum Adherence
(n=271)1 (n=131) (n=285) (n=303)
Power Dynamic aOR aOR aOR aOR
Variable: (95% Cl) (95% Cl) (95% Cl) (95% Cl)
Number of IPV events 0.81* 0.81 0.70** 0.68***
(0.67-0.98) (0.57-1.14) (0.55-0.85) (0.57-0.80)
Controls:
Disclosed Status 16.07*** 10.36* 17 89*** 10.40***
(5.11 -50.26) (1.40-76.67) (5.27 60.77) (3.37-32.12)
Age 1.11* 1.11 1.18* 1.10+
(0.98- 1.25) (0.86- 1.43) (1.03 1.35) (0.99- 1.22)
Infant age 1.00 0.87 0.94 0.88
(0.82-1.21) (0.62 1.24) (0.75-1.17) (0.61 1.26)
Parity 0.69* 0.67 0.62* 0.88
(0.48-0.98) (0.31-1.45) (0.41-0.95) (0.61 1.26)
Highest educational 0.87 0.39* 1.16 1.09
attainment (0.57- 1.32) (0.18-0.83) (0.70- 1.90) (0.75- 1.58)
Knowledge of PMTCT 1 56 1 27 1 17 1 13
(0.88-2.74) (0.57 2.79) (0.70- 1.98) (0.75 1.7-)
Wealth 0.98 0.92 1.43 1.04
(0.61-1.58) (0.40-2.10) (0.82-2.50) (0.69- 1.57)
On treatment2 3.01* 2.49+ 3.04**
(1.17-7.74) N/A (0.87-7.14) (1.35-6.88)
+ p<0.10 p<0.05 ** p<0.01 ***p<0.001
1 Different sample sizes because women were offered medication differentially
2 Reference group was women on short-course prophylaxis_______________________
78


Additionally, women with more frequent violence in the past year have reduced odds of ARV adherence during and after pregnancy (see Table 11). With each additional mean frequency score (see Chapter Three for explanation of measurement), a woman has 10% lower adjusted odds of adherence during pregnancy (p<0.05); 16% lower adjusted odds of adherence postpartum (p<0.01); and 17% lower adjusted odds of adherence to giving the infant prophylaxis (p<0.001).
Table 11 Adjusted Logistic Regression Results for the Odds of Drug Adherence by the
Frequency of Violence in the Past Year
Model 1: Model 2: Model 3: Model 4:
>80% ARV Took sdNVP >80% ARV >80% Infant
Adherence during during Adherence Prophylaxis
Pregnancy Childbirth Postpartum Adherence
(n=271)1 (n=131) (n=285) (n=303)
Power Dynamic aOR aOR aOR aOR
Variable: (95% Cl) (95% Cl) (95% Cl) (95% Cl)
Frequency of IPV in 0.90* 0.91 0.84** 0.83***
past year Controls: (0.82-0.99) (0.78 1.06) (0.76-0.93) (0.77-0.90)
Disclosed Status 15.65*** 10.24* 17 48*** 10.06***
(4.97-49.26) (1.40-74.75) (5.15 59.34) (3.27-30.93)
Age 1.11+ 1.11 1.18* 1.11+
(1.00- 1.23) (0.86- 1.43) (1.03 1.36) (1.00- 1.23)
Infant age 0.99 0.88 0.94 0.93
(0.82-1.19) (0.62 1.24) (0.75-1.17) (0.78-1.10)
Parity 0.88 0.66 0.61* 0.86
(0.57- 1.32) (0.31-1.43) (0.39-0.94) (0.60- 1.24)
Highest educational 1.08 0.39* 1.15 1.08
attainment (0.74- 1.56) (0.18-0.83) (0.70- 1.89) (0.75- 1.56)
Knowledge of PMTCT 0.87 1.28 1.20 1.13
(0.57-1.31) (0.58-2.82) (0.71-2.01) (0.75- 1.69)
Wealth 0.98 0.92 1.45 1.04
(0.61-1.58) (0.40-2.10) (0.83 -2.54) (0.69- 1.57)
On treatment 3.03* 2.48+ 2.95**
(1.18-7.83) N/A (0.86-7.14) (1.30-6.66)
+ p<0.10 p<0.05 ** p<0.01 ***p<0.001
1 Different sample sizes because women were offered medication differentially
2 Reference group was women on short-course prophylaxis_______________________
79


Not only is the number of IPV events and the frequency of IPV associated with nonadherence during and after pregnancy, but I also found strong evidence of a dose-response relationship for the majority of protocols with these two variables. Figure 18 displays the adjusted logistic regression results for the number of violent events as dummy variables. As is evidenced in the figure, each adjusted odds ratio associated with each dummy variable is greater then the prior providing evidence of a dose-response relationship between the number of IPV events and non-adherence to PMTCT medication.
Comparison Group: No IPV (i.e., zero reported events)
Figure 18 Adjusted Odds Ratio Estimates for Adherence to PMTCT Medication at Each Time Period by the Number of Violence Events
A similar trend is also seen with the frequency of IPV in the past year with the
exception of medication during pregnancy (see Figure 19). During pregnancy, there is no significant difference in the odds of adherence between having a frequency of IPV score of six to 10 and a score of greater than 10. However, for postpartum medication and infant
80


Full Text

PAGE 1

GENDER, POWER, & VERTICAL HIV PREVENTION IN URBAN ZAMBIA By KAREN MARIE HAMPANDA M.P.H., Boston University, 2010 B.A., Beloit College, 2008 A thesis submitted to the faculty of the Graduate School of the University of Colorado Denver in partial fulfillment of the requirements for the degree of Doctor of Philosophy Health and Behavioral Sciences Program 2016

PAGE 2

ii 2016 KAREN MARIE HAMPANDA ALL RIGHTS RESERVED

PAGE 3

iii This thesis for the Doctor of Philosophy degree by Karen Marie Hampanda has been approved for the Health and Behavioral Sciences Program b y Sara Yeatman Advisor Karen Spencer Chair Sheana Bull Patrick Krueger April 29 2016

PAGE 4

iv Hampanda, Karen Marie (PhD, Health and Behavioral Sciences) Gender, Power, & V ertical HIV Prevention in Urban Z ambia Thesis directed by Associate Professor Sara Yeatman ABSTRACT Biomedical and behavioral interventions can virtually eliminate the risk of vertical (i.e., mother to child) HIV transmission P regnant and breastfeeding women adhere nce to prevention of mother to child transmission (PMTCT) interventions however, remains a challenge across sub Saharan Africa. Using a concurrent mixed methods approach, including a survey and semi structured interviews, I test whether a relationship ex low power within married couple s (based on doma ins from the Theory of Gender and Power) and adherence across the PMTCT cascade of care, including drug adherence during and after pregnancy, safe infant feeding practices, and pediatric HIV testing. The results of this study indicate that intimate partner violence is particularly detrimental to PMTCT adherence Cer tain PMTCT protocols are also affected by p artner controlling behaviors, participation in household decis ions, and economic dependence, but not to the same extent as violence. W omen with low power cite a lack of partner support and an unwillingness to disclose their HIV status to the husband due to fear of violence or abandonment as reasons for low PMTCT adherence. Conversely, women with high power cite partner support and the ability to prioritiz e PMTCT, sometimes even over the marriage as enabling ad herence Based on these results, a ugmented efforts to a ddress gender power dynamics both in society and within the home are recommended to promote the health of HIV positive women and their families The form and content of this abstract are approved. I r ecommend its publication Approved: Sara Yeatman

PAGE 5

v ACKNOWLEDGEMENTS This dissertation would not have been possible without the intellectual emotional, and logistical support from several key individuals. First and foremost, a tremendous thank you goes to my advisor, Dr. Sara Yeatman. Your guidance was absolutely crucial for this research and my academic progression. Thank you for always taking the time to work through ideas with me and provide detailed feedback. In addition, I was fortunate to have an amaz ingly dedicated committee: Drs. Sheana Bull, Karen Spencer and Patrick Krueger. Thank you for your outstanding mentorship and consistent help with this study over the past several years. I deeply appreciate your time and assistance. I would also like to extend my gratitude to my research team in Zambia. First, Yusuf and support has truly been essential to the success of this project. I cannot thank you enoug h for your guidance. In addition, my sincere appreciation goes to the Lusaka District Health Office, the University Teaching Hospital, and Kanyama Health Center for allowing me into your community. I would also like to thank my research assistants in Lusaka, Christine Chewe Sakala, Grace Lungeani Phiri, Franklin M unsanje, and Bibi Lambert Manda, for your dedication to this study. I could not have asked for a better research team and owe much of the success of this study to your dedicated efforts. Z ikomo Kwambili! In addition, I would like to thank my family for their unconditional love and support during this process. My parents, Johanna and Jack McManemin, are the most generous, kind, and loving people I have ever met. Thank you for your ongoing s upport of my education and for always taking the time to help in any capacity needed, including emotional reassurance, editing my work, and financial support I honestly do not know what I would have done without you!

PAGE 6

vi Also, to my brother, John McManemin, thank you for your persistent encouragement and love. You have helped me more than you know. I n addition, I am extremely grateful to my husband, Mwiza Hampanda, for his support, love, and patience during this research. Not only have you helped me understa nd the Zambian context and translated interviews for me but you also are my best friend whose love and kindness during the project have kept me sane. Thank you so much for your assurance and comfort when things got tough and for your enthusiasm for my acc omplishments. All of your kind words and gestures have made this challenge bearable because I had you by my side at every step. Finally, this study was made possible through funding from 1) the National Institute of Mental Health (Awar d Number F31MH10734 8) and 2) the Center for Global Health at the University of Colorado Denver

PAGE 7

vii TABLE OF CONTENTS CHAPTER I. INTRODUCTION 1 The Public Health Problem: Pediatric HIV/AIDS ................................ .......................... 2 The Solution: Prevention of Mother to Child Transmission ................................ .......... 3 The Neglected Factor: Gender Power Dynamics ................................ ............................ 4 Research Questions and Hypotheses ................................ ................................ ............... 5 Chapter Summary ................................ ................................ ................................ ............ 7 II. BACKGROUND & THEORETICAL FRAMEWORK ................................ ................. 8 Prevention of Mother to Child Transmission in Zambia ................................ ................ 8 Individual Level Perspective ................................ ................................ ......................... 10 Family Systems Perspective ................................ ................................ .......................... 12 The Theory of Gender and Power ................................ ................................ ................. 13 Conceptual Model ................................ ................................ ................................ ......... 16 Gender, Power, and Vertical HIV Prevention ................................ ............................... 18 Economic Inequities ................................ ................................ ............................... 17 Household Decision Making ................................ ................................ ................. 18 Intimate Partner Violence and Control ling Behavior ................................ ............ 19 Disclosure of HIV Status ................................ ................................ ............................... 22 Chapter Summary ................................ ................................ ................................ .......... 23 III. METHODS ................................ ................................ ................................ .................... 24 Study Setting ................................ ................................ ................................ ................. 24 Research Design ................................ ................................ ................................ ............ 25 Ethical Considerations ................................ ................................ ................................ ... 26

PAGE 8

viii Formative Research and Preparation ................................ ................................ ............. 26 Sampling Technique ................................ ................................ ................................ ...... 30 Sample Description ................................ ................................ ................................ ....... 32 Survey Measures ................................ ................................ ................................ ........... 34 PMTCT Adherence ................................ ................................ ................................ 33 Gender Power Dynamics ................................ ................................ ....................... 36 Quantitative Analytic Strategy ................................ ................................ ...................... 38 Data Entry and Cleaning ................................ ................................ ........................ 38 Dependent Variables 39 Independent Variables 40 Covariates .... 43 Missing Data and Multiple Imputation ................................ ................................ .. 45 Semi Structured Interviews ................................ ................................ ........................... 48 Qualitative Analysis ................................ ................................ ................................ ...... 50 Limitations ................................ ................................ ................................ ..................... 52 Chapter Summary ................................ ................................ ................................ .......... 53 IV. THE SEXUAL DIVISION OF POWER ................................ ................................ ....... 54 Review of Methods ................................ ................................ ................................ ....... 54 Intimate Partner Violence and Controlling Behaviors ................................ .................. 54 Experiences With Specfic Forms of Violence and Control ................................ .......... 57 Predictors of Intimate Partner Violence and Control ................................ .................... 60 Positive Deviants ................................ ................................ ................................ ........... 65 Gender Asymmetry in Intimate Partner Violence ................................ ......................... 68 Chapter Summary ................................ ................................ ................................ .......... 70

PAGE 9

ix V. DRUG ADHERENCE DURING AND AFTER PREGNANCY ................................ .. 70 Review of Methods ................................ ................................ ................................ ....... 71 Adequate and Optimal Drug Adherence ................................ ................................ ....... 72 Unprotected Children at Risk of Vertical HIV Infection ................................ .............. 75 Intimate Partner Violence and Drug Adherence ................................ ........................... 75 The Severity and Frequency of Partner Violence and Drug Adherence ....................... 78 Specific Types of Partner Violence and Drug Adherence ................................ ............ 82 Partner Control and Drug Adherence ................................ ................................ ............ 87 Participation in Household Decisions and Drug Adherence ................................ ......... 87 ................................ .......................... 87 The Combined Influence of Gender Power Dynamics on Drug Adherence ................. 91 HIV Status Disclosure and Drug Adherence ................................ ................................ 93 Chapter Summary ................................ ................................ ................................ .......... 94 VI. SAFE INFANT FEEDING ................................ ................................ ............................ 95 Review of Methods ................................ ................................ ................................ ....... 97 Infant Feeding Practices ................................ ................................ ................................ 97 Intimate Partner Violence and Safe Infant Feeding ................................ .................... 100 The Severity and Frequency of Partner Viole nce and Safe Infant Feeding ................ 102 Specific Types of Violence and Safe Infant Feeding ................................ .................. 108 Other Gender Power Dyanmics and Safe Infant Feeding ................................ ........... 109 The Combined Influence of Gender Power Dynamics on Safe Infant Feeding .......... 113 HIV Status Disclosure and Safe Infant Feeding ................................ .......................... 114 Chapter Summary ................................ ................................ ................................ ........ 115

PAGE 10

x VII. PEDIATRIC HIV TESTING ................................ ................................ ...................... 116 Review of Methods ................................ ................................ ................................ ..... 117 Adherence to Pediatric HIV Testing ................................ ................................ ........... 117 Intimate Partner Violence and Pediatric HIV Testing ................................ ................. 119 The Severity and Frequency of Partner Vio lence and Pediatric HIV Testing ............ 120 Specific Types of Partner Violence and Pediatric HIV Tesitng ................................ .. 123 Other Gender Power Dynamics and Pediatric HIV Testing ................................ ........ 126 The Combined Influence of Gender Power Dynamics on Pediatric HIV Testing ...... 129 HIV Status Disclosure and Pediatric HIV Testing ................................ ...................... 131 Adherence to Other PMTCT Protocols and Pediatric HIV Testing ............................ 132 Chapter Summary ................................ ................................ ................................ ........ 134 VIII. PMTCT IN CONTEXT ................................ ................................ ............................... 135 Review of Methods ................................ ................................ ................................ ..... 135 Qualitative Sample Characteristics ................................ ................................ ............. 136 High Power and High PMTCT Adherence ................................ ................................ 137 ................................ .......................... 137 HIV Status Disclosure to Husbands ................................ ................................ .. 138 Partner Support ................................ ................................ ................................ .. 139 HIV Status Concordance ................................ ................................ ................... 140 Low Power and Low PMTCT Adherence ................................ ................................ ... 141 Prioritizing the Marriage ................................ ................................ ................... 141 Non Disclosure of HIV Status ................................ ................................ .......... 142 Lack of Partner Support ................................ ................................ .................... 144

PAGE 11

xi Discordance and Emotional Abuse ................................ ................................ ... 144 Controlling Behavior ................................ ................................ ........................ 145 ................................ ................................ ..... 145 High Power and Low PMTCT Adherence ................................ ................................ .. 146 Poor Mental Health ................................ ................................ ........................... 146 Maternal Illness ................................ ................................ ................................ 1 47 Poor Understanding of Protocols and Erratic Supply of Medication ............... 147 Family Influence ................................ ................................ ............................... 148 Low Power and High PMTCT Adherence ................................ ................................ .. 148 Family Support ................................ ................................ ................................ .. 148 Child Death ................................ ................................ ................................ ....... 149 Chapter Summary ................................ ................................ ................................ ........ 150 IX. CONCLUSIONS ................................ ................................ ................................ ......... 151 1. ............. 151 1.a. Which Power Dynamics Are the Most Detrimental? ................................ ........... 151 1.b. Which PMTCT Protocols Are the Most Affected? ................................ .............. 155 2. What is the Role of HIV Status Disclosure? ................................ ........................... 156 3. Why Do Gender Power Dynamics Affect PMTCT Adherence? ............................ 157 Final Conceptual Model ................................ ................................ .............................. 159 Addressing Gender Power Imbalances and PMTCT Adherence ................................ 160 Next Steps ................................ ................................ ................................ .................... 163 REFERENCES ................................ ................................ ................................ ...................... 164

PAGE 12

xii APPENDIX A. Consent Form 194 B. Survey Q uestionnaire ................................ ................................ ................................ .. 196 C Semi Structured Interview G uide ................................ ................................ ................ 208 D My D ata com pared to the Most R ecent ZDHS ................................ ........................... 211

PAGE 13

xiii LIST OF TABLES TABLE 1 WHO Options for PMTCT Programs in 2014 ................................ ............................... 9 2. Constructs and Questions on Gender Power Dynamics ................................ .............. 37 3. Missing Data for Key Variables ................................ ................................ .................. 46 4. Participant Characteristics ................................ ................................ ........................... 47 5. Sampling Domains for Semi structured Intervi ews ................................ ..................... 48 6. Example of Qualitative Coding Scheme ................................ ................................ ...... 52 7. Multivariate Logistic Regression Results for the Odds of Experiencing IPV and Three or More Partner Controlling Behaviors ................................ ................................ ....... 61 8. Correlation Coefficients B etween Gender Power Dynamic Measures ........................ 64 9. Intimate Partner Violence and the Adjusted Odds of Drug Adherence ....................... 77 10. Adjusted Logistic Regression Results for the Odds of Drug Adherence by the Number of Violent Events a Woman Experienced ................................ ................................ .... 78 11. Adjusted Logistic Regression Results for the Odds of Drug Adherence by the Frequency of Violence in the Past Year ................................ ................................ ...... 79 12. Adjusted Logistic Regression Results for the Odds of Drug Adherence by Injuries from Intimate Partner Violence ................................ ................................ ................... 82 13. Adjusted Logistic Regression Results for the Odds of Drug Adherence by Experiences with Emotional Violence ................................ ................................ ............................. 85 14. Adjusted Logistic Regression Results for the Odds of Drug Adheren ce by Experiences with Sexual Violence ................................ ................................ ................................ ... 86

PAGE 14

xiv 15. Adjusted Logistic Regression Results for the Odds of Drug Adherence by Experiences with Physical Violence ................................ ................................ ................................ 87 16. Partner Controlling Behaviors and the Adjusted Odds of Drug Adherence ................ 89 17. Household Decision Making and the Adjusted Odds of Drug Adherence ................................ ................................ ................................ .................... 90 18. and the Adjusted Odds of Drug Adherence ..................... 91 19. Adjusted Logistic Regression Results for the Odds of Drug Adherence by Gend er Power Dynamics in Combined Models ................................ ................................ ....... 92 20. HIV Status Disclosure and the Adjusted Odds of Drug Adherence ............................ 94 21. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices by Experiences with Intimate Partner Violence ................................ ......................... 101 22. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices by the Number of Violent Events ................................ ................................ .............. 103 2 3. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices by the Frequency of Violence in the Past Year ................................ .......................... 104 24. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices by Injuries from Violence ................................ ................................ .......................... 105 25. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices by Emotional Violence ................................ ................................ .............................. 107 26. M ultinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices by Sexual Violence ................................ ................................ ................................ .... 108 27. Multinomial Logistic Regression Results for t he Odds of Safe Infant Feeding Practices by Physical Violence ................................ ................................ ................................ .. 109

PAGE 15

xv 28. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices b y Partner Controlling Behavior ................................ ................................ ................ 110 29. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices Household Decision Making ................................ ...... 111 30. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices ................................ ................................ .................... 112 31. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices by Gender Power Dynamics ................................ ................................ ...................... 113 32. Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices by Gender Power Dynamics ................................ ................................ ...................... 115 33. Logistic Regression Results for the Odds of Pediatric HIV Testing by Experiences with Intimate Partner Violence ................................ ................................ .......................... 119 34. Adjus ted Logistic Regression Models for the Odds of Pediatric HIV Testing by The Number of IPV Events ................................ ................................ ............................... 120 35. Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by the Frequency of IPV in the Past Year ................................ ................................ ............ 121 36. Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by Injuries Experienced as a Result of Violence ................................ ................................ ......... 122 37. Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by Emotional Violence ................................ ................................ ................................ ... 123 38. Adjusted Logistic Regression Models f or the Odds of Pediatric HIV Testing by Sexual Violence ................................ ................................ ................................ ..................... 124

PAGE 16

xvi 39. Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by Physical Violenc e ................................ ................................ ................................ ....... 125 40. Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by Partner Control ................................ ................................ ................................ ....................... 127 41. Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by ................................ ....................... 128 42. Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by ................................ ................................ ......................... 129 43. Adjusted Logistic Regression Results for the Odds of Pediatric HIV Testing by Gender Power Dynamics ................................ ................................ ................................ ...... 130 44. Adjusted Logistic Regression Results for the Odds of Pediatric HIV Testing by HIV Status Disclosure ................................ ................................ ................................ ........ 132 45. The Odds of Pediatric HIV Testing by Other PMTCT Protocols .............................. 133 46. Qualitative Sample Characteristics ................................ ................................ ............ 137

PAGE 17

xv ii LIST OF FIGURES FIGURE 1. Map of Zambia ................................ ................................ ................................ ................ 2 2. The PMTCT Cascade of Care ................................ ................................ ......................... 3 3. Conceptual Model ................................ ................................ ................................ ......... 16 4. Map of Lusaka and Kanyama Compound ................................ ................................ .... 24 5. Study Timeline ................................ ................................ ................................ .............. 27 6. Children's Under Five Card ................................ ................................ .......................... 31 7. Sample Description ................................ ................................ ................................ ....... 32 8. Example of Drug Adherence Question ................................ ................................ ......... 34 9. Women's Experiences with IPV and Partner Controlling Behaviors ........................... 55 10. riences With Specific Violent Events ................................ .................. 57 11. Experiences with Specific Forms of Male Controlling Behavior ................................ 58 12. Types of Intimate Partner Violence Reported ................................ .............................. 59 13. Household Decision Making and Ex periences with IPV/Controlling Behavior .......... 66 14. IPV Victimization in Physically Violent Relationships by Gender .............................. 68 15. Adequate and Optimal Drug Adherence Levels ................................ ........................... 72 16. Number of Medication Protocols Whe re Women Achieved Adequate Adherence 73 17. P roportion of Children Not Fully Protected by PMTCT at Each Time Point .............. 74 18. Adjusted Odds Ratio Estimates for Adherence to PMTCT Medication at Each Time Period by the Number of Violence Events ................................ ................................ .. 80

PAGE 18

xviii 19. Adjusted Odds Ratio Estimates for Adherence to PMTCT Medication at Each Time Period by the Frequency of IPV in the Past Year Score ................................ .............. 81 20. Proportion of HIV infected Women Practicing Three Types of Infant Feeding Modalities ................................ ................................ ................................ .................... 97 21. Reported Food/Liquids Given to Infants Prior to Six Months Among HIV Infected Women Practicing Mixed Feeding ................................ ................................ ............ 100 22. Adjusted Multinomial Odds Ratio Estimates for Adherence to Safe Infant Feeding by the Number of IPV Events ................................ ................................ ......................... 106 23. Adjusted Multinomial Odds Ratio Estimates for Adherence to Safe Infant Feeding by the Frequency of IPV in the Past Year Score ................................ ............................ 106 24. Proportion of Children with Any Pediatric HIV Testing ................................ ........... 118 25. Proportion of Children Testing at 6 Weeks and 6 Months ................................ ........ 118 26. Adjusted Odds Ratio Estimates for Adherence to Both Recommended Pediatric HIV Tests by the Frequency of IPV in the Past Year ................................ ........................ 122 27. Final Conceptual Model ................................ ................................ ............................. 159

PAGE 19

xix LIST OF ABBREVIATIONS AIDS Acquired immune deficiency syndrome ANC Antenatal care ARV Antiretroviral ART Antiretroviral therapy CTS2 Conflict Tactics Scale HIV Human immunodeficiency virus IPV Intimate partner violence IQR Interquartile range MCH Maternal and child health NVP Nevirapine PCA Principal componen t analysis PMTCT Prevention of mother to child transmission PTSD Post traumatic stress disorder sdNVP Single dose Nevirapine STI Sexually transmitted infection UTH University Teaching Hospital VAS Visual analog scale VCT Voluntary counseling and t esting WHO World Health Organization ZDHS Zambian Demographic and Health Survey

PAGE 20

1 CHAPTER I INTRODUCTION Throughout my public health career, o ne of the most emotionally challenging events I have witnessed wa s an HIV positive mother being informed by a nurse that her child is also HIV positive T he reason this is so heart wrenching is because vertical (i.e., mother to child) HIV transmission is almost entirely preventable through biomedical and beh avioral interventions, such as antiretroviral (ARV) medication and safe infant feeding practices. Without intervention, however, approximately 40% of infants born to HIV positive women will become infected in utero, during childbirth, or while breastfeedin g (WHO, 2010c) Zambia is one of the countries hardest struck by HIV one in five pregnant women are currently living with the virus (CIDRZ, 2012) Zambia also has one of the highest global fertility rates with an average of almost six children per woma n (CIA, 2014) This translates into approximately 78,000 HIV positive women giving birth annually with 15% currently transmit ting the virus to their children (UNICEF, 2014) do to lack of access and poor adherence to PMTCT care My research took place in the capital city of Zambia, Lusaka (see Figure 1 ), which has a population of 1.8 million individuals (CIA, 2014) I chose to conduct my research here for two reasons. First, women of reproductive age living in Lusaka have disproportionately high HIV rates (CS O, 2009) T he most recent estimates indicate that the overall HIV prevalence for adults 1 in the country is 13%, yet women in Lusaka ha ve an HIV prevalence of over 19 % (CSO, 2014) Second, L usaka has the best health care in the country, making it an ideal l ocation to study non adherence to prevention of mother to child transmission ( PMTCT ) because generally, w omen have access to health care. 1 Estimate is for adults age 15 49 years

PAGE 21

2 Figure 1 Map of Zambia Interpersonal relationships including the family, are a critical influence over health behaviors (Glanz et al., 2008). W ithin families however, sexual divisions exist leading to more or less relative power between heterosexual partners One aspect that has been neglected in the PMTCT literature is the role that gender power dynamics within married couples may play on HIV in high HIV prevalence settings, such as Zambia In thi s chapter I first describe the disease burden of mother to child transmission of HIV and why gender power dynamics are a likely adherence to PMTCT, followed by my a priori research questions and hypotheses

PAGE 22

2 The Public Health Problem: Pediatric HIV/AIDS Although tremendous improvements have been made, hundreds of thousands of infants unfortunately continue to contract H IV each year I n 2013, 240 000 2 children became newly infected with HIV that is approximately 660 children every day (UNIADS, 2014) The main route of new pediatric HIV infections is from mother to child, which accounts for 15% of all new HIV infections annually a substantial proportion of the overall global HIV incidence ( Msellati, 2009; UNAIDS, 2012; WHO, 2010c) Sub Saharan Africa carries th e largest pediatric HIV burden with 94% of HIV infected children living in this region (WHO, 2013). In Zambia alone, over 10,000 infants are newly infected with HIV annually (UNAIDS, 2 014) Moreover, p ediatric HIV is a major contributor to infant and child mortality in sub Saharan Africa (Khan, Michaels, & Eley, 2006) In southern Africa, where Zambia is located, 20% of child deaths are attributable to HIV infection (Torpey et al., 2012) HIV progresses more rapi dly to AIDS in children and without treatment, 50% of infected children die by their second birthday (Zambia Ministry of Health, 2008) Every year, there are over 260 000 pediatric deaths worldwide due to AIDS related illnes ses one in seven of all AIDS related deaths (WHO, 2010d) In sub Saharan Africa, AIDS remains one of the leading causes of death for children less than five years ( Khan et al., 2006) If the epidemic of mother to child transmission of HIV is not halted, pediatric AIDS threatens to reverse years of steady progress in infant and child survival in the region (UNAIDS, 1999) Thus, eliminating the transmission of HIV from mother to child is currently one of the most pressing global health priorities 2 Estimate is from 210,000 to 280,000

PAGE 23

3 The Solut ion: Prevention of Mother to Child Transmission Fortunately, vertical HIV transmission is almost entirely preventable through interventions commencing during pregnancy and continued throughout the breastfeeding cascade of care. HIV positive women and their children includes an HIV diagnosis during antenatal care (ANC) ; maternal ARV medication during and after pregnancy; ARV medication during childbirth (if applicable) ; infant ARV prophylaxis; specific in fant feeding practices ; and pediatric HIV testing (see Figure 2 ). If HIV positive women adhere to these protocols, the risk of vertical HIV transmission is reduced to less than 5% (WHO, 2010a) Figure 2 The PMTCT Cascade of Care Major strides have been made at the political and institutional level s to increase coverage and accessibility of PMTCT services throughout sub Saharan Africa (WHO, UNAIDS, & UNICEF, 2011) In 200 3, only 3% of HIV positive pregnant women living in the region had access to ARV drugs for the purpose of PMTCT (Hardon et al., 2012). This percentage dramatically increased to 33% in 2007 and 59% in 2011 (WHO, 2010d) Source: Based on 2010 WHO PMTCT Guidel ines for Zambia

PAGE 24

4 Commendably, Z ambia is one of six cou ntries in sub Saharan Africa currently reporting PMTCT drug coverage of more than 75% (UNAIDS, 2012) During my field research, every one of the 25 p ublic health centers in Lusaka offered PMTCT care 3 Despite these admirable gains in access adherence to PMTCT among HIV positive women remains a challenge across sub Saharan Africa (Nachega et al., 2012) To date, the focused on scaling up biomedical s ervices While this is an essential component, it has left a gap in our understanding of the social and behavioral determinants PMTCT There is currently a poor understanding of why HIV positive women are not adhering to PMTCT proto cols despite their accessibility throughout much of urban sub Saharan Africa. This knowledge is critical for our ability to create effective behavioral interventions aimed at curb ing the HIV epidemic and reducing child mortality in the region. The Neglect ed Factor: Gender Power Dynamics Gender is fundamentally a matter of social embodiment and is considered one of the social forces with the greatest constraint over human agency (Connell, 1987; Rylko Bauer, Whiteford, & Farmer, 2009) For this research, I define gender as the culturally expected norms shared within a society about appropriate male and female behavior, characteristics, and roles (Blanc, 2001) Gender is an essential consideration in preventing vertical HIV because only women can pass HIV t o offspring, and, for this reason, protocols to prevent transmission are aimed exclusively at women. Paradoxically, however, HIV positive women may not hold the necessary power within the family to make independent decisions regarding adherence t o the vari ous PMTCT protocols in many settings such as Zambia. 3 Although stock outs of medication, particularly infant Nevirapine, were common

PAGE 25

5 One major theoretical shortcoming in the current literature on PMTCT adherence is a lack of examination into dynamics within heterosexual couples, particularly gender power imbalances. P olicies and cou nseling regarding PMTCT in Zambia do not adequately take into consideration dynamics within couples affected by HIV which may exert a large influence over HIV relevant health behaviors HIV positive women are assumed to have the necessary agency and power to make these health care decisions, even in cultures where large gender inequities persist in society and the home Research Questions and Hypotheses The overall aim of this study i s to examine the relationship between gender power dynamics within married couples, which stem from larger societal inequities, and adherence to PMTCT protocols across the cascade of care for HIV positive women and their children. The theoretical framework for this study comes primarily from the Theory of Gender Power and Family Systems Theory. Using a concurrent mixed methods approach, including a survey and semi structured interviews, I test whether a relationship ex low power within married couple s and adherence to PMTCT. I specifical ly analyze how intimate partner violence (IPV) and control household decision making, and economic inequity influence HIV related beh aviors. I additionally explore the role of HIV status disclosure to husbands and why gender power d ynamics may have a negative influence on HIV adherence The specific research questions and a priori hypotheses of my study are : Question 1: ouples reduce the odds of PMTCT adherence?

PAGE 26

6 Hypothesis 1: Th ere is a relationship between non adherence to PMTCT The presence of violent or controlling behavior by a husband, low household decision making power, and economic dependence on the husband will reduce the odd s of PMTCT adherence. Question 1A: Which power dynamics are the most likely to hinder PMTCT adherence? Hypothesis 1A: HIV positive women in relationships characterized by high levels of male partner controlling behavior (e.g., through violence/threat o f violence, economic control, etc.) will have the lowest PMTCT adherence. Question 1B: Which PMTCT protocols are most affected by gender power imbalances within couples? Hypothesis 1B: Gender power imba lances will have the strongest e adherence to postpartum protocols because these are the most perceptible to a spouse (i.e., maternal medication infant medication and safe infant feeding practices ). Gender power im balances will have the weakest e ffect on childbirth prot ocols because this is a single dose medication taken at the clinic without the need for a permission. Question 2: What is the role of HIV status disclosure in the relationship between gender power dynamics and PMTCT adherence? Hyp othesis 2: HIV status disclosure mediates the relationship between gender power dynamics and non adherence to PMTCT Question 3: Why do gender power dynamics within couples affect HIV positive

PAGE 27

7 Hypothes is 3: R elationships with large power inequities are characterized by less spousal communication and support, making PMTCT more arduous on the woman. Additionally, f ear of violence or abandonment prevents women from disclosing their HIV status, leading to difficulty adhering to PMTCT Chapter Summary Eliminating mother to child transmission is one of the top global HIV/AIDS priorities, but poor understanding of the social barriers to PMTCT adherence among HIV positive women is limiting progress towards thi s goal. Currently, there is l imited knowledge regarding how gender power dyanmics within married couples affect HIV adhere nce to PMTCT in sub Saharan Africa. The present study addresses this gap by testing r elative power within couples decreases adherence across the PMTCT cascade of care in Lusaka, Zambia. In the following chapter, I provide a review of the current PMTCT literature and de scribe the theoretical framework for this research Chapter Three then describes my methodology followed by my specific findings in Chapters Four through Eight

PAGE 28

8 CHAPTER II BACKGROUND & THEORETICAL FRAMEWORK Prevention of Mother to Child Transmission in Zambia Zambia has had a national PMTCT p rogram in place since 1999; however, Zambia PMTCT recommendations have significantly evolved over the past decade with growing scientific evidence Prior to 2010, Zambia and many other African countries only offered single dose Nevirapine (sdNVP) intrapartum (i.e., at the time of labor/delivery) to HIV positive women for PMTCT. This intervention occurred at the health facility in the delivery room out of sight of the husband, which enabled women to adhere to the medication witho ut having to disclose their HIV status to the husband and/ or obtain his permission. In 2010 PMTCT protocols in Zambia beca me more complicated because of improved medical knowledge, recommending an extended duration of taking combination ARVs throughout pr egnancy and postpartum, including providing the infant with prophylaxis At this same time, the World Health Organization ( WHO ) began recommending and still do recommend, that HIV positive women in low/middle income settings like Zambia exclusively breas tfeed their infants for six months, followed by the introduction of complementary fo ods and continue breastfeeding to at least one year of life (WHO, 2010b) In addition, infants who are exposed to HIV from their mothers should be tested for the virus at 6 weeks, 6 months, 12 months, and 18 months and started on antiretroviral therapy ( ART ) immediately if infected (Zambia Ministry of Health, 2010). Due to the extended duration of taking ARVs and the increased number of necessary PMTCT interventions that commenced in 2010, it is plausible that adherence to PMTCT

PAGE 29

9 became more challenging for HIV positive women This may be particularly true for women who have low relative power within the family because the current recommendations require adherence to protocols that take place outside of the health care facility and in the home Table 1 WHO Options for PMTCT Programs in 2014 Woman receives Infant receives Option Treatment (for CD4 count < 350 cells/mm) Prophylaxis (for CD4 count >350 cells/mm) A* Triple ARVs starting as soon as diagnosed, continued for life Antepartum: AZT 1 starting as early as 14 weeks gestation Intrapartum: at onset of labor, single dose NVP and first dose of AZT/3TC 2 Postpartum: daily AZT/3TC through 7 days postpartum Daily NVP from birth until 1 week after cessation of all breastfeeding; or, if not breastfeeding or if mother is on treatment, through age 4 6 weeks B Same initial ARVs for both: Daily NVP or AZT from birth through 4 6 weeks regardless of infant feeding method Triple ARVs starting as soon as diagnosed, continued for life Triple ARVs starting as early as 14 weeks gestation and continued intrapartum and through childbirth if not breastfeeding or one week after cessation of all breastfeeding B + Same treatment and prophylaxis: Daily NVP or AZT from birth through 4 6 weeks regardless of infant feeding method Regardless of CD4 count, triple ARVs starting as soon as diagnosed, continued for life 1 AZT = zidovudine (specific type of ARV) 2 AZT / 3TC = zidovudine /lamivudine (specific type of combination ARV) S ource: WHO, 2012 Programmatic Update: Use of Antiretroviral Drugs for Treating Pregnant Women and Preventing HIV Infection in Infants At the time of the study (2014), WHO recommend ations included three PMTCT options for countries to target vertical H IV transmission b ased on their individual health system capabil ities (See Table 1). Zambia opted for Option A 4 in 2 010 which included two 4 In 2014, toward s the end of my field research, Zambia began transitioning to Option B + which is what the country currently follows. Essentially this is the same regimen as the treatment option under Option A, but all HIV positive women regardless of their CD4 count or clinical stage initiate lifelong ART

PAGE 30

10 First, p regnant or breastfeeding women with a CD4 cell count 5 of less than 3 50 or WHO clinical stage three or four commenced lifelong ART Second HIV positive women who did not meet the treatment criteria were given short course ARV prophylaxis starting at 14 weeks gestation and continued until one week postpartum (WHO, 2012a) HIV exposed i nfants received Nevirapine (NVP 6 ) prophylaxis on both regimens but for differing amounts of time While effective biomedical interventions and national policies exist, health behaviors and compliance to these interve ntions is a critical, underemphasized, component of PMTCT success related behaviors is essential to achiev e the goal of global eliminating vertical HIV transmission. In the next sections, I explain the theoretical perspectives typically applied to the analysis HIV behaviors related to PMTCT, and why a focus on the familial level, in particular gender power dynamics, is imperative. Individual Level Perspective Individual level theories of healt h behavior, such as the Health Belief Model (Glanz, Rimer, & Viswanath, 2008) and the Theory of Planned Behavior (Glanz et al., 2008) have been widely used to explain HIV positive women s PMTCT related behaviors. The construct the Health Belief Model is a somewhat helpful construct in understanding women s PMTCT behaviors For example, e stablished perceived barriers in discrimination of atus being disclosed to sexual partner s family, or the community; and opposition from the husband (Kebaabetswe, 2007; Nyasulu & Nyasulu, 2011). 5 Blood test for HIV positive individuals to diagnosis HIV progression/stage of disease 6 A type of ARV drug given in syrup form to infants

PAGE 31

11 Additionally, the Theory of Planned Behavior has been used to explain individual factors that may motivate an H IV positive woman to use PMTCT. C onstructs of attitude, perceived norms, and personal agency are frequently cited as ways to understand and address PMTCT adherence (Awiti Ujiji et al., 2011; Hardon et al., 2012; Kebaabetswe, 2007; Nyasulu & Nyasulu 2011; Varga & Brookes, 2008) For example, a study in South Africa used the Theory of Planned Behavior to analyze a clinic based health education intervention aimed to increase s intention to use PMTCT in South Africa (Igumbor, Pengpid, & Obi, 2006) One important finding from the 2006 study by Igumbor et al. however, is that HIV positive women consistently reported low control beliefs regarding PMTCT and as a result, the authors recommended efforts Another study by Igumbo r and Obi (2006) recommend ed expanding and enhancing interventions that empower HIV positive women in order to improve behavioral intention to use PMTCT. researched yet c rucial component of increasing PMTCT uptake and adherence in sub Saharan Africa. PMTCT i nterventions focused only on the individual level (i.e., HIV positive woman) typically attempt to increase knowledge and behavioral intentions through education and counseling (Creek et al., 2009; Igumbor et al ., 2006) For example, Msellati (2009) recommend ed (Msellati, 2009, p. 810) Currently i n many sub Saharan African countries, however, such as Zambia, widespread PMTCT and HIV education campaigns are already in place (Besser, 2010a; MOH, 2008) but PMTCT adherence is poor nonetheless. For instance over 88 % of women knew that HIV could be transmitted by breastfeeding in the last Zambia Demographic and Health Survey

PAGE 32

12 (ZDHS), but only 21% of HIV positive mothers took the necessary ARVs while breastfeeding to prevent HIV transmission (UNICEF, 2012) Education regarding PMTCT is certainly a necessary component but not suff icient by it self for behavior change. The problem with using only individual level constructs to understand PMTCT adherence is that the sole responsibility to prevent HIV transmission is placed on the HIV infected mother, without taking into consideration the context in which her behaviors occur (Rylko Bauer et al., 2009) This grossly overestimates the power HIV positive women typically have, both in society and within intimate sexual relationships, especially in populations that are historically patrilineal 7 and hav e large gender inequalities, such as Zambia. Perhaps i t is no surprise then, that individual level interventions aimed at PMTCT adherence focused solely on educati on or behavioral intention have not been successful (see Igumbor et. al., 2006). Family Sys tems Perspective The family is arguably the most influential social unit in regard to health behaviors due to both the proximity to the indiv idual and the longevity of the se relationships (Simons Morton, Hayne, & Noelcke, 2009). Moreover, m arriage within African culture is one of the most significant social relationships (Smith & Mbakwem, 2007) I t is also relatively well established that i n many African societies, women are often not independent decisionmakers regarding the support /permission to use services (Auvinen, Suominen, & Valimaki, 2010). Given this background, it is perplexing that has not been more thoroughly considered in the analysis of PMTCT adherence (For exceptions, see Hatcher et al., 2014; Kiarie et al., 2006; and Mepham et al., 2011). 7

PAGE 33

13 Gender and power are interrelated factors that influe nce health behaviors. Unfortunately, g ender based power in intimate relationships is often unbala nced with women typically having less power (Blanc, 2001) In this study, I conceptualize power as the ability o f one person to control another, occurring at the societal, institutional, and interpersonal levels of the social ecological model (Yoder & Kah n, 1992) Within the context of this study, I use the definition put forth by Blanc ( 2001 ), stating that power in married decision making, to engage in behavior agains (p. 189) The Theory of Gender and Power The Theory of Gender and Power claims that relative power within heterosexual relationships arise s from the global dominance of men over women (Connell, 1 987). The patriarchal ordering of power and privilege is not only embedded in our social and political systems but also permeates into families, negatively affecting all members (Hare Mussen, 1978; Libow, Raskin, & Caust, 1982) There are three relevant in terrelated social structures related to this gender order: the sexual division of labor, the sexual division of power, and the structure of cathexis (Connell, 1987) These societal structures are exhibited both at the institutional level (e.g., family and relationships) and through social mechanisms (e.g., imbalances in control power and disparities in social norms and expectations ; Connell, 1987). Research on HIV/AIDS in sub Saharan Africa using the Theory of Gender and Power has primarily focused on wom Aside from increased biological vulnerability to HIV compared to men, women are also more vulnerable to heterosexual HIV acquisition for numerous social reasons, including sexist

PAGE 34

14 cultural norms psychosocial and legal factors, economic vulnerability, and attitudes regarding feminine sexuality that are conducive to unsafe sexual practices (Travers & Bennett, 1996) All of these factors combined result in women generally having less relative sexua l negotiating power within their intimate relationships compared to men. For example, women may be unable to enforce condom use with male partners or control their partner extramarital sexual practices because of gender power imbalances, increasing the l ikelihood that women will be exposed to and eventually contract HIV (Dunkle et al., 2004; Langen, 2005) Indeed, numerous studies indicate that gender power imbalances are associated with increased HIV prevalence rates for women across sub Saharan Africa (Drain, Smith, Hughes, Halperin, & Holmes, 2004; Gupta, Parkhurst, Ogden, Aggleton, & Mahal, 2008) I argue that t he Theory of Gender and Power is also a useful framework when investigating HIV related behaviors because many of the same social factors that put women at increased risk of sexual HIV acquisition likely also prevent them from adhering to PMTCT protocols. In this study, I specifically focu s on inequities of power and control within married couples which are a result of gendered social structures (Michalski, 2004) and may exert strong influence over related health behaviors. The first structure /domain from the Theory of Gende r and Power the sexual division of labor, is essentially the allocation of particular types of work to men and women, which can become a constraint on additional aspects of life due to disparities in allocation of resources (Connell, 1987). This structure is related to the organization of household work and childcare, the division between paid and unpaid work, the segregation of labor markets

PAGE 35

15 exchanges (Connell, 1987). I nequity in skills and training make the sexual division of labor a powerful discriminatory mechanism creating a social constraint for women (Connell, 1987). in many societies ofte n leaving women dependent on their male partners. The second structure, the sexual division of power, describes how gender power inequities in society and within relationships isolate women from the ability to control their lives (Travers & Bennett, 1996) T his may also include HIV control decisions and behaviors regarding PMTCT adherence. The sexual division of power is related to authority, control, and coercion, both of institutions over individuals and other individuals over individuals (e.g., husbands over wives) Masculinity in most societies is closely tied to authority, giving men increased power over women (Connell, 1987) which of course, permeates into familial life Power within families can be used to win contested d ecisions and dictates who is charged with carrying out decisions, who monitors this process, and who is accountable to whom (Broderick, 1993) The final structure, the structure of cathexis, describes affective attachments and social norms (Wingwood, Camp Kristin, Cooper, & DiClemente, 2009) At the societal level, this structure describes gender norms and is characterized by the emotional and sexual attachments between women and men (Wingwood et al., 2009) The structure of cathexis creates behavioral co nstraints because cultural gender roles assigned to women typically involve subordination to men (Travers & Bennett, 1996) Indeed many Zambian women report being counseled upon reaching puberty to be submissive and obey their future husbands (Human Right s Watch, 2007) Also related to the structure of cathexis are norms

PAGE 36

16 to maintain male dominance tends to be acceptable in society, especially where laws and customs c ombine to uphold the differential power between men and women (Michalski, 2004). Conceptual Model Based on the Theory of Gender and Power within family systems, Figure 3 presents my co nceptual model, taking into consideration the potential relationship be tween PMTCT adherence and gender power imbalances within marriages related to the structure of cathexis and the sexual division s of power and labor. I hypothesize that power imbalances within couples, conceptualized through economic inequity, household dec ision making, and IPV/control, decrease s adherence to protocols across the PMTCT cascade of care and that HIV status disclosure to husbands mediates this relationship In t he following sections I describe the hypothesized pathways in the conceptual model, highlighting w hat is known in the current literature and the gaps that this study will fill. In the conclusion of the dissertation, I provide an updated conceptual mo del detailing the hypotheses that my data ultimately supported. Figure 3 Conceptual Model

PAGE 37

17 Gender, Power, and Vertical HIV Prevention Economic Inequities The sexual division of labor in Zambia is readily apparent in the educatio nal and employment statist ics disaggregated by gender. Zambian Women generally receive less education and have fewer economic opportunities than men. For instance in the most recent Z DHS, only 8 % of Zambian women completed seco ndary education, compared to 13 % of their male counter parts In terms of employment status, only 49 % of Zambian women we re formally employed at the time of the survey compared to 74 % of men (CSO, 2014) There are also very few female politicians or women in power in the country to advocate for As a result of the sexual division of labor at the societal level, socioeconomic disparities exist between men and women at the interpersonal level. Due to income inequity within m arriages, wives are often economically reliant on their husbands, creatin g power imbalances in the relationship This notion has been termed (Kalmuss & Straus, 1990) economic dependence on the husband constrains their ability to make independent decisions on for example, the use of health care (Blanc, 2001) A woman may be at a disadvantage if she needs to ask her partner for money to pay for the transportation costs associated with getting to a health clinic for PMTCT services 8 if she does not have an independent source o f income of her own (Duff, Kipp, Wild, Rubaale, & Okech Ojony, 2010; Tuller et al., 2010). In addition, fear of abandonment has been cited by HIV positive women in sub Saharan Africa as a reason for why they have not disclosed their HIV status to their hus band (Kebaabetswe, 2007; Nyasulu & Nyasulu, 2011) which may be related both to economic 8 Eme rgency Plan for HIV/AIDS Relief (PEPFAR)/USAID and the Global Fund. Thus, there are no health care costs related to these services for HIV positive women/mothers.

PAGE 38

18 reliance and the social significance of being married If a woman is economically reliant on her husband, abandonment is a legitimate concern since she may not posses s the necessary education or skills to obtain her own income. For this reason, a woman may be highly motivated to hide the fact that she is HIV positive from her husband, which could in turn, result in poor adherence to PMTCT. Household Decision Making As a result of the aforementioned economic inequities in addition to sexist cultural practices women often have low participation in household decision making This is consistent across many countries, including Zambia. Kishor and Subaiya (2008) conducted a across 23 developing countries and found that there are no countries in which women make the majority of household decisions alone. In addition, they report that in most count ries, there is (Kishor & Subaiya, 2008) I n Zambia, for example, 24% of women on the last ZDHS reported that their husband alone makes decisions regarding visits to family and friends and 33% reported the husband alone makes decisions regarding major household purchases (CSO, 2014) Most indicators and accounts would consider Zambia a highly patriarchal society. This is apparent at both the structural and interpersonal levels. Men have considerably more c ontrol over social, political, economic, cultural, and familial institutions. This partially stems from the fact that most tribes in Zambia are patrilineal, meaning inheritance passes from the husband to his children and women traditionally left their vill ages to live with the (Barfield, 1997) In addition, t common in Zambia, wh family (usually the father) not the

PAGE 39

19 wife a set amount of money or other commodity of value, such as cows, as compensation for the bride (Barfield, 1997). I n some cases, this of ownership over the woman. I n addition, i t is well established that h usbands often serve health care, such as a woman needing permission to seek HIV testing or treatment (Heckert & Fabic, 2013; Human Rights Watch, 2007; Ngom, Debpuur, Akweongo, Adongo, & Binka, 2003) Indeed, over one fourth of Zambian women report ed that their husband alone makes decisions about (CSO, 2014) G enerally speaking, if a husband does not want the woman to use PMTCT for whatever reason, if she has low relative power in the marriage regarding hou sehold decisions she will be obliged to agree with his demands (Auvinen, Suominen, & Valimaki, 2010) Indeed, during my repeatedly made statements, such as he [the husband] is the head in reference to their use of pediatric HIV services. Intimate Partner Violence and Controlling Behavior IPV against women is one of the most compelling manifestations of unequal power in sexual relationships and the larger phenomenon of gender inequality ( Blanc, 2001 ) For this study, IPV is conceptualized as actual or threatened physical or sexua l violence or psychological/ emotional abuse directed toward s a female spouse that is part of a general strategy of power and control (Johnson, 2008; Population Council, 2008) IPV is thus an extreme expression of male domination and female subordination through the use of both violence and other cont rol tactics (Michalski, 2004). It is important to note that m ale direc ted

PAGE 40

20 IPV most certainly exists; however, because this study focuses on adherence to PMTCT protocols, the concentration of this research is on female directed IPV. In societies where there is unequal access to economic or political resources by gende r, the likelihood of IPV against women increases (Levinson, 1989). Additionally, a woman may be forced to remain in an abusive controlling relationship because of her economic dependence on the husband as discussed above or because her family would have to repay the bride price. Based on traditional Zambian patrilineal custom s in the event of a divorce, the land and the children generally go to the husband and his family which may prevent women from leaving abusive situations In Zambia, both IPV again st women and male partner controlling behaviors are normative. Almost half of all women (47%) reported on the last Z DHS that a husband is justified in beating his wife for at least one specified reason, such as burning the food or arguing with him In addi tion, almost half (47%) of all women also report ed experiencing physical or sexual violence from their spouse and 24% reported experiencing emotional violence Lastly 35% of women reported that their husband displayed three or more controlling behaviors, such as insisting on knowing where she is at all times (CSO, 2014) Although a quantitative relationship between IPV /controlling behavior and PMTCT has not been established, there is cause to believe an association exists IPV is associated with numerous negative health outcomes among women and children including reproductive outcomes such as low birth weight, preterm delivery, and maternal and infant mortality (Boy & Salihu, 2004; Emenike, Lawoko, & Dalal, 2008) IPV also increases the risk of having gy necological problems, including sexually transmitted infections (STIs) and HIV (Campbell, 2002; Harvey, 2007; Garcia Moreno, et al., 2007). In addition, IPV increases

PAGE 41

21 emotional distress, depression, and post traumatic stress disorder (PTSD) among women which are known barriers to medical adherence in general (Ellsberg, Jansen, Heise, Watts, & Garcia Moreno, 2008; Fischbach & Herbert, 1997). Lastly, IPV has been shown to hinder HIV related health behaviors, including HIV testing and ART adherence for wome own health (Maman, Campbell, Sweat, & Gielen, 2000; Maman, Mbwambo, Hogan, Kilonzo, & Sweat, 2001) A qualitative study from 2007 conducted by Human Rights Watch found that Zambian women were physical ly emotionally, and sexually abused by their husbands upon disscussing HIV testing or treatemnt or after disclosing their HIV positive s tatus to the husband. In addition, fear of abandoment and divorce in an environment where women suffer insecure property rights were major impediments to HIV positiv e their own health (Human Rights Watch, 2007). Although this provides compelling evidence that IPV negatively impacts adhernce to ARV drugs for Zambian the study did not specifically e xamine ARV adherence for the purpose of PMTCT. Based on this study, however, there is reason to believe that PMTCT would also be negatively affected by IPV, but this warrants further research to establish any conclusive associations. More recently, two qu alitative studies from South Afri ca reported that IPV or fear of IPV was a barrier for PMTCT (Hatcher et al., 2014; Mepham, Zondi, Mbuyazi, Mkhwanazi, & Newell, 2011) However, these studies did not examine PMTCT protocols besides medication, such as infan t feeding or pediatric HIV testing. In addition, the findings of the s e studies cannot be generalized to the Zambia n

PAGE 42

22 context although they do inform my hypotheses about IPV and other power dynamics and adherence cross the PMTCT cascade. To my knowledge, the only study that has used quantitativ e data to examine PMTCT and IPV in sub Saharan Afric a, Kiarie et al. (2006) who found no association between IPV and uptake of PMTCT in Kenya. However, there are several explanations for why I believe more research is needed in this area beyond the Kiarie et al. (2006) study. First, Kiarie et al. examined physical, financial, and psychological abuse but did not include questions regarding partner control, household decision making or economic dependence key factor s Second, the operational definition of PMTCT in the Kiarie et al. study was limited to medication during childbirth (i.e., sdNVP), which is the protocol most under medical personnel 9 and the least observable to the husband since it occurs within the delivery room at a clinic At the time of the Kiarie et al. research, sdNVP was the only PMTCT protocol in place in Kenya H owever, many change s have occurred since then in terms of PMTCT guideline s and thus, a more recent thorough investigation of the effect of IPV and other gender power dynamics on adherence across the PMTCT cascade is warranted. Disclosure of HIV Status A final factor relevant to the analysis of PMTCT adherence and dynamics within self disclosure to her husband that she is living with HIV. Numerous studies from sub Saharan Africa have established that PMTCT adherence improves when women have disclosed their HIV positive status to the husband /partner (Auvi nen et al., 2010; Doherty, Chopra, Nsibande, & Mngoma, 2009; Jasseron et al., 2011; Peltzer, Jones, Weiss, & Shikwane, 2011; Theuring et al., 2009) Unfortunately, HIV continues to be a highly 9 Compared to ARVs during and after pregnancy, which must be taken 1 2 times a day in the h ome

PAGE 43

23 stigmatized disease (USAID & Zambia Ministry of Health, 2009) According to Goffman (1963), stigma is an attribute that is deeply discrediting considered shameful and can therefore lead to a variety of animosities or discriminations directed at the individual. It is therefore no surprise that HIV positive women may want to hide their status, particularly if HIV positive wom e n often go to great lengths to ensure that their status remain s hidden from the husband due to perceived and often very real repercussions. HIV positive women across sub Saharan Africa ng violence or abandonment, as barrier s to disclosure (Etiebet, Fransman, Forsyth, Coetzee, & Hussey, 2004; Medley et al., 2004; Msellati, 2009) Indeed women have reported IPV after disclosure of an HIV positive status across sub Saharan Africa (Auvinen et al., 2010; Creek et al., 2006; Ezechi et al., 2009; Gaillard et al., 2002; Gielen, O'Campo, Faden, & Eke, 1997; Medley, Garcia Moreno, McGill, & Mama n, 2004; Ntaganira et al., 2008) The Zambian Ministry of Health specifically describes some of the key barriers to disclosure as : fear of abandonment, rejection, violence, upsetting family members and accusations in infidelity. The risks of disclosure include loss of economic support, blame, abandonment, physical and emotional abuse, discrimination ian Ministry of Health, 2010, p.3) Chapter Summary Women often have low relative power within the family in comparison to the husband. It is within this framework of unequal power relations that HIV positive women are expected to take preventative actions aimed at minimizing the risk of vertical HIV transmission. In order to understand how gender power dynamics influence HIV positive I designed a mixed methods study in Lusaka Zambia; the specific methods of th is study are discussed in the following chapter.

PAGE 44

24 CHAPTER III METHODS Study Setting The study setting, Kanyama Health Center, i s located next to a quarry in a busy area of Lusaka surrounded by makeshift shops selling a variety of items from fish to cloth es. I purposefully selected Kanyama Health Center as my recruitment site because of the large diverse population that it serves. Kanyama Health Center is the largest public health center in Lusaka providing primary care and serving a low socioeconomic population of 160,000 individuals who live in the neighborhoods just west of the city Th is densely populated area (see fi gure 4 ) is one of the poorest areas of Lusaka and home to the largest compound (i.e., low income housing). There is no sewer system in Kanyama compound and during the rainy season bouts of cholera and other diseases are especially common. Although HIV rates are not available for Kanyama specifically it is safe to assume they are likely above average low socioecono mic urban setting. Figure 4 Map of Lusaka and Kanyama Compound

PAGE 45

25 I was thoroughly impressed with the health center where I worked in Kanyama. H ealth care workers at Kanyama Health Center demonstrate a consistent commitment to improv ing ve an active community outreach program to raise awareness on issues, such as cholera and HIV/AIDS. In addition, community health workers regularly make home visits to conduct voluntar y rapi d HIV counseling and testing (VCT) Despite human resource constraints and occasional stock outs of supplies, the clinic is highly functional in regards to PMTCT and maternal and child h ealth (MCH) care. All HIV positive postpartum women for the study were recruited through the MCH Department in the clinic, which sees hundreds of women and children every day for routine health care Certain days of the week are dedicated to pregnant women while others days are primarily for children. The MCH department provide s numerous services, namely, ANC, HIV testing and counseling, pediatric immunizations, vitamin supplements, deworming treatment, growth monitoring, and PMTCT for women and their children Research Design I used a concurrent mixed me thods research design (Leech & Onwuegbuzie, 2006) including a verbally administered cross sectional survey (n=320) and semi structured interviews (subsample of n=32) The p articipants for this study we re HIV positive women attending the MCH department of Kanyama Health Center for routine pediatric health care. By using both quantitative and qualitative methods within the same population, I identify statistical relationships between gender power dynamics and PMTCT while drawing on the richer interview data to shed light on contextual explanations. In addition, the semi structured interviews facilitate opportunities to highlight cases that contradict my hypotheses (e.g.

PAGE 46

26 women with low power in their relationship but high PMTCT adherence and women with high p ower but low adherence) to uncover additional barriers and facilitators related to PMTCT adherence. Ethical Considerations T he sensitive nature of my research topic necessitates attention to confidentiality and appropriate handling of situations where women report violence I held a three day training for my research assistants which included an in depth discussion of research ethics. In addition, the study was designed and implemented in accordance with the WHO Ethical and Safety Rec ommendations for Research on Domestic Violence Against Wome n (WHO, 2001) Lastly, women reporting Association (YWCA) in Lusaka for counseling and victim support services. The study was approved by t he Colorado Multiple Institutional Review Board (COMIRB; Protocol 13 1979) in September 2013 and approved by Excellence in Research Ethics ( ERES ) Converge in Lusaka (Protocol 2014 JAN 010) in March 2014. Consent forms and information sheets (see Appendix A ) were translated into the local languages, which the research assistants verbally read aloud to all participants. Consenting p articipants provided written informed consent or a thumbprint (in cases where women were illiterate). Participants were compensat ed $3 (15 Zambian Kwacha) for travel expenses if they completed at least half of the questionnaire, which was meant to be an ince ntive, but not overly coercive and based on local standards. Fo rmative Research and Preparation Prior to implementing this study, I conducted significant formative research, incl uding key informant interviews (See Figure 5 ). I travelled to Lusaka during the summers

PAGE 47

27 of 2012 and 2013 to conduct interviews with both PMTCT clinicians and gender experts. I met with PMTCT physicians and nurses as well as with individuals working for non governmental organizations focused on gender The goal of these trips was to determine the relevance of my research questions, understand PMTCT care and gender p ower dynamics in Lusaka, and plan for the logistics of the study. I was able to navigate these meetings through the help of my in country mentor, a physician within the Department of Obstetrics and Gynecology at the University Teaching Hospital (UTH) in Lu saka. Figure 5 Study Timeline After the formative key informant interviews, I conducted a n expert panel review of my survey questio nnaire in order to establish its face validity in February of 2014 (Singleton & Straits, 2010) This was especially important for questions regarding PMTCT adherence because there was no previously validated instrument measur ing this outcome. The expert

PAGE 48

28 panel review included an obstetric and gynecology physician at UTH, a nursing officer with the Lusaka District Health Office MCH Division three PMTCT nurses/counselors from UTH and Kanyama Health center and the MCH coordinator at Kanyama Health Center. Each individual reviewed a draft of the questionnaire and provided feedback based on their e xpertise. I modified the survey questionnaire slightly based on their recommendations prior to the pilot study. For example, one clinician pointed out that it would be relevant to ask women if they were offered drugs for PMTCT since stock outs of medicatio n are a continual challenge in the country At this time, I also had the questionnaire translated into both Nyanja and Bemba to have standardized instruments in the most widely spoken local languages. U nfortunately I am not fluent in the local languages and thus, recruited and trained four local research assistants to administer the survey and semi structured interviews Kanayma Health Center would only approve the use of research assistants who were current employees at the clinic due to confidentialit y concerns. I worked closely with the MCH Coordinator at Kanyama Health Center to find health care workers who were a good fit for this research We ultimately selected three other nurses/counselors as research assistants for the data collection I n additi on the MCH Coordinator herself volunteered to be a research assistant All research assistants had previous experience with public health re search projects The other health care workers at the clinic were also hired to help with recruitment and were paid a small amount based on the number of HIV positive women they referred to the study during pediatric immunizations. I conducted several training s with the four selected research assistants on data collection and research ethics I created the training modules based on information from the WHO, COMIRB, and other materials from colleagues in Lusaka. In addition to my

PAGE 49

29 PowerPoint presentations, I had the re search assistants practice going through the questionnaire wit h each other as well as discuss any ques tions or concerns they had. I was very pleased with the level of competency of my research team They were highly motivated to participate in the research and also often served as my cultural translators regarding events that I observed at the clinic (or i n Lusaka in general) Following the training of research assistants, I conducted a pilot study (n=35) at Kanyama Health Center to test recruitment procedures, survey administration and participant comprehension of the questionnaire I made minor modific ations to the survey instrument and recruitment procedures after the pilot survey. For example, we originally were using research. However, during the pilot study, we realized that this was causing undo attention among women at the clinic N on eligible women (i.e., HIV negative or with an infant not of appropriate age) were asking why they were not given an invitation card, drawing too much attention to the study. Th us, we decided to change procedures and have nurses verbally invite eligible mothers to participate in the study, discarding the use of invitation cards. For data analysis of the pilot study, I conducted descriptive analysis of the data in Stata 12 to exa mine the variance in the distribution of my variables of interest. The questions measuring PMTCT adherence, my outcome s of interest, showed variation with a range from 0% to 100% in self reported medication adherence during and after pregnancy. The main in dependent variable s measuring gender power dynamics also displayed adequate variability. For example, women reported zero to six controlling behaviors from their husband and zero to five forms of physical/sexual violence from their husband. Household decis ion making

PAGE 50

30 also deviated with some women participating in no household decisions and others participating in the majority of decisions. Sampling Technique The population of interest for this study is HIV positive women with a young infant living in Lusak a who attend Kanyama Health Center for health care Based on a population rea, if 50% of the population are female and 22% are HIV positive ( USAID & ZMOH, 2009) the ove rall samplin g frame includes roughly 18,000 HIV positive women of reproductive age. 10 Recruitment for the study took place during routine pediatric immunization at Kanyama Health Center using a nonprobability purposive sampling design (Onwuegbuzie & Collins, 2007). I ch ose to recruit women during routine pediatric immunizations in order to capture HIV positive women who were both adherent and non adherent to PMTCT since immunization compliance is very high in Lusaka: only 1.7% of children have no immunizations by 12 mont hs of age (CSO, 2014) In addition, o ther researchers have affirmed that s ampling at pediatric immunization clinics offers a robust method for studies examining PMTCT (Horwood et al., 2012) Women were eligible for the study if they were married (or living with a man as if married) with a known HIV positive status over 18 years of age (legal age to provide consent in Zamb ia), and their youngest infant was between three and nine months of age. Infant age was selected in order to capture all of the ess ential PMTCT protocols, match the immunization schedule in Zambia, and limit recall bias. As a safety measure, we excluded any women who were at the clinic with their husbands to avoid potential confrontations 10 The sampling frame was less because we did not recruit anyone under the age of 18, not married, women who had an infant under three months or over 9 months of age, and women who did not use Kanyama Health Center for routine child health care.

PAGE 51

31 n the research. However, it is very rare for husbands to attend well child healthcare, and only one woman was excluded for this reason. From March to August of 2014, nurses recruited all eligible women they encountered (i.e., HIV positive with a child of the appropriate age) 11 during pediatric imm unizations N urses determined eligibility birthdate, height/weight, immunizations, medications, and PMTCT (see Figure 6 ). If a woman i s HIV positive, her card indicates If women did not have an Under Five Card, nurses used other available medical records to dete rmine Figure 6 Children's Under Five Card 11 Add itional screening questions were asked to women by the research assistants immediately following informed consent.

PAGE 52

32 Sample Description My a priori sample size calculation indicated a need for data from at least 274 participants to conduct a two tailed z test Poisson regression t o detect a rate ratio of at least 1.2 with a power of 0.80 and alpha of 0.05 (Faul, Erdfelder, Lang, & Buchner, 2007) Of the 517 women who were approached to participate 136 (24%) were ineligible because they were not married (see figure 7 ). Among the eligible women, 326 (86%) provided informed consent and 320 women ultimately completed the survey questionnaire Figure 7 Sample Description The main reasons women cited for declining to participate in the research were lack of time and unfounded fears regarding biomarkers being taken or future required participation related to the monetary reimbursement participants would receive. My research assistants reported that some women were skeptical

PAGE 53

33 Survey Measures PMTCT Adherence I measured the dependent variable s of interest, PMTCT Adherence, by creating a set of questions based on th e current PMTCT and ART adherence literature. My main references for these measures include s the 2010 WHO PMTCT Recommendations for Zambia (see Table 1 in Chapter 2 for specifics; Zambia Ministry of Health, 2010), the 2011 Malawi DHS PMTCT questions (Natio nal Statistical Office & ICF Macro, 2011) and a published review article by Simoni et al. (2006) detailing self report ed measures for ART with specific recommendations. I created a series of PMTCT questions on the survey (see Appendix B) capturing data r postpartum; infant feeding practices; and pediatric HIV testing at 6 weeks and 6 months. Despite the limitations of recall and social desirability bias, self reporting by HIV positive women is one of the most common methodologies used to collect data on PMTCT (see Mepham et al., 2011; Futterman et al., 2010; Nassali et al., 2009). In addition, studies have shown that self reported adherence to ARVs and plasma drug levels are significantly as sociated and a reliable measure (Fabbiani et al., 2015; Fletcher et al., 2005; Murri et al., 2000). Finally, the research assistants also responses. For example, if a woman reported 100% drug adherence but had numerous missed appointments indicated on her card, the research assistant would probe the woman to honestly report her adherence or explain where she was getting medication. For pediatric HIV testing, any tests the child had were written on the Under Fiv e C ard and also were indicated in a space on the survey by the research assistants during the interview

PAGE 54

34 For ARV rule out health care system factors T hus the first question women were asked was if they were offered medication during pregnancy. If women reported being offered medication during pregnancy, they were then asked if they took the medication. Many studies on AR V adherence stop at this point ; howe ver, I felt it necessary to try and capture more variability in adherence levels rather than medication during pregnancy were asked to indicate the percent of medication they took from 0 to 100% using a visual analog scale (VAS) that was verbally explained by the research assistants (see Figure 8; Simoni et al., 2006) 12 Many women find it difficult to take their HIV medicines exactly as they are told during pregnancy. What is your best guess of how muc h of the HIV medication you took while you were pregnant? We would be surprised if this were 100% for most women. Figure 8 Example of Drug Adherence Question At the time of the study, the Zambian Ministry of H ealth recommended two slightly ( see Table 1; ZMOH, 2010) Both groups of women were required to take medication throughout pregnancy, but only women on the short course prophylaxis regi men were given medication during childbirth. Thus, for the childbirth period, women were again asked if they were mother was on 12 conceptualize proportions by rephras

PAGE 55

35 was a skip pattern in the survey to a specific section for women on the treatment regimen (explained below). If women were not on treatment, they were then asked if they took the medication during childbirth (this is a single dos e medication so no continu ous measure of adherence was used ). For the postpartum time period, women on the short course prophylaxis regimen were asked if they were offered medication, if they took the medication, and how much of the medication they took from 0 to 100% during the week after giving birth (women on this regimen only take medication for seven days postpartum). W omen on the treatment regimen were also asked if they were given medication, if they took the medication, and what their adherence was from 0 to 100% since giving birth (women on the treatment regimen do n ot stop taking the medication after one week, but continue taking ARVs for life ). Lastly, both women on the treatment and the short course prophylaxis regimens were required to give prophylaxis to the infant postpartum, which we also captured on the surve y asking women if they were offered infant prophylaxis and if they gave the medication to the child. However, women on the treatment regimen were asked about their adherence during the six week period after giving birth and women on the short course pro phylaxis regimen were asked about their adherence to the infant medication since giving birth (women on short course prophylaxis continue giving the infant prophylaxis until one week after breastfeeding cessation). In addition to medication adherence, saf e infant feeding practices are also a critical component of PMTCT. For the measurement of infant feeding, we asked women if they had ever breastf ed the infant, if they were currently breastfeed ing the infant, and if they had ever given any other foods from a list of 11 possible items, such as cows milk, infant formula,

PAGE 56

36 water, fruit juice, porridge, etc (see Appendix B for all food items) Women who reported giving other food besides breast milk were asked at what age they first introduced that particular fo od to the child Using these questions, I was able to capture which women were following the PMTCT recommendation of exclusively breastfeeding the infant to six months of life. All questions regarding infant feeding came directly from the ZDHS (CSO, 2009) I measured the final PMTCT protocol, pediatric HIV testing, through a combination reporting and by using the Under Five Card. I developed these questions based on the WHO PMTCT guidelin es for Zambia, which recommend HIV positive women h ave thei r child ren tested at six weeks, six months, 12 months, and 18 months. Since the oldest infants in this study were 9 months of age, I was only able to capture the first two tests. Women were first asked if they had ever taken the child for pediatric HIV testing. on the survey whether the child was in fact tested at six weeks and six months based on the Unfortunately, I was not able to capture relia ble data regarding the test results (i.e., the because the turn around time for pediatric HIV tests was several months and the vast majority of mothers most recent results. Gender Power Dynamics I measured gender power dynamics within couples using a version of the Revised Conflict Tactics Scale (CTS2). The CTS2 is one of the most widely used gender power measurement tools worldwide and has strong reported psychometric properties (Straus, Hamby, BoneyMcCoy, & Sugarman, 1996) The version of the CTS2 used in this study came directly from the Z DHS Domestic Violence Module, which has been used in Zambia since 2007 (CSO, 2009 CSO 2014 )

PAGE 57

37 Table 2 Constructs and Questions on Gender Power Dynamics Construct Specific Questions Sexual Division of Labor Economic Inequity A. Would you say that the money you earn is more than what your husband earns, about the same, or less than what he earns? Sexual Division of Power 1. Household Decision Making A. In your current relationship, who usually makes decisions about health care for yourself? B. In your current relationship, who makes decisions about major household purchases? For example, a sofa or TV. C. In your current relationship, who makes decisions about daily household purchases? For example, food. D. In your current relationship, who has the final say on how to spend money? 2. Intimate Partner Violence Does your husband ever: 2a. Physical Violence A. Slap you? B. Twist your arm or pull your hair? C. Push you, shake you, or throw something at you? D. Punch you with his fist or something that could hurt you? E. Kick you, drag you or beat you up? F. Try to choke you or burn you on purpose? G. Threaten to attack you with a knife, gun, or other weapon? 2b. Sexual Violence A. Physically force you to have sexual intercourse with him even when you do not want to? B. Force you to perform any sexual acts that you do not want to? 2c. Emotional Violence A. Say or do something to humiliate you in front of others? B. Threaten to hurt you or someone close to you? C. Insult or make you feel bad about yourself? 3. Controlling Behaviors Please tell me if these apply to your relationship with your husband: A. He is always jealous or angry if you talk to other men? B. He frequently accuses you of being unfaithful? C. He does not permit you to meet your female friends? D. He tries to limit your contact with family? E. He insists on knowing where you are at all times? F. He does not trust you with money? Structure of Cathexis Perceptions of Wife Beating In your opinion, is a husband justified in hitting or beating his wife in the following situations: A. If she goes out without telling him? B. If she neglects the children? C. If she argues with him? D. Is she refuses to have sex with him? If she burns the food?

PAGE 58

38 The DHS Domestic Violence Module was first developed and standardized in 2000 and has been used in more than 80 surveys, including throughout sub Saharan Africa (MacQuarrie, Winter, & Kishor, 2014) I chose this instrument because it appropriately captures all of the domains from the Theory of Gender and Power, my overarching theoretical framework (see Table 2). Additionally, the formatting of questions on this instrument has several advantages over a single question approach (e.g., have you ever experien ced IPV? ), particularly in the context of cross cultural research (Kishor, 2005) By asking women about separate specific acts of violence, the violence measure is not affected by different understandings between women of what constitutes violence. For ex ample, all women are likely to agree on what constitutes being s lapped, but not all women may agree on what constitutes physical violence. Furthermore, the instrument has been praised for asking women about violence from many different angles, which encour ages disclosure because it gives women more time to think about their experiences and multiple opportunities to disclose their experiences of violence (Kishor, 2005) Quantitative Analytic Strategy Data Entry and Cleaning I created a data entry tool in CSPro 13 that was designed to capture and record responses to all of the questions on the survey. I double entered each survey and addressed any inconsistencies in the two data entry files by going back and checking the original survey for the correct value. Data cleaning consisted of running descriptive statistics and checking skip patterns to find any abnormalities such as out of range variables or response s within 13 The Cens us and Survey Processing System: a public domain software package for entering, editing, tabulating, and disseminating census and survey data.

PAGE 59

39 skip patterns (see below for description of missing data and the imputation method). For exa mple, if a question was asked when it was supposed to be skipped, I re coded the data (in a new dataset) as missing. In addition, throughout data analysis, I was continually cleaning the data and checking the original surveys when any irregularities arose. Dependent Variables self reported ARV adherence levels into yes (adherent) or no (non adherent) variable s for 1) ARV medication during pregnancy; 2) sdNVP during labor/delive ry (if applicable); 3) ARV medication postpartum; and 4) giving the infant prophylaxis. Women who reported being offered the respective medication, but did not take it (or give it to the infant) were coded as non adherent. Women who were not offered medication were excluded from the analysis. I defined adherenc e during and after pregnancy as the woman reporting she took (or gave to the infant) at least 80% of the prescribed medication doses. 14 I chose the cut off of 80% based on recent literature indicating that this is the adherence level needed to suppress HIV using the current AR V regimens that the women in my study were taking (Gordon et al., 2015; Kobin & Sheth, 2011) 15 report s of the food they gave to the infant and at what age: 1 ) exclusive breastfeeding to six months; 2) exclusive replacement formula feeding; and 3) mixed infant feeding. Women whose infant was less than six months of age were characterized as mixed feeding if the mother reported giving any food in addition to bre ast milk to the child. For infants over six months, if women 14 Except for medication during childbirth, which is a single dose. I coded women who reported taking the tablet during childbirth as adherent for this protocol. 15 I also explored the cut off of 95% and found very similar associations, indicating that my fi ndings were not particularly sensitive to the cut off point used.

PAGE 60

40 reported giving any food to the infant prior to six mo nths of age i n addition to breast milk, the mother was also classified as mixed feeding. Lastly, in order to analyze pediatric HIV testing, I created two new variables based on reporting and what was indicated on the Under Five Card: 1) if the child had any HIV testing (for all infants); and 2) if t he child had both HIV tests at six week s and six months (for infan ts ages six to nine months, n=158). If there were any reporting and the Under Five Card, I opted to use data from the Under Five Card. One possible response to the self reporting question regarding pediatric HIV testing w as that the mother attempted to bring the child for HIV testing but did not receive a test due to lack of supplies at the clinic. The women who selected this response category were not included in the analysis (n=24) Again, this was meant to exclude parti cipants who encountered health system barriers. Independent Variables The first gender power dynamic of interest, on the husband (i.e the sexual division of labor), was measured via one questio n asking women if their earnings were 1) more than the husband, 2) about the same as the husband, or 3) less than the husband. I collapsed the first and third categories (i.e., women who have the same or greater earnings) compared to women whose husband has greater earnings The second power dynamic of interest i participation in household decision making which is part of the sexual division of power In order to capture this dynamic, I created four new decision care 2) daily purchases 3) large household purchases, and 4) the final say over money. In the original questions, women were able to choose three responses regarding who primarily

PAGE 61

41 makes the respective decision s : 1) the woman alone, 2) the husband alone, or 3) they mak e the decision together. Similar to my economic dependence variable, I collapsed the first and third categories for all four variables household decision making (i.e., she makes the decision alone or with the husband) co mpared to when the husband alone makes the decision It was additionally appropriate to collapse the two categories because of the rarity of women reporting they alone make decisions. For ex ample, only 8% of women report they alone make the decision regard ing major household purchases and only 10% of women report having the fina l say over how money is spent. Finally, I summed the four household decision variables to create a final count of the number o f decisions a woman participates in from 0 to 4, which i s a standard way of examining household decision making (Upadhyay & Karasek, 2012 ; CSO, 2014 ) I analyzed the third power dynamic, IPV, by creating a series of new variables measuring different aspects of IPV. First, a lthough the data collection instrumen t I used to measure IPV (i.e., the DHS Domestic Violence Module) generally researchers dichotomize the variables into those who experience IPV (or certain forms of IPV) and those who have not experience d IPV (Alio, Nana, & Salihu, 2009; Kishor & Johnson, 2004; Lawoko, Dalal, Jiayou, & Jansson, 2007). I followed in this trend, using the module as a way to capture different forms of violence by asking women various questions related to IPV but not necissarily analyzing the qu estions as scale s I created several new binary variables to capture different aspects of IPV First, I created dummy variables for emotional, physical, and sexual violence, as well as for experiencing any IPV. In addition, to better understand the nature of IPV, I used some unique approaches with the module that went beyond simply exa mining the presence of violence

PAGE 62

42 First, I created two ordinal variables to capture the mean number of violent events a woman experienced and the frequency of violence in the past year. I calculated the mean number of violent events by summing the total number of specific emotional, physical, or sexual events a woman reported out of a total of 12 possible events (3 possible emotionally violent events, 7 possible physically viol ent events, and 2 possible sexually violent events). I also gave women a frequency of IPV score by summing how frequently each violent event occurred in the past year (often=3, sometimes=2, rarely ever=1), yielding a scale from 0 (no violent events in the past year) to 36 (experiencing every violent event often in the past year). Lastly, I created a dummy variable for any injuries as a result of violence to capture the severity of physical/sexual IPV. If a woman reported experiencing any type of injury fro m a list of four possible injuries, such as cuts, bruises, or aches, I classified her as experiencing injury from IPV. In addition to creating ordinal variables, I additionally created two sets of dummy variables to measure whether the number of violent events or the frequency of violence in the past year has a dose response relationship with the various PMTCT adherence measures. Five dummy variables were created for each measure. F irst, f or the number of IPV events, I used the following cut off points: 1 ) no IPV events; 2) 1 to 3 IPV events; 3) 4 to 6 IPV events; and 4) 7 to 9 (highest reported number) IPV events. S econd for the frequency of IPV in the past year, I used the following cut off points: 1) no IPV in the past year; 2) a frequency of IPV score from 1 to 5; 3) a frequency of IPV score from 6 to 10; and 4) a frequency of IPV score greater than 10. Despite not analyzing the these variables as scales per se I did run internal consitency checks of the IPV scales to establish how well the questio ns were inter related Generally, an

PAGE 63

43 alpha over 0.70 is considered an acceptable internal consistency (Tavakol & Dennick, 2011). In which case, only the physical violence and sexual violence scales had adequate scale reliablity coefficients in my study (al pha=0.73 for both). The scales for controlling behavior (alpha=0.66) and acceptance of wife beating (0.67) were very close to the 0.70 alpha threshold, but emotional violence (alpha=0.36) displayed very poor internal consistency. In addition to IPV, I am also interested in the effect of a behavior, another component of the sexual division of power For controlling behaviors, I created a new variable to capture if women if women experienced three or more partner controlling behaviors. First I summed the number of male controlling behaviors (from 0 to 6) and then dichotomized the variable into 1) three or more controlling behaviors or 0) 2 or less controlling behaviors. This is a standard way of examining partner controlling behaviors using the DHS data (CSO, 2014) This method was also appropriate due to the extremely small number (n=23) of women who reported no controlling behaviors. acceptance of wife b eating. The questionnaire asked women whether a husband is justified in beating his wife for five specified reasons (e.g., neglecting the children, burning the food, etc. ; see Appendix B d a new count variable summing the total number of reasons women agreed a husband is justified in beating his wife (from 0 5). Covariates Covariates include the in years (continuous variable) in months (continuous) ; the mothe education attainment (ordinal) ; parity (continuous) ; knowledge of PMTCT (count, see below) ; relative wealth (standardized index,

PAGE 64

44 see below ); PMTCT regimen (binary) ; and disclosure of HIV status to the husband (binary) All covariate questions on the survey came directly from the ZDHS, with the exception of PMTCT regimen, which I determined during the data analysis stage based on the medication a woman reported taking on the questionnaire: either lifelong ART or short course prophylaxi s (CSO, 2009) K nowledge of PMTCT was measured by four questions on the survey (CSO, 2009) that were analyzed as a count variable based on correct answers, for there any special drugs that a doctor or nurse can give to a woman infected with HIV to In order to measure socioeconomic status I generated a relative wealth index from a list of 21 possible hous ehold assets on the s urvey An alternative method would have been to simply count the number of reported assets; however, this c an lead to biased results because two individuals with very different economic resources can be assigned the same wealth score. Instead, I opted to u se the method that Filmer and Pritchett (2001) recommend for use with the DHS household assets measure principal component analysis (PCA ) Th e PCA procedure first standard izes the indicator variables, then calculates the factor coefficient scores, and la stly, for each household, multiplies the indicator values by the loadings These numbers are (Filmer & Pritchett, 2001) The estimation of relative wealth using PCA is based on the first principal component wh ich yields a wealth index that assigns a larger weight to assets that vary the most across households. Thus, luxury household assets are weighted more heavily and assets that everyone owns are weighted zero The scale was standardized prior to inclusion in the regression models.

PAGE 65

45 Missing Data and Multiple Imputation Overall, I analyzed data from 320 surveys. Surveys with more than 50% missing data were not included in the analyses (n=4). M issing data ( 2.3 %) were imputed using multivariate chained equations in Stata 12 (Raghunathan, Lepkowski, Hoewyk, & Solenberger, 2001) 16 D ata converged indicating that the multivariate chained model w as a good fit for the data set (StataCorp, 2009) See Table 3 for the extent of missi n g data a mong my key variables. In order to determine possible auxiliary variables (those that explain missing values ), I generated indicator variables coded as 1 if the value was missing and 0 if the value was present. Next, I ran a series of logistic regre ssion models with t he possible auxiliary variables (Acock, 2010) V ariable s that were significa ntly associated with missing data were included in the imputation m odel as auxiliary variables in addition to standard controls such as age, education, and wealth I included the dependent variables in the imputation model to help predict missing values among the independent variables but did not use them in the final statistical models because skip patterns were used in the survey and the missing values w ere genuinely not applica ble to the outcome questions. For example, if a woman was not offered medication dur ing pregnancy, she should not have answered the question on adherence to medication during pregnancy. A total of 20 auxiliary variables were included in the final imputatio n models. Through chained imputation, 20 datasets were created and pooled for missing values among the variables I used the impute d dataset in all of my subsequent statistical models. The statistical analyse s I used were specific to each outcome, which is discussed in detail in 16 Imputed data were analyzed using maximum likelihood estimations, which is specifically discussed in the methods sections of the proceeding chapters

PAGE 66

46 the following chapters. Table 4 displays the participant characteristics of the survey sample after multiple imputation Table 3 Missing Data for Key Variables (n=320) Variable No. Missing Percent Missing Dependent Variables Pregnancy adherence* 49 15% Childbirth adherence* 10 3% Postpartum adherence* 35 11% Infant prophylaxis adherence* 16 5% Infant feeding 1 0.3% HIV test at 6 weeks 4 1% HIV test at 6 months* 16 5% Independent Variables Acceptance wife beating scale 11 3% Emotional violence scale 1 0.3% Physical violence scale 1 0.3% Sexual violence scale 1 0.3% Controlling behavior scale 0 0% Health care decision 3 0.9% Minor purchases decision 11 3% Large purchases decision 4 1% Final say over money decision 3 0.9% Economic dependence 10 3% Missing due to skip patterns (e.g., women were not offered that medication or child was not over 6 months of age)

PAGE 67

47 Table 4 Participant Characteristics Demographic Characteristics: Total (n=320) Median (IQR) or % Age (years) 28.0 (24, 34) Age of infant (months) 6.0 (3, 9) Parity 3.0 (2, 4) Completed primary education 72.0% Completed secondary education 13.7% Job past 12 months 63.4% Household had electricity 1 74.9% PMTCT knowledge 95.6% Relationship Characteristics Length of relationship (years) 5.0 (2, 9) Husband tested for HIV 80.3% Discordant couple (n=254) 2 35.1% Disclosed to husband HIV positive status 91.6% Sexual Division of Labor Greater or about the same earnings as the husband 23.7% Sexual Division of Power Any IPV 60.9% Any controlling behaviors 92.80% Ever hit partner 8.5% Daily household items 73.8% Major household items 55.6% Health care for the woman 67.2% Final Say over money 37.8% Structure of Cathexis: Believe a husband is justified in beating his wife for at least one reason 64.9% 1 Example of household asset possession. There were 21 possible assets that a participant could own, which were then converted into a standardized wealth index for the multivariate analyses. 2

PAGE 68

48 Semi Structured Interviews Concurrently with the survey, my research assistants conducted semi structured interviews with a subsample of 32 women in order to explore confirmatory and contradictory cases regarding gender power dynamics within couples and PMTCT adherence I chose to use a subsample of participants for the qualitative interviews because i n mixed methods research it is appropriate to use the same individuals for both forms of data collection so the data may be more easily interpreted (Creswell & Plano Clar k, 2011) During the survey questionnaire, m y research assistants were asked to purposively select participants for the semi structured interviews based on four previously established domains of interest (see Table 5 ). Table 5 Sampling Domains for Semi structured Interviews 1. Women with high power in their sexual relationship and high PMTCT adherence 3. Women with low power in their sexual relationship and high PMTCT adherence 2. Women with high power in their sexual relationship and low PMTCT adherence 4. Women with low power in their sexual relationship and low PMTCT adherence There were no cut off points based on survey responses for the research assistants to use at the time of qualitative sampling; 17 however, I held a formal training session prior to data collection to explain the four domains of interest. For example, I explained that power was operationalized as the combination of partner violence and control, as well as household decisio n adherence across the PMTCT cascade. Although the criteria were relatively subjective and 17 Cut off points were established post hoc and described in further detail below

PAGE 69

49 based on how the research assistant interpreted the four categories, this techni que aimed to include individuals with different informative experiences (Shenton, 2004) Domains 1 and 4 (i.e., women with high power/high adherence and women with low power/low adheren ce) are the groups that provide support for my hypotheses r egarding g ender power dynamics. These two groups help explain why gender power dynamics, such as IPV decrease power/low adherence and with low power/hig h adherence), however, uncover additional barriers (Domain 2) and facilitators (Domain 3) re lated to PMTCT adherence unrelated to gender power dynamics within couples The qualitative i nterviews were conducted using a semi structured interview guide (see Appendix C ) that I developed based on my formative research in Lusaka and with the guidance of my local research team. On the interview guide, research assistants indicated say it was difficult for I also included several questions specific to the to ask follow up questions or continue interesting discussions that helped inform the research questions The interviews took place in the same location on the same day immediate ly following the survey with the selected subsample of participants. Prior to beginning the interview, the research assistants went through a second inform ed consent process with each woman, asking her to stay fo r additional time and for consent to voice r ecord the interview.

PAGE 70

50 The interviews were also co nducted in the local languages. The interviews continued until the research team felt that we had a diverse sampling of women who fell into one of the four domains. 18 I later transcribed the interviews into Mi crosoft Word while one of my research assistant s listened to the audio recordings and verbally translated them into English. Interviewing women on the same day immediately following the survey had several notable strengths and limitations. First, since w omen were already recruited into the study and present, there was no hassle of tying to do any follow up. Indeed, this was my rationale for the decision to conduct the interviews on the same day as the survey However, by using this method, women were aske d to stay for an additional 45 minutes to one hour on top of the 30 45 minute s they spent taking the survey. It is possible that selection bias may have occurred among women who had controlling partners and needed to account for their whereabouts without d isclosing their status and/or use of PMTCT to the husband. These women may have opted not to participate in order to get home sooner and not raise suspicions from the husband. Unfortunately, we did not collect data on the response rates of women for the qu alitative interviews or why women declined the interview (if any did). In hindsight, this would have been useful information. Qualitative Analysis For the semi structured interviews, I used a content analysis method (Green & Thorogood, 2014) in Atlas.ti within a predetermined thematic framework This framework was in the form of the four original power/adherence dimensions discussed above. In addition to this deductive coding, I also incorporated inductive approach es within the four previously es tablished dimensions to determine significant themes This technique has been 18 We also unfortunately had a time constraint with how long I was planning to stay in Zambia, which limited the number of women we could interview.

PAGE 71

51 referred to as a hybrid approach and has the advantage of allowing both data driven and theory driven analysis (Fereday & Muir Cochrane, 2006) For the first step of my qualita tive analysis I placed each interview into one of the four previously established domains questionnaire and determining appropriate cut off points. I labeled women who reported less than 80% medication adherenc in addition to experiencing IPV and participating in less than two household decisions. This coding sche me provided the following distributions with the four domains : high adherence/high power (n=12) low adherence/low power (n=12) high adherence/low power (n=4) and low adherence/high power (n=4). After indexing all if the interviews into one of the four power/adherence domains, I then used a more iterative approach. For the first step, familiarization, I carefully read th rough each transcript (Green & Thorogood, 2014) of interviews in each particular domain While reading through the transcripts, I a ssigned tentative open codes to chunks of data that were relevant to my understanding of adh erence and gender power dynamics I also simultaneously wrote memos for myself regarding my emerging hypotheses about the relationships between codes After I established a comprehensive code list, I re read through the transcripts assigning additional codes to quotes (Daly, Kellehear, & Gliksman, 1997) I continued this processes of re reading the trasnscripts in Atlas.ti and assigning codes to quotes as well as creating new codes until I acheleived theoretical saturation (LeCompte & Schensul, 2010) for each of the four power/adherence domains

PAGE 72

52 Once all of the data had been coded in Atlas.ti, I arranged the relaevant quotes via a chart in Microsoft W o rd. I placed each quote with it s affliated code into one of the four ad herence/ power survey responses). Finally, I created additional codes (i.e., axial codes) to represent the speci fic themes within that particular domain of interest (see Table 6). I used these themes to organize my presentation of the data within each domain (see Chapter Eight) and present illustrative quotes. Table 6 Example of Qualitative Coding Scheme Quote Domain Open Code Axial Code me about giving the child medication, not even one day. I am the only one that Low power/ low adherence Relationship dynamic Lack of partner support Limitations The study has limitations to note. First, it is c ross sectiona l and cannot establish causality or the timing of events. There is no way to determine for example, whe ther non adherence or violence occurred first. Second, the results are primarily based on self reporting, which is vulnerable to recall and social desirability bias es. Third, questions non representative, limiting the generali zability of findings outside of low socioeconomic women in urban Zambia Lastly, due to the relatively small sample size, there may have been inadequate statistical power to detect all significant quantitative relationships.

PAGE 73

53 Chapter Summary This chapter details the specific methods used in this study, including a survey and semi structured interviews with over 300 HIV positive postpartum women attending a large public heal th center in Lusaka, Zambia. In the following chapters, I report on the results of my quantitative and qualitative analys e s, starting with an in depth description of gender power dynamics within couples and moving int o significant statistical relationships between power dynamics and adherence to PMTCT. Finally, I discuss important findings from the semi structured interviews which help inform the statistical associations

PAGE 74

54 CHAPTER IV THE SEXUAL DIVISION OF POWER Globally, large gender power inequities exist both within society and with in intimate relationships with women generally having less power The Theory of Gender and Power, which I discussed in depth in Chapter Two, views inequities between men and women thr ough three key domains: 1) The structure of c athexis (i .e., unequal gender norms), 2) the sexual division of power, and 3) the sexual division of l abor (Connell, 1987) Although these domains often overlap and perpetuate one an other, I specifically focu s on the sexual d i vision of p ower because is the main explanatory mechanism of interest in this study regarding adherence to PMTCT. Moreover, IPV/control is one of the most compelling manifestations of power differenti als with couples, which is why I particularly focus on interpersonal power constructs in this chapter. Review of Methods I measured gender power dynamics within intimate relationships using the ZDHS Domestic Violence Module (CSO, 2009; see Chapter Three for Details) The statistical methods used in this chapter include descriptive statistics simple and multivariate logisitic regression and linear regression in Stata 12 partner controlling behaviors as the dependen t variables of interest. In timate Partner Violence and Controlling Behaviors A large proportion of particip ants in this study experienced IPV (61%) and partner controlling behavior (93%). Only 5% of women report no experiences of violent or controlling events from their current husbands (see Figure 9). 19 Not only is IPV and partner controlling behavior shockingly prevalent, but there is also a large overlap between IPV and 19 Characteristics of these positive deviants are discussed at th e end of this chapter.

PAGE 75

55 controlling behavior among the HIV positive women in this study : 59% of women experience d both IPV and partner controlling behavior Notably, almost all of the IPV women experience d occurred with the concomitance of controlling behaviors: only 2% of women report IPV without partner controlling behaviors. Figure 9 Women's Experiences with IPV & Partner Controlling Behaviors (n=320) V iolence that occurs in combination with controlling behavior deserves particular attention because of the nature and repercussions of these behaviors. IPV that includes partne r control has been termed coercive controlling violence, as opposed to situational couple violence, which results from conflicts between partners that occasionally escalate into violence (Kelly & Johnson, 2008) Notably, coercive controlling violence t ends t o occur more frequently, results in more physical injuries, and result s in greater psychological distress compared to situational couple violence (Johnson, 2006; Pico Alfonso et al., 2006) Importantly, t h e levels of IPV and pa rtner control in my sample are much greater than what is reported in the most recent ZDHS (see Appendix D for a detailed comparison of my 2% 34% 59% 5% IPV only Controlling behaviors only IPV and controlling behaviors No IPV or controlling behaviors

PAGE 76

56 data and the 2013 2014 ZDHS). The 2014 ZDHS reports that 38% of women in Lusaka and 47% of Zambia n women experience IP V (CSO, 2014) In addition, 65% of women in Lusaka and 74% of Zambian women on the ZDHS rep ort that their partner displays at least one type of controlling behavior. There are two factors that likely account for the greater prevalence of IPV /control found in my sample compared to the ZDHS : 1) socioeconomic status and 2) HIV infection. First, the women in my sample a re from one of the most impoverished areas of Lusaka and low socioeconomic status is an established risk factor for IPV (Heise & Garcia Moren o, 2002; WHO/LSHTM, 2010) The WHO Multi and Domestic Violence, which surveyed women in 10 countries from various regions, including sub Saharan Africa, report that higher socioeconomic status i s generally protective of phys ical and sexual partner violence (Abramsky et al., 2011) Indeed, in the most recent ZDHS, women in the lower wealth quintiles report greater levels of IPV and partner control than women in the wealth ier quintiles (CSO, 2014) Secondly the women in my sam ple a re all HIV positive which is a n additional known risk factor for IPV ( Campbell et al., 2008) Across sub Saharan Africa, women report IPV following disclosure of an HIV positive status to male partners (Kilewo et al., 2001) Moreover, a recent longitudinal study in South Africa also establishe d that women who experience IPV a re at greater risk of incident HIV infection (Jewkes, Dunkle, Nduna, & Shai, 2010) Another cross sectional study report s that in addition to IPV, male control i s associated with an HIV positive diagnosis among women attending ANC in South Africa (Dunkle et al., 2004) Given these associations, it is not surprising that the low

PAGE 77

57 socioeconomic HIV pos itive women in my study report greater levels of IPV /cont rol than women in the general Zambian population. Experiences With Specfic Forms of Violence and Control Differentiating between the causes, consequences, and nature of IPV and control is essential in order to develop an appropriate understanding and ability to design effective interventions In terms of experiences with specific violent events, the most commonly reported event women experienc e d in this study i s the husband/partner insulting her or making her feel bad about herself, followed by being s lapped and phy sically forced to have unwanted sex (see Figure 10 ). Being choked or threatened with a weapon wa s extremely rare in this cohort of women Lastly, o nly 8% of women report ever hitting their husband Figure 10 0% 5% 10% 15% 20% 25% 30% 35% 40% Ever hit partner MALE-DIRECTED: Forced you to perfrom a sexual act Physically forced you to have sex SEXUAL VIOLENCE: Threatened you with a weapon Tried to choke you Kicked you, dragged you, or beat you up Punched you with his fist of something else Pushed you, shook you, or threw something at you Twisted your arm or pulled your hair Slapped you PHYSICAL VIOLENCE: Say/do something to humiliate you Threaten to hurt you or someone you love Insult you or make you feel bad EMOTIONAL VIOLENCE:

PAGE 78

58 The most commonly reported male controlling behaviors women were : the husband is 11). The high prevalence of male controlling behavior highlights the strong patriarchal cultural in Zambia something that I regularly observed while conducting this research In addition, there are also strict gender norms (i.e., the structure of cathexi s) that condemn women socializing with non familial men which may explain the high prevalence of women reporting their husband is always jealous if she talks to other men Figure 11 Experiences with Specific Forms of Male Cont rolling Behavior (n=320) In addition to controlling behavior, c lassifying IPV into three different types (i.e., emotional, physical, and sexual) helps provide a better understanding of the power dynamics within these relationships. T he most prevalent form of IPV in this study i s emotional violence, with 40% of women reporting at least one emotionally violent ev ent, such as the husband humiliating her in public or threatening to hurt her (see Figure 12) Often, emotional 0% 20% 40% 60% 80% 100% Does not trust you with money Insists on knowing where you are at all times Limits contact with family Does not permit you to meet friends Frequently acceses you of being unfaithful Always jealous if you talk to other men

PAGE 79

59 violence is no t evaluated in studies examining IPV (see Abramsky et al., 2011; Harling, Msisha, & Subramanian, 2010; Pallitto et al., 2013) My research highlights that this could be problematic because it overlooks a large group of women who are for all intents and pu rposes experiencing IPV. In addition, o ver one third of women also report sexual violence and over 30% report physical violence. The most frequently repo rted combinations of violence a re experiencing only emotional violence or experiencing all three type s of IPV There are also a large pr oportion of women who either report experiencing emotional and physical violence together or only sexual violence. Figure 12 Types of Intimate Partner Violence Reported (n=320)

PAGE 80

60 Predictors of Intimate Partner Violence and Control While t reating IPV and partner controlling behavior as dependent variables, several significant associations emerge (See Table 7). First, younger women are more likely to experience male controlling behaviors: fo r every year of age, women have 8% lower adjusted odds of experiencing three or more partner controlling behaviors (p<0.05). Age, however, is not associated with experiencing IPV. To my knowledge, no other research from sub Saharan Africa has established younger age as a risk factor for partner controlling behaviors. T his may being much older among the younger women in my study 20 Indeed, some research has indicated that yo unger age is a risk factor for IPV, while others have found null results (Jewkes, 2002) An additional risk factor for partner controlling behavior among women in this study is greater parity. For each additional child, women have 31% higher adjusted o dds of experiencing three or more partner controlling behaviors (p<0.05). T o my knowledge, t his association has not been directly explored in the literature around IPV and partner control. However, there is evidence that parental stress, child health outco mes, and IPV are interrelated (Huth Bocks & Hughes, 2007) It is plausible that families with more children have more stressors, which could lead to increased violent/controlling behaviors, although parity and IPV are not specifically associated in this st udy. In addition, there is a marginally month of infant age, women have 13% higher adjusted odds of experiencing three or more controlling behaviors (p=0.08). 20

PAGE 81

61 Table 7 Multivariate Logistic Regression Results for the Odds of Experiencing IPV and Three or More Partner Controlling Behaviors Participant/Relationship Characteristics Model 1 Model 2 Experienced Intimate Partner Violence aOR (95% CI) Experienced Three of More Controlling Behaviors aOR (95% CI) Age (years) 1.0 (0.93 1.07) 0.91* (0.84 0.98) Infant age (months) 0.98 (0.87 1.06) 1.13 + (0.98 1.29) Parity 1.10 (0.69 1.40) 1.31* (1.00 1.72) Highest educational attainment 0.97 (0.75 1.23) 1.35* (1.00 1.80) Wealth index score 1.21 (0.89 1.65) 0.65* (0.46 0.91) Length of relationship 0.94+ (0.88 1.01) 0.98 (0.91 1.06) Disclosed HIV status 0.88 (0.24 3.28) 0.88 (0.20 3.78) 1 1.78 (0.81 3.94) 1.02 (0.42 2.48) Discordant couple 1 0.99 (0.53 1.85) 1.10 (0.55 2.23) Number of acceptable reasons to beat a wife 1.31** (1.12 1.53) 1.80*** (1.52 2.14) Experienced IPV N/A 2.88** (1.56 5.31) Experienced 3+ partner controlling behaviors 2.85** (1.55 5.21) N/A Number of household decisions the woman participates in 0.88 (0.71 1.12) 0.90 (0.67 1.17) Equal or greater earnings than the husband 1.20 (0.60 2.40) 0.85 (0.38 1.94) + p <0.10 p<0.05 ** p<0.01 ***p<0.001 1

PAGE 82

62 educational attainment, 21 she has 35% higher adjusted odds of experiencing three or more partner controlling behaviors (p<0.05). This is PV (Abramsky et al., 2011; Garcia Moreno, Jansen, Ellsberg, Heise, & Watts, 2006; R. Jewkes, 2002) much less is known about the relationship with partner controlling behaviors. Indeed, education is negatively associated with IPV in my study, although the relationship was not education may mak e the husband feel emasculated, and as a result, he attempt s to enforce his dominance in the relationship through controlling beh aviors. There is evidence from low and middle income countries, such as India, Bangl adesh, and Nicaragua, that report a positive (which is related to educational attainment) and IPV (Kishor & Johnson, 2004; Vyas & Watts, 2009b) but this association was not established in my models. In contrast, wealth i s negatively associated with partner controlling behaviors. For each standard deviation on the wealth index, women have 35% lower adjusted odds of experiencing t hree or more partner controlling behaviors (p<0.05) As mentioned earlier in this chapter, the most recent ZDHS also reports that women in the lower wealth quintiles have hi gher levels of partner control ( CSO, 2014) It is unclear, however, why such an ass ociation exists in this study between wealth and partner control, but not with IPV, which has been reported elsewhere (Abramsky et al., 2011 ; CSO, 2014 ) 21 Ordinal scale: no education, some primary education, completed primary, some secondary, completed secondary, some college, completed college, and graduate education

PAGE 83

63 Not surprisingly, greater endorsement of wife beating as an acceptable respon se for sions i s positively associated with experiencing both IPV (p<0.01) and three or more partner controlling behaviors (p<0.001). Due to the cross sectional nature of my research, it is not feasible to establish if IPV /control or acc eptance of wife beating occ urred fi rst, but I suspect that there is a bidirectional relationship. Women who are accepting of IPV may be more likely to end up in violent /controlling relationsh ips, while women who experience IPV /control Overall, t he HIV positive women in my study endorse wife beating to an even greater extent than the women in the most recent ZDH S: 65% of my participants agree that a husband is justified in beating his wife for at least one reason compared to only 47 % of Zambian women on the 2014 ZDHS (CSO, 2014) However, my participants also experience higher levels of IPV, which may account for their greater acceptance of wife beating. Interestingly, there is evidence from sub Saharan Africa that women tend to endorse wife beating to an even greater extent than men. Uthman, Lawako, & Moradi (2009) report that in sub Saharan African co untries where IPV is an accepted response to women's transgressions, women find more justification for the practice than men. Inde ed, the m ost recent ZDHS found that 47% of Zambian women agree a husband is justified in beating his wife for at least one speci fied reason, compared to only 31 % of Zambian men who endorsed this belief (CSO, 2014) highlighting the gendered interpretations of the structure of cathexis. Lastly, experiencing IPV and the partner displaying three or more controlling behaviors are significantly associated with each other in the multivariate models (p<0.01). Surprisingly, h owever, household decision making and are not associated with either IPV or partner control in the multivariate models. This may be

PAGE 84

64 because there is a large amount of collinearity between the gender power measures. Indeed, at the bivariate level, each of the four gender power measures (i.e., IPV, partner control, hou sehold decision making, and significantly correlated with one another (see Table 8). Table 8 Correlation Coefficien ts Between Gender Power Dynamic Measur es (n=320) IPV 3 or More Controlling Behaviors Number of Decisions the Woman Participates in Earnings Greater or Equal IPV 1.00 3 or More Controlling Behaviors 0.40*** 1.00 Number of Decisions the Woman Participates in 0.24*** 0.32*** 1.00 Greater or Equal to 0.16**** 0.18*** 0.11*** 1.00 *** p <0.001 Despite null results in this study in the models from Table 7, it is important to note that n umerous other studies from various global settings have reported a connection between in household decisions and female directed IPV (Gage & Hutchinson, 2006; Hindin & Adair, 2002; Murphy & Meyer, 1991; Murray A. Straus, Gelles, & Steinmetz, 1980) Research from th e United States (Murphy & Meyer, 1991; Murray A. Straus et al., 1980) Asia (Hindin & Adair, 2002), and the Caribbean (Gage & Hutchinson, 2006) has reported that autonomy in decision making generally has a negative association with IPV. Lastly, r ese arch from Zambia and Kenya found that women who lack autonomy in household decisions are more likely to justify IPV than their more empowered peers (Lawoko, 2006, 2008)

PAGE 85

65 Positive Deviants Only 5% of participants report no IPV or partner control (n=16; see figure 9 ) Upon compa rison with women who experience IPV/controlling behavior, two significant differences emerge at the bivariate level First, women who have not experie nce d IPV/controlling behavior have been in their relationship l onger than women who e xperience IPV/cont rolling behavior: 10.3 years compared to 6.4 years, respectively (p<0.01). Second, women who have not experience d IPV/controlling behavior universally disclosed they wer e HIV infected to their husband (100%), compared to just over 90% of women who experience d IPV/controlling behavior (p<0.01). Although not statistic ally significant, women who have not experience d I PV/controlling behavior also report a higher prevalence of discordant relationships (i.e., when one partner is HIV positive, in this case the woman, and the other is HIV negative). In contradiction, other studies have found an increased risk of IPV among HIV positive women in discordant relationships (Emusu et al., 2009; Ezechi et al., 2009; Shuaib et al., 2012; Were et al., 201 1) A possible explanation for my finding is tha t discordance in my study was based on reporting of the which may or may not in fact be his actual HIV status. As a result, my null finding m a y be based on Type II error. If I had biomarker information regarding the actual HIV status of the husband, I may have found quantifiably different results. Optimistically, it is also possible that this is an indication that HIV is becoming less stigmatized and that HIV positive women in d iscordant relationships are not at greater risk of IPV/control from HIV negative partners.

PAGE 86

66 Lastly the women who did not experience any IP V/controlling behaviors report more income or about the same as their husband (instead of the partner having a hig he r income), but the difference s are not statistically si gnificant. Likewise, women who have not experience d I PV/controlling behavior report greater participation in all household decisions (see Figure 13) ; but again, these differences a re not statistically significant. However, this is possibly due to small sample size and inadequate statistical power. Given the consistent trend, I would like to see results using a larger sample size because I suspect with more statistical power, a significant relationship m ay be present. Figure 13 Household Decision Making and Experiences with IPV/Controlling Behavior (N=320) 75% 63% 75% 56% 67% 55% 74% 37% 0% 20% 40% 60% 80% Own health care Major household decisions Daily household decisions Final say over money No IPV or Controlling Behaviors (n=16) IPV or Controlling Behaviors (n=304)

PAGE 87

67 Gender Asymmetry in Intimate Partner Violence There is an ongoing debate among sociological scholars regarding whether IPV is primarily female directed and a result of patriarchal structures or conversely, gender symmetrical with men and women perpetrating and experiencing IPV at similar levels ( Stra us, 2008) Studies from various settings although heavily focused in the United States report gender symmetry and that men and women commit violence equally in heterosexual relationships (Anderson, 2002; Capaldi & Owen, 2001; Moffit t, Robins, & Caspi, 2001; Straus, 2008; Straus & Ramirez, 2002) Feminist scholars, on the other hand, argue that IPV is a form of male domination over women in order to maintain power and control (Dobash & Dobash, 1979; Michalski, 2004; Stark & Flitcraft, 1996; Yllo, 1993) The Theory of Gender and Power supports the feminist perspective and postulates that men perpetrate greater violence against women than vice versa (i.e., part of the sexual division of power). Indeed, studies from South Africa (Gass, Stein, Williams, & Se edat, 2011) ; Australia (de Vries Robbe, March, Vinen, Horner, & Roberts, 1996; Roberts, O'Toole, Raphael, Lawrence, & Ashby, 1996) the Netherlands (Romkens, 1997) the United Kingdom (Dobash & Dobash, 2004) and South Korea (Lee, Stefani, & Park, 2014) su pport the argument that men do in fact account for the majority of IPV perpetration. Unfortunately, the only measure for male directed IPV on the instrument I used is one question asking women if they have ever hit thei r partner. This limited my ability t o create a comprehensive picture of what type of violence/control is occurring towards men in these relationships. However, using my physical violence against women measurement 22 and the 22 This measurement includes seven items on the CTS2. If women reported yes to any of the items, they are considered to have experienced physical IPV.

PAGE 88

68 one question regarding whether a woman ever hit her partner I am able to make some inferences about the symmetry/asymmetry of physical violence in these marriages In the vast majority (77%) of relation ships where physical violence i s reported (n=118), it i s only female directed (see Figure 14). S urprisingly, m ore women re port male directed physical violence compared to bidirectional violence. If we are to examine only o ne item from the CTS2 that asks nt of difference between male and female directed physical IPV decreases. However, even comparing just these two questions, women still report their husband hitting them at 3.5 times the rate they hit him (i.e., 28% compared to 8%, respectively ; data not s hown in figure ). Based on these findings, my data finds overwhelming support for gender asymmetry in physical violence which is by and large perpetrated by men against their HIV positive female spouses. Figure 14 IPV Victimization in Physically Violent Relationships by Gender (n=118) 12% 77% 11% Male-directed physical IPV only Female-directed physical IPV only Bidirectional physical IPV

PAGE 89

69 Chapter Summary This chapter explores HIV and partner controlling behaviors. I also establish factors associated with these e xperiences and the characteristic s of the women who did not experience any IPV or male controlling behaviors Lastly, this chapter indicates that physical violence i s gender asymmetrical with women more likely to be victimized than their male partners In the next several chapters, I examine how gender power dynamics within couples including different forms of IPV, male controlling behavior, and household decision making behaviors related to PMTCT, beginning with drug adherence during and after pregnancy in the following chapter

PAGE 90

70 CHAPTER V DRUG ADHERENCE DURING AND AFTER PREGNANCY PMTCT depends critically on HIV ARVs during and after pregnancy Suboptimal adherence by HIV positive pregnant and breastfeeding women not only increases the risk of mother to child transmission but also increases the likelihood of maternal HIV related disease progression and drug resistance for both the mother and the infant (Nachega et al., 2007; Ngoma et al., 2015; M. Onono et al., 2015) Recent studies suggest that HIV positive individuals need to take at least 70 80% of prescribed ARV doses to adequately suppress the virus (Kobin & Sheth, 2011; Martin et al., 2008; Nachega et al., 2007) However, the optimal standa rd is generally considered to be at least 95% of ARV doses in order to maximize health outcomes, including reducing vertical HIV transmission (Alexander et al., 2003; Garcia de Olalla et al., 2002; Gross et al., 2006; Harrigan et al., 2005; Howard et al., 2002; Paterson et al., 2000; Press, Tyndall, Wood, Hogg, & Montaner, 2002; Wood et al., 2003; Wood et al., 2006) WHO recently stated that in order to achieve the global goal of eliminating mother to child transmission, it is crucial to es t ablish current trends in ARV adherence during and after pregnancy as well as a clear understanding of the barriers to adherence (WHO, 2010c) Nachega et al. (2012) confirms in their recent review that adherence to ARVs during and after pregnancy remains a challenge, incl uding in sub Saharan Africa However, studies on a dherence to ARVs for PMTCT vary vastly across the region with reports as high as 98% in urban Kenya (Imbaya, Odhiambo Otieno, & Okello Agina, 2008) to as low as 38% in rural Uganda (Barigye et al., 2010) R ecent research from Zambia indicates that PMTCT medication adherence ranges from 63% to 79% (Conkling et al., 2010; Megazzini et al.,

PAGE 91

71 2010; Stringer et al., 2010) T hese estimates however, are based on sdNVP at the time of labor/delivery. Little is known about HIV positive Zambian newer combination ARV regimens during and after pregnancy and what social barriers hinder adherence. In this chapter, I answer a n important research question: i s there a relationship between gender power dynamics within couples and non adherence to ARVs for the purpose of PMTCT? First, I describe HIV levels to ARVs during and after pregnancy. Second, I estimate the number of children at risk for vertical HIV transmission among the mother baby pairs in this study b y combining the women who reported inadequate adherence with those who were not offered medication The final sections of this chapter focus on how non adherence to ARVs during and after pregnancy is affected by gender power dynamics including IPV, male controlling behaviors, participation in household decision making, and economic dependence on the husband Review of Methods W omen who were offered ARVs for PMTCT were s ubsequently asked questions regarding their drug adherence during and after pregnancy I n this chapter, I employ descriptive statistics to show levels of adequate (>80%) and optimal (>95%) drug adherence. I also use multivariate logistic regression models to determine the adjusted odds of adequate adherence to each ARV power dynamics Due to the collinear nature of my gender power dynamic measures (see Chapter Four), I first examine the effect on adherence of each power dynamic as a separate independent variable in models both with and without HIV status disclosure. I then estimate the models again combining all of the power dynamic measures.

PAGE 92

72 Adequate and Optimal Drug Adherence Of the women w ho were offered medication during pregnancy (n=271), 88% report adequate ARV adherence ( i.e., took >80% of the medication doses ), but only 62% report optimal ARV adherence ( i.e., took >95% of doses ; see Figure 15) Among the women on short course prophylaxis who were offered sdNVP during childbirth (n=131), 92% report taking the me dication (i.e., were adherent). During postpartum, 91% of women who were offered medication (n=285) report adequate adherence, but only 70% report optimal adherence. Las tly, adherence to giving the infant prophylaxis during the postpartum period i s adequate among 86% of women and optimal among 72% of women who were offered infant prophylaxis (n=303). My research thus supports the growing body of literature indicating that achieving high PMTCT adherence remains a challenge, including among low socio economic urban women Zambia Childbirth medication is a single dose that the woman either took or did not take. Figure 15 Adequate and Optimal Drug Adherence Levels 0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% Adequate Optimal Took Adequate Optimal Adequate Optimal Pregnancy (n=271) Childbirth* (n=131) Postpartum (n=284) Infant Prophylaxis (n=303)

PAGE 93

73 Figure 16 additionally displays the number of medication protocols where women achieved adequate adherence out of a possible total of four protocols 23 Thirty nine percent of women (n=125) report inadequate adher ence to at least one of the medication protocols and 5% ( n=17) report inadequate adherence to all of the medi cation protocols However, the majority of women (61%) report being adherent to all four of the medication protocols having their child completely protected across the PMTCT cascade of care. Includes medication during pregnancy, childbirth 24 postpartum, and giving the infant prophylaxis Figure 16 Number of Medication Protocols Where Women Achieved Adequate Adherence (n=320) 23 Pregnancy, childbirth, postpartum, and infant prophylaxis 24 Women on lifelong treatment (opposed to shor t course prophylaxis) were considered adherent during childbirth if they were adherent during pregnancy because this would be protective for the infant during birth; Women on lifelong treatment were not offered any specific additional medication intrapartu m to protect the infant unlike women on the short course regimen.

PAGE 94

74 Unprotected Children at Risk of Vertical HIV Infection Figure 17 displa ys the percent of children who we re unprotected through PMTCT during each time period as a result of poor adherence or lack of access to drugs. It is not clear whether the women in my study who report not being offered ARVs were in fact not offered drugs by the health center or if they did not take the medication despite being offered it. I suspect this group constitutes a combination of women who 1) were genuinely not offered any medication 2) were offered medication but did not understand what the medication was for and 3) were offered medication but decided n ot to take the medication and did not want to admit this in the i nterview. Nonetheless, infants we re unprotected through PMTCT in any of these scenarios. Includes women who were non adherent to either their medication or giving the infant prophylaxis (or both) Figure 17 Proportion of Children Not Fully Protected by PMTCT at Each Time Point (n=320)

PAGE 95

75 During pregnancy, one quarter of infants (n=81) were not fully protected through PMTCT. During childbirth, 21% of infants (n=67) were unprotected. During postpartum 28% of infants (n=90) were not fully protected through maternal medication and infant prophylaxis. Of the 90 children during the postpartum period who we re not fully protected by both medication pr otocols for the woman and the infant, over one third (n=32) we re not protected by either medication protocol putting the child at high risk for vertical HIV transmission through breastfeeding By examining these gaps in coverage and adherence it becom es clear why mother to child transmission is still a public health concern in Zambia Moreover, my study only captures women coming to the clinic for well child care, which may exclude women who are the least adherent those who do not come to the clinic for any medical care. Addressing both health system problems (e.g., stock outs of drugs and heath care workers missing HIV positive patients ), in addition to HIV drug adherence will be critical to achiev ing the PMTCT targets in Zambia In order to adequately address HIV positive however, a thorough understanding of the key social barriers is necessary, including the role of gender power dynamics within couples, such as IPV. Intimate Partner Violence and Drug Adherenc e As I hypothesized, IPV is negatively associated with adherence to all PMTCT medication, except sdNVP during childbirth (see Table 9). After controlling for status disclosure and other covariates, w omen who experienced IPV have 75% lower adjusted odds of adherenc e during pregnancy (p<0.05); 89% lower adjusted odds of adher ence postpartum (p<0.05); and 91% lower adjusted odds of adherence to giving the infant prophylaxis (p<0.001) compared to women who have not experience d IPV.

PAGE 96

76

PAGE 97

77 My findings regarding IPV and non adherence to ARVs for PMTCT are supported by qualitative research from South Africa (Hatcher et al., 2014; Mepham et al., 2011) My findings, however, run contrary to what Kiarie et a l. (2006) report in Kenya, who found n o association be tween IPV and uptake of PMTCT. H owever, at the time of the Kiarie et al. study PMTCT in Kenya was limited to sdNVP duri ng childbirth In my analysis, sdNVP intrapartum i s also not significantly associated wi th IPV. Yet, experiencing IPV i s associated with non adherence to PMTCT for all other medication protocols in the PMTCT cascade of care. There are several reasons why sdNVP during childbirth may not be as vulnerable to non adherence as other PMTCT medication protocols. First, sdNVP durin g labor/delivery is easier to conceal from a male men are generally not present in the delivery room. Second, if HIV positive women come to deliver in a health center, they should receive this dose o f medication by the health care personnel whereas the other protocols require constant refills of ARV prescriptions and daily consumption within the home which may be challenging if women have limited power within their household or are concealing their HIV status Although the current PMTCT regimens are more effective than only sdNVP at childbirth (WHO, 2010a) the extended duration of taking ARVs is likely more difficult for HIV positive mothers to follow, especially those in violent relationships and with less autonomy. Alternatively, there may be a relationship between IPV and sdNVP, but this study did not have enough power to detect such an association given the small sample size of women on the short course prophylaxis regimen who were offered the medication intrapartum (n=131).

PAGE 98

78 The Severity and Frequency of Partner Violence and Drug Adherence Not only does IPV reduce the odds of adherence to drugs across the PMTCT cascade, but the number of violent events and the frequency of violence in the pa st year also affect adherence The women in this study who experienced more violent events i n their current relationship have reduced odds of adherence to all PMTCT medication protocols except sdNVP during childbirth (see Table 10 ). For each additional vio lent event, women have 1 9 % lower adjusted odds of adheren ce during pregnancy (p<0.05); 3 0 % lower adjusted odds of adher ence postpartum (p<0.01); and 32% lower adjusted odds of adherence to giving the infant prophylaxis (p<0.001) Table 10 Adjusted Logisti c Regression Results for the Odds of Drug Adherence by the Number of Violent Events a Woman Experienced Model 1: Model 2: Model 3: Model 4: Power Dynamic Variable: >80% ARV Adherence during Pregnancy (n=271) 1 a OR (95% CI) Took sdNVP during Childbirth (n=131) a OR (95% CI) >80% ARV Adherence Postpartum (n=285) a OR (95% CI) >80% Infant Prophylaxis Adherence (n=303) a OR (95% CI) Number of IPV events 0.81* (0. 67 0.98 ) 0.81 (0 .57 1.14 ) 0.70** (0.55 0.85 ) 0. 68*** (0. 57 0.80 ) Controls: Disclosed Status 16.07 *** ( 5.11 50.26 ) 10.36 (1.40 76.67 ) 17.89 *** (5.27 6 0.77 ) 10.40 *** (3.37 32.12 ) Age 1. 11* (0.98 1.25 ) 1.11 (0.86 1.43) 1.18 (1.03 1.35 ) 1.10 + (0.99 1.22 ) Infant age 1.00 (0.82 1.21 ) 0.87 (0.62 1.24 ) 0.94 (0.75 1.17) 0.88 (0 .61 1.26 ) Parity 0.69 (0.48 0.98 ) 0.67 (0.31 1.45 ) 0.62 (0.41 0.95 ) 0.88 (0.61 1.26) Highest educational attainment 0.87 (0.57 1.32 ) 0.39 (0. 18 0.83 ) 1.16 (0.70 1.90 ) 1.09 (0.75 1.58 ) Knowledge of PMTCT 1.56 (0.88 2.74 ) 1.27 (0.57 2.79 ) 1.17 (0.70 1.98 ) 1.13 (0.75 1.7 ) Wealth 0.98 (0.61 1.58) 0.92 (0.40 2.10 ) 1.43 (0.82 2.50 ) 1.04 (0.69 1.57 ) On treatment 2 3.01 (1.17 7.74 ) N/A 2. 49 + (0.87 7.14 ) 3.04* (1.35 6.88) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Different sample sizes because women were offered medication differentially 2 Reference group was women on short course prophylaxis

PAGE 99

79 Additionally, women with more freque nt violence in the past year have reduced odds of ARV adherence during and after pregnancy (see Table 11 ) With each additional mean frequency score (see Chapter Three for explanat ion of measurement), a woman has 10% lower adjusted odds of adherence during pregnancy (p<0.05) ; 16% lower adjusted odds of ad herence postpartum (p<0.01); and 17% lower adjusted odds of adherence to giving the infant prophylaxis (p<0.001). Table 11 Adjusted Logistic Regression Results for the Odds of Drug Adherence by the Frequency of Violence in the Past Year Model 1: Model 2: Model 3: Model 4: Power Dynamic Variable: >80% ARV Adherence during Pregnancy (n=271) 1 a OR (95% CI) Took sdNVP during Childbirth (n=131) a OR (95% CI) >80% ARV Adherence Postpartum (n=285) a OR (95% CI) >80% Infant Prophylaxis Adherence (n =303) a OR (95% CI) Frequency of IPV in past year 0. 90* (0. 82 0.99 ) 0. 91 (0. 7 8 1. 06 ) 0.8 4** (0. 76 0.93 ) 0. 83*** (0. 77 0.90 ) Controls: Disclosed Status 15.65 *** ( 4.97 49.26 ) 10.24 (1. 40 74.75 ) 17. 48 *** (5.1 5 59.34 ) 10.06 *** (3. 27 30. 93 ) Age 1 .11 + ( 1.00 1.2 3 ) 1.11 (0.86 1.43) 1.18* (1.03 1.36) 1.1 1 + ( 1.00 1.23) Infant age 0.9 9 (0. 82 1. 19 ) 0.8 8 (0.6 2 1.2 4 ) 0.94 (0.75 1.17) 0.93 (0.78 1.10) Parity 0. 88 (0. 57 1.32 ) 0.6 6 (0.31 1.4 3 ) 0.6 1 (0.39 0.9 4 ) 0.8 6 (0.6 0 1.2 4 ) Highest educational attainment 1.08 (0. 74 1. 56 ) 0.3 9 (0.1 8 0.8 3 ) 1.1 5 (0. 70 1.8 9 ) 1. 08 (0.7 5 1 .56 ) Knowledge of PMTCT 0.87 (0. 57 1.31 ) 1. 28 (0.5 8 2.82 ) 1. 20 (0.7 1 2.0 1 ) 1. 13 (0. 75 1. 69 ) Wealth 0.98 (0.6 1 1.5 8 ) 0.9 2 (0.4 0 2.1 0 ) 1. 45 (0.83 2.5 4 ) 1.0 4 (0.6 9 1.5 7 ) On treatment 3.03* (1. 18 7.83 ) N/A 2. 48 + (0.86 7. 14 ) 2. 9 5* (1. 30 6. 66 ) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Different sample sizes because women were offered medication differentially 2 Reference group was women on short course prophylaxis

PAGE 100

80 Not only is the number of IPV events and the frequency of IPV associated with non adherence during and after pregnancy, but I also found strong evidence of a dose response relationship for the majority of protocols with these two variables. Figure 18 displ ays the adjusted logistic regression results for the number of violent events as dummy variables As is evidenced in the figure, each adjusted odds ratio associated with each dummy variable is greater then the prior providing evidence of a dose response relationship between the number of IPV events and non adherence to PMTCT medication. Comparison Group: No IPV (i.e., zero reported events) Figure 18 Adjusted Odds Ratio Estimates for Adherence to PMTCT Medication at Each Time Period by the Number of Violence Events A similar trend is also seen with the frequency of IPV in the past year with the exception of medication during pregnancy (see Figure 19). During pregnancy, there is no significant difference in the odds of adherence between having a frequency of IPV score of six to 10 and a score of greater than 10. However, for postpartum medication and infant 0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 Pregnancy Posptartum Infant Prophylaxis 1 to 3 events 4 to 6 events 7 to 9 events

PAGE 101

81 prophylaxis, a dose response relationship with the frequency of IPV in the past year is supported by the dummy variable data. Comparison Group: IPV score of ze ro (i.e., women who either experienced no IPV or experienced no IPV in the past year) Figure 19 Adjusted Odds Ratio Estimates for Adherence to PMTCT Medication at Each Time Period by the Frequency of IPV in the Past Year Score L astly, experiencing any injuries as a result of physical/sexual violence i s associated with non adherence, but only for the postpartum protocols, including giving the infant prophylaxis (see Table 12). I found evidence, however, of a marginally significant relationship between experiencing injuries and adherence during pregnancy. Women who experienced injuries as a result of violence have 53% lower adjusted odds of adheren ce during pregnancy (p=0.09), 69% lower adjusted odds of adherence postpartum (p<0.05) and 81% lower adjusted odds of adherence to giving the infant prophylaxis (p<0.001) compared to women who did not experience any injuries. 0 0.2 0.4 0.6 0.8 1 1.2 1.4 Pregnancy Posptartum Infant Prophylaxis score of 1 to 5 score of 6 to 10 score of >10

PAGE 102

82 Table 12 Adjusted Logistic Regression Results for the Odds of Drug Adherence by Injuries from Intimate Partn er Violence Model 1: Model 2: Model 3: Model 4: Power Dynamic Variable: >80% ARV Adherence during Pregnancy (n=271) 1 a OR (95% CI) Took sdNVP during Childbirth (n=131) a OR (95% CI) >80% ARV Adherence Postpartum (n=285) a OR (95% CI) >80% Infant Prophylaxis Adherence (n=303) a OR (95% CI) Any injuries from IPV 0.47 + (0.19 1.14) 0. 35 (0.0 7 1.72 ) 0.31* (0. 12 0.86 ) 0. 19*** (0. 08 0.41 ) Controls: Disclosed Status 16.47 *** ( 5.29 51.31 ) 9.14 (1. 20 69.42 ) 16.57 *** (5. 07 54.14 ) 10.06 *** (3. 33 30. 36 ) Age 1.1 0 + (0.9 8 1.2 4 ) 1.11 (0.86 1.4 2 ) 1.1 6 (1.0 1 1.3 2 ) 1. 09 (0.9 8 1.2 1 ) Infant age 1.00 (0.8 3 1.2 1 ) 0.8 6 (0.6 1 1.2 2 ) 0.9 5 + (0.7 7 1.17) 0.93 (0.7 9 1.10) Parity 0. 71 + (0. 50 1.02 ) 0. 71 (0.3 4 1.46) 0.6 7 (0. 45 1.01 ) 0. 92 (0.6 5 1. 31 ) Highest educational attainment 0.8 8 (0.57 1.3 4 ) 0.36* (0.16 0.81) 1.1 8 (0. 72 1. 94 ) 1.1 3 (0.7 7 1.6 4 ) Knowledge of PMTCT 1.5 5 (0.8 8 2.7 2 ) 1.3 5 (0.5 8 3. 14 ) 1. 26 (0.7 3 2. 12 ) 1. 31 (0.8 7 1. 97 ) Wealth 0.98 (0.61 1.5 7 ) 0.94 (0.42 2.1 1 ) 1. 39 (0.8 1 2. 38 ) 1.0 0 (0.6 6 1.5 2 ) On treatment 2.99 (1.1 6 7. 74 ) N/A 2. 42 + (0. 86 6.80 ) 2.8 4 (1.2 5 6. 46 ) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Different sample sizes because women were offered medication differentially 2 Reference group was women on short course prophylaxis Specific Types of Partner Violence and Drug Adherence Interestingly, physical violence has a less pronounced affect on adherence to the medication protocols compared to emotional, and especially, sexual violence. This underscores the importance of examining IPV as a multi dimensional phenomenon. Table 13 displays the association between emotion al violence and adherence to each medication protocol before and after controlling for HIV status disclosure. After adjusting for status disclosure and other covariate s, women who experienced IPV have 59% lower adjusted odds

PAGE 103

83 of adherence during pregnancy ( p= 0.053 ); 90% lower adjusted odds of adher ence postpartum (p<0.001); and 90% lower adjusted odds of adherence to giving the infant prophylaxis (p<0.001) compared to women who have not experience emotional IPV. Sexual viol ence, on the other hand, is the type of IPV that has the greatest impac t on non adherence (see Table 14 ). After adjusting for covariates, including HIV status disclosure, women who experienced sexual violence have 66% lower adjusted odds of adheren ce during pregnancy (p<0.05), 79% l ower adjusted odds of adherence to sdNVP during childbirth (p=0.07 ), 74% lower adjusted odds of adher ence postpartum (p<0.05), and 66% lower adjusted odds of adherence to giving t he infant prophylaxis (p<0.01) compared to women who have not experienced sex ual IPV in their current relationship. Lastly, the on ly medication protocol that is negatively affected by physical violence is giving the infant prophylaxis (see Table 15). Women who e xperienced physical violence have 62% lower adjusted odds of adherence to giving the infant prophylaxis (p<0.001) compared to women who have not experienced physical violence and net of HIV status d isclosure and other covariates Medication adherence during pregnan cy, childbirth, or postpartum a re not significantly affected by the presence of physical violence in the relationship.

PAGE 104

84

PAGE 105

85

PAGE 106

86

PAGE 107

87 Partner Control and Drug Adherence Mal e partner control is also associated with non adherence to PMTCT medication, but only during the postpartum period (see Table 16). For maternal medication postp artum, however, the relationship is no longer significant after adjusting for HIV status disclosure. Conversely, HIV status disclosure has a negligible affect on the association between partner control and adherence to giving the infant prophylaxis. After adjusting for HIV status disclosure and other covariate s, women whose partner displays three or mo re controlling behaviors have 74% lower adjusted odds of adherence to medication postpartu m (p=0.06) and 9 4% lower adjusted odds of adherence to giving the infant prophylaxis (p<0.001) c ompared to women who experience less than three partner controlling behaviors. Participation in Household Decisions and Drug Adherence In addition, w pation in household decisions i s positively associated with medication adhere nce postpartum (see Table 17). Controlling for HIV status disclosure also did not affect the strength of association. For each additional household decision that a woman participates, she has 49% higher adjusted odds of adherence to taking medication postpartum (p<0.05) and a 114% higher adjusted odds of adherence to giving the infant prophylaxis (p<0.001) after adjusting for HIV status disclosure. Lastly, contrary to my hypothesis, greater earnings are negatively associated with adherence to infant prophylaxis (see Table 18). After adjusting for status disclosure, women with g reater or equal earnings have 67% lower adjusted odds of adherence to giving the infan t prophylaxis (p<0.01). More research is needed to determine why greater earnings reduces adherence; I suspect however, that IPV may mediate this relationship.

PAGE 108

88

PAGE 109

89

PAGE 110

90

PAGE 111

91 The Combined Influence of Gender Power Dynamics on Drug Adherence Table 19 displays the logistic regression models that include all gender power measures combined. In this analysis, IPV appears to be driving the relationship between gender power dynamics and non adherence to medication during and after pregnancy. In addition, by combining all of the gender power dynamics into the models two divergent findings emerge compared to the separate models discussed earlier in the chapter. First, partici pation in household decisions i adhe rence to drugs postpartum but did remain associated with infant prophylaxis. Second, are no longer associated with adherence to giving the infant prophylaxis as they were i n the separate model in Table 18 In the combined mode l s, women who experienced IPV have 80% lower adjusted odds of adherenc e during pregnancy (p<0.05), 88% lower adjusted odds of adher ence postpartum (p<0.05), and 83% lower adjusted odds of adherence to giving t he infant prophylaxis (p<0.05) compared to women who did not experience IPV and net of other power dynamics and covariates. Partner controlling behaviors and low household decision makin g ability, on the other hand, a re only associated with non adherence to giving the infant prophylaxis in the comb ined models (see Table 19). Women who experienced three or more p artner controlling behaviors have 86% lower adjusted odds of adherence to giving t he infant prophylaxis (p<0.05) compared to women who experienced two or less controlling behaviors net of oth er power dynamics In addition, for each household de cision that a woman participates in, sh e has 88% higher adjusted odds of adherence to giving the infant prophylaxis (p<0.01).

PAGE 112

92 Table 19 Adjusted Logistic Regression Results for the Odds of Drug Adherence by Gender Power Dynamics in Combined Models Power Dynamic Variable: Model 1: Model 2: Model 3: Model 4: >80% ARV Adherence during Pregnancy (n=271) 1 aOR (95% CI) Took sdNVP during Childbirth (n=131) aOR (95% CI) >80% ARV Adherence Postpartum (n=285) aOR (95% CI) >80% Infant Prophylaxis Adherence (n=303) aOR (95% CI) Any IPV 0.20 (0.05 0.75 ) 0. 24 (0.02 3.92 ) 0.12* (0.02 0.71) 0.17* (0.04 0.70) Three or more partner controlling behaviors 1.4 4 (0.47 4.40 ) 3.00 (0.30 30.50 ) 0.52 (0.12 2.34) 0.14* (0.03 0.67) Number of household decisions women participate 1.20 (0.83 1.74) 1.41 (0.77 2.61) 1.33 + (0.87 2.03) 1.88** (1.30 2.73) Greater or equal earnings than husband 2.48 (0.77 8.01) 0.40 (0.06 2.60) 2.44 (0.69 8.62) 0.60 (0.25 1.43) Controls: Disclosed s tatus 18.56 *** (5.96 56.65) 12.95 (1.45 115.80 ) 20.24*** (5.54 73.87) 15.81*** (4.17 60.00) Age 1.1 3 + ( 1.00 1.27 ) 1.09 (0.87 1.38 ) 1.19* (1.03 1.38) 1.11 + (0.99 1.24) Infant age 0.98 (0.80 1.19 ) 0.9 1 (0.65 1.27 ) 0.96 (0.77 1.20) 0.96 (0.80 1.15) Parity 0. 67 (0.46 0.98 ) 0.65 (0.30 1. 38 ) 0.61* (0.39 0.95) 0.93 (0.64 1.36) Highest educational attainment 0.88 (0.56 1.37 ) 0.37 (0.16 0.84 ) 1.25 (0.74 2.12) 1.35 (0.88 2.07) Knowledge of PMTCT 1.71 + (0.94 3.13 ) 1.37 (0.59 3.04 ) 1.30 (0.74 2.27) 1.27 (0.79 2.02) Wealth 0.98 (0.60 1.59 ) 0.89 (0.37 2.20 ) 1.28 (0.73 2.26) 0.81 (0.50 1.30) On treatment 2 2.79 (1.06 7.33 ) N/A 2.12 (0.74 6.09) 2.89* (1.22 6.83) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Different sample sizes because women were offered medication differentially 2 Reference group was women on short course prophylaxis

PAGE 113

93 Although IPV appears to exert a greater influence than controlling behavior in the combined models, it is extremely difficult to tease these two factors apart because, as discussed in Chapter Four, there is high interconnectedness. In this study, it appears that IPV and partner controlling behaviors are overlapping variables, but also represent independent risk factors for non adherence specifically to giving the infant prophylaxis. More research is needed particularly qualitative studies, exploring why factors such as partner controlling behaviors and limited participation in household decision making have a greater affect on adherence to giving the infant medication compared to women taking medication th emselves. HIV Status Disclosure and Drug Adherence Although HIV status disclosure does not appear to mediate the relationship between gender power dynamics and adherence to PMTCT medication as I hypothesized, 25 it nonetheless is an important factor for wo to drug protocols Women who disclosed their HIV posi tive status to their husband have significantly higher adjusted odds of adherence to medication across the PMTCT cascade in all of the models in this chapter and net of gender power dynam ics and other covariates (see Tables 9 19). Table 20 highlights the association between HIV status self disclosure without including gender power dynamics in the models. Women who disclosed their HIV positive status to the husband have 20.0 times higher adjusted odds of adherence during pregnancy (p<0.001), 13.6 times higher adjusted odds of adherence to taking sdNVP during childbirth (p<0.05), 20.2 times higher adjusted odds of adherence postpartum (p<0.001), and 13.1 times higher adjusted odds of adhere nce to giving the infant prophylaxis (p<0.001) compared to women who did not disclose their HIV status. 25 With few exceptions, the association between the various gender power dynamics and non adherence to medication after including status disclosure.

PAGE 114

94 Table 20 HIV Status Disclosure and the Adjusted Odds of Drug Adherence Variable: Model 1 : Model 2 : Model 3 : Model 4 : >80% ARV Adherence during Pregnancy (n=271) 1 aOR (95% CI) Took sdNVP during Childbirth (n=131) aOR (95% CI) >80% ARV Adherence Postpartum (n=285) aOR (95% CI) >80% Infant Prophylaxis Adherence (n=303) aOR (95% CI) Disclosed s tatus 19. 97*** (6.53 61.06) 13.64* (1.87 99.21) 20.24*** (6.32 64.81) 13.05*** (4.54 37.47) Age 1.10+ (0.98 1.24) 1.11 (0.88 1.40) 1.15* (1.01 1.30) 1.08 (0.98 1.19) Infant age 1.01 (0.84 1.22) 0.92 (0.66 1.27) 0.97 (0.79 1.19) 0.97 (0.83 1.14) Parity 0.71+ (0.50 1.01) 0.68 (0.34 1.38) 0.68+ (0.46 1.01) 0.93 (0.66 1.30) Highest educational attainment 0.87 (0.58 1.31) 0.40* (0.19 0.84) 1.15 (0.71 1.85) 1.06 (0.75 1.51) Knowledge of PMTCT 1.48 (0.85 2.58) 1.28 (0.57 2.91) 1.13 (0.67 1.90) 1.09 (0.74 1.61) Wealth 0.99 (0.63 1.57) 0.90 (0.40 2.02) 1.41 (0.84 2.38) 1.08 (0.73 1.59) On treatment 2 3.36* (1.32 8.54) N/A N/A N/A + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Different sample sizes because women were offered medication differentially 2 Reference group was women on short course prophylaxis Chapter Summary This chapter explores the effect of various gender power dynamics both separately has a with more frequent and severe forms of IPV exerting a greater influence. Sexual violence, in particular, is detrimental to drug adherence during and after pregnancy Partner control, household decision making, and of the postpartum drug protocols bu t not to the same extent as IPV In addition, HIV status disclosure emerged as an important predictor of drug adherence. The following chapter explores if there are similar trend s in non adherence regarding safe infant feeding practices

PAGE 115

95 CHAPTER VI SAFE INFANT FEEDING In the literature on PMTCT adherence to safe infant feeding is often neglected, despite the risk of HIV transmission through breast milk One of the aims of this research project was to extend the analysis of adherence across the entire PMTCT cascade, including safe infant feeding practices among HIV positive mothers in Lusaka. In addition to ARVs during and after pregnancy adherence to a ppropriate infant feeding is critical for preventing vertical HIV transmission and infant survival HIV transmission through breastfeeding is estimated to be responsible for 30% to 60% of all HIV infections in children (Breastfeeding and HIV International Transmission Study Group, 2004) Indeed, Torpey et al. (2010) reported that postpartum HIV transmission due to suboptimal infant feeding is a major contributor to mother to c hild transmission specifically in Zambia. Despite the 5 20% risk of vertical HIV transmission from breast milk (without ARV intervention; WHO, 2000) breastfeeding remains a safer infant feeding practice than exclusive replacement formula feeding for HIV exposed infants in low and middle income co untries This is because although the infants who do not receive breast milk cannot contract HIV, they often suffer from increased morbidity and mortality namely, from gastrointestinal and respiratory infection s (Bahl et al., 2005; WHO, 2010b) Thus, HIV positive mothers in settings like Zambia are encouraged to exclusively breastfeed for the first six months of the life followed by the introduction of complementary foods and continued breastfeeding for at least 12 months. This feeding modality off ers the best balance of protection from infant mortality as well as mother to child transmission of HIV (WHO, 2010b)

PAGE 116

96 Early m ixed infant feeding, which includes giving the infant both br east milk and other liquids or food within the first six months of l ife is the most d angerous infant feeding practice due to the increased risk of both mother to child transmission and other infant morbidities, particularly diarrhea l disease (Coovadia et al., 2007; Coutsoudis et al., 2001; Coutsoudis, Pillay, Spooner, Kuhn & Coovadia, 1999; Iliff et al., 2005; Kuhn et al., 2007). E xclusive breastfeeding during the first six months of life i s associated with more than a 40% decreased risk of HIV transmission compared with mixed infant feeding (Coovadia et al., 2007; Iliff e t al., 2005). Previous research idicates that w hen infants are fed breast milk in addition to other food/liquids prior to six months of life (i.e., mixed feeding), intestinal disruption increases the risk of HIV transmission, compared to when the infant only recives breast milk (Kuhn et al., 2007). Unfortunately, HIV infected women tend to introduce other food and wean their infants significantly earlier than HIV uninfected women (Omari et al., 2003) Indeed, rates of exclusive breastfeeding a mong HIV positive women specifically range from only 19% to 48% at four months of life according to various studies across sub Saharan Africa (Maru & Haidar, 2009; Omari et al., 2003; Poggensee et al., 2004) The objective of this chapter is to determine how gender power dynamics within couples affect adherence to safe infant feeding practices. I suspect ed that gender power dynamics within couples negatively affect adherence to exclusive breastfeeding because infant feeding is something that a woman truly cann ot hide from her husband if he is unaware of her HIV status or disapproves of the medically rec ommended infant feeding practices I begin this chapter by highlighting the proportion of women in this study who report adherence to safe infant feeding, fo llowed by an analysis of the relationship between various

PAGE 117

97 practices (i.e., exclusive breastfeeding for six months). Review of Methods I measured infant feeding practices th e main outcome of interest in this chapter, using questions directly from the ZDHS (CSO, 2009) asking women which foods they gave the child and at what age The statistical methods I use in this chapter include descriptive and multinomial logistic regress ion analysis with infant feeding practice s as the dependent variable of interest and gender power dynamics as the independent variables of interest Infant Feeding Practices Among the women in this study, the most c ommon infant feeding practice i s exclusive breastfeeding, with 60% of women reporting only giving breast milk to their infants prior to six months of life (see Figure 20 ) Only 3% of women in this study report exclusive replacement feeding (i.e., never having breastf e d the infant). Lastly a large portion (37%) of women report mixed infant feeding, putting their infants at increased risk for HIV transmission. Figure 20 Proportion of HIV infected Women Practicing Three Types of Infant Feeding Modalities (n=320) Exclusive breastfeeding 60% Exclusive replacement feeding 3% Mixed feeding 37%

PAGE 118

98 A vast literature purports that i n sub Saharan Africa, breastfeeding is the most common infant feeding practice, but exclusive breastfeeding for six months is not generally considered normative (Doherty, Chopra, Nkonki, Jackson, & Persson, 2006; Kuhn et al ., 2007; Maru & Haidar, 2009; Omari, Luo, Kankasa, Bhat, & Bunn, 2003; Poggensee et al., 2004) This may be changing, however; based on the most recent ZDHS, 73% of women in the overall population of Zambia reported exclusive breastfeeding for six months ( CSO, 2014). This surprisingly high proportion may be an indication that cultural norms are changing with recent medical advice. In Zambia, exclusive replacement feeding with commercial infant formula is only cceptable, feasible, affordable, sustainable, and safe (ZMOH, 2010) Given the low socioeconomic status of my sample, I would be surprised if any of the women in this study actually had the necessary resources to meet the AFASS criteria. Indeed, Chisenga e t al. (2011) recently reported that Zambian women often opt for replacement feeding without meeting the appropriate AFASS conditions. My research supports the hypothesis that HIV positive women tend to introduce supplemental food earlier to their infants (compared to the general population of Zambian women). There are two specific reasons that likely contribute to why HIV positive women introduc e complementary foods earlier than uninfected women : 1) fear of HIV transmission through breast milk and 2) HIV related stigma. Unfortunately, in many sub Saharan African settings, HIV infected mothers are not given evidence based information that breast milk is significantly better than formula (WHO, 2010b) Health care workers often present HIV transmission through breast milk as a certainty instead of as a probability and underemphasize the risks of formula feeding (Buskens, Jaffe, & Mkhatshwa, 2007) This

PAGE 119

99 leaves many HIV infected mothers confused about th e risks associated with the various infant feeding options and many subsequently opt for mixed feeding, the most dangerous infant feeding practice. Additionally, HIV related stigma prevent s HIV infected women from feeding infants in ways that are outside of perceived positive status from household members including the husband ( Fadnes et al., 2010; Farquhar et al., 2001; Sibeko et al., 2009) Without disclosure of an HIV positive status, women may find it dif ficult to justify why they are practicing exclusive breastfeeding to family members. In point of fact, f amily i nfluence has been associated with HIV regarding how to feed their infant (Buskens et al., 2007; Doherty et al., 2007) Frequently cited family members with strong influence over infant feeding decisions in sub Saharan grandmother (Bezner Kerr, Dakishoni, Shumba, Msachi, & Chirwa, 2008; Buskens et al., 2007; Desclaux & Alfieri, 2009; Doherty, Chopra, Nkonki, Jackson, & Persson, 2006; Nankunda, Tumwine, Nankabirwa, & Tylleskar, 2010) Across sub Saharan Africa, studies report that it is common to give infants water, tea, porridge, and ot her foods in addition to breast milk as early as the first few weeks of life (Doherty, Chopra, Nkonki, Jackson, & Persson, 2006) Actually in this study, plain water and porridge were the two most commonly reported food items given to infants prior to si x months (see Figure 21 women cite were vegetable oil (given to infants to treat a perceived upset stomach) and a commercial drink called Super Shake (a smoothie l ike beverage with questionable nutritional content).

PAGE 120

100 Figure 21 Reported Food/Liquids Given to Infants Prior to Six Months Among HIV Infected Women Practicing Mixed Feeding (n=169) One limitation of my research is that some of the infants in this study were less than six months of age. Mothers may have introduced complementary foods to their infant prior to six months, but we would not be able to capture feeding practices after the s urvey was completed Thus, my calculation on the proportion of women practicing exclusive breastfeeding may be slightly overestimated in this cohort Intimate Partner Violence and Safe Infant Feedin g Similar to drug adherence, experiencing IPV negatively practice safe infant feeding (see Table 21). After adjusting for HIV status disclosure and other covariates, women who experienced IPV have 2.9 times higher adjusted odds of practicing mixed infant feeding rather than exclusive breastfeeding (p<0.001) compared to 0% 10% 20% 30% 40% 50% 60% 70% Other Porridge Honey Infant formula Fruit juice Sugar salt water Herbs Gripe water Sugar/glucose water Plain water Cow's milk

PAGE 121

101 women who did not experience IPV. One potential explanation for this association is that the stress related to IPV may physically inhibit the release of oxytocin, the hormone that facilitates the flow of breast milk (Law rence & Lawrence, 1985), making exclusive breastfeeding challenging, if not impossible In addition, factors such as partner support may mediate the relationship, which is discussed in further detail in Chapters Eight and Nine. Table 21 Multinomial Logist ic Regression Results for the Odds of Safe Infant Feeding Practices by Experiences with Intimate Partner Violence (n=320) Reference Group: Exclusive Breastfeeding (n=192) Power Dynamic Variable: ERF 2 (n=10) RRR 3 (95% CI) Mixed Feeding (n=118) RRR (95% CI) w/o status disclosure w/status disclosure w/o status disclosure w/status disclosure Any IPV 1.47 (0.37 5.78) 1.42 (0.36 5.67) 3.16*** (1.84 5.44) 2.94*** (1.70 5.08) Controls: Disclosed s tatus ----0.55 (0.05 6.13) ----0.32* (0.12 0.82) Age 1.21* (1.04 1.42) 1.21* (1.03 1.42) 1.02 (0.96 1.08) 1.01 (0.95 1.07) Infant age 1.01 (0.76 1.36) 1.01 (0.76 1.36) 1.15* (1.03 1.28) 1.15* (1.03 1.28) Parity 0.50* (0.26 0.98) 0.51* (0.26 1.00) 1.18 (0.96 1.45) 1.22+ (0.98 1.50) Highest educational attainment 0.82 (0.40 1.69) 0.82 (0.40 1.69) 0.94 (0.74 1.19) 0.93 (0.73 1.19) Knowledge of PMTCT ----3 ----3 0.57** (0.74 1.19) 0.57** (0.41 0.78) Wealth 1.45 (0.66 3.22) 1.48 (0.66 3.32) 1.08 (0.83 1.41) 1.14 (0.87 1.50) On treatment 2 1.89 (0.35 10.20) 1.95 (0.36 10.61) 0.73 (0.43 1.24) 1.14 (0.87 1.50) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Different sample sizes because women were offered medication differentially 2 Reference group was women on short course prophylaxis 3 Data could not converge (likely due to small sample size)

PAGE 122

102 Karamagi et al. (2007) also report that lifetime IPV i s associated with infant feeding practices in Uganda (not specific to HIV positive women), but only in their unadjusted analysis. Two factors may account for why I found a relationship in my adjusted models. First, I examined IPV with current sex ual partnership, whereas Karamagi et al. (2007) used a measure of lifetime IPV. C urrent ongoing violence is likely more detrimental than lifetime experiences with IPV. Second, my sample includes only HIV infected women, who have unique experiences regardin g IPV and infant feeding decisions due to the stigma associated with being HIV positive. T he Severity and Frequency of Partner Violence and Safe Infant Feeding In addition infant feeding practices a re also adverse ly affected by the number of violent even ts a woman has experienced (see Table 22) For each additional violent event a woman experienced, she has 17% higher adjusted odds of mixed feeding rather than exclusive breastfeeding (p<0.05). The frequency of IPV in the past year also increases the likel ihood that a woman will practice mixed infant feeding rather than exclusive breastfeeding (see Table 23). For each additional mean frequency score of violence (see Chapter Three for explanation of measurement), a woman has 8% higher adjusted odds of mixed feeding rather than exclusive breastfeeding (p<0.05). I did not, however, find any evidence of a relationship between injuries as a result of violence and an increased risk of practicing mixed feeding rather than exclusive breastfeeding (RRR 1.35, p=0.29; see Table 24).

PAGE 123

103 Table 22 Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices by the Number of Violent Events (n=320) Reference Group: Exclusive Breastfeeding (n=192) Gender Power Variable E xclusive replacement feeding ( n=10) RRR (95% CI) Mixed f eeding (n=118) RRR (95% CI) Number of IPV Events 1.03 (0.70 1.53 ) 1.17* (1.03 3.34 ) Controls: Disclosed s tatus 0.53 (0.04 6.65) 0.34* (0.13 0.88) Age 1.20* (1.03 1.42) 1.00 (0.94 1.07) Infant age 1.01 (0.76 1.36) 1.15* (1.03 1.28) Parity 0.51* (0.26 1.00) 1.23 + (1.00 1.52) Highest educational attainment 0.83 (0.41 1.71) 0.95 (0.75 1.20) Knowledge of PMTCT ___ 3 0.58** (0.42 0.81) Wealth 1.48 (0.63 3.32) 1.17 (0.90 1.53) On treatment 2 1.91 (0.35 10.43) 0.74 (0.44 1.25) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Different sample sizes because women were offered medication differentially 2 Reference group was women on short course prophylaxis 3 Data could not converge (likely due to small sample size)

PAGE 124

104 Table 23 Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices by the Frequency of Violence in the Past Year (n=320) Reference Group: Exclusive Breastfeeding (n=192) Gender Power Variable E xclusive replacement feeding (n=10) RRR 3 (95% CI) Mixed f eeding (n=118) RRR (95% CI) Frequency of violence in the past year 1.03 (0.86 1.23 ) 1.08* (1.01 1.14 ) Controls: Disclosed s tatus 0.57 (0.04 7.44) 0.34* (0.13 0.90) Age 1.21* (1.03 1.41) 1.00 (0.94 1.07) Infant age 1.01 (0.76 1.36) 1.15* (1.03 1.28) Parity 0.51 + (0.26 1.00) 1.23* (1.00 1.52) Highest educational attainment 0.83 (0.40 1.70) 0.95 (0.75 1.20) Knowledge of PMTCT ___ 3 0.59** (0.43 0.81) Wealth 1.49 (0.66 3.37) 1.17 (0.90 1.53) On treatment 2 1.94 (0.35 10.57) 0.75 (0.44 1.26) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Different sample sizes because women were offered medication differentially 2 Reference group was women on short course prophylaxis 3 Data could not converge (likely due to small sample size)

PAGE 125

105 Table 24 Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices by Injuries from Violence (n=320) Reference Group: Exclusive Breastfeeding (n=192) Gender Power Variable E xclusive replacement feeding (n=10) RRR 3 (95% CI) Mixed f eeding (n=118) RRR (95% CI) Injuries as a result of violence 0.31 (0.03 3.12) 1.36 (0.77 2.38) Controls: Disclosed s tatus 0.31 (0.02 3.97) 0.29* (0.11 0.75) Age 1.20* (1.03 1.41) 1.00 (0.94 1.07) Infant age 1.02 (0.76 1.37) 1.14* (1.02 1.27) Parity 0.52+ (0.27 1.01) 1.22+ (0.99 1.50) Highest educational attainment 0.87 (0.43 1.76) 0.94 (0.75 1.19) Knowledge of PMTCT ___ 3 0.57** (0.43 0.81) Wealth 1.34 (0.60 2.98) 1.17 (0.89 1.52) On treatment 2 1.82 (0.34 9.86) 0.71 (0.42 1.19) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Different sample sizes because women were offered medication differentially 2 Reference group was women on short course prophylaxis 3 Data could not converge (likely due to small sample size) Unlike Chapter Five, a dose response relationship between the number of IPV events and non adherence to safe infant feeding is not supported by my data (see Figure 22). Although an overall relationship exists between the number of IPV events and unsafe inf ant feeding practices (see Table 22), each subsequent dummy variable does not increase the adjusted odds of mixed infant feeding father than exclusive breastfeeding. A dose response relationship with the frequency of violence in the past year, on the other hand, is supported

PAGE 126

106 by the dummy variables each adjusted odds ratio associated with each dummy variable for the frequency of IPV score is greater than the prior dummy variable (see Figure 23). Comparison Group: No IPV (i.e., zero reported events) Fi gure 22 Adjusted Multinomial Odds Ratio Estimates for Adherence to Safe Infant Feeding by the Number of IPV Events (n=320 ) Comparison Group: Frequency of IPV score of zero Figure 23 Adjusted Multinomial Odds Ratio Estimates for Adherence to Safe Infant Feeding by the Frequency of IPV in the Past Year Score (n=320) 0 0.5 1 1.5 2 2.5 3 1 to 3 events 4 to 6 events 7 to 9 events RRR of mixed feeding rather than exclusive breastfeeding 0 0.5 1 1.5 2 2.5 score of 1 to 5 score of 6 to 10 score of >10 RRR of mixed feeding rather than exclusive breastfeeding

PAGE 127

107 Specific Types of Violence and Safe Infant Feeding Different types of IPV also have differing affects on adherence t o safe infant feeding with emotional and sexual violence having the most pronounced effect. Physical violence, on the other hand, is not associated with unsafe infan t feeding practices After adjusting for HIV status disclosure and other covariates, women who experienced emotional violence h ave 85% higher adjusted odds of practicing mixed infant feeding rather than exclusive breastfeeding (p<0.05) compared to women who did not experience emotional IPV (see Table 25). Table 25 Multinomial Logistic Regressi on Results for the Odds of Safe Infant Feeding Practices by Emotional Violence (n=320) Reference Group: Exclusive Breastfeeding (n=192) Power Dynamic Variable: ERF 2 (n=10) RRR 3 (95% CI) Mixed Feeding (n=118) RRR (95% CI) w/o status disclosure w/status disclosure w/o status disclosure w/status disclosure Emotional Violence 1.40 (0.34 5.36) 1.33 (0.32 5.32) 1.95** (1.19 3.20) 1.85* (1.12 3.06) Controls: Disclosed Status ----0.54 (0.05 5.88) ----0.28** (0.11 0.72) Age 1.21* (1.04 1.42) 1.21* (1.03 1.41) 1.02 (0.96 1.08) 1.00 (0.94 1.07) Infant age 1.01 (0.76 1.36) 1.01 (0.76 1.36) 1.15* (1.03 1.28) 1.14* (1.03 1.28) Parity 0.51* (0.26 0.99) 0.51+ (0.26 1.00) 1.19 (0.97 1.45) 1.23+ (1.00 1.51) Highest educational attainment 0.83 (0.41 1.70) 0.83 (0.40 1.70) 0.96 (0.76 1.21) 0.95 (0.75 1.20) Knowledge of PMTCT ----3 ----3 0.59** (0.44 0.81) 0.59** (0.43 0.81) Wealth 1.47 (0.66 3.28) 1.49 (0.66 3.37) 1.10 (0.85 1.42) 1.17 (0.89 1.52) On treatment 2 1.87 (0.35 10.07) 1.92 (0.35 10.45) 0.68 (0.41 1. 13) 0.73 (0.43 1.23) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Different sample sizes because women were offered medication differentially 2 Reference group was women on short course prophylaxis 3 Data could not converge (likely due to small sample size)

PAGE 128

108 In addition, women who experienced sexual violence have 2.2 times higher adjusted odds of practicing mixed infant feeding rather than exclusive breastfeeding (p<0.01; see Table 26). Physical violence displayed a marginally significant relationship with infant feeding prior to including HIV status disclosure in the model (p=0.08) but th is relationship disappeared after adding HIV status disclosure as a covariate (p=0.22; see Table 27). Table 26 Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices by Sexual Violence (n=320) Reference Group: Exclusive Bre astfeeding (n=192) Power Dynamic Variable: ERF 2 (n=10) RRR 3 (95% CI) Mixed Feeding (n=118) RRR (95% CI) w/o status disclosure w/status disclosure w/o status disclosure w/status disclosure Sexual Violence 1.18 (0.28 5.02) 1.13 (0.26 4.99) 2.41** (1.45 3.99) 2.21** (1.32 3.70) Controls: Disclosed Status ----0.52 (0.04 5.95) ----0.31* (0.12 0.80) Age 1.21* (1.04 1.42) 1.21* (1.03 1.41) 1.01 (0.95 1.08) 1.00 (0.94 1.07) Infant age 1.01 (0.75 1.35) 1.01 (0.75 1.35) 1.13* (1.01 1.25) 1.12* (1.08 1.25) Parity 0.51* (0.26 0.98) 0.51* (0.26 0.99) 1.18 (0.96 1.44) 1.21+ (0.98 1.50) Highest educational attainment 0.83 (0.41 1.71) 0.83 (0.41 1.70) 0.95 (0.75 1.19) 0.94 (0.74 1.19) Knowledge of PMTCT ----3 ----3 0.60** (0.44 0.83) 0.60** (0.43 0.82) Wealth 1.44 (0.66 3.17) 1.48 (0.66 3.29) 1.07 (0.82 1.39) 1.13 (0.87 1.48) On treatment 2 1.85 (0.34 9.97) 1.91 (0.35 10.37) 0.68 (0.41 1.15) 0.73 (0.43 1.23) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Different sample sizes because women were offered medication differentially 2 Reference group was women on short course prophylaxis 3 Data could not converge (likely due to small sample size)

PAGE 129

109 Table 27 Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices by Physical Violence (n=320) Reference Group: Exclusive Breastfeeding (n=192) Power Dynamic Variable: ERF 2 (n=10) RRR 3 (95% CI) Mixed Feeding (n=118) RRR (95% CI) w/o status disclosure w/status disclosure w/o status disclosure w/status disclosure Physical Violence 0.77 (0.14 4.17) 0.72 (0.13 4.13) 1.60+ (0.95 2.70) 1.40 (0.82 2.41) Controls: Disclosed Status ----0.46 (0.04 5.29) ----0.29* (0.11 0.73) Age 1.21* (1.04 1.42) 1.21* (1.03 1.42) 1.02 (0.96 1.08) 1.00 (0.94 1.7) Infant age 1.01 (0.75 1.35) 1.01 (0.75 1.35) 1.15* (1.03 1.28) 1.14* (1.02 1.27) Parity 0.50* (0.26 0.98) 0.51* (0.26 1.00) 1.19+ (0.97 1.46) 1.23+ (1.00 1.51) Highest educational attainment 0.85 (0.41 1.73) 0.84 (0.41 1.72) 0.96 (0.77 1.21) 0.95 (0.75 1.20) Knowledge of PMTCT ---3 ----3 0.59*** (0.43 0.81) 0.59** (0.43 0.81) Wealth 1.39 (0.62 3.11) 1.42 (0.63 3.20) 1.09 (0.84 1.41) 1.16 (0.89 1.51) On treatment 2 1.77 (0.33 9.58) 1.83 (0.34 9.95) 0.66 (0.40 1.11) 0.71 (0.42 1.18) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Different sample sizes because women were offered medication differentially 2 Reference group was women on short course prophylaxis 3 Data could not converge (likely due to small sample size) Other Gender Power Dyanmics and Safe Infant Feeding In addition to IPV, I also examined the effect of other power dynamics within couples on adherence to safe infant feeing, including partner control, household decision making, and ever, none of these variables a re significantly associated with infant feeding in either the models with or without HIV status disclosure as a covariate (see Tables 28 30).

PAGE 130

110 Table 28 Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices by Partner Controlling Behavior (n=320) Reference Group: Exclusive Breastfeeding (n=192) Power Dynamic Variable: Exclusive Replacement Feeding (n=10) RRR 3 (95% CI) Mixed Feeding (n=118) RRR (95% CI) w/o status disclosure w/status disclosure w/o status disclosure w/status disclosure Three or more partner controlling behaviors 1.60 (0.41 6.25) 1.57 (0.40 6.22) 1.46 (0.87 2.44) 1.34 (0.79 2.26) Controls: Disclosed Status ----0.59 (0.05 6.67) ----0.27** (0.11 0.70) Age 1.22* (1.04 1.42) 1.21* (1.03 1.42) 1.02 (0.96 1.08) 1.01 (0.94 1.07) Infant age 1.01 (0.75 1.35) 1.01 (0.75 1.35) 1.33* (1.02 1.26) 1.13* (1.02 1.26) Parity 0.50* (0.26 0.97) 0.50* (0.26 0.99) 1.17 (0.95 1.43) 1.21+ (0.98 1.49) Highest educational attainment 0.81 (0.39 1.66) 0.81 (0.39 1.67) 0.94 (0.75 1.19) 0.94 (0.74 1.18) Knowledge of PMTCT ----3 ----3 0.60** (0.44 0.82) 0.60** (0.43 0.82) Wealth 1.49 (0.68 3.28) 1.51 (0.67 3.36) 1.12 (0.85 1.44) 1.17 (0.90 1.53) On treatment 2 1.93 (0.36 10.41) 1.96 (0.36 10.68) 0.66 (0.40 1.11) 0.71 (0.42 1.19) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Different sample sizes because women were offered medication differentially 2 Reference group was women on short course prophylaxis 3 Data could not converge (likely due to small sample size)

PAGE 131

111 Table 29 Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Making (n=320) Reference Group: Exclusive Breastfeeding (n=192) Power Dynamic Variable: ERF 2 (n=10) RRR 3 (95% CI) Mixed Feeding (n=118) RRR (95% CI) w/o status disclosure w/status disclosure w/o status disclosure w/status disclosure Number of household decisions women participate in 1.13 (0.67 1.89) 1.15 (0.68 1.94) 0.89 (0.73 1.07) 0.91 (0.75 1.11) Controls: Disclosed Status ----0.47 (0.04 5.22) ----0.27** (0.11 0.69) Age 1.22* (1.04 1.43) 1.21* (1.03 1.42) 1.01 (0.95 1.08) 1.00 (0.94 1.06) Infant age 1.01 (0.76 1.35) 1.01 (0.76 1.35) 1.14* (1.03 1.27) 1.14* (1.02 1.27) Parity 0.49* (0.25 0.97) 0.50* (0.25 0.99) 1.20+ (0.98 1.46) 1.23* (1.00 1.52) Highest educational attainment 0.83 (0.41 1.69) 0.83 (0.41 1.68) 0.96 (0.76 1.21) 0.95 (0.75 1.20) Knowledge of PMTCT ----3 ----3 0.61** (0.45 0.82) 0.60** (0.44 0.82) Wealth 1.40 (0.62 3.16) 1.43 (0.63 3.24) 1.12 (0.86 1.46) 1.18 (0.90 1.55) On treatment 2 1.73 (0.32 9.36) 1.78 (0.33 9.72) 0.65+ (0.39 1.08) 0.69 (0.42 1.17) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Different sample sizes because women were offered medication differentially 2 Reference group was women on short course prophylaxis 3 Data could not converge (likely due to small sample size)

PAGE 132

112 Table 30 Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding (n=320) Reference Group: Exclusive Breastfeeding (n=192) Power Dynamic Variable: ERF 2 (n=10) RRR 3 (95% CI) Mixed Feeding (n=118) RRR (95% CI) w/o status disclosure w/status disclosure w/o status disclosure w/status disclosure Have greater or equal earnings than husband 2.08 (0.47 9.29) 2.07 (0.46 9.27) 1.27 (0.73 2.22) 1.25 (0.71 2.19) Controls: Disclosed Status ----0.55 (0.05 5.92) ----0.26** (0.10 0.66) Age 1.23* (1.05 1.44) 1.22* (1.04 1.44) 1.02 (0.96 1.08) 1.00 (0.94 1.07) Infant age 0.99 (0.74 1.33) 0.99 (0.74 1.33) 1.13* (1.02 1.26) 1.13* (1.02 1.26) Parity 0.48* (0.25 0.94) 0.49* (0.25 0.96) 1.17 (0.96 1.44) 1.22+ (0.99 1.50) Highest educational attainment 0.81 (0.39 1.65) 0.81 (0.39 1.66) 0.96 (0.76 1.21) 0.95 (0.75 1.20) Knowledge of PMTCT ----3 ----3 0.61** (0.45 0.83) 0.60** (0.44 0.82) Wealth 1.41 (0.63 3.14) 1.43 (0.63 3.24) 1.07 (0.83 1.38) 1.14 (0.88 1.49) On treatment 2 1.79 (0.33 9.64) 1.85 (0.34 10.02) 0.63+ (0.38 1.04) 0.68 (0.41 1.14) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Different sample sizes because women were offered medication differentially 2 Reference group was women on short course prophylaxis 3 Data could not converge (likely due to small sample size)

PAGE 133

113 The Combined Influence of Gender Power Dynamics on Safe Infant Feeding In the combined adjusted multinomial model the only power dynamic that i s significantly associated with safe infant feeding practices is experiencing IPV. W omen who experienced IPV have 3.0 higher adjusted odds of practicing mixed infant feeding rather than excl usive breastfeeding (p<0.001; see Table 31) compared to women who did not experience IPV and net of other power dynamics and covariates including status disclosure Table 31 Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices by Gender Power Dynamics (n=320) Reference Group: Exclusive Breastfeeding (n=192) Power Dynamic Variable: ERF 2 (n=10) RRR 3 (95% CI) Mixed Feeding (n=118) RRR (95% CI) Any IPV 1.28 (0.30 5.57) 2.97*** (1.61 5.48) Three or more partner controlling behaviors 1.87 (0.37 9.52) 0.80 (0.43 1.48) Number of household decisions women participate in 1.33 (0.72 2.44) 0.96 (0.78 1.19) Have greater or equal earnings than husband 2.00 (0.44 9.02) 1.06 (0.59 1.92) Controls: Disclosed Status 0.56 (0.05 6.45) 0.32* (0.12 0.83) Age 1.24* (1.05 1.48) 1.00 (0.94 1.07) Infant age 1.00 (0.75 1.34) 1.16** (1.04 1.30) Parity 0.46* (0.23 0.93) 1.21 + (0.98 1.50) Highest educational attainment 0.78 (0.38 1.60) 0.94 (0.73 1.20) Knowledge of PMTCT ___ 3 0.57** (0.41 0.79) Wealth 1.32 (0.57 3.05) 1.12 (0.84 1.50) On treatment 2 1.87 (0.45 10.12) 0.77 (0.45 1.32) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Different sample sizes because women were offered medication differentially 2 Reference group was women on short course prophylaxis 3 Data could not converge (likely due to small sample size)

PAGE 134

114 HIV Status Disclosure and Safe Infant Feeding Disclosure positive status to her husband is associated with reduced odds of unsafe infant feeding. Table 32 displays that women who di sclosed their HIV status have 74% lower adjusted odds of practicing mixed infant feeding rather than exclusive breastfeedi ng (p<0.01). This association i s also apparent in the models including gender power dynamics displayed above. However, I did not find any evidence that HIV status disclosure is a med iating variable in the relationship between gender power dynamics and non adherence to the safe infant feeding. HIV status disclosure also does not appear over exclusive breastfeeding (see Table 32); however, this may be a result of the small sample size of women practicing exclusive replacement feeding (n=10) leading to low statistical power. D practicing a speci fic infant feeding strategy which in some contexts may be outside of perceived cultural norms M y study provides evidence from Zambia to the growing body of literature disclosure of an HIV infected st atu s (to husbands or others) and following the recommended infant feeding practices in sub Saharan Africa (Doherty et al., 2007; Doherty, Chopra, Nkonki, Jackson, & Greiner, 2006; Fadnes et al., 2010; Farquhar et al., 2001; Madiba & Letsoalo, 2013; M. A. Onon o, Cohen, Jerop, Bukusi, & Turan, 2014; Sibeko et al., 2009)

PAGE 135

115 Table 32 Multinomial Logistic Regression Results for the Odds of Safe Infant Feeding Practices by Gender Power Dynamics (n=320) Reference Group: Exclusive Breastfeeding (n=192) ERF 2 (n=10) RRR 3 (95% CI) Mixed Feeding (n=118) RRR (95% CI) Disclosed Status 0.53 (0.48 5.72) 0.26** (0.10 5.72) Age 1.21* (1.03 1.41) 1.00 (0.94 1.07) Infant age 1.01 (0.76 1.35) 1.14* (1.02 1.26) Parity 0.51* (0.26 1.00) 1.23+ (1.00 1.51) Highest educational attainment 0.83 (0.40 1.69) 0.95 (0.75 1.20) Knowledge of PMTCT ----3 0.60** (0.44 0.82) Wealth 1.48 (0.66 3.31) 1.15 (0.88 1.50) On treatment 2 1.87 (0.35 10.16) 0.68 (0.41 1.14) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Different sample sizes because women were offered medication differentially 2 Reference group was women on short course prophylaxis 3 Data could not converge (likely due to small sample size) Chapter Summary Mixed infant feeding prior to six months of life is common among the women in this study increasing the risk of vertical HIV transmission. The gender power dynamic with the strongest association to unsafe infant feeding i s IPV. Sexual and emotional violence in particular are the type s of IPV that are especially detrimental to safe infant feeding practices Similar to drug adherence, the severity a nd frequency of IPV also reduce the odds of safe infant feeding. This chapter provides add itional support for my hypo thesi s that gender power dynamics influence HIV to protocols across the PMTCT cas cade In the following chapter, I examine whether pediatric HIV testing the last step in the cascade, is similarly affected by gender power dynamic s

PAGE 136

116 CHAPTER VII PEDIATRIC HIV TESTING When PMTCT fails and infants do acquire HIV early infant diagnosis through pediatric HIV testing is critical for an HIV (Becquet & Mofenson, 2008; Edmonds et al., 2011; Prendergas t et al., 2008) Once diagnosed, early initiation of ART can reduce infant mortality by more than 75% among HIV positive infants (Violari et al., 2008) In Zambia, without ART intervention one third of HIV infected infants die by their first birthday and half die by their second birthday (ZMOH, 2010) While PMTCT interventions have tremendously reduced the incidence of pediatric HIV, there is still a large population of children living with HIV in sub Saharan Africa who desperately need pediatric HIV test ing to diagnosis them In Zambia alone, an estimated 120,000 children 26 are HIV positive (UNICEF, 2015). Unfortunately, r ates of pediatric HIV testing remain very low in settings like Zambia. In 2010, WHO estimated that only 28% of HIV ex posed infants in lo w and middle income countries had been tested for HIV within the first two months of life (WHO, 2011) 27 In t he previous chapters I establish that gender power dynamics negatively impact adherence to ARV medication and safe infant feeding practices In addition, studies across sub Saharan Africa report that IPV is associated with decreased HIV testing among women (Maman et al., 2000; Maman et al., 2001) It is plausible that the same barriers that exist for other PMTCT protocols m ay hinder mothers from bringing their children in for pediatric HIV testing On the other hand, pediatric HIV testing is unique compared to the other PMTCT protocols because it is only required periodically and thus 26 Aged 0 to 14 years 27 Rates of test ing for Zambia specifically were not available

PAGE 137

117 does not require daily diligence like m edication or infant feeding. It also takes place at the health center in contrast to medicatio n or infant feeding, which takes place inside the home and are much more visible to family members, most notably the husband The aim s of this chapter are to 1) estimate the rates of pediatric HIV testing among the mother bay pairs in this study and 2) determine if there is a relationship between gender power dynamics within couples and adherence to pediatric HIV testing at six weeks and six months In addition, I examine the relationship between adherence to other PMTCT protocols, including medication a nd safe infant feeding, and pediatric HIV testing. This has significant implications because the women who are non adherent to PMTCT are the population that needs pediatric HIV testing for their infants the most. Review of Methods Pediatric HIV testing, the main outcome of interest in this chapter, was measured using a self reported question on the survey, as well as what was written Under Five Card ( see Chapter Three for details ). For the statistical analysis in this this chapter, I use descriptive and multivariate logistic regression models I also downloaded a user written command for Stata ( (Ender) ) to examine the mediating effect of HIV status disclosure in the relationship between physical violence and the child having any pediatric HIV testing. 28 Adherence to Pediatric HIV Testing Figure 24 displays the proportion of children who received any pediatric HIV test ing, excluding those who were not tested due to missing supplies at the clinic (n=24). A high proportion of infants (79%) were tested for HIV at least once However the rate s of 28 Physical violence in particular emerged indicted a potential mediation with HIV status disclosure in the nested models (see Table 39)

PAGE 138

118 adherence to both pediatric HIV tests were dramatical ly lower (see Figure 23) Of the infants over six mon t hs of age (n=158), only 34% received both the recommended HIV tests at six weeks and six months The low levels of six month pediatric HIV testing is especially concerning given that recent research from Zambia indicates that there is a higher proportion of vertical HIV infections occurring later in the postpartum period (Ngoma et al., 2015; Torpey et al., 2012) Figure 24 P roportion of Children with Any Pediatric HIV Testing (n=296) Figure 25 P roportion of Children Testing at 6 Weeks and 6 Months (n=158) Not Tested 21% Tested 79% Not Tested 66% Tested 34%

PAGE 139

119 Intimate Partner Violence and Pediatric HIV Testing Similar to the previous chapters, IPV significantly reduces the odds that a child will be tested for HIV in both the models with and without HIV status disclosure (see Table 33). IPV, however, does not appear to reduce the odds of the child having both recommended tests. Yet, I interpret this finding with caution because it may be attributable to the small sample size of older infants eligible for both tests (n=158). Table 33 Logistic Regression Results for the Odds of Pediatric HIV Testing by Experiences with Intimate Partner Violence Power Dynamic Variable: Child Ever Tested (n=296) 1 aOR Tested at Both 6 Weeks and 6 Months (n=158) 2 aOR w/o status disclosure w/status disclosure w/o status disclosure w/status disclosure Any IPV 0.36** (0.18 0.72) 0.46* (0.22 0.96) 0.57 (0.28 1.15) 0.78 (0.31 1.28) Controls: Disclosed Status -----19.35*** (6.26 59.83) ----3.12 (0.60 16.21) Age 1.03 (0.95 1.11) 1.07 (0.98 1.17) 0.95 (0.86 1.05) 0.95 (0.86 1.05) Infant age 1.30*** (1.13 1.51) 1.40*** (1.19 1.65) 1.52** (1.11 2.07) 1.56** (1.14 2.15) Parity 1.04 (0.80 1.35) 0.97 (0.73 1.29) 1.13 (0.80 1.59) 1.12 (0.80 1.57) Highest educational attainment 1.11 (0.84 1.47) 1.13 (0.85 1.52) 1.05 (0.73 1.49) 1.06 (0.74 1.51) Knowledge of PMTCT 1.35+ (0.98 1.87) 1.38+ (0.97 1.95) 0.86 (0.55 1.33) 0.86 (0.55 1.34) Wealth 1.17 (0.84 1.63) 1.00 (0.97 1.44) 1.24 (0.87 1.79) 1.19 (0.82 1.72) On treatment 3 1.16 (0.62 2.18) 0.94 (0.47 1.88) 1.02 (0.48 2.15) 0.91 (0.43 1.96) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Excluding women who brought the child for testing, but clinic was missing supplies 2 Excluding women who brought the child for testing, but clinic was missing supplies and children less than six months of age 3 Reference group was women on short co urse prophylaxis After adjusting for HIV status disclosure and other covariates, women who experienced IPV have 54% lower adjusted odds of the child ever being tested for HIV (p<0.05). Including HIV status disclosure in the model slightly reduces the level of

PAGE 140

120 significance for the association between IPV and any pediatric HIV testing, but did not completely account for the association again indicating that status disclosure does not meet the definition of being a mediating factor. The Severity and Fre quency of Partner Violence and Pediatric HIV Testing In contrast to the findings in Chapter Five and Six I did not find evidence of an association between pediatric HIV testing and the number of violent events a woman experienced (see Table 34) The frequency of violence in the past year, however, is marginally associated with reduced odds of the child having both HIV tests (p=0.093; see Table 35). Table 34 Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by The Number of IPV Events Model 1 Model 2 Power Dynamic Variable: Child Ever Tested (n=296) 1 aOR Tested at Both 6 Weeks and 6 Months (n=158) 2 aOR Number of IPV Events 0.87 (0.77 1.05) 0.84 (0.69 1.04) Controls: Disclosed Status 19.22*** (6.10 60.57) 2.76 (0.52 14.72) Age 1.08 (0.98 1.17) 0.95 (0.86 1.05) Infant age 1.39*** (1.18 1.63) 1.52** (1.11 2.09) Parity 0.95 (0.72 1.26) 1.11 (0.79 1.56) Highest educational attainment 1.12 (0.84 1.50) 1.05 (0.73 1.50) Knowledge of PMTCT 1.36+ (0.96 1.93) 0.84 (0.54 1.32) Wealth 0.98 (0.68 1.40) 1.17 (0.81 1.69) On treatment 3 1.00 (0.50 1.98) 0.88 (0.41 1.89) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Excluding women who brought the child for testing, but clinic was missing supplies 2 Excluding women who brought the child for testing, but clinic was missing supplies and children less than six months of age 3 Reference group was women on short cou rse prophylaxis

PAGE 141

121 Table 35 Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by the Frequency of IPV in the Past Year Power Dynamic Variable: Child Ever Tested (n=296) 1 aOR Tested at Both 6 Weeks and 6 Months (n=158) 2 aOR Frequency of IPV Score 0.95 (0.88 1.03) 0.92+ (0.83 1.01) Controls: Disclosed Status 18.84*** (5.95 59.66) 2.71 (0.50 14.58) Age 1.08 (0.99 1.17) 0.95 (0.86 1.05) Infant age 1.39*** (1.18 1.63) 1.52* (1.11 2.09) Parity 0.95 (0.71 1.26) 1.11 (0.79 1.56) Highest educational attainment 1.12 (0.84 1.50) 1.05 (0.73 1.51) Knowledge of PMTCT 1.36+ (0.96 1.92) 0.84 (0.54 1.32) Wealth 0.97 (0.68 1.40) 1.16 (0.80 1.68) On treatment 3 0.99 (0.50 1.97) 0.86 (0.40 1.87) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Excluding women who brought the child for testing, but clinic was missing supplies 2 Excluding women who brought the child for testing, but clinic was missing supplies and children less than six months of age 3 Reference group was women on short course prophylaxis Not only is there a marginally significant relationship between the frequency of IPV in the past year and adherence to both pediatric HIV tests, but there is also evidence of a dose response relationship. Each odds ratio associated with the frequency of IPV in the past year dummy variable is greater than the prior dummy variable (see Figure 26) Lastly, there is also a relationship between IPV related injuries and the child having any HIV testing (see Table 36). Women who experien ced IPV related injuries have 54% lower adjusted odds of the child receiving any pediatric HIV testing (p<0.05) compared to women who did not experience injuries from IPV This relationship, however, did not emerge for both HIV tests at six weeks and six months, which may be attributable again, to the small sample size.

PAGE 142

122 Figure 26 Adjusted Odds Ratio Estimates for Adherence to Both Recommended Pediatric HIV Tests by the Frequency of IPV in the Past Year (n=158) Table 36 Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by Injuries Experienced as a Result of Violence Model 1 Model 2 Power Dynamic Variable: Child Ever Tested (n=296) 1 aOR Tested at Both 6 Weeks and 6 Months (n=158) 2 aOR Injury from violence 0.46* (0.22 0.93) 0.60 (0.26 1.43) Controls: Disclosed Status 18.30*** (5.84 57.31) 3.08 (0.59 16.20) Age 1.07 (0.98 1.17) 0.95 (0.86 1.05) Infant age 1.39*** (1.17 1.63) 1.56** (1.14 2.14) Parity 0.98 (0.74 1.30) 1.11 (0.79 1.57) Highest educational attainment 1.14 (0.84 1.52) 1.06 (0.74 1.52) Knowledge of PMTCT 1.45* (1.02 2.07) 0.91 (0.58 1.42) Wealth 0.97 (0.67 1.39) 0.93 (0.80 1.68) On treatment 3 0.92 (0.46 1.85) 0.93 (0.44 1.99) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Excluding women who brought the child for testing, but clinic was missing supplies 2 Excluding women who brought the child for testing, but clinic was missing supplies and children less than six months of age 3 Reference group was women on short cou rse prophylaxis 0% 20% 40% 60% 80% 100% score of 1 to 5 score of 6 to 10 score of >10 pediatric HIV testing at six weeks and six months

PAGE 143

123 Specific Types of Partner Violence and Pediatric HIV Tesitng Different types of IPV differential ly testing. After adjusting for HIV status disclosure and other covariates, women who e xperienced emotional IPV have 64% lower adjusted odds of their child having any HIV testing (p<0.01; see Table 37). In addition, emotional violence also has a marginally significant relationship with the child having both HIV tests at six weeks and six mon ths (p=0.06). Sexual violence, on the other hand, is not associated with reduced odds of pediatric HIV testing in ether the models including or excluding HIV status disclosure (see Table 38). Table 37 Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by Emotional Violence Power Dynamic Variable: Child Ever Tested (n=296) 1 aOR Tested at Both 6 Weeks and 6 Months (n=158) 2 aOR w/o status disclosure w/status disclosure w/o status disclosure w/status disclosure Emotional Violence 0.34** (0.18 0.63) 0.36** (0.19 0.70) 0.44* (0.21 0.95) 0.48+ (0.22 1.02) Controls: Disclosed Status ----21.61*** (6.94 67.30) ----3.03 (0.59 15.68) Age 1.03 (0.95 1.11) 1.07 (0.98 1.17) 0.94 (0.85 1.04) 0.94 (0.85 1.04) Infant age 1.30*** (1.13 1.50) 1.41*** (1.19 1.66) 1.55** (1.13 2.12) 1.59** (1.15 2.18) Parity 1.03 (0.80 1.35) 0.96 (0.72 1.28) 1.14 (0.80 1.61) 1.13 (0.80 1.60) Highest educational attainment 1.07 (0.81 1.41) 1.11 (0.83 1.49) 1.08 (0.75 1.56) 1.09 (0.76 1.57) Knowledge of PMTCT 1.37+ (1.00 1.88) 1.38+ (0.98 1.96) 0.82 (0.52 1.28) 0.83 (0.53 1.29) Wealth 1.13 (0.81 1.58) 0.96 (0.66 1.39) 1.23 (0.86 1.78) 1.18 (0.82 1.72) On treatment 3 1.17 (0.61 2.22) 0.94 (0.46 1.89) 0.98 (0.46 2.07) 0.88 (0.41 1.90) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Excluding women who brought the child for testing, but clinic was missing supplies 2 Excluding women who brought the child for testing, but clinic was missing supplies and children less than six months of age 3 Reference group was women on short course prophylaxis

PAGE 144

124 Table 38 Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by Sexual Violence Power Dynamic Variable: Child Ever Tested (n=296) 1 aOR Tested at Both 6 Weeks and 6 Months (n=158) 2 aOR w/o status disclosure w/status disclosure w/o status disclosure w/status disclosure Sexual Violence 0.61 (0.33 1.13) 0.88 (0.44 1.75) 0.56 (0.27 1.19) 0.59 (0.28 1.24) Controls: Disclosed Status ----22.42*** (7.18 69.90) ----3.38 (0.66 17.21) Age 1.03 (0.96 1.12) 1.08+ (0.99 1.17) 0.95 (0.86 1.05) 0.95 (0.86 1.05) Infant age 1.31*** (1.13 1.50) 1.40*** (1.19 1.64) 1.53** (1.12 2.09) 1.58** (1.15 2.17) Parity 1.02 (0.79 1.32) 0.96 (0.73 1.27) 1.12 (0.80 1.58) 1.12 (0.79 1.57) Highest educational attainment 1.11 (0.85 1.46) 1.13 (0.85 1.51) 1.02 (0.72 1.45) 1.04 (0.72 1.48) Knowledge of PMTCT 1.30 (0.95 1.80) 1.24+ (0.95 1.91) 0.85 (0.54 1.33) 0.86 (0.55 1.34) Wealth 1.16 (0.84 1.60) 1.35+ (0.95 1.91) 1.23 (0.86 1.77) 1.17 (0.81 1.70) On treatment 3 1.29 (0.69 2.40) 0.99 (0.69 1.41) 0.99 (0.47 2.10) 0.88 (0.41 1.89) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Excluding women who brought the child for testing, but clinic was missing supplies 2 Excluding women who brought the child for testing, but clinic was missing supplies and children less than six months of age 3 Reference group was women on short course prophylaxis Lastly, in the model without HIV status disclosure, physical violence is significantly associated with the odds of the child ever being tested for HIV (see Table 39). In the first model without status disclosure, women who experienced physical IPV have 59 % lower adjusted odds of the child being tested for HIV (p< 0.001) compared to women who have not experience physica l IPV. Once status disclosure i s added to the mode l, however, physical violence i s no longer significantly associated with pediatric HIV testing (p=0.08).

PAGE 145

125 Table 39 Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by Physical Violence Power Dynamic Variable: Child Ever Tested (n=296) 1 aOR Tested at Both 6 Weeks and 6 Months (n =158) 2 aOR w/o status disclosure w/status disclosure w/o status disclosure w/status disclosure Physical Violence 0.41** (0.22 0.77) 0.54+ (0.28 1.07) 0.67 (0.31 1.46) 0.76 (0.40 1.69) Controls: Disclosed Status ----19.82*** (6.40 61.32) ----3.28 (0.63 17.12) Age 1.03 (0.95 1.11) 1.07 (0.98 1.17) 0.95 (0.86 1.04) 0.95 (0.86 1.05) Infant age 1.29** (1.11 1.48) 1.29*** (1.18 1.63) 1.49* (1.10 2.04) 1.55** (1.23 2.13) Parity 1.02 (0.79 1.33) 0.96 (0.72 1.27) 1.11 (0.79 1.56) 1.10 (0.78 1.55) Highest educational attainment 1.09 (0.83 1.43) 1.12 (0.84 1.49) 1.03 (0.72 1.47) 1.05 (0.73 1.49) Knowledge of PMTCT 1.38* (1.00 1.91) 1.40+ (0.99 1.98) 0.87 (0.56 1.35) 0.87 (0.56 1.36) Wealth 1.13 (0.82 1.56) 0.98 (0.68 1.40) 1.24 (0.86 1.78) 1.18 (0.82 0 1.71) On treatment 3 1.21 (0.64 2.27) 0.98 (0.49 1.95) 1.08 (0.52 2.26) 0.96 (0.45 2.03) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Excluding women who brought the child for testing, but clinic was missing supplies 2 Excluding women who brought the child for testing, but clinic was missing supplies and children less than six months of age 3 Reference group was women on short co urse prophylaxis This is the first time in my statistical analysis that HIV status disclosure emerged as a potential mediating factor and warranted further exploration. 29 Through b inary mediation analysis (Baron & Kenny, 1986) I found that the proportion of the total effect mediated by HIV status disclosure on the relationship between physical IPV and any pediatric HIV testing 29 In all of the models in prior chapters adding status disclosure did n ot dramatically change the relationship between the independent power dynamic variable of interest and adherence

PAGE 146

126 is 50%. However, this proportion is not statistically significant (z score: 1.20 p= 0.230 95%CI 0 .32 1.3 4 ). 30 There a re severa l plausible explanations for this null finding. First, there may not be enough statistical power to detect a significant relationship. Alternatively, my proposed theoretical framework may be incorrect with HIV status disclosure not serving as a me diating HIV status disclosure, IPV, particularly physical IPV, and pediatric HIV testing in order to establish more conclusive findings. Other Gender Power Dynamic s and Pediatric HIV Testing Partner control does not appear to have as pronounced an effect on adherence to pediatric HIV testing as IPV. Prior to adjusting fo r HIV status disclosure there i s a marginally significant relationship between experiencing thre e or more controlling behaviors and pediatric HIV testing (OR 0.58, p= 0.10), which disappeared after the inclusion of status disclosure ( OR 0.7, p=0.32; see Table 40). In addition, experiencing three of more partner controlling behaviors is marginally asso ciated with the odds that the child will be tested at six weeks and six months both before and after adjusting for status disclosure (OR 0.48, p=0.06 and OR 0.50, p=0.08 respectively). ently associated with pediatric HIV testing in the models both including and excluding HIV status disclosure (see Table 41). After adjusting for HIV status disclosure and other covariates, for each additional decision that a woman participates in, she has 38% higher adjusted odds of the child ever bein g tested for HIV (p<0.05) and 39% higher adjusted odds of the child having both tests at six weeks and six months (p<0.05). 30 Based on correlation coefficients

PAGE 147

127 Table 40 Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testin g by Partner Control Power Dynamic Variable: Child Ever Tested (n=296) 1 aOR Tested at Both 6 Weeks and 6 Months (n=158) 2 aOR w/o status disclosure w/status disclosure w/o status disclosure w/status disclosure Three or more partner controlling behaviors 0.58+ (0.30 1.11) 0.71 (0.35 1.42) 0.48+ (0.22 1.02) 0.50+ (0.23 1.08) Controls: Disclosed Status ----21.93*** (7.11 67.63) -----3.22 (0.63 16.55) Age 1.03 (0.95 1.11) 1.08 (0.99 1.17) 0.92 (0.83 1.03) 0.93 (0.83 1.03) Infant age 1.30*** (1.13 1.50) 1.40*** (1.19 1.65) 1.59** (1.16 2.18) 1.64** (1.19 2.25) Parity 1.05 (0.81 1.36) 0.97 (0.73 1.29) 1.20 (0.84 1.72) 1.20 (0.84 1.71) Highest educational attainment 1.13 (0.86 1.48) 1.15 (0.86 1.53) 1.06 (0.74 1.52) 1.08 (0.75 1.55) Knowledge of PMTCT 1.34+ (0.97 1.85) 1.37+ (0.96 1.94) 0.88 (0.57 1.37) 0.88 (0.57 1.38) Wealth 1.11 (0.80 1.54) 0.96 (0.67 1.38) 1.22 (0.85 1.77) 1.17 (0.80 1.70) On treatment 3 1.21 (0.65 2.28) 0.98 (0.49 1.96) 1.01 (0.47 2.14) 0.89 (0.41 1.92) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Excluding women who brought the child for testing, but clinic was missing supplies 2 Excluding women who brought the child for testing, but clinic was missing supplies and children less than six months of age 3 Reference group was women on short course prophylaxis

PAGE 148

128 Table 41 Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by Power Dynamic Variable: Child Ever Tested (n=296) 1 aOR Tested at Both 6 Weeks and 6 Months (n=158) 2 aOR w/o status disclosure w /status disclosure w/o status disclosure w/status disclosure Number of household decisions women participate in 1.44** (1.14 1.81) 1.38* (1.07 1.77) 1.40* (1.03 1.91) 1.39* (1.02 1.89) Controls: Disclosed Status -----21.43*** (6.86 67.00) -----3.34 (0.65 17.04) Age 1.04 (0.96 1.13) 1.08+ (0.99 1.18) 0.95 (0.86 1.05) 0.95 (0.86 1.05) Infant age 1.30*** (1.12 1.50) 1.40*** (1.18 1.65) 1.51** (1.11 2.07) 1.56** (1.13 2.14) Parity 1.01 (0.78 1.30) 0.95 (0.72 1.26) 1.11 (0.79 1.57) 1.11 (0.78 1.56) Highest educational attainment 1.10 (0.84 1.46) 1.14 (0.85 1.53) 1.06 (0.74 1.52) 1.07 (0.74 1.54) Knowledge of PMTCT 1.35+ (0.97 1.86) 1.36+ (0.95 1.93) 0.87 (0.56 1.35) 0.88 (0.56 1.36) Wealth 1.04 (0.75 1.46) 0.90 (0.62 1.31) 1.09 (0.74 1.61) 1.05 (0.71 1.56) On treatment 3 1.20 (0.64 2.26) 0.95 (0.48 1.90) 1.01 (0.48 2.13) 0.89 (0.42 1.92) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Excluding women who brought the child for testing, but clinic was missing supplies 2 Excluding women who brought the child for testing, but clinic was missing supplies and children less than six months of age 3 Reference group was women on short co urse prophylaxis also have a marginally significant relationship After adjusting for HIV status disclosure and other covariates, women who have equal or greater earnings than their husba nd have 94% higher adjusted odds of adherence to both pediatric HIV tests (p=0.09) compared to women who earn less than the husband.

PAGE 149

129 Table 42 Adjusted Logistic Regression Models for the Odds of Pediatric HIV Testing by Power Dynamic Variable: Child Ever Tested (n=296) 1 aOR Tested at Both 6 Weeks and 6 Months (n=158) 2 aOR w/o status disclosure w/status disclosure w/o status disclosure w/status disclosure Have greater or equal earnings than husband 1.09 (0.53 2.53) 1.19 (0.54 2.63) 1.91+ (0.90 4.09) 1.94+ (0.90 2.20) Controls: Disclosed Status ----23.37*** (7.61 71.78) ----3.63 (0.72 18.29) Age 1.04 (0.96 1.12) 1.08+ (0.99 1.18) 0.96 (0.67 1.06) 0.96 (0.87 1.07) Infant age 1.29*** (1.12 1.48) 1.39*** (1.18 1.63) 1.51* (1.11 2.06) 1.55** (1.13 2.13) Parity 1.03 (0.79 1.33) 0.96 (0.72 1.27) 1.05 (0.74 1.49) 1.05** (0.74 1.48) Highest educational attainment 1.10 (0.84 1.44) 1.13 (0.85 1.50) 0.99 (0.69 1.42) 1.00 (0.70 1.45) Knowledge of PMTCT 1.32+ (0.96 1.81) 1.35+ (0.95 1.92) 0.86 (0.55 1.35) 0.87 (0.57 1.36) Wealth 1.15 (0.83 1.58) 0.98 (0.68 1.40) 1.23 (0.85 1.77) 1.16 (0.80 1.67) On treatment 3 1.33 (0.72 2.47) 1.02 (0.52 2.03) 1.06 (0.51 2.24) 0.94 (0.44 2.01) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Excluding women who brought the child for testing, but clinic was missing supplies 2 Excluding women who brought the child for testing, but clinic was missing supplies and children less than six months of age 3 Reference group was women on short course prophylaxis The Combined Influence of Gender Power Dynamics on Pediatric HIV Testing The combined models with all gender power measures for this chapter display distinctly different results than what I found in the previous combined models of Chapters Five and Six. T he o nly gender power dynamic that i s significantly associated w ith any ped iatric HIV testing in the combined multivariate models household decisions (see Table 43). For each additional decision that a woman participates

PAGE 150

130 in she has 34% higher adjusted odds of the child receiving any pediatric HIV test ing (p<0.05) net of other power dynamics and covariates. Table 43 Adjusted Logistic Regression Results for the Odds of Pediatric HIV Testing by Gender Power Dynamics Power Dynamic Variable: Child Ever Tested (n=296) 1 aOR Tested at Both 6 Weeks and 6 Months (n=158) 2 aOR Any IPV 0.46+ (0.20 1.05) 0.77 (0.34 1.69) Three or more partner controlling behaviors 0.91 (0.49 2.43) 0.53 (0.22 1.23) Number of household decisions women participate in 1.34* (1.03 1.75) 1.42+ (1.00 2.02) Have greater or equal earnings than husband 1.54 (0.67 3.53) 3.01* (1.27 7.12) Controls: Disclosed Status 18.97 *** (6.03 59.67 ) 3.14 (0.59 16. 92 ) Age 1.08 (0.99 1.18) 0.94 (0.84 1.05) Infant age 1.40*** (1.18 1.65 ) 1.54 (1.11 2.14 ) Parity 0.95 (0.71 1.26 ) 1.11 (0.76 1.61 ) Highest educational attainment 1.14 (0.85 1.54) 1.02 (0.70 1.50 ) Knowledge of PMTCT 1. 37 + (0.96 1.96 ) 0.90 (0.57 1.39 ) Wealth 0.92 (0.63 1.33 ) 1.01 (0.68 1.52 ) On treatment 3 0. 88 (0.43 1.78 ) 0.78 (0.35 1.72 ) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Excluding women who brought the child for testing, but clinic was missing supplies 2 Excluding women who brought the child for testing, but clinic was missing supplies and children less than six months of age 3 Reference group was women on short course prophylaxis Conversely, the only gender power measure significantly associated with the chil d receiving both of the recommended tests at six weeks and six months is economic independence (see Table 43). Women with greater or equal earnings than the husband have 2.9 times higher adjusted odds of the child receiving both HIV tests (p<0.05) compared to

PAGE 151

131 women whose husband has greater earnings and net of other power dynamics and covariates. I also found marginally significant relationships between IPV and any pediatric HIV testing (p= 0.07 ) as well as between household decision making and the child hav ing both pediatric HIV tests (p=0.05 ). My finding s are supported by a recent study by Cook et al. (2011) which found that an independent maternal source of income increases the odds of pediatric HIV testing among HIV exposed children in Mozambique In ad dition, there is evidence from the literature on maternal and child health indicating that when women have control over economic resources instead of men they tend to invest them more into the family and child well being (The World Bank, 2011) HIV Status Disclosure and Pediatric HIV Testing In support of the findings from the previous chapters, HIV status disclosure significantly increases the odds of the child having any H IV testing (see Table 44). Women who disclosure they are HIV positive to the husband have 23.2 times higher adjusted odds of the child having any pediatric HIV testing (p<0.001) compared to women who did not disclose their status. This finding adds support to recent research, which states of stigma and involuntary HIV status disclosure to her family or the community are ba rriers to pediatric HIV testing and the continuation of HIV care for HIV positive mothers and their children (Braitstein et al., 2011; Donahue, Dube, Dow, Umar, & Van Rie, 2012) However, I did not fin d support for a relationship between status disclosure and the likelihood of the child having both pediatric HIV tests, which as discussed previously, may be due to the small sample size.

PAGE 152

132 Table 44 Adjusted Logistic Regression Results for the Odds of Ped iatric HIV Testing by HIV Status Disclosure Model 1 Model 2 Variable: Child Ever Tested (n=296) 1 aOR Tested at Both 6 Weeks and 6 Months (n=158) 2 aOR Disclosed Status 23.20*** (7.55 71.30) 3.62 (0.71 18.38) Age 1.08+ (0.99 1.18) 0.95 (0.86 1.05) Infant age 1.39*** (1.19 1.64) 1.58** (1.15 2.16) Parity 0.96 (0.73 1.27) 1.10 (0.78 1.54) Highest educational attainment 1.13 (0.85 1.50) 1.05 (0.73 1.50) Knowledge of PMTCT 1.35+ (0.96 1.92) 0.87 (0.56 1.35) Wealth 0.98 (0.69 1.41) 1.18 (0.82 1.71) On treatment 3 1.03 (0.52 2.04) 0.95 (0.45 2.03) + p<0.10 p<0.05 ** p<0.01 ***p<0.001 1 Excluding women who brought the child for testing, but clinic was missing supplies 2 Excluding women who brought the child for testing, but clinic was missing supplies and children less than six months of age 3 Reference group was women on short course prophylaxis Adherence to Other PMTCT Protocols and Pediatric HIV Testing Unfortunately, b u t not unexpectedly, women who were non adherent to ARVs and safe infant feeding practices were also less likely to be adherent to pediatric HIV testing in the multivariate models Essentially, this indicates that the same children who are at the greatest r isk of contracting HIV are the least likely to be diagnosed. Table 45 displays the interaction between the PMTCT protocols described in the previous chapters and the main outcome of this chapter, pediatric HIV testing.

PAGE 153

133 Table 45 The Odds of Pediatric HIV Testing by Other PMTCT Protocols Child Ever Tested Tested at Both 6 weeks and 6 months PMTCT Protocol aOR 1 (95% CI) aOR 1 (95% CI) ARV a dherence >80% during Pregnancy 2 5.83*** (2.17 15.65) 11.85* (1.16 121.18) Took sdNVP during Childbirth 3 6.93* (1.03 46.36) 0.41 (0.02 7.60) ARV a dherence >80% during Postpartum 4 8.72*** (2.80 27.12) 11.93* (1.02 139.83) Infant NVP prophylaxis a dherence >80% during Postpartum 5 6.55*** (2.67 16.03) 2.96 (0.80 10.93) Exclusive b reastfeeding 6 2.32* (1.16 4.64) 2.55* (1.20 5.43) 1 Ten separate models using the protocol listed as the main independent variable and c ontrolling for wealth HIV status disclosure and PMTCT regimen (when applicable) 2 n=251 & n=132, respectively (observations dropped include women that were not offered medication and women who reported the clinic was missing supplies for pediatric testing) 3 n=120 & n=55, respectively 4 n=264 & n=147, respectively 5 n=279 & n=152, respectively 6 n=296 & n=158, respectively When women have adequate adherence to medication across the PMTCT cascade, they are also more likely to take the ir child for pediatric HIV te sting, and in some cases, have the child tested at both six weeks and six months. Women with adequate ad herence duri ng pregnancy have 5.8 times higher adjusted odds of their child having any pediatric HIV testing (p<0.0 0 1) and 11.9 times higher adjusted odds of the child hav ing had both tests at six weeks and six months (p<0.05). Women on the short c ourse regimen who took sdNVP have 6.9 times higher adjusted odds of the child having any HIV testing (p<0.05). W omen who with adequate adherence to thei r medication postpartum ha ve 8.7 times higher adjusted odds of the child having any pediatri c HIV testing (p<0.001) and 11.9 times higher adjusted odds of the child having both tests (p <0.05 ). Wo men with adequate adherence to givin g the infant prophylaxis have 6.6 times higher adju sted odds of the child having any HIV testing

PAGE 154

134 (p<0.001). Lastly, w om e n who followed safe infant feeding guideli nes have 2.3 times higher adjusted odds of the child having any ped iatric testing (p<0.05) and 2.6 times higher adjusted odds of the chi ld having both tests (p<0.05). Chapter Summary This chapter highlights the relationship level power dynamics associated with the last step in the PMTCT cascade pediatric HIV testing IPV, particularly emotional IPV, e conomic dependence, and low household decision making ability have the most pronounced detrimental effect on pediatric HIV testing. Of particular concern i s the finding that the infants at greatest risk for HIV due to maternal non adherence to PMTCT protocols are also the least likely t o be tested for HIV. In the following chapter, I use my low power within couples affects adherence to all of the aforementioned PMTCT protocols including ARVs during and aft er pregnancy, safe infant feeding, and pediatric HIV testing

PAGE 155

135 CHAPTER VIII PMTCT IN CONTEXT From the quantitative analysis described in Chapters F ive through Seven, I establish a compelling association bet and non adherence to protocols across the PMTCT cascade. Although these findings contribute to the literature surrounding gender, power, and HIV/AIDS, there are limitations with quantitative data in regard to understanding why such associations exist. In order to address this limitation, I conducted semi structured interviews with a sub sample of women to help explain the associations established from the quantitative data as well as to uncover additional barriers and facil itators that we re not captured on the survey For example, some of the women we interviewed we re able to achieve high PMTCT adherence despite having low power in their relationship. Understanding both barriers and facilitators to PMTCT will help enable effective interventions aimed at imp roving adherence. I originally hypothesize d erence to PMTCT through several mechanisms, such as fear of HIV status disclosure, lack of spousal support, and poor mental health. In this chapter, I explore the data from the semi structured interviews regarding PMTCT related behaviors and the factor s in their lives that influence adherence I also connect the qualitative themes to the quantitative findings in order to present a comprehensive picture of g ender power dynamics within couples and PMTCT adherence Review of Methods While conducting the survey we recruited a sub sample of women to participate in the semi structured interview s based on four previously established domains: 1) women with

PAGE 156

136 high power and high adherence; 2) women with low power and low adherence; 3) women with high power and low adherence; and 4) women with low power and high adh erence. Women were assigned to one of these four categories post hoc based on their responses to the s urvey questionnaire (see Chapter Three for details) The interviews were conduc ted in the local languages and were audio recorded. I use d a thematic framework approach to analyze the translated transcripts, including both deductive and inductive coding ste ps. Qualitative Sample Characteristics Table 46 displays the characteristics of the 32 women who participated in the semi structured interviews, stratified by the power/adherence domains. T he average age of interview participants is 27 years with an average infant age of 5 months. Sixty eight percent of participants completed primary education while 19% completed secondary education. Women own on average 6.7 household assets (out of 21 possible items) Slightly less than one fourth of women a re in a discordant relationship with an HIV negative partner (accordin g to the woman) and 80% report disclosing their HIV positive sta tus to their husband Sixty percent of women report experiencing IPV in their current relationship, while on average wome n report that their partner display s 3.2 controlling behaviors. Lastly, on average women report participating in 50% of the household decisions 31 31 House hold decisions include : healthcare for herself; minor purchases; major purchases; and the final sa y over money.

PAGE 157

137 Table 46 Qualitative Sample Characteristics (n=32) Total High Adherence/ High Power (n=12) Low Adherence/ Low Power (n=12) High Adherence/ Low Power (n=4) Low Adherence/ High Power (n=4) Demographic Characteristics Age (years) 26.9 27 28.1 26.8 23.3 Infant age (months) 4.8 5.5 5.1 5.0 3.8 Parity 2.8 3.2 3.0 2.8 2.3 Completed primary education 68% 75% 81% 67% 50% Completed secondary education 19% 25% 18% 24% 10% Number of assets 1 6.7 8.6 6.5 7.3 4.3 Job past 12 months 79% 73% 80% 79% 82% Sexual Relationship Characteristics Discordant couple 23% 8% 33% 18% 32% Disclosed HIV status 80% 91% 54% 84% 93% Any IPV 59% 58% 63% 66% 50% Mean no. of controlling behaviors 3.2 2.7 3.6 3.3 3.1 No. of decisions women participate 2 2.0 2.5 1.6 1.8 2.1 Greater or equal earnings as husband 33% 25% 30% 50% 25% Total 100% 37.5% 37.5% 12.5% 12.5% 1 Out of a possible list of 21 items 2 Out of four possible decisions High Power and High PMTCT Adherence W omen high power and high adherence repeatedly discuss ed wanting (and being able) to prioritize Many women in this domain understand why high adherence is important and made the conscious decision to follow protocols appropriately in order to protect their children. For example, one 29 yea r old woman e xplains:

PAGE 158

138 [with PMTCT] I am always mindful that at the time I stop taking my drugs, I might bring problems to my breastfeeding child [I take the drugs] so that the HIV will continue to be suppressed in my In addition to women understanding the importance of PMTCT, many women with high power and high adherence also realize the importance of taking medication for their own health motivating high adherence : and I wan ted my child not to be infected with the HIV T hat is why I just accepted. The other thing, I looked at 20 year old woman Additional ly, women in this domain often state d they woul d rather le ave their husband than stop PMTCT When women with high power and hig h adherence we r e asked what t hey would do if the husband did not want them to include statements such as: ARVs then it is the end of our marriage because I do not see myself having a healthy life so what is the 20 year old woman Statements like these are an indication of relative powe r within the couple, and iter ated in correlation with high PMTCT adherence. HIV Status Disclosure to Husbands In addition, women with high power and high adherence often cite HIV status disclosure to the husband as an important component of their PMTCT adherence, similar to the quan titative findings. Disclosure wa s mentioned over and over again as enabling be tter adherence because women do not have to hide their medication from their partner. For example, one 29 year old mother explains:

PAGE 159

139 is easy to give when he is there or when he is not there because when you have told someone the truth you can give anytime Partner Support One significant theme that substantially adds to the quantitative findings is that a PMTCT higher power in the relationship. W omen with high adherence/high power often discuss ed their adherence to various PMTCT protocols. For example, t his co me s in the form of reminding women to take medication: was in the protection of the baby bed without taking the medication. Even if he had gone somewhere with his friends a few minutes before 20 hours you would see him coming back to tell me to take the medication S o it w as from this encouragement that I got used 20 year old woman Second, several women with high power and high adhere nce report that their husbands additionally help them give the medication to the child or collect medication, wh ich aids in their ability to be adherent. One 25 year old woman states: child] and sometimes when I fall asleep he gives the medication O ther researchers have affirmed that partner support is an important factor related to PMTCT adherence (Auvinen et al., 2010; Kiarie et al., 2006) a finding that i s corroborated by the women in my study. More generally, other researchers also report that increased social support among HIV positive women is linked to improved clinical outcomes (Bateganya, Amanyeiwe, Roxo, & Dong, 2015; Gaede et al., 2006) My work adds to this literature by indicating that (e.g., an absence of violent/controlling behavior) is related to male partner support which i n turn, promotes better PMTCT adherence.

PAGE 160

140 In addition, women in my study also report that partner support came in the form of husbands accompanying them to he alth care visits, which results in better adh erence to PMTCT protocols because husbands were aware of the medical recommendations and could promote them in the home For example, a woman with an infant below six months of age wanted to early mix feed the child, but the husband encouraged her to exclu sively breastfeed based on what they were tol d together at the health center: had some abdominal pains because she used to cry a lot and when my husband saw that I was doing th at, he was not happy. He said that no I should not be giving anything besides breast milk to the baby and that even that orange juice can be dangerous for the baby. I said I am doing this because the baby has abdominal pains, but he said o, stop, don that orange juice before 6 months. So I stopped giving orange juice and 20 year old woman Other researchers report that participation in PMTCT improves the ( see Kiarie et al., 2006) ; however, little is known about participation in PMTCT care may be beneficial because he is able to remind or even insist that the mother follow th e medically recommended advice. In addition, as discussed above, men who participate in PMTCT related care may also be more likely to feel comfortable getting medication or gi ving medication to the infant if they have received instructions directly from he alth care workers. HIV Status Concordance The last theme that arose for women with women in the high power and high adherence domain is HIV status concordance (i.e., both partners having HIV) Alt hough the HIV status of husbands d id not predict adherence i n my quantitative analysis, it wa s a re

PAGE 161

141 occurring theme in the semi structured interviews. Women whose partner is also HIV positive report having additional support because they are able to remind each other about medication and clinic visits, r esulting in bette r adherence. One woman explains how having a positive husband has helped her: medication. When I get back from the clinic, I tell my husband what my CD4 count is and m 32 year old woman Low Power and Low PMTCT Adherence Conversely, the narratives of women with low power and low adherence have distinctly contrasting themes to the women with high power and high adherence One of my research assista nts explains this well when she said after an interview with a woman with poor adherence : S he [the participant] had so many social issues and was so angry about the husband that she had too many things to deal with [to be adherence to PMTCT] Women wi th low power and low adherence gi ve an abundance of explanations for why adherence was difficult, ranging from lack of HIV status disclosure due to fear of abandonment or violence to abusive controlling partners and having a lack of social support within t he home Prioritizing the Marriage In contrast to the women with high power in their relationship and high adherence, women with low power and low adherence made statements indicating that keeping their marriage was their priority even over keeping the child or themselves healthy This was especially salient for women on their second or third marriage who did not want to be

PAGE 162

142 abandoned again which is understandable given the strong cultural significance of marriage for Zambian women. One 34 year old woman on her second marriage explains: Non Discl osure of HIV Status their HIV positive s tatus, which negatively affects PMTCT adherence One woman explains why she had not brought her child for any PMTCT care or pediatric HIV testing: Y es it is true, I was not bringing my child to the clinic because I was scared of my husband. From the time I was pregnant, I have been taking ARVs, but I have not told my husband that I am taking ARVs. No wonder that I could not bring the child to the clinic. I was scared that he would ask me why I am bringing the child to the clinic 29 year old woman In addition, fear of violence wa s often stated by the women in this study as an other reason why they do not disclose their HIV status to the husband. For example, o ne 25 year old woman states: this man has a hot temper because when he is angry, he always beats me up. With him, even a little mistake he will beat you up T hat is w hy I was scared to disclose my status to him because this would be a very big issue with The stories from the women in my study help explain the struggles of PMTCT adherence while living with a partner that does not know she is HIV positive. An exa mple of how lack of disclosure affects adherence i s that some women report they c an not take their medication or give medication to the infant in the presence of their husband, having to wait until he i s not around The following quote from a 29 year old wo man with low power and low adherence is illustrative:

PAGE 163

143 appointments because I did not find [a] chance to sneak to this place. It has not been easy for me. At some point I even stopped taking drugs for some time. For me to find time to come here, I have to make sure that he [the husband] has gone somewhere very far and that is how I sneak and come to Several women discuss ed ways of attempting to hide their medication or lie about its purpose to avoid disclosure of their status to the husband, which is an interesting way that some women cope with their situation and exert power. Women mention ed telling the husband that the infant HIV prophylaxis was for cough s or stomachaches In addition, women discuss ed bedrooms. Some women also mention ed house. One 28 year old woman explains that : i f the husband is still around and suffer s During the interview above, the interviewer goe s on to ask the mother if she still goes its that she does not go there during heavy rains house so the baby does not get pr ophylaxis. In order for husbands to be involved in PMTCT care which is important for adherence, they must know the woman is HIV positive. Thus, women who do not disclose their status to male partners not only have to hid e the ARV medication, which makes a dherence difficult, but they also do not benefit from the soc ial support of having a husband involved in the health related behaviors and decisions regarding PMTCT. Unfortunately,

PAGE 164

144 however, status disclosure alone may not be enough to garner partner support in all types of relationships. Lack of Partner Support T here i s a noticeable difference in the amount of partner support women with high power and high adherence discuss ed compared to women with low power and low adherence Women with low power often st ate d that they missed doses of medication because they a re unable to go to the clinic for refills, either because they we re too busy or they we re sick and d id not have anyone to help them. They also discuss ed having the sole responsibility of PMTCT with no help from their partner even if he kn ew HIV status. These interviews a re characterized with statements, such as: not even one day. I am the only one that gives the chi 31 year old woman Discordance and Emotional Abuse Although HIV status disclosure appears to improve adherence among certain relationship s namely those where women have high power, status disclosure can also perpetuate violence for HIV positive women with low power. This is especially salient when the couple is HIV discordant (i.e., husband is HIV negative). Among the women who se partner knew their status several participants state d that serodiscordance within the relationship l ed to emotional abuse, which wa s a barrier to PMTCT adherence. For example, two women with poor adherence explain: [PMTCT] is difficult for me because the person I live with is always tormenting me on the same issues [being HIV 23 year old woman

PAGE 165

145 but it seems he does no 34 year old woman Controlling Behavior Anot her recurring theme that emerged among the women with low power and low adherence i ng violent behavior with their own health behaviors. For example, in order for women to be seen at Kanyama Health Center, they must arrive early morning and wait in a long queue. One woman explains why she has difficulties leaving her house at an early eno ugh time to get to the clinic : [the husband] will say, repare food for me before you do anything. He is self employed, but he always gives me problems before he goes for work. He always lingers in the house and will not leave the house until he is given something to eat. That is when I can find time to leave. Even these scars that you see on my face, he 29 year old woman Following the Advice Lastly, women with low power and low adherence often report listening to their regarding PMTCT which is problematic when t not medically accurate, increasing the risk of vertical HIV transmission For exa mple, a woman who wa s not using any postp artum PMTCT interventions (i.e., medication f or herself o r the infant) states : M because I had taken them when I was pregnant and I was not sick and the 21 year old woman

PAGE 166

146 This problem may be partiall y alleviated if husbands accompanied their wives to antenatal and PMTCT care where they would also hear the app ropriate medical instru ctions; y et another reason to promote partner involvement in PMTCT. High Power and Low PMTCT Adherence Poor Mental Health Despite having high power within their relationships, some women a re still unable to obtain high PMTCT adherence. One particularly salient theme that emerge d among women in this domain i s poor mental health related to having an HIV positive st atus. Some women report ed suicidal thoughts after being diagnosed with HIV, which in turn negatively affect ed their ability to be adherent to PMTCT One 25 year old woman explain s : Poor mental health may thus not only be a concern for women in violent relationships but also for HIV positive women in general due to the internalized stigma associated with being HIV positive. I originally assumed that poor mental health is a mediating factor in the relationship between gender power dynamics and non adherence; however, it may in fact be a more far reaching concern for t he general population of HIV positive women not just those experiencing violence related to the stigma of living with HIV Currently, Zambia and other countries throughout sub Saharan Africa are moving towards Option B+, which is a WHO policy that a dvocates initiating all pregnant or breastfeeding women onto lifelong ART as soon as they are diagnosed with HIV (WHO,

PAGE 167

147 2014) One concern with this policy that health workers in Malawi 32 have raised which my qualitative research also supports, is that wome n often are not adequately prepared to start life long treatment immediately and need time to psychologically adjust before commencing and achieving high ARV adherence (Bedell et al., 2014) This underscores the importance of quality HIV and PMTCT counseli ng, especially for newly diagnosed women as well as quality referral systems for mental health care Maternal Illness Sickness wa s also mentioned as a barrier to PMTCT adherence among women with high power but low adherence. Sickness include s potential o pportunistic infections resulting in wome n not feeling well, in addition to side effects from the ARV medication. The quote below is illustrative of this problem: and I felt dizzy and vomiting and then [on] the se cond [day] 27 year old woman Therefore, adherence to PMTCT medication can suffer when women do not have supportive partners to help them get medication when they are sick (i.e., women with low power and low adherence), but can also be unrelated to dynamics within the couple, such as women experiencing unpleasant side effects of the ARVs and opting to stop taking them Poor Understanding of Protocols and Erratic Supply of Medication Non adherence despite high power in the co uple i s also attributable to poor understanding of the PMTCT protocols, especially breast feeding. A compounding factor i s an erratic supply of infant prophyl axis at the clinic. Women on the sh ort course prophylaxes regimen we re supposed to provide ARV prophylaxis to the ir infant as long as she/he 32 Malawi was the first country to introduce Option B+ and has been leading the way in research regarding this new PMTCT policy.

PAGE 168

148 breastfed but when the medication is out of stock, the child i s essentially u nprotected because the mother i s not on ARVs herself past one week postpartum. Numerous women explain ed that they were not follow ing the exclusi ve breastfeeding recommendations because they we re worried about transmitting the virus to their child. For example one 21 year old woman explains: major reason why I stopped breastfee ding the baby and the other reason was that the medication had started running out of stock at the clinic and so I thought I just stop breastfeeding my baby so that he does not get infected with Family I nfluence Family influence has previously been associated with HIV decisions regarding how to feed their infant (Buskens et al., 2007; Doherty et al., 2007) The present study also provides support that the family, incl uding the husband and others, can h ave a negative or positive influence over PMTCT adherence, particularly in rega rd to infant feeding decisions, as is de scribed in the conversation below : Interviewer: baby at 3 months of age. Is there any reason why you did not wait until 6 Participant: so when my mother saw that the baby is crying she is the one that told me that the baby does not get satisfied with breast milk alone and she is t 29 year old woman Low Power and High PMTCT Adherence Family Support The women who have low power in their relationship but achieve d high PMTCT adherence all had one key factor in common: support from other family members. Despite

PAGE 169

149 having low relative power to the husband women with encou raging family members (e.g., aunt s or sisters) report ed being able to achieve high PMTCT adherence, which is highlighted in the following quote: heartbroken that now I have to take this medication for life. Why me? But my sisters were the ones that keep advising me and encouraging me to concentrate o n my kids. But right now I am just fine. I have even stopped 25 year old woman Child Death An additional theme that emerge d among some women with low power and high adherence i s the motivation to follow PMTCT instructions with their youngest child because an older child ha d died. This is a particularly interesting theme because it is something that the survey would never have been able to capture. Women who were non adherent with an older HIV exposed child due to low relati ve power within the couple or for other reasons, who se child subsequently died report being They we re often able to overcome their low power within the couple because of such determination and the guilt associa ted with loo sing a prior child A 23 year old woman explains her medication adherence during pregnancy despite low power and emotional abuse from the husband: so even though my husband used to call me names, he called me a prostitute, but I still used to hide and come here to get the medication to protect my unborn child one year but then he passed away so now eve n if he [the husband] refuses M any of the women in this study (40%) had a previous child die, which is unfortunate and devastating for the women but may motivate them to have better adherence to PMTC T

PAGE 170

150 medical protocols with their youngest child in order to prevent another child death This is also an additional example of some w their behavior s even in adverse household environments Chapter Summary This chapter uses data from the semi structured interviews to shed light on the mechanisms underlying the relationship between gender power dynamics and PMTCT adherence. Major themes that emerge d are the importance of HIV status disclosure and partner supp ort as well as the negative effect of poor mental health. Two facilitating factors that enable relative power are family support and having a previous child die. In the following chapter, I summarize my original research aims and discuss recommendations to address non adherence to PMTCT based on my quantitative and qualitative findings.

PAGE 171

151 CHAPTER IX CONCLU SIONS At the beginning of this study, I set out to answer three specific research questions regarding gender power dynamic s within couples and HIV PMTCT protocol s in Lusaka, Zambia. Based on the findings discussed in the previous chapters, I summarize below the answer to each of my original research questions. 1. A ffect PMTCT Adherence? My findings indicate, without a doubt, that yes, couples negatively affects adherence to PMTCT protocols. I have established a compelling association between gender power dynamics within couples (i.e., IPV, partner control, household decision making, and economic dependence) and non adherence to various protocols across the PMTCT cascade However, there is a multifaceted relationship between specific gender power dyn amics and particular PMTCT protocols. 1. a. Which Power Dynamics A re the Most Detrimental ? adherence across the PMTCT cascade of care. Women who experienced IPV from their current husband have reduced odds of PMTCT adherence to almost every necessary protocol, with the exception of sdNVP during childbirth. Given the high prevalence of both HIV and IPV against women i n Zambia, this finding is extremely important. I additiona lly fou nd evidence of a dose response a ffect of IPV, where women who experience more violent events and a greater frequency of violence in the past year have reduced odds of adherence to the majority of PMTCT protocols. Finally, the severity of physical/se xual violence indicated through injuries is also detrimental to medication adherence during and after pregnancy and

PAGE 172

152 the child having any pediatric HIV testing but is not associated with safe infant feeding The relationship between experiencing IPV and non adherence remains significant even after adjusting for HIV status disclosure and other gender power dynamics, with the exception of pediatric HIV testing. 33 M y research also underscores the importance of examining IPV as a multidimensional phenomenon. I ndeed, d ifferent types of IPV have differing levels of effect on non adherence to PMTCT Emotional violence in particular has the most pronounced effect on non adherence reducing the odds of adherence to all protocols except sdNVP. 34 My qualit ative data a dditionally support the link between emotional violence and non adherence to PMTCT, with women stating, for example, that it is difficult to adhere when they were by their husband. I suspect that emotional violence in particular increased emotional distress and potentially depression, which are known barriers to medical compliance (Ellsberg et al., 2008; Fischbach & Herbert, 1997; Murray et al., 2006). Further research is neede d to formally estblish a connection, however, between IPV, poor mental health, and non adherence to HIV care, including PMTCT. infant feeding practices, but did not affect the odds of pediatric HIV testing. P hysical violence, on the other hand, has a much less pronounced effect on PMTCT than emotional or 33 In the combined gender power dynamic models, IPV is no longer significantly associated with the child having any pediatric HIV testing like it is prior to including the three other gender power dynamic measures. 34 Emotional violence has a marginally sign ificant relationship with adherence during pregnancy pediatric HIV testing at both six weeks and six months after adjusting for status disclosure.

PAGE 173

153 sexual violence. Women who experienced physical violence only had reduced odds of adherence to medication postpartum and t he child receiving any pediatric HIV testing 35 A is also as an important predictor for certain PMTCT protocols Male controlling behavior has not received the same attention in the literature as IPV, particularly physical and sexual IPV (Abramsky et al., 2011; Harling et al., 2010; Pallitto et al., 2013) which may limit our ability to fully understand and address gender power dynamics within couples affected by HIV. This is the first study to my knowledge to exami ne an HIV related behavioral outcome and its relationship to male partner controlling behavior in addition to IPV. The findings of this study indicate that w omen who experienced three or more partner controlling behaviors have reduced odds of adherence to medication postpartum (marginally significant), to giving the infant prophylaxis, and any pediatric HIV testing (marginally significant) after adjusting for HIV status disclosure and other covariates. More qualitative research is called for however, in o rder to better understand why both IPV and controlling behavior impact certain protocols, while only IPV is significantly associated with others. Indeed, there is a strong correlation between these two variables almost all IPV (98%) occurred with the con comitance of controlling behavior among my participants This is an i ndication of a specific typology of IPV coercive controlling violence, which tends to result in more injuries and psychological distress compared to situational couple violence. Althoug h there is a strong correlation between IPV and control, my data indicate that IPV is likely driving the associations with PMTCT adherence. For example, in the combined models with all gender power dynamics, the relationship between partner control and non adherence disappears with most of the PMTCT protocols, with the exception of giving the 35 The association with pediatric HIV testing is only marginally significant.

PAGE 174

154 infant prophylaxis, while most of the relationships remain significantly associated with IPV. Due to the high collinearity of these two variables, however, I make my c onclusions regarding IPV and control with caution it may not be possible tease apart the separate effects of spousal violence or control because of their strong overlap. ld decision making. The number of household decisions that a woman participates in is not consistently associated with adherence across the PMTCT but was significantly associated hold decision making is associated with higher odds of adherence to postpartum medication, giving the infant prophylaxis, the child having any pediatric HIV testing, and the child being tested at both six weeks and six months. These relationships remained significant after adjusting for HIV status disclosure, but disappeared in the combined models with all gender power dynamics, with the exception of infant prophylaxis and pediatric HIV testing 36 37 was the gender power dynamic with the fewest significant relationships to PMTCT adherence However, this variable was unique in that it displayed a negative association with adherence to infant prophylaxis but a positive association with pediatric HIV test ing. T his variable is also the only gender power dynamic that remained significantly associated with pediatric HIV testing at both the recommended times in the combined mo del with all other power dynamics. to believe that different barriers exist for PMTCT protocols that take place in the home (i.e., 36 In the combined models househ old decision making is significantly associated with any pediatric HIV testing and marginally associated with the child have both recommended tests. 37 Measured as a binary variable with 1) women who have equal or greater earnings as the husband, or 0) wom en whose husband has greater earnings.

PAGE 175

155 medication and infant feeding ) versus those that take place at the health center (i.e., pediatric testing). Thus, may provide the necessary resources for women to travel to the clinic periodically for procedures, such as pediatric HIV testing; however, their higher earning may also emasculate men in th e home leading to increased IPV/control and subsequently, reduced adherence to some medication. Again, greater research i certain PMTCT protocols while h indering others. 1. b. Which PMTCT Protoc ols A re the Most Affected? In support of my a priori hypothesis, PMTC T protocols during the postpartum time period are the most affected by gender power imbalances In particular, w to giving the infant prophylaxis i s negatively affected by all gender power dynamics in the separate regression models and by the vast majority of power dynamics in the combined regression models. 38 I suspect that it may be easier for women to make autonomous decisions regarding their own medication than it is for giving the infant medication women may be approval. Indeed, recent research from Tanzanian found that mothers preferred to take medication themselves postpartum rat her than give medication to their infant (Ngarina et al., 2014). Under Option A in Zambia, infant prophylaxis syrup i s given twice da il y for at least six weeks postpartum and for some, throughout the entire breastfeeding period. 39 This extended duration of giving medication to the infant may explain why this protocol was the 38 income. 39 Women on lifelong treatment gave infant NVP for six weeks while women on short course prophylaxis g ave it until one week after breastfeeding cessation.

PAGE 176

156 most vulnerable to non adherence when women have limited household decision making ability or experience high levels of partner control /violence Fortunately, w recent transition to the Option B + policy the required length of giving infant prophylaxis is now six weeks for all HIV exposed infants which may help improve adherence to this protocol. Of course, prioritizing gender power dynamics should still be recommended despite these promising policy improvements as they do not address the core issues preventing PMTCT adherence 2. What is the Role of HIV Status Disclosure? My a priori hypo thesis wa s that HIV status disclosure is a mediating factor in the relationship between IPV and non adherence to PMTCT. H owever, my data ultimately did not find support for this hypothesis. The relationship between gender power dynamics and non adherence to PMTCT remains significant even after c ontrolling for HIV status disclosure in the quantitative models. Nonet heless, my findings indicate that HIV status disclosure i s an extremely important predictor of adherence to PMTCT across the cascade of care. This finding adds support from the Zambian context to what other researchers have found regarding the importance of status disclosure for PMTCT medication adherence (Auvinen et al., 2010; Jasseron et al., 2011; Theuring et al., 2009) and safe infant feeding (Doherty et al., 2007; Doherty, Chopra, Nkonki, Jackson, & Greiner, 2006; Fadnes et al., 2010; Farquhar et al., 2001; Madiba & Letsoalo, 2013; M. A. Onono et al., 2014; Sibeko et al., 2009) Additionally, my findings are the first to my knowledge to report a significant quantitative relationsh ip between HIV status disclosure to husbands and adher ence to pediatric HIV testing. N ot only is HIV status disclosure significantly associated with all

PAGE 177

157 PMTCT protocols (both in the models with and without gender power dynamics) in the quantitative data, b ut it i s also a re occurring theme in the semi structured interviews. Unfortunately, the interviews also suggest that status disclosure can have a negative affect on PMTCT adherence as a result of emotional abuse among certain types of couples, namely, ser odiscordant partners where women have low power in the relationship Overall, disclosure rates to husbands among the participants in the study were surprisingly high (92% ) Two recent studies however, have also reported similarly high rates of HIV statu s disclosure to male partners (Jasseron et al., 2011; Shamu et al., 2014) Notably, in the Jasseron et al. (2011) study, only two thirds of single women disclosed they were HIV positive to the father of their child whereas 94% of women who were living wit h the father disclosed their status. All of the women in my study were either married or living with a man as if married, which is likely why I found such high rates of HIV status disclosure. In addition, Kanyama Health Center ha s been actively promoting c ouples HIV testing during ANC I generally observed about half of the pregnant women attending ANC on any given day accompanied by their male partner and tested together for HIV This is promising, given the significa nt role of status disclosure for PMTCT adherence. 3. Why D o Gender Power Dynamics Affect PMTCT Adherence ? My final research aim wa PMTCT related behavior. Not only do the semi structured interviews help explain why gender power dynamics within couple s affect PMTCT a dherence, but they also establish additional facilitators and barriers related to adher ence. I hypothesize d that fear of violence or abandonment from the husband would prevent women from disclosing their status, which would subsequently negatively impact P MTCT adherence. This theme did indeed come out in the interviews. Women

PAGE 178

158 mention fear of the ir husband knowing their status as a critical barrier to adherenc e. These women often prioritize keeping their marriage above all else, even at the expense of PMTCT adherence. I also hypothesize d based on the previous literature, that women with low relative power within a couple will also have less s upportive partners, which in turn would negati vely affect adherence. Again, this i s an important theme that emerged in the qualitative analysis. Women with high power in their relationship an d high PMTCT adherence discuss numerous aspects of partner support, which ena ble d high PMTCT adherence. Convers ely, women with low power and low adherence report receiving little social support from their spouses. Poor mental health is another important theme in the semi structured interviews, which is talked about by women in relation to and aside from dynamics with their husband s. Women with low power discuss verbal abuse from husbands, often related to being in a discordant relatio nship, which negatively impacts their ability to b e adheren t. Women with high power, however, also discuss being depressed as a resu lt of their HIV positive diagnosis, which led to suboptimal adherence Finally, two key facilitating factor s we re discussed in the semi structured interviews, which help highlight why some women with low power a re able to navigate PMTCT adherence despite u nfavorable dynamics with the husband First, supportive family members are which can offset some of the negative affects of their low power in the marriage A second highly motivating factor for women to be adherent is if they had an older child pass away. Women discuss believing the child died as a result of their non adherence and with their youngest child they were determined not to let that happen again despite low power in the relationship.

PAGE 179

159 ab ility to be adherent despite violence or other gender power dynamics in the relationship is regards to PMTCT, even if it means lying or hiding drugs from the husband. Final Conceptual Model By combining my quantitative findings with the themes that emerged in the qualitative data, I updated my a priori conceptual model to display the critical factors linked to cascade of care (Figure 27 ). First, constructs from the Theory of Gender and Power provide the structural framework, followed by critical interpersonal process, including gender power dynamics. Lastly, intrapersonal factors both in relation to and aside fr om gender power dynamics provide the most proximal influences over care. Figure 27 Final Conceptual Model

PAGE 180

160 Addressing Gender Power Imbalances and P MTCT Adherence a large concern in its own right. My s tudy indicates, moreover, that gende r power dynamics within couples also play s a significant role in maternal non adherence to PMTCT putting HIV exposed infants at increased risk of contracting the virus Thus, g reater attention is needed both at the policy and institutional levels prioritizing IPV against HIV positive women in order to achieve optimal health for women and their families. Many of the risk factors associated with IPV one of the most critical gender power dynamics in terms of PMTCT adherence, are symptoms of persistent gender inequity. For or victimization at the individual level (WHO, 2012b) which could be addressed through greater efforts to empower girls and young women through school and community based interventions. At the relationship level, male dominance, which is related to the st ructure of cathexis (i.e., unequal gender norms), is a risk factor associated with IPV perpetration (WHO/LSHTM, 2010) and boys to promote nonviolence and gender equality including helping wo men gain economic opportunities and have decision making power within the household, will also be key intervention s starting with young school aged children to evolve the gendered cultural norms and the overall status of women in society and in the home In addition, at the facility level, PMTCT counseling should attempt to evaluate home circumstances and determine who may be at risk for low compliance because of low power within their household Questions regarding the home environment and whe ther IPV and controlling behavior occur should be standard in PMTCT counseling. An

PAGE 181

161 appropriate referral system between PMTCT services and mental healt h and victim support services should also be prioritized at health clinics, such as Kanyama. Another poten tial intervention that has been shown to improve PMTCT adherence in sub Saharan Africa is positive pregnant and breastfeeding women (Besser, 2010) although more research is needed to establish whether such mentorship programs can counter the effects of r during ANC eliminating the problem of non disclosure to male partners among women as well as counseling related to IPV and conflict resolution. Currently, Kanyama Health Center is actively trying to encourage couples to test together during ANC, but th ere is no counseling regarding conflict resolution or referrals to mental health specialists. The family centered approach should also include HIV testing for male partners, initiation onto ART for both partners if HIV positive and linkages to family plan ning for the couple Indeed, during the semi structured interviews women often mentioned that having a partner also taking medication was helpful as they were able to support one another through reminders and discussion of HIV related health issues. Wett stein et al. (2012) report that using a family centered approach in Ivory Coast helped achieve over 90% initiation of PMTCT among HIV positive women At the structural level, although much harder to implement and evaluate, a systems approach should inclu de creating equitable gender norms reducing poverty improving the low social and economic status of women reforming legal frameworks (e.g., marital rape

PAGE 182

162 establishing community sanctions against IPV and progressing wo (WHO, 2012b) in some cases actually increase their risk of experiencing IPV (Kishor & Johnson, 2004; Vyas & Watts, 2009b) which could inadvertently hurt HIV relate d outcomes, such as PMTCT adherence similar changes for men, can threaten masculine ideals held by male partners (Farr, 2013; Macmillan & Gartner, 1999). Interventions aimed at add ressing power dynamics within couples affected by HIV/AIDS should thoroughly establish when and where providing opportunities, such as microfinance, may in turn decrease adherence to PMTCT and HIV care due to increased exposure to IPV. There appears to be a delicate balance that must be met for impoverished women in sub Saharan Africa, where they are able to access some economic resources but do not cause too much tension and resentment within the household. On a positive note, based on a comparison of the 2007 and 2013 2014 ZDHS, it For example, in the 2007 ZDHS, only 37% of women reported participating in all four hou sehold decisions, which increased to 53% in 2013 2014. In 2007, 62% of Zambian women agreed a husband is justified in beating his wife for at least one specified reason, which fell to 47% in 2013 2014. Lastly, in 2007, 54% of Zambian women reported experie ncing some form of IPV, which was reduced to 47% in 2013 2014 (CSO, 2009; CSO, improve the status of women is moving the country in a positive direction. I ndeed, i n 2011, Zambia implemented two positive legal acts: the Anti Gender Based Violence Act the Domestic Violence Act n addition,

PAGE 183

1 63 National Gender Policy Hopefully the country will continue prioritizing gender equality both on paper and through sustained on the ground efforts, especially given the far non adherence to PMTCT. Next Steps During the formative phases of my research, I was asked repeatedly by colleag ues data, takes off, and is never heard from again by the communities she/he worked with). In addition to publishing articles, I also have a strong sense of oblig ation to disseminate my findings with health practitioners and policy makers in Lusaka. Within the next year, I plan to meet with Zambian stakeholders at various governmental levels to highlight my research findings and make policy recommendations. By init iating this project, I made a commitment to continue helping the community I worked with in any capacity that I can, which at this point means sharing my accumulated knowledge with anyone who is inte rested and could benefit from the findings. I also pla n to initiate a follow up research project within my next position Through the processes of completing this dissertation research, I have become increasingly interested related behaviors, especially women in violent/controlling relationships. I would like to develop a larger study aimed at establishing what relationship level factors hurt or promote mental health and how this in turn affects various HIV related health behaviors, such as HIV testing a nd treatment, and PMTCT. This information can then be used to develop and test promising interventions aimed at retention in HIV care and overall quality of life

PAGE 184

164 REFERENCES Abramsky, T., Watts, C. H., Garcia Moreno, C., Devries, K., Kiss, L., Ellsberg, M., Heise, L. (2011). What factors are associated with recent intimate partner violence? findings from the WHO multi country study on women's health and domestic violence. BMC Public Health, 11 109. doi:10.1186/1471 2458 11 109 Acock, A. C. (2010). A Gentle Introduction to S tata (3rd ed.). College Station, Tex.: Stata Press. Alexander, C. S., Asselin, J. J., Ting, L. S., Montaner, J. S., Hogg, R. S., Yip, B., Harrigan, P. R. (2003). Antiretroviral concentration s in untimed plasma samples predict therapy outcome in a population with advanced disease. J Infect Dis, 188 (4), 541 548. doi:10.1086/376835 Anderson, K. L. (2002). Perpetrator or victim? Relationships between intimate partner violence and well being. Jour nal of Marriage and the Family, 64 (4), 851 863. Auvinen, J., Suominen, T., & Valimaki, M. (2010). Male participation and prevention of human immunodeficiency virus (HIV) mother to child transmission in Africa. Psychology, Health & Medicine, 15 (3), 288 313 doi:10.1080/13548501003615290 Awiti Ujiji, O., Ekstrom, A. M., Ilako, F., Indalo, D., Wamalwa, D., & Rubenson, B. (2011). Reasoning and deciding PMTCT adherence during pregnancy among women living with HIV in Kenya. Cult Health Sex, 13 (7), 829 840. doi:1 0.1080/13691058.2011.583682 Bahl, R., Frost, C., Kirkwood, B. R., Edmond, K., Martines, J., Bhandari, N., & Arthur, P. (2005). Infant feeding patterns and risks of death and hospitalization in the first half

PAGE 185

165 of infancy: multicentre cohort study. Bull World Health Organ, 83 (6), 418 426. doi:/S0042 96862005000600009 Barfield, T. (Ed.) (1997) The Dictionary of Anthropology. Malden, MA: Blackwell Publishing. Barigye, H., Levin, J., Maher, D., Tindiwegi, G., Atuhumuza, E., Nakibinge, S., & Grosskurth, H. (2010). Operational evaluation of a service for prevention of mother to child transmission of HIV in rural Uganda: barriers to uptake of single dose nevirapine and the role of birth reporting. Trop Med Int Health, 15 (10), 1163 1171. doi:10.1111/j.1365 3156.2010.0 2609.x Baron, R. M., & Kenny, D. A. (1986). The moderator mediator variable distinction in social psychological research: conceptual, strategic, and statistical considerations. J Pers Soc Psychol, 51 (6), 1173 1182. Bateganya, M., Amanyeiwe, U., Roxo, U., & Dong, M. (2015). Impact of support groups for people living With HIV on clinical outcomes: a systematic review of the literature. J S368 S374. Becquet, R., & Mofenson, L. M. (2008). Early antiretroviral therapy of HIV infected infants in resource limited countries: possible, feasible, effective and challenging. AIDS, 22 (11), 1365 1368. doi:10.1097/QAD.0b013e32830437f5 Bedell, R., Berman, J., Fabi an Cataldo, Landes, M., Lettow, M. v., & Oosterhout, J. v. (2014). Targeting Research for the Elimination of Mother to Child HIV Transmission + Era Retrieved from DignitasInternational.org:

PAGE 186

166 Besser, M. (2010). Preventing mother to chil d HIV transmission in Africa using new paradigms in health care delivery. Retrieved from http://www.m2m.org/media/publications.html Bezner Kerr, R., Dakishoni, L., Shumba, L., Msachi, R., & Chirwa, M. (2008). "We grandmothers know plenty": breastfeeding, complementary feeding and the multifaceted role of grandmothers in Malawi. Soc Sci Med, 66 (5), 1095 1105. doi:10.1016/j.socscimed.2007.11.019 Bhandari, N., Kabir, A. K., & Salam, M. A. (2008). Mainstreaming nutrition into maternal and child health program mes: scaling up of exclusive breastfeeding. Matern Child Nutr, 4 Suppl 1 5 23. doi:10.1111/j.1740 8709.2007.00126.x Blanc, A. K. (2001). The effect of power in sexual relationships on sexual and reproductive health: an examination of the evidence. Stud Fam Plann, 32 (3), 189 213. Boy, A., & Salihu, H. M. (2004). Intimate partner violence and birth outcomes: a systematic review. Int J Fertil Womens Med, 49 (4), 159 164. Braitstein, P., Songok, J., Vreeman, R. C., Wools Kaloustian, K. K., Koskei, P., Walus una, L., Yiannoutsos, C. (2011). "Wamepotea" (they have become lost): outcomes of HIV positive and HIV exposed children lost to follow up from a large HIV treatment program in western Kenya. J Acquir Immune Defic Syndr, 57 (3), e40 46. doi:10.1097/QAI .0b013e3182167f0d Broderick, C. (1993). Understanding Family Process: Basics of Family Systems Theory Thousand Oaks, CA: Sage Publications.

PAGE 187

167 Buskens, I., Jaffe, A., & Mkhatshwa, H. (2007). Infant feeding practices: realities and mind sets of mothers in Sou thern Africa. AIDS Care, 19 (9), 1101 1109. doi:10.1080/09540120701336400 Campbell, J., Baty, M., Ghandour, R., Stockman, J., Francisco, L., & Wagman, J. (2008). The intersection of intimate partner violence against women and HIV/AIDS: a review. Int J Inj C ontr Saf Promot, 15 (4), 221 231. doi:10.1080/17457300802423224 Campbell, J., Jones, A. S., Dienemann, J., Kub, J., Schollenberger, J., O'Campo, P., Wynne, C. (2002). Intimate partner violence and physical health consequences. Arch Intern Med, 162 (10) 1157 1163. Capaldi, D. M., & Owen, L. D. (2001). Physical aggression in a community sample of at risk young couples: gender comparisons for high frequency, injury, and fear. J Fam Psychol, 15 (3), 425 440. Chopra, M., & Rollins, N. (2008). Infant feedin g in the time of HIV: rapid assessment of infant feeding policy and programmes in four African countries scaling up prevention of mother to child transmission programmes. Arch Dis Child, 93 (4), 288 291. doi:10.1136/adc.2006.096321 CIA. (2014). Worldfactboo k: Zambia. Retrieved from https://www.cia.gov/library/publications/the world factbook/geos/za.html CIDRZ. (2012). Prevention of mother to child HIV transmission (PMTCT). Centre for Infectious Disease Research in Zambia. Retrieved from http://www.cidrz.o rg/pmtct Conkling, M., Shutes, E. L., Karita, E., Chomba, E., Tichacek, A., Sinkala, M., Allen, S. A. (2010). Couples' voluntary counselling and testing and nevirapine use in antenatal

PAGE 188

168 clinics in two African capitals: a prospective cohort study. J In t AIDS Soc, 13 10. doi:10.1186/1758 2652 13 10 Connell, R. (1987). Gender and power : society, the person, and sexual politics Stanford, CA: Stanford University Press. Cook, R. E., Ciampa, P. J., Sidat, M., Blevins, M., Burlison, J., Davidson, M. A., Moon, T. D. (2011). Predictors of successful early infant diagnosis of HIV in a rural district hospital in Zambezia, Mozambique. J Acquir Immune Defic Syndr, 56 (4), e104 109. doi:10.1097/QAI.0b013e318207a535 Coovadia, H. M., Rollins, N. C., Bland, R. M., Little, K., Coutsoudis, A., Bennish, M. L., & Newell, M. L. (2007). Mother to child transmission of HIV 1 infection during exclusive breastfeeding in the first 6 months of life: an intervention cohort study. L ancet, 369 (9567), 1107 1116. doi:10.1016/S0140 6736(07)60283 9 Coutsoudis, A., Pillay, K., Kuhn, L., Spooner, E., Tsai, W. Y., & Coovadia, H. M. (2001). Method of feeding and transmission of HIV 1 from mothers to children by 15 months of age: prospective c ohort study from Durban, South Africa. AIDS, 15 (3), 379 387. Coutsoudis, A., Pillay, K., Spooner, E., Kuhn, L., & Coovadia, H. M. (1999). Influence of infant feeding patterns on early mother to child transmission of HIV 1 in Durban, South Africa: a prospe ctive cohort study. South African Vitamin A Study Group. Lancet, 354 (9177), 471 476. Creek, T. L., Alwano, M. G., Molosiwa, R. R., Roels, T. H., Kenyon, T. A., Mwasalla, V., Kilmarx, P. H. (2006). Botswana's Tebelopele voluntary HIV counseling and t esting network: use and client risk factors for HIV infection, 2000 2004. J Acquir Immune Defic Syndr, 43 (2), 210 218. doi:10.1097/01.qai.0000230525.71717.5d

PAGE 189

169 CSO. (2009). Zambia Demographic and Health Survey 2007 Calverton, Maryland, USA. CSO. (2014). Za mbia Demographic and Health Survey 2013 14 Rockville, Maryland, USA. Dahl, V., Mellhammar, L., Bajunirwe, F., & Bjorkman, P. (2008). Acceptance of HIV testing among women attending antenatal care in south western Uganda: risk factors and reasons for test refusal. AIDS Care, 20 (6), 746 752. doi:10.1080/09540120701693990 Daly, J., Kellehear, A., & Gliksman, M. (1997). The public health researcher: A methodological approach. Mel bourne, Australia: Oxford University Press. de Vries Robbe, M., March, L., Vine n, J., Horner, D., & Roberts, G. (1996). Prevalence of domestic violence among patients attending a hospital emergency department. Aust N Z J Public Health, 20 (4), 364 368. Desclaux, A., & Alfieri, C. (2009). Counseling and choosing between infant feeding options: overall limits and local interpretations by health care providers and women living with HIV in resource poor countries (Burkina Faso, Cambodia, Cameroon). Soc Sci Med, 69 (6), 821 829. doi:10.1016/j.socscimed.2009.06.007 Dobash, R. P., & Dobash, R E. (1979). Violence against wives : a case against the patriarchy New York: Free Press. Working on a puzzle. British Journal of Criminology, 44 (324 349). Doherty, T., Cho pra, M., Jackson, D., Goga, A., Colvin, M., & Persson, L. A. (2007). Effectiveness of the WHO/UNICEF guidelines on infant feeding for HIV positive women: results from a prospective cohort study in South Africa. AIDS, 21 (13), 1791 1797. doi:10.1097/QAD.0b01 3e32827b1462

PAGE 190

170 Doherty, T., Chopra, M., Nkonki, L., Jackson, D., & Greiner, T. (2006). Effect of the HIV epidemic on infant feeding in South Africa: "When they see me coming with the tins they laugh at me". Bull World Health Organ, 84 (2), 90 96. doi:/S0042 96862006000200008 Doherty, T., Chopra, M., Nkonki, L., Jackson, D., & Persson, L. A. (2006). A longitudinal qualitative study of infant feeding decision making and practices among HIV positive women in South Africa. J Nutr, 136 (9), 2421 2426. D oherty, T., Chopra, M., Nsibande, D., & Mngoma, D. (2009). Improving the coverage of the PMTCT programme through a participatory quality improvement intervention in South Africa. BMC Public Health, 9 406. doi:10.1186/1471 2458 9 406 Doherty, T., Sanders, D., Goga, A., & Jackson, D. (2011). Implications of the new WHO guidelines on HIV and infant feeding for child survival in South Africa. Bull World Health Organ, 89 (1), 62 67. doi:10.2471/BLT.10.079798 Donahue, M. C., Dube, Q., Dow, A., Umar, E., & Van Rie A. (2012). "They have already thrown away their chicken": barriers affecting participation by HIV infected women in care and treatment programs for their infants in Blantyre, Malawi. AIDS Care, 24 (10), 1233 1239. doi:10.1080/09540121.2012.656570 Drain, P K., Smith, J. S., Hughes, J. P., Halperin, D. T., & Holmes, K. K. (2004). Correlates of national HIV seroprevalence: an ecologic analysis of 122 developing countries. J Acquir Immune Defic Syndr, 35 (4), 407 420. Duff, P., Kipp, W., Wild, T. C., Rubaale, T., & Okech Ojony, J. (2010). Barriers to accessing highly active antiretroviral therapy by HIV positive women attending an antenatal

PAGE 191

171 clinic in a regional hospital in western Uganda. J Int AIDS Soc, 13 37. doi:10.1186/1758 2652 13 37 Dunkle, K. L., Jewkes, R. K., Brown, H. C., Gray, G. E., McIntryre, J. A., & Harlow, S. D. (2004). Gender based violence, relationship power, and risk of HIV infection in women attending antenatal clinics in South Africa. Lancet, 363 (9419), 1415 1421. doi: 10.1016/S0140 6736(04)16098 4 Edmonds, A., Yotebieng, M., Lusiama, J., Matumona, Y., Kitetele, F., Napravnik, S., Behets, F. (2011). The effect of highly active antiretroviral therapy on the survival of HIV infected children in a resource deprived se tting: a cohort study. PLoS Med, 8 (6), e1001044. doi:10.1371/journal.pmed.1001044 Ellsberg, M., Jansen, H. A., Heise, L., Watts, C. H., & Garcia Moreno, C. (2008). Intimate partner violence and women's physical and mental health in the WHO multi country st udy on women's health and domestic violence: an observational study. Lancet, 371 (9619), 1165 1172. doi:10.1016/S0140 6736(08)60522 X Emenike, E., Lawoko, S., & Dalal, K. (2008). Intimate partner violence and reproductive health of women in Kenya. Int Nurs Rev, 55 (1), 97 102. doi:10.1111/j.1466 7657.2007.00580.x Emusu, D., Ivankova, N., Jolly, P., Kirby, R., Foushee, H., Wabwire Mangen, F., Ehiri, J. (2009). Experience of sexual violence among women in HIV discordant unions after voluntary HIV counsell ing and testing: a qualitative critical incident study in Uganda. AIDS Care, 21 (11), 1363 1370. doi:10.1080/09540120902883077 Etiebet, M. A., Fransman, D., Forsyth, B., Coetzee, N., & Hussey, G. (2004). Integrating prevention of mother to child HIV transmi ssion into antenatal care: learning from the

PAGE 192

172 experiences of women in South Africa. AIDS Care, 16 (1), 37 46. doi:10.1080/09540120310001633958 Evans, C., & Ndirangu, E. (2009). The nursing implications of routine provider initiated HIV testing and counsellin g in sub Saharan Africa: a critical review of new policy guidance from WHO/UNAIDS. Int J Nurs Stud, 46 (5), 723 731. doi:10.1016/j.ijnurstu.2008.11.003 Ezechi, O. C., Gab Okafor, C., Onwujekwe, D. I., Adu, R. A., Amadi, E., & Herbertson, E. (2009). Intimate partner violence and correlates in pregnant HIV positive Nigerians. Archives of Gynecololgy and Obstetrics, 280 (5), 745 752. doi:10.1007/s00404 009 0956 9 Fadnes, L. T., Engebretsen, I. M., Moland, K. M., Nankunda, J., Tumwine, J. K., & Tylleskar, T. (201 0). Infant feeding counselling in Uganda in a changing environment with focus on the general population and HIV positive mothers a mixed method approach. BMC Health Serv Res, 10 260. doi:10.1186/1472 6963 10 260 Farquhar, C., Mbori Ngacha, D. A., Bosire R. K., Nduati, R. W., Kreiss, J. K., & John, G. C. (2001). Partner notification by HIV 1 seropositive pregnant women: association with infant feeding decisions. AIDS, 15 (6), 815 817. Farr, L. (2013). The role of men in the economic and social developme nt of women: implications for gender equality. The World Bank Research Observer, 28 (1), 22 51. doi:10.1093/wbro/lks010 Faul, F., Erdfelder, E., Lang, A. G., & Buchner, A. (2007). G*Power 3: a flexible statistical power analysis program for the social, beha vioral, and biomedical sciences. Behavioral Research Methods, 39 (2), 175 191.

PAGE 193

173 Fereday, J., & Muir Cochrane, E. (2006). Demonstrating Rigor Using Thematic Analysis: A Hybrid Approach of Inductive and Deductive Coding and Theme Development. International Jo urnal of Qualitative Methods, 5 (1), 80 92. Filmer, D., & Pritchett, L. (2001). Estimating Wealth Effect Without Expenditure Data or Tears: An Application to Educational Enrollments in States of India. Demography, 38 115 132. Fischbach, R. L., & Herbert, B. (1997). Domestic violence and mental health: correlates and conundrums within and across cultures. Soc Sci Med, 45 (8), 1161 1176. Gaede, B., Majeke, S., Modeste, R. R. M., Naidoo, J. R., Titus, M. J., & Uys, L. R. (2006). Social support and health beh aviour in women living with HIV in KwaZulu Natal. SAHARA J: Journal of Social Aspects of HIV/AIDS, 3 (1), 362 368. doi:10.1080/17290376.2006.9724862 Gage, A. J., & Hutchinson, P. L. (2006). Power, control, and intimate partner sexual violence in Haiti. Arch Sex Behav, 35 (1), 11 24. doi:10.1007/s10508 006 8991 0 Gaillard, P., Melis, R., Mwanyumba, F., Claeys, P., Muigai, E., Mandaliya, K., Temmerman, M. (2002). Vulnerability of women in an African setting: lessons for mother to child HIV transmission pr evention programmes. AIDS, 16 (6), 937 939. Garcia de Olalla, P., Knobel, H., Carmona, A., Guelar, A., Lopez Colomes, J. L., & Cayla, J. A. (2002). Impact of adherence and highly active antiretroviral therapy on survival in HIV infected patients. J Acquir Immune Defic Syndr, 30 (1), 105 110. Garcia Moreno, C., Jansen, H. A., Ellsberg, M., Heise, L., & Watts, C. H. (2006). Prevalence of intimate partner violence: findings from the WHO multi country study on women's

PAGE 194

174 health and domestic violence. Lancet, 368 (9543), 1260 1269. doi:10.1016/S0140 6736(06)69523 8 Garcia Moreno, C., & Stockl, H. (2009). Protection of sexual and reproductive health rights: addressing violence against women. Int J Gynaecol Obstet, 106 (2), 144 147. doi:10.1016/j.ijgo.2009.03.053 Gass J. D., Stein, D. J., Williams, D. R., & Seedat, S. (2011). Gender differences in risk for intimate partner violence among South African adults. J Interpers Violence, 26 (14), 2764 2789. doi:10.1177/0886260510390960 Gielen, A. C., O'Campo, P., Faden, R. R., & Eke, A. (1997). Women's disclosure of HIV status: experiences of mistreatment and violence in an urban setting. Women Health, 25 (3), 19 31. doi:10.1300/J013v25n03_02 Glanz, K., Rimer, B. K., & Viswanath, K. (2008 ). Health behavior and health education : theory, research, and practice (4th ed.). San Francisco, CA: Jossey Bass. Green, J., & Thorogood, N. (2014). Qualitative methods for health research (Third edition. ed.). Los Angeles: SAGE. Gross, R., Yip, B., Lo R e, V., 3rd, Wood, E., Alexander, C. S., Harrigan, P. R., Hogg, R. S. (2006). A simple, dynamic measure of antiretroviral therapy adherence predicts failure to maintain HIV 1 suppression. J Infect Dis, 194 (8), 1108 1114. doi:10.1086/507680 Gupta, G. R ., Parkhurst, J. O., Ogden, J. A., Aggleton, P., & Mahal, A. (2008). Structural approaches to HIV prevention. Lancet, 372 (9640), 764 775. doi:10.1016/S0140 6736(08)60887 9

PAGE 195

175 Hardon, A., Vernooij, E., Bongololo Mbera, G., Cherutich, P., Desclaux, A., Kyaddond o, D., Obermeyer, C. (2012). Women's views on consent, counseling and confidentiality in PMTCT: a mixed methods study in four African countries. BMC Public Health, 12 26. doi:10.1186/1471 2458 12 26 Hare Mussen, R. T. (1978). A feminist approach to family systems. Family Proess, 17 181 194. Harling, G., Msisha, W., & Subramanian, S. V. (2010). No Association between HIV and Intimate Partner Violence in 10 Developing Countries. PLoS One, 5 (12), 1 8. doi:10.1371 Harrigan, P. R., Hogg, R. S., Dong, W. W., Yip, B., Wynhoven, B., Woodward, J., Montaner, J. S. (2005). Predictors of HIV drug resistance mutations in a large antiretroviral naive cohort initiating triple antiretroviral therapy. J Infect Dis, 191 (3), 339 347. doi:10.1086/427192 Harvey, A ., Garcia Moreno, C., & Butchart, A. (2007). Primary prevention of intimate partner violence and sexual violence: Background paper for WHO expert meeting May 2 3, 2007 Hatcher, A. M., Woollett, N., Pallitto, C. C., Mokoatle, K., Stockl, H., MacPhail, C., Garcia Moreno, C. (2014). Bidirectional links between HIV and intimate partner violence in pregnancy: implications for prevention of mother to child transmission. J Int AIDS Soc, 17 19233. doi:10.7448/ias.17.1.19233 Heckert, J., & Fabic, M. S. (201 3). Improving Data Concerning Women's Empowerment in Sub Saharan Africa. Stud Fam Plann, 44 (3), 319 344. doi:10.1111/j.1728 4465.2013.00360.x

PAGE 196

176 Heise, L., & Garcia Moreno, C. (2002). Violence by intimate partners Geneva. Hindin, M. J., & Adair, L. S. (2002 ). Who's at risk? Factors associated with intimate partner violence in the Philippines. Soc Sci Med, 55 (8), 1385 1399. Horwood, C., Vermaak, K., Butler, L., Haskins, L., Phakathi, S., & Rollins, N. (2012). Elimination of paediatric HIV in KwaZulu Natal, S outh Africa: large scale assessment of interventions for the prevention of mother to child transmission. Bull World Health Organ, 90 (3), 168 175. doi:10.2471/BLT.11.092056 Howard, A. A., Arnsten, J. H., Lo, Y., Vlahov, D., Rich, J. D., Schuman, P., S choenbaum, E. E. (2002). A prospective study of adherence and viral load in a large multi center cohort of HIV infected women. AIDS, 16 (16), 2175 2182. Human Rights Watch. (2007). Hidden in the mealie meal: Gender based abuses and Zambia Retrieved from Retrieved from http://www.unhcr.org/refworld/docid/4768e6292.html Huth Bocks, A. C., & Hughes, H. M. (2007). Parenting Stress, Parenting Behavior, and Journa l of Family Violence, 23 (4), 243 251. doi:10.1007/s10896 007 9148 1 Igumbor, O. J., Pengpid, S., & Obi, C. L. (2006). Effect of exposure to clinic based health education interventions on behavioural intention to prevent mother to child transmission of HIV infection. Journal of Social Aspects of HIV/AIDS Research Alliance, 3 (1), 394 402. Iliff, P. J., Piwoz, E. G., Tavengwa, N. V., Zunguza, C. D., Marinda, E. T., Nathoo, K. J., Humphrey, J. H. (2005). Early exclusive breastfeeding reduces the risk of postnatal HIV 1 transmission and increases HIV free survival. AIDS, 19 (7), 699 708.

PAGE 197

177 Imbaya, C. L., Odhiambo Otieno, G. W., & Okello Agina, B. M. (2008). Adherence levels to antenatal regimens of NVP and AZT among HIV + women receiving PMTCT treatment at Pumwani Maternity Hospital in Nairobi, Kenya Paper presented at the AIDS 2008 XVII International AIDS Conference: Abstract no. MOPE0519. Jasseron, C., Mandelbrot, L., Dollfus, C., Trocme, N., Tubiana, R., Teglas, J. P., Warszawski, J. (2011). Non Disclosure of a Pregnant Woman's HIV Status to Her Partner is Associated with Non Optimal Prevention of Mother to Child Transmission. AIDS Behav doi:10.1007/s10461 011 0084 y Jewkes, R. (2002). Intimate partner violenc e: causes and prevention. Lancet, 359 (9315), 1423 1429. doi:10.1016/s0140 6736(02)08357 5 Jewkes, R. K., Dunkle, K., Nduna, M., & Shai, N. (2010). Intimate partner violence, relationship power inequity, and incidence of HIV infection in young women in Sout h Africa: a cohort study. Lancet, 376 (9734), 41 48. doi:10.1016/S0140 6736(10)60548 X Johnson, M. P. (2006). Conflict and control: gender symmetry and asymmetry in domestic violence. Violence Against Women, 12 (11), 1003 1018. doi:10.1177/1077801206293328 Johnson, M. P. (2008). A typology of domestic violence : intimate terrorism, violent resistance, and situational couple violence Boston, MA; Hanover, NH: Northeastern University Press ; Published by University Press of New England. Kalmuss, D., & Straus, M. (1990). Wife's marital dependency and wife abuse In M.Straus & R.Gelles (Eds.), Physical violence in American families: Risk factors and adaptations to violence New Brunswick, NJ: Transaction.

PAGE 198

178 Kebaabetswe, P. M. (2007). Barriers to participation in the prevention of mother to child HIV transmission program in Gaborone, Botswana a qualitative approach. AIDS Care, 19 (3), 355 360. doi:10.1080/09540120600942407 Kelly, J., & Johnson, M. (2008). Differentiation among types of intimate partner violence: resear ch update and implications for interventions. Fam Court Rev, 46 476 499. Kiarie, J. N., Farquhar, C., Richardson, B. A., Kabura, M. N., John, F. N., Nduati, R. W., & John Stewart, G. C. (2006). Domestic violence and prevention of mother to child transmis sion of HIV 1. AIDS, 20 (13), 1763 1769. doi:10.1097/01.aids.0000242823.51754.0c Kilewo, C., Massawe, A., Lyamuya, E., Semali, I., Kalokola, F., Urassa, E., Biberfeld, G. (2001). HIV counseling and testing of pregnant women in sub Saharan Africa: expe riences from a study on prevention of mother to child HIV 1 transmission in Dar es Salaam, Tanzania. J Acquir Immune Defic Syndr, 28 (5), 458 462. Kimani Murage, E. W., Wekesah, F., Wanjohi, M., Kyobutungi, C., Ezeh, A. C., Musoke, R. N., Griffiths, P. (2014). Factors affecting actualisation of the WHO breastfeeding recommendations in urban poor settings in Kenya. Matern Child Nutr doi:10.1111/mcn.12161 Kishor, S. (2005). Domestic violence measurement in the demographic and health surveys: The histor y and the challenges Retrieved from Geneva Switzerland: Kishor, S., & Johnson, K. (2004). Profiling Domestic Violence: A Multi Coutry Study Calverton, Maryland. Kishor, S., & Subaiya, L. (2008). Understanding Women's Empowerment: A Comparative Analysis of Demographic and Health Surverys (DHS) Data Calverton MA.

PAGE 199

179 Kobin, A. B., & Sheth, N. U. (2011). Levels of adherence required for virologic suppression among newer antiretroviral medications. Ann Pharmacother, 45 (3), 372 379. doi:10.1345/aph.1P587 Kuhn, L., Sinkala, M., Kankasa, C., Semrau, K., Kasonde, P., Scott, N., Thea, D. M. (2007). High uptake of exclusive breastfeeding and reduced early post natal HIV transmission. PLoS One, 2 (12), e1363. doi:10.1371/journal.pone.0001363 Langen, T. T. (2005) Gender power imbalance on women's capacity to negotiate self protection against HIV/AIDS in Botswana and South Africa. Afr Health Sci, 5 (3), 188 197. doi:10.5555/afhs.2005.5.3.188 Lawoko, S. (2006). Factors associated with attitudes toward intimate partn er violence: a study of women in Zambia. Violence Vict, 21 (5), 645 656. Lawoko, S. (2008). Predictors of attitudes toward intimate partner violence: a comparative study of men in Zambia and Kenya. J Interpers Violence, 23 (8), 1056 1074. doi:10.1177/0886260507313972 Lawrence, R. A., & Lawrence, R. A. (1985). Breastfeeding, a guide for the medical profession (2nd ed.). St. Louis: Mosby. Lee, M., Stefani, K. M., & Park, E. C. (2014). Gender specific differences in risk for intimate partner v iolence in South Korea. BMC Public Health, 14 415. doi:10.1186/1471 2458 14 415 Leech, N. L., & Onwuegbuzie, A. J. (2006). A typology of mixed methods research designs. Qual Quant, 43 265 275. doi:DOI 10.1007/s11135 007 9105 3 Leshabari, S. C., Koniz Boo her, P., Astrom, A. N., de Paoli, M. M., & Moland, K. M. (2006). Translating global recommendations on HIV and infant feeding to the local

PAGE 200

180 context: the development of culturally sensitive counselling tools in the Kilimanjaro Region, Tanzania. Implement Sci 1 22. doi:10.1186/1748 5908 1 22 Libow, J. A., Raskin, P. A., & Caust, B. L. (1982). Feminism and family systems therapy: Are they interconnected? American Journal of Family Therapy, 10 3 10. Macmillan, R., & Gartner, R. (1999). When she brings home t he bacon: labor force participation and the risk of spousal violence against women. Journal of Marriage and Family, 61 (4), 947 958. Madiba, S., & Letsoalo, R. (2013). HIV disclosure to partners and family among women enrolled in prevention of mother to ch ild transmission of HIV program: implications for infant feeding in poor resourced communities in South Africa. Glob J Health Sci, 5 (4), 1 13. doi:10.5539/gjhs.v5n4p1 Maman, S., Campbell, J., Sweat, M. D., & Gielen, A. C. (2000). The intersections of HIV a nd violence: Directions for future research and interventions. Social Science & Medicine, 50 (4), 459 478. Maman, S., Mbwambo, J., Hogan, N. M., Kilonzo, G. P., & Sweat, M. (2001). Women's barriers to HIV 1 testing and disclosure: Challenges for HIV 1 volu ntary counselling and testing. AIDS Care, 13 (5), 595 603. doi:10.1080/09540120120063223 Martin, M., Del Cacho, E., Codina, C., M, T., De Lazzari, E., Mallolas, J., & al., e. (2008). Relationship between adherence level, type of the antiretroviral regimen, and plasma HIV type 1 RNA viral load: a prospective cohort study. AIDS Res Hum Retroviruses, 24 1263 1268.

PAGE 201

181 Maru, Y., & Haidar, J. (2009). Infant feeding practices of HIV positive mothers and its determinants in selected health institutions of Addis Aba ba, Ethiopia. Ethiop J Health Dev, 232 (2), 107 114. Medley, A., Garcia Moreno, C., McGill, S., & Maman, S. (2004). Rates, barriers and outcomes of HIV serostatus disclosure among women in developing countries: Implications for prevention of mother to child transmission programmes. Bull World Health Organ, 8 2 (4), 299 307. Megazzini, K. M., Sinkala, M., Vermund, S. H., Redden, D. T., Krebs, D. W., Acosta, E. P., Stringer, J. S. (2010). A cluster randomized trial of enhanced labor ward based PMTCT services to increase nevirapine coverage in Lusaka, Zambi a. AIDS, 24 (3), 447 455. doi:10.1097/QAD.0b013e328334b285 Mepham, S., Zondi, Z., Mbuyazi, A., Mkhwanazi, N., & Newell, M. L. (2011). Challenges in PMTCT antiretroviral adherence in northern KwaZulu Natal, South Africa. AIDS Care, 23 (6), 741 747. doi:10.108 0/09540121.2010.516341 Michalski, J. H. (2004). Making sociological sense out of trends in intimate partner violence The social structure of violence against women. Violence Against Women, 10 (6), 652 675. doi:Doi 10.1177/1077801204265018 Moffitt, T. E., Robins, R. W., & Caspi, A. (2001). A couples analysis of partner abuse with implications for abuse prevention policy. Criminology and Public Policy, 1 5 26. Msellati, P. (2009). Improving mothers' access to PMTCT programs in West Africa: a public health perspective. Soc Sci Med, 69 (6), 807 812. doi:10.1016/j.socscimed.2009.05.034

PAGE 202

182 Msuya, S. E., Mbizvo, E. M., Hussain, A., Uriyo, J., Sam, N. E., & Stray Pedersen, B. (2008). Low male partner participation in antenatal HIV counselling and testing in northern Tanzania: implications for preventive programs. AIDS Care, 20 (6), 700 709. doi:10.1080/09540120701687059 Murphy, C. M., & Meyer, S. L. (1991). Gender, power and violence in marriage. The Behavior Therapist, 14 95 100. Murray, L. K., Haworth, A., Semrau, K., Singh, M., Aldrovandi, G. M., Sinkala, M., Bolton, P. A. (2006). Violence and abuse among HIV infected women and their children in Zambia: a qualitative study. J Nerv Ment Dis, 194 (8), 610 615. doi:10.1097/01.nmd.0000230662.01953.bc Nachega, J. B ., Hislop, M., Dowdy, D. W., Chaisson, R. E., Regensberg, L., & Maartens, G. (2007). Adherence to nonnucleoside reverse transcriptase inhibitor based HIV therapy and virologic outcomes. Ann Intern Med, 146 (8), 564 573. Nachega, J. B., Uthman, O. A., Ander son, J., Peltzer, K., Wampold, S., Cotton, M. F., Mofenson, L. M. (2012). Adherence to antiretroviral therapy during and after pregnancy in low income, middle income, and high income countries: a systematic review and meta analysis. AIDS, 26 (16), 203 9 2052. doi:10.1097/QAD.0b013e328359590f Nankunda, J., Tumwine, J. K., Nankabirwa, V., & Tylleskar, T. (2010). "She would sit with me": mothers' experiences of individual peer support for exclusive breastfeeding in Uganda. Int Breastfeed J, 5 16. doi:10.1186/1746 4358 5 16 National Statistical Office, & ICF Macro. (2011). Malawi Demographic and Health Survey 2010 Zomba, Malawi and Calverton, Maryland, USA.

PAGE 203

183 Ngom, P., Debpuur, C., Akweongo, P., Adongo, P., & Binka, F. N. (2003). Gate keeping and wo men's health seeking behaviour in Navrongo, northern Ghana. Afr J Reprod Health, 7 (1), 17 26. Ngoma, M. S., Misir, A., Mutale, W., Rampakakis, E., Sampalis, J. S., Elong, A., Silverman, M. S. (2015). Efficacy of WHO recommendation for continued brea stfeeding and maternal cART for prevention of perinatal and postnatal HIV transmission in Zambia. J Int AIDS Soc, 18 19352. doi:10.7448/IAS.18.1.19352 Ntaganira, J., Muula, A. S., Masaisa, F., Dusabeyezu, F., Siziya, S., & Rudatsikira, E. (2008). Intimate partner violence among pregnant women in Rwanda. BMC Womens Health, 8 17. doi:10.1186/1472 6874 8 17 Nyasulu, J. Y., & Nyasulu, P. (2011). Barriers to the Uptake of Prevention of Mother to Child Transmission (PMTCT) Services in Rural Blantyre and Balaka Districts, Malawi. Journal of Rural and Tropical Public Health, 10 48 52. Omari, A. A., Luo, C., Kankasa, C., Bhat, G. J., & Bunn, J. (2003). Infant feeding practices of mothers of known HIV status in Lusaka, Zambia. Health Policy Plan, 18 (2), 156 162. Onono, M., Owuor, K., Turan, J., Bukusi, E. A., Gray, G. E., & Cohen, C. R. (2015). The role of maternal, health system, and psychosocial factors in prevention of mother to child transmission failure in the era of programmatic scale up in western Kenya: a case control study. AIDS Patient Care STDS, 29 (4), 204 211. doi:10.1089/apc.2014.0181

PAGE 204

184 Onono, M. A., Cohen, C. R., Jerop, M., Bukusi, E. A., & Turan, J. M. (2014). HIV serostatus and disclosure: implications for infant feeding practice in rural south Nyanza Kenya. BMC Public Health, 14 390. doi:10.1186/1471 2458 14 390 Pallitto, C. C., Garcia Moreno, C., Jansen, H. A., Heise, L., Ellsberg, M., Watts, C., Domestic, V. (2013). Intimate partner violence, abortion, and unintended pregnancy: results from the WHO Multi country Study on Women's Health and Domestic Violence. Int J Gynaecol Obstet, 120 (1), 3 9. doi:10.1016/j.ijgo.2012.07.003 Paterson, D. L., Swindells, S., Mohr, J., Brester, M., Vergis, E. N., Squier, C., Singh, N. (2000). Adherence to p rotease inhibitor therapy and outcomes in patients with HIV infection. Ann Intern Med, 133 (1), 21 30. Peltzer, K., Jones, D., Weiss, S. M., & Shikwane, E. (2011). Promoting male involvement to improve PMTCT uptake and reduce antenatal HIV infection: a clu ster randomized controlled trial protocol. BMC Public Health, 11 778. doi:10.1186/1471 2458 11 778 Pico Alfonso, M. A., Garcia Linares, M. I., Celda Navarro, N., Blasco Ros, C., Echeburua, E., & Martinez, M. (2006). The impact of physical, psychological, and sexual intimate male partner violence on women's mental health: depressive symptoms, posttraumatic stress disorder, state anxiety, and suicide. J Womens Health (Larchmt), 15 (5), 599 611. doi:10.1089/jwh.2006.15.599 Piwoz, E. G., Ferguson, Y. O., Bentle y, M. E., Corneli, A. L., Moses, A., Nkhoma, J., Kazembe, P. (2006). Differences between international recommendations on breastfeeding in the presence of HIV and the attitudes and counselling messages of

PAGE 205

185 health workers in Lilongwe, Malawi. Int Breas tfeed J, 1 (1), 2. doi:10.1186/1746 4358 1 2 Poggensee, G., Schulze, K., Moneta, I., Mbezi, P., Baryomunsi, C., & Harms, G. (2004). Infant feeding practices in western Tanzania and Uganda: implications for infant feeding recommendations for HIV infected mot hers. Trop Med Int Health, 9 (4), 477 485. doi:10.1111/j.1365 3156.2004.01214.x Population Council. (2008). Sexual and Gender Based Violence in Africa: A Literature Review Retrieved from Retrieved from http://www.popcouncil.org/pdfs/AfricaSGBV_LitReview.pdf : Prendergast, A., Mphatswe, W., Tudor Williams, G., Rakgotho, M., Pillay, V., Thobakgale, C., Goulder, P. (2008). Early virological suppression with three class antiretroviral therapy in HIV infected African infants. AIDS, 22 (11), 1333 1343. doi:10.1097/QAD.0b013e32830437df Press, N., Tyndall, M. W., Wood, E., Hogg, R. S., & Montaner, J. S. (2002). Virologic and immunologic response, clinical progression, and highly active antiretroviral therapy adherence. J Acquir Immune Defic Syndr, 31 Suppl 3 S112 117. Roberts, G. L., O'Toole, B. I., Raphael, B., Lawrence, J. M., & Ashby, R. (1996). Prevalence study of domestic violence victims in an emergency department. Annals of Emergency Medicine, 27 (6), 741 753. Romkens, R. (1997). Prevalence of wife abuse in the Netherlands: Combining quantitative and qualitative methods in survey research. J Interpers Violence, 12 99 125. Rylko Bauer, B., Whiteford, L. M., & Farmer, P. (2009). Global health in times of violence (1st ed.). Santa Fe: School for Advanced Research Press.

PAGE 206

186 Sarker, M., Sanou, A., Snow, R., Ganame, J., & Gondos, A. (2007). Determinants of HIV counselling and testing participation in a prevention of mother to child transmission programme in rural Burkina Faso. Trop Med Int Health, 12 (12), 1475 1483. doi:10.1111/j.1365 3156.2007.01956.x Shamu, S., Zarowsky, C., Shefer, T., Temmerman, M., & Abrahams, N. (2014). Intimate partner violence after disclosure of HIV test results among pregnant women in Harare, Zi mbabwe. PLoS One, 9 (10), e109447. doi:10.1371/journal.pone.0109447 Shenton, A. K. (2004). Strategies for Ensuring Trustworthiness in Qualitative Research Projects. Education for Information, 22 63 75. Shuaib, F. M., Ehiri, J. E., Jolly, P., Zhang, Q., Em usu, D., Ngu, J., Wabwire Mangen, F. (2012). Sexual violence and associated factors among women in HIV discordant and concordant relationships in Uganda. Int J Adolesc Med Health, 24 (2), 125 133. doi:10.1515/ijamh.2012.019 Sibeko, L., Coutsoudis, A., Nzuza, S., & Gray Donald, K. (2009). Mothers' infant feeding experiences: constraints and supports for optimal feeding in an HIV impacted urban community in South Africa. Public Health Nutr, 12 (11), 1983 1990. doi:10.1017/S1368980009005199 Simoni, J. M., Kurth, A. E., Pearson, C. R., Pantalone, D. W., Merrill, J. O., & Frick, P. A. (2006). Self report measures of antiretroviral therapy adherence: A review with recommendations for HIV research and clinical management. AIDS Behav, 10 (3), 227 245. doi:10.1007 /s10461 006 9078 6 Simons Morton, B. G., Hayne, D., & Noelcke, E. (2009). Social Influences: The Effects of Socialization, Selection, and Social Normative Processes on Health Behavior. In R. J.

PAGE 207

187 DiClemente, R. A. Crosby, & M. C. Kegler (Eds.), Emerging Theo ries in Health Promotion Practice and Research San Francisco, CA: John Wiley & Sons Inc. Singleton, R., & Straits, B. C. (2010). Approaches to social research (5th ed.). New York, NY: Oxford University Press. Smith, D. J., & Mbakwem, B. C. (2007). Life pr ojects and therapeutic itineraries: Marriage, fertility, and antiretroviral therapy in Nigeria. AIDS, 21 (suppl 5), S37 S41. Spangler, S. A., Onono, M., Bukusi, E. A., Cohen, C. R., & Turan, J. M. (2014). HIV positive status disclosure and use of essential PMTCT and maternal health services in rural Kenya. J Acquir Immune Defic Syndr, 67 Suppl 4 S235 242. doi:10.1097/qai.0000000000000376 Stark, E., & Flitcraft, A. (1996). Women at risk : domestic violence and women's health Thousand Oaks: Sage Publication s. Straus, M. A. (2008). Dominance and symmetry in partner violence by male and female university students in 32 nations. Children and Youth Services Review, 30 252 275. Straus, M. A., Gelles, R. J., & Steinmetz, S. K. (1980). Behind closed doors : viole nce in the American family (1st ed.). Garden City, N.Y.: Anchor Press/Doubleday. Straus, M. A., & Ramirez, I. L. (2002). Gender symmetry in prevalence, severity, and chronicity of physical aggression against dating partners by university students in Mexico and the USA. Paper presented at the the annual meeting of the International Society for Research on Aggression, Montreal, Canada. Stringer, E. M., Ekouevi, D. K., Coetzee, D., Tih, P. M., Creek, T. L., Stinson, K., Team, P. S. (2010). Coverage of nevirapine based services to prevent mother to child HIV

PAGE 208

188 transmission in 4 African countries. JAMA, 304 (3), 293 302. doi:10.1001/jama.2010.990 International Journal of Medical Educatio n, 2 53 55. doi:10.5116/ijme.4dfb.8dfd The World Bank. (2011). Gender Equality and Development Washington DC. Theuring, S., Mbezi, P., Luvanda, H., Jordan Harder, B., Kunz, A., & Harms, G. (2009). Male involvement in PMTCT services in Mbeya Region, Tanza nia. AIDS Behav, 13 Suppl 1 92 102. doi:10.1007/s10461 009 9543 0 Torpey, K., Kasonde, P., Kabaso, M., Weaver, M. A., Bryan, G., Mukonka, V., Colebunders, R. (2010). Reducing pediatric HIV infection: estimating mother to child transmission rates in a program setting in Zambia. Journal of Aquired Immune Deficiency Syndrome, 54 (4), 415 422. doi:10.1097/QAI.0b013e3181e36616 Torpey, K., Mandala, J., Kasonde, P., Bryan Mofya, G., Bweupe, M., Mukundu, J., Welsh, M. (2012). Analysis of HIV early infan t diagnosis data to estimate rates of perinatal HIV transmission in Zambia. PLoS One, 7 (8), e42859. doi:10.1371/journal.pone.0042859 Travers, M., & Bennett, L. (1996). AIDS, Women, and Power. In L. Sherr, C. Hankins, & L. Bennett (Eds.), AIDS as a Gender I ssues: Psychosocial Perspectives (pp. 64 77). Bristol, PA: Taylor & Francis, Ltd. Tuller, D. M., Bangsberg, D. R., Senkungu, J., Ware, N. C., Emenyonu, N., & Weiser, S. D. (2010). Transportation costs impede sustained adherence and access to HAART in a clinic population in southwestern Uganda: a qualitative study. AIDS Behav, 14 (4), 778 784. doi:10.1007/s10461 009 9533 2

PAGE 209

189 Tuthill, E., McGrath, J., & Young, S. (2014). Commonalities and differences in infant feeding attitudes and practices in the context of HIV in sub Saharan Africa: a metasynthesis. AIDS Care, 26 (2), 214 225. doi:10.1080/09540121.2013.813625 Upadhyay, U. D., & Karasek, D. (2012). Women's empowerment and ideal family size: an examination of DHS empowerment measures in Sub Saharan Africa. Int Perspect Sex Reprod Health, 38 (2), 78 89. doi:10.1363/3807812 UNAIDS. ( 1999). Prevention of HIV Transmission from Mother to Child: Strategic Options Geneva, Switzerland. UNAIDS. (2012). Regional Fact Sheet 2012: Sub Saharan Africa. Retrieved from http://www.unaids.org/sites/default/files/en/media/unaids/contentassets/docum ents/ep idemiology/2012/gr2012/2012_FS_regional_ssa_en.pdf UNAIDS. (2014). The Gap Report. Retrieved from http://www.unaids.org/en/resources/documents/2014/20140716_UNAIDS_gap_repor t UNAIDS. (2015). How AIDS Changed Everything. Retrieved from http://www .unaids.org/sites/default/files/media_asset/MDG6Report_en.pdf UNIADS. (2014). Children and HIV Fact Sheet Retrieved from Geneva, Switzerland: http://www.unaids.org/sites/default/files/media_asset/FactSheet_Children_en.pdf UNICEF. (2014). Elimination of n ew HIV Infections among children by 2015 and keeping their mothers alive in low and middle income countries. Retrieved from data.unicef.org/hiv aids/emtct UNICEF. (2015). Zambia: Prevention of Mother to_Child Transmission. UNICEF Zambia Fact Sheets. Re trieved from http://www.unicef.org/zambia/5109_8456.html

PAGE 210

190 USAID & Zambia Ministry of Health. (2009). The 2007 Zambia Demographic and Health Survey Retrieved from Retrieved from http://www.measuredhs.com/pubs/pdf/FR211/FR211[revised 05 12 2009].pdf: Varga, C., & Brookes, H. (2008). Factors influencing teen mothers' enrollment and participation in prevention of mother to child HIV transmission services in Limpopo Province, South Africa. Qual Health Res, 18 (6), 786 802. doi:10.1177/1049732308318449 Violari, A ., Cotton, M. F., Gibb, D. M., Babiker, A. G., Steyn, J., Madhi, S. A., McIntyre, J. A. (2008). Early antiretroviral therapy and mortality among HIV infected infants. New England Journal of Medicine, 359 (21), 2233 2244. doi:10.1056/NEJMoa0800971 Vyas S., & Watts, C. (2009a). How does economic empowerment affect women's risk of intimate partner violence in low and middle income countries? A systematic review of published evidence. Journal of International Development, 21 (5), 577 602. Vyas, S., & Watt s, C. (2009b). How does economic empowerment affect women's risk of intimate partner violence in lowland middle income countries? A systematic review of published evidence. Journal of International Development, 21 577 602. Were, E., Curran, K., Delany Mo retlwe, S., Nakku Joloba, E., Mugo, N. R., Kiarie, J., Partners in Prevention, H. T. S. T. (2011). A prospective study of frequency and correlates of intimate partner violence among African heterosexual HIV serodiscordant couples. AIDS, 25 (16), 2009 2018. doi:10.1097/QAD.0b013e32834b005d

PAGE 211

191 WHO. (2000). Collaborative Study on the role of breastfeeding on the prevention of infant mortality: Effect of breastfeeding on infant and child mortality due to infectious diseases in less developed countries: a poo led analysis. The Lancet, 355 (9202), 451 455. WHO. (2010a). Antiretroviral Drugs for Treating Pregnant Women and Preventing HIV Infection in Infants: Recommendations for a Public Health Approach 2010 Version Retrieved from France: WHO. (2010b). Guidelines on HIV and infant feeding: Prinicples and recommendations for infant feeding in the context of HIV and a summary of evidence Geneva, Switzerland. WHO. (2010c). Towards the elimination of mother to child transmission of HIV Retrieved from Gene va, Switzerland: http://www.who.int/ WHO. (2010d). Towards universal access: Scaling up priority HIV/AIDS interventions in the health sector Geneva, Switzerland http://www.who.int/hiv/pub/2010progressreport/report/en/index.html WHO. (2011). Global HIV/AI DS Response Progress Report 2011 Retrieved from Geneva, Switzerland: http://www.who.int/hiv/pub/progress_report2011/en/index.html WHO. (2012a). Programmatic Update: Use of Antiretroviral Drugs for Treating Pregnant Women and Preventing HIV Infection in In fants. Executive Summary Retrieved from http://www.who.int/hiv/PMTCT_update.pdf WHO. (2012b). Understanding and Addressing Violence Against Women. Retrieved from http://apps.who.int/iris/bitstream/10665/77432/1/WHO_RHR_12.36_eng.pdf WHO. (2014). Implemen tation of Option B + for Prevention of Mother To Child Transmission of HIV: the Malawi experience Brazzaville, Republic of Congo.

PAGE 212

192 WHO, UNAIDS, & UNICEF. (2011). Global UNAIDS response: Epidemic update and health sector progress towards universal Access, pr ogress report 2011 Retrieved from http://www.unaids.org/en/media/unaids/contentassets/documents/unaidspublication/2 011/20111130_ua_report_en.pdf WHO/LSHTM. (2010). Preventing intimate partner and sexual violence against women: taking action and generating evidence Retrieved from Geneva/London: Wingwood, G. M., Camp, C., Kristin, D., Cooper, H., & DiClemente, R. J. (2009). The Theory of Gender and Power: Constructs, Variables, and Implications for Developing HIV Interventions for Women In R. J. DiClemente R. A. Crosby, & M. C. Kegler (Eds.), Emerging Theories in Health Promotion Practice and Research (2 ed., pp. 393 414). San Francisco, CA: John Wiley & Sons, Inc. Wood, E., Hogg, R. S., Yip, B., Harrigan, P. R., O'Shaughnessy, M. V., & Montaner, J. S. (2003). Effect of medication adherence on survival of HIV infected adults who start highly active antiretroviral therapy when the CD4 + cell count is 0.200 to 0.350 x 10(9) cells/L. Ann Intern Med, 139 (10), 810 816. Wood, E., Hogg, R. S., Yip, B., Moore, D ., Harrigan, P. R., & Montaner, J. S. (2006). Impact of baseline viral load and adherence on survival of HIV infected adults with baseline CD4 cell counts > or = 200 cells/microl. AIDS, 20 (8), 1117 1123. doi:10.1097/01.aids.0000226951.49353.ed Yllo, K. (Ed .) (1993). Through a feminist lens: Gender, power, and violence ewbury Park, CA: Sage.

PAGE 213

193 Yoder, J. D., & Kahn, A. S. (1992). Toward a Feminist Understanding of Women and Power. Psychology of Women Quaterly, 16 381 388. Zambia Ministry of Health. (2008). N ational Protocol Guidelines Integrated prevention of mother to child transmission of HIV/AIDS Zambia Ministry of Health. (2010). 2010 National protocol guidelines: Integrated prevention of mother to child transmission of HIV ZMOH. (2010). PMTCT 2010 National Protocol Guidelines: Integrated Prevention of other to Child Transmission of HIV

PAGE 214

194 APPENDIX A: CONSENT FORM A1: English version Study Title: Gender, Power, and Perinatal HIV among Patients and the University Teaching Hospital and Lusaka Di strict Health Clinics Principal Investigator: Karen Hampanda You are asked to be in this research study because you are a mother of a young infant. The study wants to know how your husband/partner influences your decisions about the use of certain heal th care. We will use this information to make health care services for mothers and babies better. If you agree to take part in the study, you will be one of 320 women that will be asked about 50 questions health and taking drugs. Also, we will ask about how your husband/partner feels about it. This should take about 30 minutes. This will help us learn how to help mothers and babies have better health. Thirty of you will be invited to stay for another 30 minutes for us to learn more about problems in taking the drugs. We will ask some sensitive questions about HIV/AIDS and disturbances at home. This ma y make you uncomfortable. There may be other risks like distress, the researchers have not thought of. We will try our best to protect your privacy Your name will not be written on the answer sheet. All papers will be kept in a locked area that only t he research team has access to. We will make every effort to keep information confidential. You do not have to be in this study if you do not want to be. You do not have to answer any question you do not want to, and you may stop the study at any point if you want No matter what you do, you will still get all the care in this clinic. If you have questions, you can call Karen Hampanda, at 0974 073 794 You can call and ask questions at any time. You may also go to the office at UTH in the maternity uni t. You may have questions about your rights as someone in this study. If you have questions, you can call ERES (the Research Ethics Board). Their number is 0955 155 633, or 0955 155 634, or 0966 765 503. Their office is at, 33 Joseph Mwilwa Road, Rh odes Park, Lusaka. You will get reimbursement for your travel expenses if you join the study.

PAGE 215

195 The purpose of this study has been explained to me and I understand the purpose of the study. I further understand that: If I agree to take part in this study I can withdraw at any time without having to give an explanation and that taking part in this study is purely voluntary. I a gree to take part in both the survey and interview Signed/Thumbprint___________ Date _____ ( Participant) Signed____ _______ Date ______ ( Witness ) Signed___________ Date ______ ( Researcher) Information sheet to be given to the participant and consent to be kept by resear cher

PAGE 216

196 APPENDIX B: SURVEY QUESTIONNAIRE

PAGE 217

197

PAGE 218

198

PAGE 219

199

PAGE 220

200

PAGE 221

201

PAGE 222

202

PAGE 223

203

PAGE 224

204

PAGE 225

205

PAGE 226

206

PAGE 227

207

PAGE 228

208 APPENDIX C : SEMI STRUCTURED INTERVIEW GUIDE

PAGE 229

209

PAGE 230

210

PAGE 231

211 APPENDIX D : MY DATA COMPARED TO THE MOST RECENT ZDHS Percent of Women Reporting Intimate Partner Violence and Controlling Behaviors Emotional Violence Physical Violence Sexual Violence Any injuries Ever Hit partner Displays 3 or more controlling behaviors N Kanyama 1 (My study) 40.3% 31.6% 43.4% 56.6% (n=157) 3 8.5% 60.6% 320 Lusaka 1 (DHS) 20.6% 34.0% 9.3% Not reported 8.1% 31.9% 1,642 Zambia 2 (DHS) 24.0% 38.8% 16.7% 35.7% (n=3,663) 4 8.5% 34.9% 8, 571 1 Currently married women age 18 49 2 Ever married women age 15 49 3 Currently married women who experienced spousal violence 4 Ever married women who experienced spousal violence Making health care Major household purchases Daily household purchases N Kanyama (My study) 67.2% 55.6% 73.8% 320 Lusaka (DHS) 88.3% 84.2% 94.0% 1,780 Zambia (DHS) 74.0% 66.3% 86.1% 9,859