Citation
Individual and neighborhood effects on active lifestyles and social isolation in a sample community-dwelling elderly

Material Information

Title:
Individual and neighborhood effects on active lifestyles and social isolation in a sample community-dwelling elderly a socio-ecological study
Creator:
King, Diane Karen
Publication Date:
Language:
English
Physical Description:
xx, 235 leaves : ; 28 cm

Subjects

Subjects / Keywords:
Older people -- Social conditions -- Colorado -- Denver ( lcsh )
Physical fitness for older people -- Colorado -- Denver ( lcsh )
Older people -- Health and hygiene -- Colorado -- Denver ( lcsh )
Older people -- Dwellings -- Colorado -- Denver ( lcsh )
Neighborhoods ( lcsh )
Social isolation -- Colorado -- Denver ( lcsh )
Neighborhoods ( fast )
Older people -- Dwellings ( fast )
Older people -- Health and hygiene ( fast )
Older people -- Social conditions ( fast )
Physical fitness for older people ( fast )
Social isolation ( fast )
Colorado -- Denver ( fast )
Genre:
bibliography ( marcgt )
theses ( marcgt )
non-fiction ( marcgt )

Notes

Bibliography:
Includes bibliographical references (leaves 220-235).
General Note:
Department of Health and Behavioral Sciences
Statement of Responsibility:
by Diane Karen King.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
All applicable rights reserved by the source institution and holding location.
Resource Identifier:
71752928 ( OCLC )
ocm71752928
Classification:
LD1193.L566 2006d K56 ( lcc )

Downloads

This item has the following downloads:


Full Text
J
INDIVIDUAL AND NEIGHBORHOOD EFFECTS ON ACTIVE LIFESTYLES
AND SOCIAL ISOLATION IN A SAMPLE OF COMMUNITY-DWELLING
ELDERLY:
A SOCIO-ECOLOGICAL STUDY
by
Diane Karen King
B.S., State University of New York at Buffalo, 1981
M.B.A., State University of New York at Buffalo, 1983
M.S., Rush University, 1998
A thesis submitted to the
University of Colorado at Denver
in fulfillment of the requirements for the degree of
Doctor of Philosophy
Health and Behavioral Sciences
2006


2006 by Diane Karen King
All rights reserved.


This thesis for the Doctor of Philosophy
degree by
Diane Karen King
has been approved
/'l bv.
Deborah S.K. Thomas
3, Z OO(o
Date


King, Diane Karen (Ph.D., Health and Behavioral Sciences)
Individual and Neighborhood Effects on Active Lifestyles and Social Isolation in a
Sample of Community-dwelling Elderly: A Socio-ecological Study.
Thesis directed by Professor Craig R. Janes
ABSTRACT
There is strong evidence that maintaining an active lifestyle into old age
provides important physical and mental health benefits. Understanding the
individual and neighborhood factors that facilitate or inhibit active lifestyles in this
important population segment will contribute needed data toward the development of
multi-level, multi-disciplinary interventions to prevent decline. The overall aim of
this study was to investigate the affect of neighborhood contextual factors on activity
engagement and social isolation in elderly, community-dwelling adults. Methods:
Neighborhood and individual data were collected from adults, age 65 years and
older, residing in distinct neighborhoods within the city and county of Denver,
Colorado. Eight neighborhoods were selected to participate based upon the number
of potential participants and variability with regard to neighborhood walkability,
socio-economic status, proportion of adults over 65, and crime statistics. A list of
patients age 65 or older was obtained from electronic medical records, geocoded, and
matched to the 8 selected neighborhoods using a Geographic Information System. A
randomly selected sample (N=871) from the address-matched patient list were
invited by letter to participate in the study. 190 seniors completed and returned the
survey, providing data regarding perceptions of their neighborhood including access
to resources, safety from traffic and crime, and social cohesion. They also reported
on their weekly engagement in physical and social activities, loneliness and social
support. Individual health data were extracted from electronic medical charts.
Environmental data on both structural and social features of each neighborhood were
collected using published data as well as an environmental audit tool adapted from
the Systematic Pedestrian and Cycling Environment Scan (SPACES) instrument and
the Neighborhood Brief Observational Tool (NBOT). Results: Neighborhood
structural variables that maximized walking efficiency and mixed land-use
significantly predicted greater walking for errands (p<.05). Social capital variables
were indirectly associated with more frequent community engagement, more total
IV


and moderate-level PA and less loneliness, and their effects were mediated by
perceptions of safety from crime and social cohesion (p<.05). Conclusions:
Addressing social factors within neighborhood environments by enhancing
perceptions of safety from crime and social cohesion may be just as important as
providing walkable communities with convenient destinations.
This abstract accurately represents the content of the
recommend its publication.
Signed_
I
v


DEDICATION
I dedicate this dissertation to my husband Gene and my family for your
incredible support and understanding, my cohort-member Bonnie, who kept me
going, and my terrific committee members, for your patience and mentorship
throughout this process.


ACKNOWLEDGEMENT
This dissertation was made possible with support from The Robert Wood Johnson
Foundation, Active Living Research Program.


CONTENTS
Figures....................................................... xvi
Tables......................................................... xviii
CHAPTER
1. INTRODUCTION................................................ 1
The Health Problem...................................... 1
Conceptual Model and Study Framework.................... 3
Organization of the Dissertation........................ 4
2. THEORETICAL BACKGROUND...................................... 5
Theories of Aging............... ...................... 5
Theories of Individual Behavior......................... 6
Social Cognitive Theory........................... 6
Human Occupation Perspective...................... 8
Theories of Social Cohesion............................. 10
Social Identity and Integration................... 10
Social Capital.................................... 12
Theories of Social Ecology.............................. 16
Social-Ecological Perspective..................... 16
Analyzing multi-level data, concepts & challenges. 18
vm


3. REVIEW OF THE LITERATURE.................................... 22
Environmental Influence of Health Behaviors............. 22
Importance of Activity to the Health of Older Adults.... 26
Social Support, Loneliness and Health of Older Adults... 30
Social Support.................................... 30
Loneliness........................................ 34
Culturally-Mediated Perspectives on Loneliness.... 37
Contribution of This Study.............................. 40
Understanding How Context Affects Lifestyles for
Older Adults............................................ 40
4. RESEARCH DESIGN AND METHODS................................ 41
Research Questions & Specific Aims...................... 42
Measurable Obj ectives................................ 43
Hypotheses............................................. 43
Research Question 1: What components of the environmental
context enable or inhibit physical and social activity
in older adults?..................................... 43
Research Question 2: What are the Pathways through which
Neighborhood and Individual Factors Influence Outcomes?. 44
Research Question 3: Do Objective Health Outcomes
Cluster by Neighborhood?................................ 45
Study Design............................................ 45
IX


Participant and Neighborhood Sampling Frame................ 46
Participant Sampling Frame.......................... 46
Geocoding........................................... 47
Neighborhood Sampling Frame......................... 48
Neighborhood Selection............................. 51
Participant Selection & Inclusion Criteria.......... 53
Data Collection and Measures............................... 56
Individual Demographics, Health & Fitness Data............. 57
Indicators of Coronary Vascular Disease Risk........ 58
Chronic Disease Score............................. 58
Individual Perceptions of Neighborhood Environment....... 59
Self-efficacy for Transport......................... 59
The Neighborhood Environment Walkability Scale.... 60
Neighborhood Cohesion Scale......................... 60
Outcome Measures for Active Lifestyle and Social Isolation... 61
Community Health Activities Model Program for
Seniors (CHAMPS).................................. 61
The Loneliness Scale................................ 62
The Social Provisions Scale......................... 62
Pilot of Survey Instruments......................... 63
x


Participant and Neighborhood Sampling Frame................ 46
Participant Sampling Frame.......................... 46
Geocoding........................................... 47
Neighborhood Sampling Frame......................... 48
Neighborhood Selection............................. .51
Participant Selection & Inclusion Criteria........... 53
Data Collection and Measures................................ 56
Individual Demographics, Health & Fitness Data.............. 57
Indicators of Coronary Vascular Disease Risk... 58
Chronic Disease Score................................ 59
Individual Perceptions of Neighborhood Environment....... 59
Self-efficacy for Transport......................... 59
The Neighborhood Environment Walkability Scale.... 60
Neighborhood Cohesion Scale.......................... 60
Outcome Measures for Active Lifestyle and Social Isolation... 61
Community Health Activities Model Program for
Seniors (CHAMPS)..................................... 61
The Loneliness Scale................................. 62
The Social Provisions Scale......................... 62
Pilot of Survey Instruments.......................... 63
x


Data Management........................................... 66
Strategies Used to Overcome Implementation
Barriers............................................ 66
Maximizing Survey Response Rate & Completion...... 67
Maximizing Accuracy in Characterizing
Neighborhoods....................................... 67
Assuring Participant Privacy and Avoiding
Location Uncertainty.............................. 68
5. RESULTS OF ANALYSES........................................ 69
Research Questions........................................ 69
Analyses Used............................................. 69
Section 1. Characterizing Denver Neighborhoods........... 74
Reported Data........................................... 75
Comparison of Selected and Non-Selected
Neighborhoods....................................... 75
Measured Data........................................... 78
Field Audit: Inter-Rater Reliability Analysis....... 78
Functionality: Walking.............................. 82
Functionality: Traffic.............................. 84
Safety: Traffic..................................... 86
Safety: Personal.................................... 88
xi


Aesthetics
90
Destinations: Land-Use................................. 93
Destinations: Facilities............................... 95
Destinations: Recreation............................... 96
Destinations: Public Courtesies........................ 98
Social Capital: Incivilities........................... 100
Social Capital: Territoriality......................... 102
Social Capital: Stability.............................. 105
Social Capital: Vitality............................... 106
Subjective Assessment................................ 108
Discussion................................................. 110
Section 2. Characterizing Study Participants................... 115
Pilot Study.................................................... 116
Main Study..................................................... 119
Participation Rate............................................. 119
Participant Characteristics.................................... 120
Sociodemographic Variables............................. 120
Objective Health Variables............................. 122
Participant Perceptions....................................... 126
Community Mobility..................................... 126
xn


Neighborhood Environment for Walking................ 128
Neighborhood Social Cohesion........................ 130
Discussion................................................ 131
Section 3. The Built Environment and
Activity Engagement in Seniors............................ 135
Individual Level Outcomes................................. 136
Active Lifestyles: Weekly Activity Frequency........ 136
Active Lifestyles: Weekly Energy Expenditure...... 138
Results of Multi-Level Hierarchical Modeling.............. 139
Research Question 1: Neighborhoods and Active
Lifestyles: Does Context Matter?.................... 139
Hypothesis 1: Summary of Results.......................... 153
Results of Mediation Analyses............................. 153
Research Question 2: What Are the Potential Causal
Pathways through which Neighborhood and Individual
Factors Influence Outcomes?......................... 153
Perceived Safety from Crime and Social Cohesion as
Mediators of PA..................................... 164
Gender as a Moderating Variable........................... 166
Hypothesis 2: Summary of Results.......................... 167
Discussion................................................ 168
Section 4. Social Capital, Social Cohesion and
Social Isolation in Seniors............................... 172
xm


Individual Level Outcomes................................ 173
Social Isolation: Perceived Loneliness and
Social Support...................................... 173
Results of Multi-Level Hierarchical Modeling.............. 175
Research Question 1: Neighborhood Influence on
Social Isolation: Does Context Matter?............ 175
Results of Mediation Analyses............................. 178
Research Question 2: What Are the Potential Causal
Pathways through which Neighborhood and Individual
Factors Influence Loneliness?..................... 178
Gender as a Moderating Variable......................... 182
Hypothesis 2: Summary of Results.......................... 182
Discussion................................................ 183
Section 5. The Relationship of Neighborhood to Health... 186
Participant and Non-Participant Comparison................ 187
Health Characteristics of Combined Sample................. 188
Research Question 3: Do Objective Health Outcomes
Cluster by Neighborhood?.......................... 192
Neighborhood Factors and Cardiovascular Disease Risk.... 194
Hypothesis 3: Summary of Results.......................... 196
Discussion................................................ 197
6. CONCLUSIONS AND FUTURE DIRECTIONS............................ 199
Answers to the Research Questions......................... 199
xiv


Question 1: What Components of the Environmental
Context Enable or Inhibit Physical and Social Activity
in Older Adults?............................. 199
Question 2: What are the Potential Causal Pathways
through which Neighborhood and Individual Factors
Influence Outcomes?............................ 200
Question 3: Do objective health outcomes cluster by
neighborhood?.................................. 201
Limitations of the Study............................. 202
Generalizability.............................. 202
Culture and Ethnicity.......................... 203
Spatial Autocorrelation........................ 203
Contributions of the Study.......................... 203
Neighborhood Characterization.................. 204
Safety and Social Cohesion..................... 205
Revised Conceptual Model....................... 206
Implications for Future Research and Intervention.... 207
GLOSSARY......................................................... 209
APPENDICES
A: NEIGHBORHOOD AUDIT & DATA COLLECTION FORM.... 212
B: SAMPLE NEIGHBORHOOD MAPS................................ 217
BIBLIOGRAPHY..................................................... 220
xv


FIGURES
Figure
1.1. Conceptual Model............................................ 3
4.1. Conceptual Model............................................ 41
4.2. Geocoding results for active KPCO members ages 65 and
older........................................................ 48
4.3. Final Eight Selected Study Neighborhoods.. ................. 52
5.1.1. Conceptual Model: Neighborhood Characteristics.............. 74
5.2.1. Conceptual Model: Participant Characteristics................. 115
5.3.1. Conceptual Model: Neighborhood Effects on PA.................. 135
5.3.2. Mediation effect: yard maintenance with social cohesion....... 157
5.3.3. Mediation effect: window bars with social cohesion............ 158
5.3.4. Mediation effect: litter with social cohesion................. 159
5.3.5. Mediation effect: violent crime rate with social cohesion. ... 160
5.3.6. Mediation effect: yard maintenance with perceived safety...... 162
5.3.7. Mediation effect: window bars with perceived safety........... 163
5.3.8. Mediation effect: incivilities with perceived safety.......... 165
5.3.9. Mediation effect: window bars with perceived safety........... 166
5.3.10. Conceptual Model: Direct and Mediated Effects of Neighborhood
Variables on Activity Engagement............................. 168
xvi


5.4.1. Conceptual Model: Neighborhood Effects on Social Isolation... 172
5.4.2. - Mediation effect: window bars with perceived safety.......... 180
5.4.3. Mediation effect: window bars with perceived social cohesion. 181
5.4.4. Conceptual Model: Direct and Mediated Effects of Neighborhood
Variables on Social Isolation................................. 182
6.1. Revised Conceptual Model..................................... 206
XVII


TABLES
Table
4.1. Summary of Constructs and Measures................................ 64
5.1.1. Denver Neighborhood Socio-Demographic Comparison.................. 76
5.1.2. Independent Samples Test between KP and Non-KP Neighborhoods... 77
5.1.3. Neighborhood Variation: Comparison of Denver Neighborhoods with
Selected Neighborhoods, highest and lowest on selection criteria. 78
5.1.4. Inter-rater Reliability Results........................... ....... 81
5.1.5. The Built Environment for Walking: Sidewalk Functionality....... 82
5.1.6. The Built Environment for Walking: Traffic Calming.............. 85
5.1.7. The Built Environment for Walking: Traffic Safety............. 87
5.1.8. The Built Environment for Walking: Personal Safety............ 89
5.1.9. The Built Environment for Walking: Aesthetics................... 91
5.1.10. The Built Environment for Walking: Land Use..................... 94 .
5.1.11. The Built Environment for Walking: Destination Types............ 95
5.1.12. The Built Environment for Walking: Parks and Recreation......... 97
5.1.13. The Social Environment for Walking: Public Courtesies............. 99
5.1.14. The Social Environment for Walking: Physical Incivilities...... 101
5.1.15. The Social Environment for Walking: Territoriality............. 103
5.1.16. The Social Environment for Walking: Stability................ 105
xviii


5.1.17. The Social Environment for Walking: Vitality.................... 107
5.1.18. Subjective Walking Rating....................................... 109
5.1.19. Summary of Neighborhood Characteristics......................... 112
5.2.1. Participant Recruitment Results by Neigh........................ 119
5.2.2. Sociodemographic Characteristics of Participants by Neigh....... 121
5.2.3. Participant Cardiac Risk Metrics &
Chronic Disease Score by Neighborhood.......................... 123
5.2.4. Self-Reported Mobility by Neighborhood.......................... 127
5.2.5. Perceptions of Walking Environment (NEWS Scale results)........... 129
5.2.6. Perceptions of Social Cohesion.................................. 130
5.3.1. Frequency of Weekly Activities (CHAMPS results)................. 137
5.3.2. Weekly PA (CHAMPS results)........................................ 138
5.3.3. Multilevel Analysis: Frequency of Walking for Errands............. 140
5.3.4. Multilevel Analysis: Frequency of Home-Based PA......... ......... 142
5.3.5. Multilevel Analysis: Frequency of Community-Based Activity...... 144
5.3.6. Multilevel Analysis: Weekly calorie expenditure for total PA...... 146
5.3.7. Multilevel Analysis: Weekly calorie expenditure for Mod PA........ 148
5.3.8. Community-based Activity (Self Efficacy for Walking as Mediator)... 156
5.3.9. Community-based activity and yard maintenance
(with social cohesion)........................................... 157
5.3.10. Community-based activity and window bars (with social cohesion).... 158
xix


5.3.11. Community-based activity and litter (with social cohesion)....... 159
5.3.12. Community-based activity and violent crime rate
(with social cohesion)............................................ 160
5.3.13. Community-based activity and yard maintenance
(with perceived safety)........................................... 161
5.3.14. Community-based activity and window bars
(with perceived safety)......................................... 162
5.3.15. Total PA and incivilities (with perceived safety)................ 164
5.3.16. Total PA and window bars (with perceived safety)................. 165
5.4.1. Self-Reported Loneliness and Social Support
(LASA and SPS results)........................................... 174
5.4.2. Loneliness, Social Support, Social Cohesion and PA level......... 175
5.4.3. Multilevel Analysis: Loneliness.................................... 176
5.4.4. Loneliness (with perceived safety from crime)...................... 179
5.4.5. Loneliness (with perceived social cohesion)........................ 181
5.5.1. Characteristics of participants compared with non-participants
for sex, age, CDS and CVD....................................... 187
5.5.2. Combined Participant and Non-Participant Objective Health Metrics
by Neighborhood................................................... 189
5.5.3. Proportion of Participants and Non-Participants (combined) at higher
MI risk, by Neighborhood.......................................... 191
5.5.4. Multilevel Analysis: Combined sample (N=781) BMI................. 193
5.5.5. Multilevel Analysis: Combined sample (N=781) CVD................. 195
xx


CHAPTER 1
INTRODUCTION
The Health Problem
Maintaining an active lifestyle into old age, including engagement in
meaningful activities such as walking or gardening, has known physical and mental
health benefits. These include reduced risks of illness and depression, retention of
functional fitness, and decreased mortality (Glass, de Leon, Marottoli, & Berkman,
1999; Takano, Nakamura, & Watanabe, 2002). In addition, an understudied and
equally important area of inquiry is the potential relationship between community-
based physical activity and social isolation in the elderly.
Research conducted on physical activity engagement and use of discretionary
time among older adults has shown that as individuals age, the time spent engaged in
outdoor and active leisure activities declines (Dallosso et al., 1988; Lee & King,
2003). While physiological realities of aging cannot be overlooked as contributing
to the decline in activities that require greater energy expenditure (Shephard, 1997),
ecological features of the surrounding neighborhoods as well as individual self-
efficacy with regard to driving or use of alternative modes of transportation may also
contribute to an individuals tendency to remain at home as opposed to getting out.
Longer life spans will likely result in longer retirements accompanied by
decreased participation in what society deems productive or valuable activities
and roles (Glass et al., 1999), despite the potential to maintain health and
1


independence into old age (Rakowski, 1997). The biological process of senescence,
accompanied by an inactive lifestyle, often results in chronic illness and disability,
leading to decreased activity, increased isolation, and loneliness (Khaw, 1997), thus,
preserving ability to maintain an active lifestyle into old age may be one way to
simultaneously maintain physical and mental health and diminish loneliness.
Theories of aging that do not reflect the contextual physical and social
environments in which behavior occur, limit our ability to predict behavior change
across the lifespan (Rakowski, 1997). There is a need for theory and models that
include the interactions among environmental and individual factors that affect the
behavior of the elderly. Current research has shown that as individuals age their
activity space (Cromley & McLafferty, 2002) often shrinks to the locale of their
home or immediate neighborhood (Herzog, Ofstedal, & Wheeler, 2002; Lee & King,
2003; Porter, 1994). Thus, more studies of individual and neighborhood factors that
affect physical activity engagement in the elderly are needed to develop models that
can inform multi-level, multi-disciplinary interventions to prevent decline in older
adults.
Diez-Roux (Diez-Roux, 1998) emphasizes the importance of developing a
causation model that extends across multiple levels when attempting to study the
inter-relationships among potentially important variables. The conceptual model
presented below summarizes the hypothesized relationships between neighborhood
and individual factors targeted in this study.
2


Conceptual Model and Study Framework
Neighborhood f Level 2)
Vital Statistics; Socio-demographics, Crime Rates
Structural Factors: Walking Paths, Traffic, Land Use, Aesthetics
Social Factors: Social Capital, Safety, Vitality
Individual (Level 11
Characteristics; Socio-demographics, Health
Perceptions; Self Efficacy, Pedestrian Safety, Crime Safety,
Access to Resources, Social Cohesion
Active Lifestyle
Weekly Activity Frequency
\ Weekly Energy Expenditure
Social Isolation
Perceived Loneliness
Social Support
Figure 1.1. Conceptual Model
This study provides a multi-level analysis of environmental and individual
factors that affect active lifestyles and social isolation of adults, age 65 years and
older. Particular attention was paid to structural variables in the neighborhood, such
as presence and condition of sidewalks, walking paths, parks and traffic, as well as
the existence of amenities that are desirable to older adults, such as banks,
pharmacies, markets, health and recreation centers (Paul, 1993), and individual
perceptions of neighborhood resources, pedestrian safety and social cohesion. Data
on social isolation were also collected and analyzed, given its known association
3


with health decline (Herzog et al., 2002; Seeman, 1996) and the potential role of an
active life style in its prevention or diminution.
Organization of the Dissertation
This dissertation is organized to follow the authors path of exploration
beginning with the theoretical basis for the study (Chapter 2) and a comprehensive
review of the literature (Chapter 3). The research questions, specific aims,
hypotheses, study design, selection of participants and neighborhoods, and data
collection methods and measures are detailed next (Chapter 4). Results and their
relevant discussion are presented in Chapter 5, which has been sub-divided into five
sections in order to focus the reader on key areas components of the conceptual
model: 1. Characterizing Denver Neighborhoods; 2. Characterizing Study
Participants; 3. The Built Environment and Activity Engagement in Seniors; 4.
Social Capital, Social Cohesion and Loneliness in Seniors; and 5. Neighborhood
Context and Health. The dissertation concludes with a summary of the key study
findings, along with study limitations and areas for future research (Chapter 6). A
glossary with definitions of key terms used is provided along with Appendices that
contain maps, measures and copies of study materials.
4


CHAPTER 2
THEORETICAL BACKGROUND
' Theories of Aging
Traditional theories of aging and lifelong development are primarily stage-
based perspectives that begin at infancy, progress through childhood and then stop
after adolescence (Chapman, 2000). Few acknowledge phenomena that predict
behavior during old age; and those that do tend to describe more traditional views of
aging, including detachment from society, reflection on the past, and preparation for
death as the primary developmental tasks(Boeree, 1997). Current beliefs about
aging suggest a need for a life course perspective that includes the desire of many
adults to continue to lead productive and socially integrated lives into old age, while
coping with the challenges of retirement, illness, and disability (Antonovsky, 1979;
Mendes de Leon, Glass, & Berkman, 2003). This proposal draws on several
theoretical perspectives to create a model that integrates neighborhood and
individual factors that may impact the phenomenon of active aging and its positive
health effects. These include theories of individual behavior, i.e., social cognitive
theory and the human occupation perspective; theories of social cohesion, i.e.,
Durkheims theory of social integration, and various perspectives on social capital;
and theories of social ecology.
5


Theories of Individual Behavior
Social Cognitive Theory
Albert Bandura developed his theory of social development in the 1960s and
1970s, which later evolved into what is currently known as Social Cognitive theory
(SCT) (Bandura, 2001). SCT follows an agency perspective that assumes human
beings act with intention and forethought, and are capable of adjusting behavior
based on social and environmental feedback as well as on self-reflection. The theory
is developmental in that behavior is adopted and revised through observational
learning, self regulation, and reciprocal determinism (Bandura & Wood, 1989).
Observational learning, or modeling, involves attention to modeled behaviors
resulting in actions that are similar to the behavior Observed. Self-regulation is a
response to the reaction of others with regard to the behavior, which may result in
the behavior being repeated, modified, or stopped. Reciprocal determinism
emphasizes that cognition, actions and the environment interact and may also impact
activity engagement.
Within the framework of Social Cognitive Theory, there are three primary
cognitive processes: self-efficacy, outcome expectations, and goals. Self-efficacy,
which is an indicator of an individuals confidence in his or her own ability, related
to how much control he or she believes to have over events and potential threats, and
outcome expectations, which is a belief about the likelihood of success and/or the
perceived consequences of failure when the specific behavior is attempted, are key
6


components of a reciprocal process that produces goal-directed behavior (Bandura,
1997; Clark & Dodge, 1999). When individuals desire to achieve a behavioral goal
they evaluate their own beliefs and assumptions as well as information and advice
from external sources. They will then either attempt the behavior or not, based on
the conclusion they reach regarding their likely success and/or the perceived
consequence of failure. An important point to emphasize about self-efficacy is that it
is situation and behavior-specific, as opposed to a trait or generalized state. Thus, it
would be possible to have high self-efficacy regarding engaging in exercise within
ones own house, and low self-efficacy regarding walking for exercise outside, in the
neighborhood. For instance, if an older adult believes her neighborhood is unsafe
and confirms this with external information, (e.g., broken glass in the garden or
loitering teenagers on the comer), she might limit going outside. The reciprocal
effect from her behavior may be a deterioration of her yard that engenders more
littering, graffiti, and loitering around the vicinity of her house. Thus, individual
characteristics, cognition, behavior, and environmental factors maintain a dynamic
interaction that leads to multi-directional outcomes.
Many older adults report that maintaining their independence and living their
lives on their own terms are the most important determinants of their life satisfaction
(Langlois et al., 1997; Letvak, 1997; Porter, 1994). Because self-monitoring of
behavior is an important component of social cognitive theory (Bandura & Wood,
1989; Baronowski, Perry, & Parcel, 1997), interventions aimed at improving
7


perceptions of autonomy and control often target increasing self-efficacy for a
specific behavior or set of behaviors. Self-efficacy also contributes to autonomy and
control by helping to shape aspirations and goals as well as predicting outcomes
(Senecal, Nouwen, & White, 2000). Bandura and Locke (Bandura & Locke, 2003),
in a review of nine large-scale meta-analyses across a range of activities, found
consistent evidence that beliefs about self-efficacy are a significant factor in both
motivation and performance, and that perceived self-efficacy is a strong and
independent predictor of future performance (Bandura & Locke, 2003; McAuley,
Jerome, Marquez, Elavsky, & Blissmer, 2003).
For older adults in urban settings, the notion of pedestrian self-efficacy is as
important to maintaining independence as driving in more suburban settings (Fonda,
Wallace, & Herzog, 2001). In a study of older pedestrians in New Haven,
Connecticut, a telephone survey found that 11.4% of residents aged 72 and older
reported problems crossing the street (Langlois et al., 1997). Because of this
connection between mobility and independence, understanding perceived self-
efficacy regarding mobility (walking, driving, and use of alternative modes of
transportation) is included in this proposal.
Human Occupation Perspective
, The Human Occupation perspective is the major tenet for the occupational
therapy discipline and is derived from multiple frames of references including
cognitive behavioral theories that emphasize learning and adaptation, developmental
8


theories where roles and priorities change along a continuum, and socio-ecological
systems theories that emphasize how human occupation is effected by individual
skills, roles, values, interests as well as the environmental context in which activities
are performed (Dutton, Levy, & Simon, 1993). These three theories combine to
produce the key assumption that is the core of the Human Occupation perspective:
human beings, throughout their lifespan, have an innate desire for mastery of their
environment that produces an occupational nature. Mastering ones environment
requires the ability to continuously change and adapt, not only for survival, but also
for the purpose of self-actualization (Hopkins & Smith, 1993). Having an
occupational nature encompasses a wide range of purposeful behaviors that
occupational therapists typically group into the broad domains of activities of daily
living, leisure, and work (Kielhofiier, 1997; Kielhofiier, Burke, & Igi, 1980). For
older adults, activities of daily living include activities that are basic and
instrumental, such as grooming, eating, taking medication, and driving. Leisure
activities may include social activities with friends and family, group recreation, as
well as solitary pastimes such as gardening, reading, or engagement in hobbies.
Work may include home management activities such as shopping, banking, and
household maintenance, as well as caring for grandchildren or spouses, educational
activities, and vocational activities such as volunteer work and paid work.
Self-identity is fashioned from the many occupational roles an individual
assumes in order to perform a wide variety of activities. For example, one individual
9


may simultaneously hold the varied roles of worker, parent, grandparent, spouse, and
friend. As individuals age, some or all of these roles may be lost through retirement,
children leaving home, and the death of spouses and friends. Thus, retention of
meaningful roles and occupation is theorized as preserving self-efficacy and health
(Bryant, Corbett, & Kutner, 2001). Because Human Occupation theory specifically
targets activity engagement as healthful, it is an important component of an Active
Lifestyles model of aging. While Active Lifestyle, for the purpose of this study, is
operationalized as total energy expenditure, questions about community-based
activities (both active and sedentary) are included in the proposed measures, and will
be examined separately to better understand the mix of activities performed by this
population.
Theories of Social Cohesion
Social Identity and Integration
Emile Durkheim, in 1897, discussed how individuals integrate into society in
his study of suicide as a social phenomenon (Durkheim, 1897). He theorized that
healthy individuals formed attachments with groups and the strength of those
attachments defined their own identity as individuals or as group members.
Durkheim concluded that while suicide is an individual act, it was also a social act,
resulting from disturbance to the social equilibrium. In his studies of suicide rates in
different European countries as well as rates between different religious groups,
Durkheim found that suicide was relatively stable within a society, and highly
10


variable among societies. Suicide could be anti-social or egoistic, when an
individual has low attachment to society, or altruistic, when an individuals group
identity supersedes his or her own and death is seen as way to advance the group. In
addition, he observed that suicides occurred in societies that were very disorganized
and unregulated. These he called anomic, and were akin to egoistic suicide, but
the former differs in its emphasis on societys deficiency in regulating individuals as
opposed to the individuals own low attachment.
Durkheims theory emphasizes the importance of social integration and
cohesion to individual mortality (Berkman, Glass, Brissette, & Seeman, 2000).
McMillan and Chavis (McMillan & Chavis, 1986) define social cohesion as having
four elements: 1) membership; 2) influence; 3) integration and fulfillment of needs;
and 4) shared emotional connection or history. The ability to trust and expect
reciprocity from neighbors has been shown to be a positive predictor of social
cohesion (Macinko & Starfield, 2001; Sampson, Raudenbush, & Earls, 1997).
Having a sense of belonging, as well as being able to adapt and participate in life
within a defined community, has known benefits to the health of older adults
including improved prognosis for recovery from illness, improved immune function,
and decreased mortality (Patrick & Wickizer, 1995; Seeman, 1996).
Durkheims theory of social integration is reflected in studies of how
communities maintain order and enforcement of social norms. For instance,
Sampson and Raudenbush studied neighborhoods and violent crime and found that
11


neighborhood norms of reciprocity, trust, and a willingness to intervene in order to
prevent social disturbances, such as truancy or teenage loitering, were significantly
associated with reduced violent crimes. The authors termed the combined
phenomenon of social cohesion and informal social control as collective efficacy,
a concept that is related to, but not synonymous with Banduras use of the same term
to describe individual beliefs about the collective capability, shared knowledge and
skills of a group to achieve desired outcomes (Bandura, 1997). Sampson and
Raudenbushs measure of social cohesion will be used in this study (Sampson et al.,
1997).
Social Capital
Society is not the sum of individuals just as population health status is not
equal to the composition of individual risk factors, but is determined by collective
characteristics of communities and societies (Kawachi & Berkman, 2000).
According to Macinko and Starfield (Macinko & Starfield, 2001) the social capital
construct has a long history of use, beginning with Hanifan in 1920, who defined it
as good will, fellowship, sympathy and social intercourse among the individuals
and families who make up a social unit. The works of Marx, Durkheim, Bourdieu,
and the more contemporary Coleman and Putnam, have engendered diverse
perspectives on social capital, with definitions ranging from the individual to more
structural/political concepts. For instance, social capital has been defined as
resources that accrue to individuals via social connections, group level norms,
12


cohesion and participation (Coleman, 1988) and also as the existence of
organizations and resources within a community that can be accessed by its
members. The question of whether a tight community with strong values, norms
and sanctions also has strong social capital has engendered much discussion
(Coleman, 1988; Putnam, 1995). From a Durkheimian perspective, attachment to a
social group may be protective against despair and other negative health
consequences (Durkheim, 1897). From a Marxist viewpoint, communities that are
poorer in assets may experience demoralization, resentment of the society at large,
and consequently lead to negative health outcomes (Kawachi & Berkman, 2000;
Waitzkin, 1988).
Bourdieu examined how possession of the right credentials, education,
occupation, or taste, may dictate whether or not individuals have wealth, legitimacy
or status within society, and can access the form of capital most valuable to them,
i.e., economic, cultural, or social capital. His notion of habitus, i.e., the typical
customs, attitudes and behaviors practiced by groups or individuals, as a
determination of their position within a larger society, may explain why
marginalized groups who experience strong social cohesion often lack the resources
enjoyed by those who are integrated into the mainstream. Bourdieu emphasized
economic capital as the foundation of all types of capital (Bourdieu, 1984).
Coleman, like Bourdieu, viewed social capital as membership in social
groups, an especially important aspect of adolescent life, where approval from peers
13


is of preeminent importance. However, his theory does not account for other types
of capital that may be even more important to ensuring success in life versus mere
acceptance (Coleman, 1988).
Putnam views social capital as residing in communities, including not only a
Durkheimian sense of belonging, but also a sense of trust, reciprocity, and
cooperation, and is a proponent of activism to achieve these goals (Putnam, 1995).
His view shares many of the elements of a true structural perspective, where social
capital is an element of social cohesion that has to do with specific features of social
structures that provide resources for individuals and facilitate collective action
(Bourdieu, 1984; Kawachi & Berkman, 2000; Putnam, 1995). These features
include trust, available social organizations, norms and sanctions, and information
channels. Structuralists consider social capital as ecological and external to
individuals, as opposed to social networks that are often measured at the individual
level (Kawachi & Berkman, 2000). From this perspective, social capital is a public
good that all members of society can benefit from, although individual contributors
may only reap a small part of the benefits themselves. The lack of agreement as to
whether social capital resides in individuals, groups, or communities, presents
challenges for measuring the concept of social capital as well as its effect on health
outcomes, and is a chief reason for criticism (Hawe & Shiell, 2000; Macinko &
Starfield, 2001; Morrow, 1999).
14


The mechanism that links social capital to health outcomes can be evaluated
in several ways. First, compositional effects of social capital (e.g., attributes of the
individuals who reside in a given community) indicate that individuals who are more
socially isolated and less integrated, reside in areas where social capital is low.
Second, contextual effects associated with the shared environment (e.g., at the state
or neighborhood level) suggest that social capital influences health-related behaviors
and access to services and amenities. Health-related behaviors may be affected due
to promotion of rapid diffusion of health information, and increased likelihood that
healthy behavioral norms are adopted. Access to services and amenities include
transportation, recreation, neighborhood health resources, and existence of local
groups who can lobby for services. Psychosocial processes affected include
provision of support, trust and reciprocity. Third, social capital may influence policy
on a broader scale, as it has been shown that states that have low levels of
interpersonal trust are less likely to invest in human security and social safety net
programs (Kawachi & Berkman, 2000).
The definition of social capital is still evolving, and it is unclear at present as
to the best way to operationalize it for practical application (Hawe & Shiell, 2000).
Measurement of social capital is primarily of aggregate variables that are made up of
individual responses to social surveys and more integral variables that involve direct
social observation of neighborhoods (e.g., number of establishments that still accept
personal checks may be an indicator of trust). Thus, while conceptually interesting,
15


further honing of the definition and its unique contribution to our understanding of
its potential impact on health outcomes is warranted. For the purpose of this study,
the aspects of social capital that will be examined include individual perceptions of
access to resources, social cohesion and safety from crime, that may impact their
activity levels, as well as individual perceptions of loneliness and social support, and"
objective neighborhood factors such as existence of resources, reported crime rate,
and observed public courtesies (such as public benches and transportation) and
incivilities (such as vacant lots, home security features, litter and graffiti).
Theories of Social Ecology
Social-Ecological Perspective
The notion of reciprocal relationships between environmental factors and
behavior is a key part of social cognitive theory (Bandura, 1986; Bandura, 1997),
human occupation theory, and other theories of social ecology that delineate multiple
levels of influence on individual behavior (Dzewaltowski, 1997; Saelens, Sallis, &
Frank, 2003; Sallis & Owen, 1996). The impact of ecological factors on health is a
key tenet of public health and was popularized in the 1800s. More recently,
theorists and researchers in a variety of disciplines are acknowledging that there are
multiple levels of influence on behavior. This is in contrast to earlier psychological
and sociological perspectives that greatly emphasized individual agency over
social/environmental influence. Psychologists, who made early contributions to the
social-ecological perspective included Kurt Lewin, who in the 1930s theorized that
16


perceptions of the external environment had an important influence on behavior,
Roger Barker in the late 60s who believed environments had a direct influence on
behavior, and Urie Bronfenbrenner in the late 70s who described how individual
behavior and development are influenced by multiple levels (Brown, 1998; Bull &
Shlay, 2005, In press). Brofenbrenner developed a multi-level model that
emphasized interactive systems, described as microsystems (e.g., an individuals
immediate environment such as their family, home and neighborhood),
mesosystems, that link individuals with other settings important to their day-to-day
lives (e.g., school and work), exosystems, which affect individuals indirectly (e.g.,
community organizations, spouse or parents workplace), and macro systems that
create the larger cultural environment (e.g., policy, media, and norms)
(Brofenbrenner, 1979; Brown, 1998).
Sallis and Owen (Sallis & Owen, 1996) outline five common tenets of
ecological models: 1) there are multiple dimensions that influence health behaviors;
2) these various dimensions interact with each other; 3) within the environment there
are multiple levels of influence; 4) environmental factors directly influence
behaviors; and 5) different environmental factors will influence specific behaviors
differently, so ecological models should specify which factors influence which
behaviors.
A socio-ecological perspective was applied in this study to evaluate both
neighborhood and individual variables that may influence individual levels of
17


physical activity and loneliness. Multi-level modeling, a method that has been used
in past research to better explain the multiple influences on individual behavior (Li,
Fisher, Bauman et al., 2005), was used to increase understanding of the complex
associations among neighborhoods, individuals and individual behavior.
Analyzing Multi-Level Data. Concepts and Challenges
Contextual analysis is the study of the effects of collective or group
characteristics on individual level outcomes (Diez-Roux, 1998). Sometimes called
multi-level analysis it measures the impact of ecological independent variables on
individual dependent variables. Group-level variables (also known as ecological;
macro-level; contextual; or aggregate) illustrate that individual risk for disease
depends not just on individual factors (e.g., income; health) but also on community
factors (resources; safety). For instance, some studies have shown that high levels of
community unemployment are associated with increased individual stress levels
regardless of the individuals actual employment status (Diez-Roux, 1998).
There is a conceptual distinction between group-level and aggregate
individual-level variables. Group level variables can be derived or integral in nature.
Derived variables are analytical or aggregate in that they summarize composites of
individual characteristics and are usually presented as means, percentages, medians,
and other distribution variables. Derived variables often have analogues at both
levels (e.g., individual income and mean neighborhood income) but measure very
different constructs (i.e., the purchasing power of an individual versus the resources
18


available to all individuals in that neighborhood). Derived variables may influence
integral variables. Integral variables describe characteristics of the group, not
derived from its individual members (e.g., regulations, availability of health care,
political systems, population density). Integral variables have no analogues at the
individual level (Diez-Roux, 1998). An illustration of these various level variables is
as follows: the number of community organizations to which each individual
belongs (individual level); the percentage of persons in the community who belong
to at least one organization (group level, derived); the number of organizations in the
community (group level, integral).
The analytical challenge of examining multi-level data and drawing the
correct inferences is to avoid fallacies that arise when data collected at one level are
used to draw conclusions about a different level (Diez-Roux, 1998). The often-cited
Ecological or aggregate fallacy involves drawing inferences at the individual level
based upon data collected at the group-level. For example: the finding that increased
per capita income at the country level is associated with increased motor vehicle
deaths, does not mean that individuals with higher income have more motor-vehicle
deaths. It may mean that wealthier countries as a whole contain more drivers. The
flip side is the Atomistic fallacy that involves drawing inferences at the group level
based on data collected at the individual level. For example: even if within
countries, increased individual income is associated with decreased coronary heart
disease (CHD) mortality, at the country level, increased per capita income may be
19


associated with increased CHD death rates. Such inferential fallacies can be
overcome if the data collected, and analyses conducted, match the level at which the
inferences are to be drawn. Two other fallacies that support using a multi-level
approach include the psychologists (individualistic) fallacy and the sociologistic
fallacy. The former assumes that social phenomena can be understood by looking at
individual characteristics and the latter that assumptions about individuals can be
drawn from studying social phenomena (Diez-Roux, 1998). Thus it is important to
include relevant factors from multiple levels to avoid these latter two fallacies and
account for confounders that may have an independent effect on the dependent
variable and are not a part of the hypothesized causal chain.
Selection of the appropriate level data sources, measures and construct is an
important process. While the level of the construct and the level of the data source
do not have to be the same (e.g., surveying individuals about a group construct such
as their perceptions of community cohesion), it is important to justify aggregating
individual level data to perform a group-level analysis (Hofmann, 2004). In
addition, defining the contextual unit and variables appropriately is another key
challenge of designing socio-ecological studies (Diez-Roux, 1998). For instance,
specifying the boundaries of a community or neighborhood is often subjective.
Most individuals belong to a variety of overlapping social and spatial contexts (e.g.,
neighborhood, work, school, organizational affiliations, and virtual communities)
whose effects may be difficult to separate. Another issue that may arise is that
20


individual-level outcomes within groups may be correlated; even after group and
individual-level variables are controlled for, due to other shared contextual variables.
Developing models that illustrate the linkages between social structure and
health outcomes are important to developing approaches to study these processes
(Marmot, 2000). For older adults, social and spatial phenomena that exist within
their immediate surroundings are important due to an assumed narrowing of activity
space, in many cases. This study will be limited to the neighborhoods in which the
selected individuals reside. The neighborhood boundaries used in this proposal are
the Statistical Neighborhoods defined by the City and County of Denver, linked to
Census Tracts. To address the ecological fallacy, data were collected at both the
neighborhood level and the individual level to avoid making inferences about
individuals based on neighborhood demographics. A two-level hierarchical
modeling approach was used to determine whether the participants activity and
social isolation outcomes were affected by the physical and social characteristics of
their neighborhood.
21


CHAPTER 3
REVIEW OF THE LITERATURE
Environmental Influence of Health Behaviors
The assumption that environmental variables may affect behavior and health
outcomes is the rationale for community-based public health programs, education,
and interventions (Patrick & Wickizer, 1995). Support for this assumption is found
in studies that control for individual sociodemographic variables, access to health
care, and other individual-level variables, yet disparities in health outcomes (De
Bourdeaudhuij, Sallis, & Saelens, 2003; Saelens, Sallis, Black, & Chen, 2003;
Saelens, Sallis, & Frank, 2003; Takano et al., 2002) are still present. In a review of
studies that looked at the association between neighborhood built environmental
variables and physical activity, significant differences in walking and cycling were
found based upon street layout and the mix of residential and commercial space
(Saelens, Sallis, & Frank, 2003). These findings were replicated by Saelens, et al.
(Saelens, Sallis, Black et al., 2003) who observed that neighborhoods designed
around car use contributed to a more sedentary lifestyle. In their study of the
relationship between physical activity and neighborhood walkability, they found that
residents in neighborhoods with higher residential density, a mix of residential and
commercial amenities, and gridlike street patterns recorded more weekly minutes of
moderate level activity than those residing in neighborhoods with long, winding
streets and few retail establishments. However, few studies (Fisher, Li, Michael, &
22


Cleveland, 2004; Li, Fisher, & Brownson, 2005; Michael, Green, & Farquhar, 2005)
have focused on older adult activity within neighborhood contexts. A recent study of
582 community-dwelling seniors recruited from 56 Portland neighborhoods did find
significant increases in levels of individual physical activity that were associated
with neighborhood-level variables as well as neighborhood social cohesion after
controlling for individual-level variables (Fisher et al., 2004). These findings were
promising, and the importance of social cohesion to both activity engagement and
social isolation were specifically explored within this study.
When studying the link between neighborhoods and health, many factors
come into play. Environmental toxins, noise, overcrowding, deterioration of
buildings and sidewalks, and housing quality have all been suggested as impacting
the health of older adults (Krause, 1996). Neighborhood deterioration has also been
linked to social isolation of older adults due to fear of crime and reduced
neighborhood cohesion (Krause, 1993). Alternatively, neighborhoods with walkable
parks and tree-lined streets were found to increase longevity in a prospective study of
Japanese elderly (Takano et al., 2002). Developing models that illustrate the
relationships among neighborhood contextual factors such as land use,
transportation, and senior services, and individual and group-level data, is an
important step to developing relevant health interventions (Diez Roux, 2004). An
example of differential access to services and resources for physical activity was
illustrated in a study by Estabrooks, et al. (Estabrooks, Lee, & Gyurcsik, 2003). This
23


study found that high SES neighborhoods had a significantly greater number of free
resources for physical activity than did the low and medium SES neighborhoods.
Despite the importance of the physical environment to active lifestyles, it is
unknown whether the actual environment or individuals perceptions about their
environment have the greater influence on physical activity (Brownson et al., 2004).
Measures of residents perceptions of their neighborhoods walkability, safety,
aesthetics, access to recreation facilities, and other amenities, have been shown to
have good test-retest reliability, particularly with regard to specific elements of the
built environment, such as estimating distances to destinations and presence of
sidewalks. Determining the validity of those perceptions using Geographic
Information Systems (GIS) and neighborhood audits have shown mixed results
(Kirtland et al., 2003; Pikora et al., 2002). Kirtland, et al. (Kirtland et al., 2003)
compared perceptions of supports for physical activity, such as existence and
condition of sidewalks and public recreation facilities; street lighting; and overall
aesthetics, with objective environmental measures using a sample of 1237 residents
of Sumter County, South Carolina. Self-reported physical activity frequency and
duration, and perceptions of the immediate neighborhood (i.e., surroundings within a
.5-mile radius or 10-minute walk), and the community (i.e., destinations within a 10-
mile radius or 20-minute drive), were included. Geographic Information Systems, a
computer-based tool that allows the user to capture, store, retrieve, analyze and
visually display database information that is spatial in nature (Cromley &
24


McLafferty, 2002), was then used to create a map of objective supports and barriers
to physical activity, and to confirm distances between respondent residences and
specific locations. Overall agreement between perceptions and objective data for
neighborhood items was highest for access to sidewalks, access to public recreation
facilities, safety/crime, equitable public spending, and streetlights. Perceptions of
access to recreation facilities were highest for those reporting at least some physical
activity as opposed to those categorized as inactive. Community items showed the
highest agreement for access to malls. Other community items, such as public pools,
trails and parks showed low agreement, suggesting that perceptions may better match
reality for more proximal destinations, and/or those that are accessed more
frequently. Thus, inclusion of both perceptual and objective measures is
recommended to adequately characterize features of neighborhoods that may affect
active lifestyles (Brownson et al., 2004; Kirtland et al., 2003).
Along with the increased interest in the importance of context to health,
discussions of how best to measure, analyze and interpret socio-ecological effects are
actively evolving. Oakes (Oakes, 2004) cautions researchers about making causal
inferences with regard to neighborhood effects on health when observational data are
used alone, without intervention or treatment. Oakes reasoning is that inferences
about neighborhood residents that are based on neighborhood factors are inherently
confounded, since the neighborhood characteristics are not independent of the
resident characteristics. Diez Roux (Diez Roux, 2004) disagrees that resident and
25


neighborhood characteristics are necessarily interdependent, though she stresses the
need to directly measure objective and integral neighborhood characteristics that are
independent of resident characteristics. She also emphasizes the importance of not
characterizing neighborhoods using group level demographics as proxies for other
attributes, e.g., using neighborhood socio-economic status to indicate lack of
neighborhood resources or high crime rates, that could be more directly measured
(Diez Roux, 2004; Oakes, 2004).
The present study evaluated the effect of both physical and social
characteristics of neighborhoods on older adult physical activity and social isolation
by characterizing neighborhoods using both independent, objective neighborhood
variables such as population density, street lay out, and parks in addition to
potentially interdependent variables such as violent crime rates, land use mix, and
neighborhood aesthetics.
Importance of Activity to the Health of Older Adults
Studies of successful or robust aging dispute the traditional Western
notion that disengagement and separation from society are normal and inevitable
conditions of aging (Garfein & Herzog, 1995). In a study of well-elderly,
participants in a group program designed around the Occupational Therapy (OT)
principle that activities of human occupation (i.e., self-care, work, play and leisure
activities) are beneficial to physical and mental health and well-being, did achieve
more benefits in life satisfaction, health perceptions, and overall mental health than
26


those who participated in either a no-treatment control group or a social activity
group (Clark et al., 1997). The OT group targeted basic activities of daily living
(ADLs) such as grooming and instrumental ADLs such as using transportation and
shopping. The OT group also focused on preserving independence by providing skill
building in the areas of safety, joint protection, energy conservation, exercise and
nutrition. Participants in the OT study were ethnically diverse, implying that focus
on purposeful activity and maintenance of independence is valued cross-culturally.
However, all participants, by nature of the programs design, needed to be motivated,
healthy and mobile enough to travel to the program.
Dallosso, et al. (Dallosso et al., 1988) used an activity inventory for older
adults in England to assess participation in four categories of activities: outdoor
productive; indoor productive; leisure; and walking. A large sample (n=1042) of
both men and women age 65 and older was interviewed. Only 53% performed
outdoor activities, including gardening (88%), with a small percent performing car
maintenance and house repairs. Leisure physical activities were least reported,
including social walking (22.4%) and cycling (5.5%). Significantly more men
reported participation in outdoor and leisure activities, including walking.
Lee & King (Lee & King, 2003) investigated how differential use of
discretionary time among older adults affects energy expenditure. They evaluated
data from two studies designed to increase physical activity among older adults, to
determine the impact on use of discretionary time. The Community Health
27


Activities Model Program for Seniors (CHAMPS) questionnaire was used to assess
weekly activities of varying metabolic equivalent (MET values). The findings were
that the majority of activities performed were less than 3 METs, i.e., required less
than 3 times the resting energy expenditure. Activity choices varied by gender with
women spending more time engaged in social activities such as visiting, helping
others, and attending meetings. Thus, activities selection in old age correspond to
their physical demands, underscoring the importance of limiting environmental
restrictions to promote more physically active lifestyles.
Herzog, et al. (Herzog et al., 2002) categorized activities as productive or
helping, i.e., activities that produce a good or a service; educational or intellectual,
including activities such as reading, taking courses, or using computers, and leisure,
including both formal pastimes such as classes, group travel and outings, and
informal pastimes such as attending movies or concerts, walking, knitting or
cooking.
Everard, et al. (Everard, Lach, Fisher, & Baum, 2000) present the results of
survey research conducted with a convenience sample of244 members of an
organization for older adults. All participants were community dwelling, aged 65
and older. The purpose of the survey was to study the association between
engagement in activities and function. Male gender was associated with higher
physical health scores. Performance of instrumental ADLs and high-demand leisure
activities was positively associated with physical health. Maintenance of low-
28


demand leisure activities was independently associated with better mental health.
Interestingly, this was the only activity item that was significantly associated with
increased mental health (p =.0001). Because this is a cross-sectional study it is
unclear if better physical health promotes maintenance of more physically
demanding activities, or if performance of these activities promotes better physical
health. The finding that maintenance of low-demand leisure was associated with
better mental health, also cross-sectional, is interesting since it implies that even
sedentary activities may provide some important health benefits, particularly to older
adults whose mobility may be impaired. The mechanism for this was not explored,
but it is consistent with other studies that have found associations between
performance of social, sedentary activities and greater life expectancy (Glass et al.,
1999; Sugisawa, Liang, & Liu, 1994).
This latter finding is important for many older adults who are isolated or
homebound due to chronic disease or disability, caregiving responsibilities for an
afflicted spouse, or limitations with regard to transportation and financial resources,
making external socialization and activities difficult. Emphasis on how solitary
pursuits may be employed to defray loneliness need more attention (Rane-Szostak &
Herth, 1995). For example, crafts, listening to music, gardening, and reading have
been cited by some older adults as fulfilling activities that stave off loneliness. Such
activities, as suggested by Csikszentmihalyis theory of optimal experience, provide
a state flow (Csikszentmihalyi, 1991,1997). Csikszentmihalyi defines flow as
29


the times when individuals are fully absorbed in activities that cause them to lose all
sense of time and provide them with a feeling of satisfaction. He proposes that such
flow experiences enhance feelings of control, mastery, and enjoyment and usually
entail concentration, clear goals, immediate feedback, effortless involvement, an
altered sense of time, and a chance for completion.
The present study focused on engagement in a wide variety of activities,
particularly those that promoted greater energy expenditure, as a way to measure
active lifestyles in older adults. While the CHAMPS was used to measure
physical activity level, sedentary social and leisure activities were also analyzed,
since they may provide health benefits, regardless of MET level. Individual activity
mix was quantitatively and qualitatively analyzed to determine the frequency of
engagement in specific categories of activities and are included in the analysis.
Social Support. Loneliness and Health of Older Adults
Social Support
Longitudinal studies have shown that people who are socially isolated have
two to five times the risk of dying from all causes, after adjusting for age, sex,
chronic diseases, use of alcohol and tobacco, self-rated health, and functional
limitations (Penninx et al., 1997; Seeman, 1996). In addition, both human and
animal studies have shown that presence of social support during stressful situations
reduces blood pressure and enhances immune function (Berkman et al., 2000;
Cacioppo et al., 1998; Seeman, 1996). Social support maybe emotional, i.e.,
30


providing sympathy, caring and love, or instrumental, i.e., providing tangible help or
assistance (Berkman et al., 2000), and has been shown to speed recovery from heart
attack and increase physical function and psychological adjustment after a stroke
(Berkman & Glass, 2000). Social networks, similarly, provide access to material
resources, such as job opportunities, housing, health care referrals, and other
benefits. Networks operate through micro psychosocial and behavioral processes
such as social support, influence, engagement, attachment, and access to resources
and material goods. They also operate through macro social forces such as labor
markets, economic pressures, organizational relations, culture, social change,
industrialization and urbanization (Berkman & Glass, 2000). Assessments of social
networks focus on network structure and strength of ties. Structural factors include
range or size (number of members); density (extent of connections); boundedness
(degree to which networks are defined); and homogeneity (extent to which
individuals are similar). Strength of individual ties is measured by frequency of
contact (number and type); multiplexity (number of transaction types); duration (how
long individuals know one another); and reciprocity (extent that transactions are
even). Social networks may help promote social participation, engagement in
activities and a sense of identity, since they provide opportunity for hosting and
attending social gatherings, group recreation, and rituals (Berkman et al., 2000).
Older adults who live alone are becoming increasingly common (Letvak,
1997). Maintaining connections to family and community are often difficult for
31


those who lack transportation, perform caregiving duties for an ailing spouse, or who
have a physical or mental disability. Social relationships may impact health in a
variety of ways (Heaney & Israel, 1997; Letvak, 1997; Stewart, 1989), particularly
for older adults who are at highest risk for isolation. In a prospective cohort study of
2812 community-dwelling adults age 65 and older, in-home and telephone
interviews were conducted over a twelve-year period to determine if there was a
relationship between social ties, social support and cognitive function (Bassuk,
Glass, & Berkman, 1999). Participants were assessed on cognitive performance, and
a composite index of their social connections, activities, and emotional support,
called the social disengagement index. After adjustment for age and baseline
cognitive scores, higher social disengagement scores were significantly associated
with a higher probability of cognitive decline at three, six, and twelve-year follow-
up. Participants with no social ties had about twice the odds of cognitive decline as
those with five or more social ties. In addition, the social disengagement index
predicted a higher mortality rate across all follow-ups. The relationship between
disengagement and subsequent decline held for those with the best initial cognitive
performance scores. The summary measure provided a strong and consistent
predictor of cognitive decline that was not present with any of the individual scale
items. The authors concluded that having multiple opportunities for social contact
and activity may be more important than any particular type of social support. This
32


may explain why many of the interventions used to improve social support in order
to ameliorate loneliness in the elderly, have shown mixed success.
Numerous studies show that social support reduces perceptions of loneliness
(Chin-Sang & Allen, 1991; Fratiglioni, Wang, Ericsson, Maytan, & Winblad, 2000;
Mullins, Elston, & Gutkowski, 1996; Penninx et al., 1997; Sugisawa et al., 1994;
Tilvis, Pitkala, Jolkkonen, & Strandberg, 2000), however, these differ in type of
support measured. For instance, a longitudinal study of older Swedish adults,
Lennartsson & Silverstein (Lennartssoh & Silverstein, 2001) investigated the
relationship between engagement in social, leisure and productive activities and
mortality. Activities were looked at along two continuums: solitary to social; and
sedentary to active. In this way they sought to provide insight into whether activities
that were social, physical, or neither social nor physical, had different effects. It was
found that frequency of contact with children, grandchildren, other relatives and
marital status were not significant predictors of mortality, so those activities were not
included. In addition, when the impact of various activities and organizational
participation were examined using multivariate analysis adjusted for gender, age,
education, functional health, circulatory/heart problems and smoking, the only
significant relationship that remained was that older men appeared to benefit from
participating in solitary but active pursuits such as carpentry, gardening, or other
hobbies. No benefits were found for social activities, group involvement, religious
participation, or family contact for this study. The mechanisms at work were not
33


investigated and no measures of social-psychological variables were used, so one can
only hypothesize possible causes for the outcome. In addition, since survival was
solely looked at as the outcome instead of overall health or quality of life, the impact
of these various activities on physical and mental health is unknown.
On the other hand, a separate study of Japanese elders (Sugisawa et al., 1994)
found that social participation (i.e., organizational attendance) was significantly
related to decreased mortality. Similar to the Swedish study, marital status and
number of social contacts showed little effect on mortality for this group. Also, in a
Netherlands-based study of mortality in the elderly, a large social network was
associated with less loneliness and lower mortality (Penninx et al., 1997). Others
have suggested that having a relationship that provides emotional support (i.e.,
closeness or intimacy) is the most important factor to reducing loneliness (Mullins et
al., 1996; Russell, Cutrona, de la Mora, & Wallace, 1997), although emotional
support alone was not found to be a strong predictor of prevention of cognitive
decline in the above-mentioned study on social disengagement (Bassuk et al., 1999).
Loneliness
Loneliness is defined as a perceived lack or loss of a meaningful network of
interpersonal contacts and/or companionship (Bergman-Evans, 1994; Penninx et al.,
1997; Walker & Beauchene, 1991). Social isolation is not synonymous with
loneliness (Mullins et al., 1996), but it may contribute to it by reducing the
opportunities for social contact and activities. Since loneliness is a subjective state,
34


the factors that contribute to it are many. They include loss of employment, income,
mobility, hearing, vision, health, independence, enjoyed pastimes, relatives, friends,
spouses, and pets (Bergman-Evans, 1994; Chin-Sang & Allen, 1991; Rane-Szostak
& Herth, 1995; Russell et al., 1997).
Loneliness has been implicated as a major contributor to declines in mental
and physical health, self-care and nutritional inadequacy, and increased depression
and suicide in older adults (Forbes, 1996; Khaw, 1997; Rane-Szostak & Herth, 1995;
Russell et al., 1997; Sugisawa et al., 1994; Walker & Beauchene, 1991).
Longitudinal studies have shown that people who are lonely are at greater risk of
dementia leading to institutionalization, than those who have satisfying, albeit
infrequent, contacts with friends/relatives (Fratiglioni et al., 2000; Russell et al.,
1997; Tilvis et al., 2000). While theories of healthy aging indicate that loneliness
and its accompanying health declines are not an inevitable component of aging
(Bonneux, Barendregt, & Van der Maas, 1998; Garfein & Herzog, 1995; Khaw,
1997), the social and cultural factors that contribute to loneliness are becoming
increasingly common. First, norms that foster geographically dispersed, nuclear
families have made it more difficult for older adults to maintain the community ties
and social participation that are important to sustained health (Chin-Sang & Allen,
1991; Forbes, 1996). Second, longer life expectancy for both men and women may
result in longer retirements accompanied by decreased participation in What are
perceived as productive or valuable activities and roles (Glass et al., 1999).
35


Third, despite the potential to maintain health and independence into old age
(Rakowski, 1997), the biological process of senescence often results in chronic
illness and disability, leading to increased isolation and loneliness (Khaw, 1997).
Fourth, reductions in health services that are reimbursable by Medicare and
decreased length of hospital stays in favor of home health, have created a situation
where many older adults are sent home alone to manage their own health, or the
health of an ailing spouse, thus constraining access to services, social support and
valued activities (Bergman-Evans, 1994; Forbes, 1996; Gustafson et al., 1998;
McTavish et al., 1995).
Studies disagree about the contribution of health, income, increased age,
gender, ethnic status, rural/urban dwelling, existence of children, or education as risk
factors for loneliness. A few have shown a tendency for more loneliness among
people who perceive their health as poor (as opposed to their actual health status),
people with disabilities, perceived poverty (as opposed to their actual income level),
and female gender have been implicated as risk factors (Mullins et al., 1996). While
these results seem to be inconsistent among studies, what is interesting to note, is
that the existence of a few good friends seem to be more critical to avoiding
loneliness then the existence of a spouse or children. Thus friends, coupled with a
sense of utility and perceived control seem to be consistently important factors
(Chin-Sang & Allen, 1991; Garfein & Herzog, 1995; Mullins et al., 1996; Porter,
1994; Sugisawa et al., 1994).
36


Culturally-Mediated Perspectives on Loneliness
The emphasis on youth, independence, and autonomy in the United States
provides some distinct cross-cultural challenges for those whose cultural background
and traditions esteem aging and integrate the elderly into the extended family unit.
Preservation of traditional roles for the elderly varies and the impact on perceived
health and well-being is dramatic. Individuals of Hispanic descent are diverse in
ethnicity and culture, yet extended families are still common among Hispanic groups
in the United States. In many families Hispanic grandparents play a key role in
helping to raise, or in some cases, raising, their grandchildren (Bagley, Angel,
Dilworth-Anderson, Liu, & Schinke, 1995).
In a qualitative study of 83 Latino elders (65 years and older), social support
from families was reported as the most important source of well being, regardless of
living arrangement (Beyene, Becker, & Mayen, 2002). Participants were
interviewed three times over 12-18 months. 62% of the participants had been in the
United States for more than 20 years. While most appreciated the economic and
health services available to seniors in the United States, many felt that respect for
~ elders was lacking, even among young Hispanics. All participants felt that feeling
old was a function of losing their role or purpose in life, as opposed to chronology.
Those with close relationships with their children and grandchildren said that they
did not perceive themselves to be old, despite multiple chronic illnesses in all cases,
and severe disability in some. The three greatest fears expressed by these Latino
37


elders were outliving their usefulness and having to go to a rest home, losing their
autonomy and being lonely.
Respect for elders and the important role of the elderly is also a cultural norm
among Asian families. For instance, Chinese nuclear families in China and
America often include one or two elderly grandparents who assist in child rearing.
Adult children are considered the preferred source of social support to elderly
parents, and are expected to provide care when they are ill. This tradition of family
elder-care has also been noted in American families of Vietnamese, Cambodian, and
Laotian descent (Bagley et al., 1995). Similarly, large extended families and clans
are central to most Native American cultures, with customs of cooperation, shared
resources and distinct roles. However, for some older Native Americans, who live in
poverty with inadequate sanitation, lack of finances, and little access to health
services, neglect and isolation accompanied by loneliness and despair, is similar to
that experienced by socially isolated nomlndian elders (Mercer, 1996; Williams &
Ellison, 1996).
For African American families, large kin networks that are not necessarily
defined by blood relationship are usual (Bagley et al., 1995; Chin-Sang & Allen,
1991). African American elders play important roles with regard to childcare,
wisdom and experience. In rural communities, even elders with dementia are
deemed able to perform roles within the kin system (Bagley et al., 1995; Gaines,
1989). However, many African American elders, particularly in urban settings, find
38


themselves increasingly alone and isolated from friends and relatives. In a
qualitative study of thirty elderly Black women, church participation, attendance at
senior citizens centers, and affiliation with friends and other older adults were cited
as the major coping strategies for loss of kin and loneliness (Chin-Sang & Allen,
1991). These women also emphasized that nurturing and helping others provided a
sense of security, camaraderie, love, and belonging.
This study did not specifically target any one ethnic or cultural group, thus
limiting its ability to reveal cultural factors that affect active lifestyles and social
isolation. It is recognized that this is a limitation of the study, given the varied
cultural views of aging, and the unique challenges experienced by those whose
traditional roles are altered by life in the U.S. It is also probable that ethnicity and
culture may have an independent effect on active lifestyles and social isolation, apart
from other neighborhood factors. In addition, it is also probable that in some
neighborhoods, group ethnic or cultural identity affects neighborhood characteristics
(Diez Roux, 2004). Analysis of self-reported ethnic and racial identity was
performed and compared to neighborhood demographics from Census Tract data to
better interpret the study results. However, possible cultural influences on
perceptions and health behavior, as well as neighborhood characteristics, were
limited by the small proportion of ethnic and racial minorities who participated in the
study.
39


Contribution of this Study
Understanding How Context Affects Active Lifestyles for Older Adults
The research discussed above (Garfein & Herzog, 1995; Rakowski, 1997)
disputes the traditional notion that disengagement and separation from society are
normal and inevitable conditions of aging. A primary goal of the present research
was to understand how the specific physical and social content of an individuaTs
activity space, and his or her perceptions of self-efficacy, neighborhood cohesion,
and access to resources, were associated with his or her ability to be physically and
socially active. This information has the potential to make an important contribution
to a more socio-ecological model of aging, which includes maintaining active
lifestyles as a desirable goal and a possible means to prevent social isolation.
One innovation and strength of this study was the ability to access address
information for a cohort of Kaiser Permanente HMO members over the age of 65,
which was matched to participants electronic medical records. This allowed for
efficient random selection of a large population of older adults, and also permitted
the use of GIS to integrate complex information, in order to better understand the
relationship between the environment, individual attributes and perceptions, and
broader indicators of health such as BMI and number of co-morbidities. In addition,
the relationship between physical activity and social isolation was specifically
examined, and its implications for intervention and prevention of the health declines
associated with inactivity and loneliness are discussed.
40


CHAPTER 4
RESEARCH DESIGN & METHODS
The purpose of this study was to determine if neighborhood factors affect
active lifestyles and social isolation of adults, age 65 years and older, either directly,
or indirectly via individual perceptions of self-efficacy with regard to mobility,
safety, and access to desirable amenities (such as banks, pharmacies, markets, health
and recreation centers) (Paul, 1993). The study also investigated whether a
relationship exists between active lifestyles and social isolation.
Neighborhood fLevel 21
Vital Statistics; Socio-demographics, Crime Rates
Structural Factors: Walking Paths, Traffic, Land Use, Aesthetics
Social Factors; Social Capital, Safety, Vitality
Individual fLevel 11
Characteristics: Socio-demographics, Health
Perceptions: Self Efficacy, Pedestrian Safety, Crime Safety,
Access to Resources, Social Cohesion
Active Lifestyle
Social Isolation
Weekly Activity Frequency I l Perceived Loneliness
Weekly Energy Expenditure / \ Social Support
Figure 4.1. Conceptual Model
41


Research Questions and Specific Aims
The key research questions to be answered by the study are:
Question 1: What components, if any, of the environmental context enable or inhibit
active lifestyles and social isolation in older adults?
Aim 1: Use GIS to select neighborhoods that vary with regard to walkability,
crime rate, SES and proportion of adults ages 65+.
Aim 2: Survey a sample of community dwelling older adult members of
Kaiser Permanente Colorado (KPCO) who reside in the selected
neighborhoods, regarding their perceptions of their neighborhoods resources,
safety, and social cohesion, and on the outcome variables of activity and
social isolation.
Aim 3: Use hierarchical regression modeling to analyze the contextual
effects on senior activity engagement and social isolation.
Question 2: What are the potential causal pathways through which neighborhood
and individual factors influence these two outcomes?
Aim 4: Conduct mediation analyses to test for intervening variables between
the initial neighborhood variables and outcome variables
Question 3: Do objective health outcomes cluster by neighborhood?
Aim 5: Use GIS to map individual health data, obtained from KPCO
electronic medical records and individual self-report (e.g., BMI), to reveal
spatial patterns.
42


Aim 6: Revise conceptual models and develop hypotheses for future studies.
Measurable Objectives
Measurable objectives of the study included: identifying neighborhood and
individual variables associated with a more active lifestyle (operationalized as
greater frequency of specific types of activity and energy expended in physical
activity) and less social isolation (operationalized as lower perceived loneliness and
greater perceived social support). The relationship between active lifestyle and
social isolation was also explicitly examined.
Hypotheses
Research Question 1: What components of the environmental
context enable or inhibit physical and social activity in older adults?
Hypothesis 1: Neighborhood structural and social factors, including those
that contribute to: walkability; aesthetics, desirable destinations,
opportunities for exercise, numerous public courtesies, few incivilities, less
territoriality-, high vitality, and high stability, as measured by walking audits
and reported neighborhood data, will be associated with:
1.1. greater frequency of walking for errands;
1.2. greater frequency of community-based activity-engagement;
1.3. higher energy expenditure, and
1.4. less social isolation in adults ages 65 years and older.
43


Research Question 2: What are the Pathways through which
Neighborhood and Individual Factors Influence Outcomes?
i
Hypothesis 2: Specific environmental factors will affect specific behaviors
differently, and the associations may be direct, indirect, or both.
2.1. higher frequency of walking for errands will be associated with:
greater walkability features', numerous public courtesies, and numerous
desirable destinations and will be partially mediated by perceived self-
efficacy for walking, perceived pedestrian safety and perceived access
to resources.
2.2. higher frequency of engagement in community-based activities
will be associated with: greater walkability features, numerous public
courtesies, numerous desirable destinations and fewer incivilities, less
territoriality-, higher vitality, and higher stability, and will be partially
mediated by perceived self-efficacy for walking, perceived pedestrian
safety and perceived social cohesion.
2.3. higher energy expended in physical activity will be associated
with: walkability, desirable destinations , opportunities for exercise ,
high vitality; and will be partially mediated by perceived self-efficacy
for transport (walking versus driving) and access to resources.
2.4. lower perceived loneliness will be associated with: neighborhood
aesthetics; desirable destinations', opportunities for exercise, less
44


individual territoriality; greater neighborhood vitality and will be
mediated by perceived self-efficacy for transport, safety from crime and
social cohesion.
Research Question 3: Do Objective Health Outcomes
Cluster bv Neighborhood?
Hypothesis 3: Neighborhood factors conducive to more active lifestyles and
lower perceived social isolation, are associated with better individual health
indicators (i.e., lower BMI, fewer comorbidities, cardiovascular risk) in Kaiser
Permanente members, ages 65 years and older.
Study Design
A cross-sectional survey and neighborhood observation design was
conducted to understand the association of neighborhood contextual factors with
older adult physical activity engagement and social isolation. Multi-level regression
models were used to analyze results. Methods included finalizing, piloting, and
refining the individual and environmental assessments, and use of GIS to select
neighborhoods and geocode participant address information in order to identify
residents of the neighborhoods of interest. Mailing of study invitations, consent
forms and surveys to randomly selected neighborhood residents followed, and data
collection activities concluded with environmental assessments of neighborhoods.
45


Participant and Neighborhood Sampling Frame
According to Bryk and Raudenbush (Bryk & Raudenbush, 1992), an optimal
sample size for a 2-level model considers the relative cost of sampling level 1 units
(i.e., individuals) versus level 2 units (i.e., neighborhoods) and amount of variability
at each level. Since the time and cost of sampling more neighborhoods is high
compared with sampling more individuals per neighborhood, and variation within
neighborhoods is expected to be relatively large, the number of level 1 units will be
substantially greater than the number of level 2 units. In addition, since the rule of
thumb for survey research is to have at least 10 subjects for each variable measured,
a level 1 sample size of n=200 should be adequate. The smaller level 2 sample size
of eight neighborhoods will be selected to vary on walkability, crime rates, SES and
proportion of residents aged 65 years and older, thus providing adequate power to
detect whether neighborhood factors explains variability in individual activity and
isolation.
Participant Sampling Frame
ARC GIS geocoding software was used to match an address list of 13,928
KPCO members, ages 65 and older, to Denver City and County streets. This list
consisted of all KPCO members who met the following criteria:
1) Currently are active KPCO members
2) Age 65 or older as of July, 2004
3) Reside in Denver City or County
46


4) Have a full street address with zipcode and phone number
5) Registered for an outpatient clinic visit within the past 12 months
6) Have NOT been admitted or discharged from a Skilled Nursing Facility
within the past 12 months.
The list, obtained from KPCO electronic medical records (EMR), was geocoded and
overlaid onto Denver City and County street maps that delineated 77 distinct Denver
statistical neighborhoods. Street map data were obtained from esri.com tigerline
files.
Geocoding
Sixty-nine percent, or 9629 records, matched to street file coordinates. Of
those addresses that matched, 9543 (68%) were identified as falling within the
boundaries of a Denver city or county neighborhood. The others were within or on
the border of Adams, Arapahoe, or Jefferson counties.
Of the unmatched addresses, 124 had only a P.O. Box, with no street address
provided. For the remaining 4175 records, two address matching services available
on the Internet were used to attempt to determine if they were within the Denver
county limits (getzips.com, 2005; mapquest.com, 2005). 3614 records appeared to
contain a street address and/or zipcode that fell outside of the Denver county limits.
The remaining 561 records appeared to have a Denver county zipcode, but may not
have matched due to errors within the street address, such as misspellings or
unrecognized formatting, or errors within the street database, such as missing address
47


range information due to new residential developments. Given their relatively small
number and the large amount of effort it would take to correct these addresses, no
further attempt to match them was made. Thus, by eliminating the addresses that fell
outside of the Denver county limits, the match rate for just Denver City and County
addresses improved from 68% to 93.7%.
j 1
Figure 4.2. Geocoding results for active KPCO members, ages 65 and older.
Neighborhood Sampling Frame
Neighborhoods in the City and County of Denver were selected based on
having at least 80 active independent living KPCO members ages 65 or older plus
variation in 1) violent crime rate per 1000 persons (includes homicides, sexual
assaults, aggravated assaults, and other assaults); 2) average household income;
48


3) proportion of adult residents age 65 years or older; and 4) walkability (determined
using both population density and street connectivity data).
Neighborhood data were obtained from the Piton Foundation
(PitonFoundation, 2004) and web page for the Denver Department of Safety
(CCDOS, 2005). Year 2000 Census data were used for three of the main criteria:
average household income, proportion of adult residents 65 or older and population
density. The latter was calculated by dividing the total population number for a
particular neighborhood by km2. High walkability neighborhoods, were selected
based upon a high population density number plus street connectivity (which was
dichotomized as high if street layouts were gridlike and low if street layouts were
winding or mixed).
2003 Denver Department of Safety statistics (CCDOS, 2005) were used to
determine neighborhood violent crime rates (calculated as the sum of total
homicides, sexual assaults, violent assaults and other assaults per resident population
from Census 2000). Violent crime rate was selected instead of total crime rate since
property and other forms of non-violent crimes encompass such a broad range of
infractions, and are less likely to be associated with perceptions of safety among
older adults (Ferraro & LaGrange, 1992; Ziegler & Mitchell, 2003). In addition the
measure of social cohesion used in this study showed an association between higher
cohesion levels and lower violent crime rates (Sampson et al., 1997).
49


Of the 77 Denver city and county neighborhoods, 5 had too small a resident
population in the year 2000, and therefore lacked census data. The remaining 72
neighborhoods were sorted separately based on each of the 4 variables and a top
stratum and bottom stratum were obtained for each criterion. This was to determine
the potential range for each of the four strata, if all 72 neighborhoods were included.
Neighborhoods with less than 80 KPCO non-cohabitating (i.e., only one
member per household was counted) members residing there were deleted and the
remaining 47 neighborhoods were resorted using the same criteria. Again the top
and bottom five neighborhoods were identified, and these were compared with the
total neighborhood strata. While omitting neighborhoods with fewer than 80 KPCO
members tended to eliminate neighborhoods that were highest in violent crime (none
of the top five high crime neighborhoods appeared on the KPCO member list),
lowest in average household income (only two of the five lowest income
neighborhoods appeared on the KPCO member list), lowest in walkability (only two
of the five neighborhoods designated as least walkable appeared on the KPCO
member list), and lowest in proportion of elderly residents (only one of the five
neighborhoods with the lowest proportion of elderly residents was on the KPCO
member list), the remaining range of neighborhoods was relatively broad and
thought to provide adequate diversity on key characteristics. Thus, the decision to
use a discrete sample was a trade-off between having a well-defined denominator
with access to contact and health data and overall generalizability.
50


Neighborhood Selection
Random selection of the eight study neighborhoods was accomplished using
a computerized research randomizer (Urbaniak & Pious, 2005) that utilizes a
JavaScript random number generator, which produced eight sets of 5 unique random
numbers, ranging from 1 to 50. These were assigned to each neighborhood within
strata that matched the four main selection variables: 1) violent crime rate per 1000
persons; 2) average household income; 3) proportion of adult residents age 65 years
or older; and 4) walkability.
Neighborhoods were then re-sorted accordingly, resulting in a randomly
ordered, ranked list within each stratum. The neighborhood assigned the top ranking
for each stratum (using the random order) was selected to participate in the study. If
a neighborhood had already been selected, the next neighborhood on the list was
chosen.
Once all 8 neighborhoods were selected, the means of the selected
neighborhoods were compared with those of the un-selected neighborhoods for each
of the 4 neighborhood variables. In addition, 2000 census data on the proportion of
non-whites living in these neighborhoods were included to assess whether selected
neighborhoods were representative of Denvers ethnic/racial population.
Comparisons were made using an independent samples t-test. In general, selected
neighborhoods showed no significant differences from un-selected neighborhoods on
each of the variables, with the exception of the more walkable neighborhoods. This
51


is most likely due to the fact that a much larger proportion of Denver City and
County neighborhoods had a grid-like street pattern., regardless of population
density. The more suburban-like street layout, of windy streets and cul du sacs, were
only observed in 17 neighborhoods. Since' both higher population density and street
layout are associated with greater walkability, the selected neighborhoods were
drawn from those with the highest population density, so are not representative of
neighborhoods with lower population density and grid-like street layout. This is
acceptable since it is important that both types of neighborhoods be represented for
comparison purposes. Figure 4.3. illustrates the geographic location of the selected
neighborhoods.
Figure 4.3. Final eight selected study neighborhoods labeled by selection strata
52


Participant Selection and Inclusion Criteria
Kaiser Permanente Colorado (KPCO) was selected due to its cohort of more
than 51,000 active patients, ages 65 or older, in the Denver/Boulder vicinity and
accessible contact information and electronic health data. The fact that all patients
have at least some level of health coverage controls for overly wide differences in
perceived access to health care due to factors beyond the scope of their environment
or transportation. The investigators access to this list of individuals and access to a
very rich electronic medical record (EMR) dataset, (subject to HIPAA regulations
and agreements) were also important considerations in selecting a KPCO sample.
KPCOs Research Review Committee and Institutional Review Board approved the
study design and the use of member database to contact potential participants.
To achieve a study sample of approximately 25 participants per
neighborhood (n=200) a list of patients, ages 65 and older, was drawn from KPCOs
electronic medical records system, identifying all active adult patients, ages 65 and
older. Additional inclusion criteria were: completed a clinic visit within the 12
months prior to the date the list was drawn, were not residing in or discharged from a
Senior Nursing Facility, were residing in a zip code that corresponded to a Denver
City and County statistical neighborhood, and had been at that address for at least 6
months prior. Using GIS and street address data, members whose residences fell
within the boundaries of one of the eight selected neighborhoods were identified. A
subset of 50 individuals per neighborhood were randomly selected using random
53


number tables. These 400 individuals were invited to participate in the study
initially, with the goal more than a 50% (N=200 participants) return rate, which
seemed to be a reasonable estimation for a targeted sample of elderly (Foler, 1988).
Slightly more women than men were expected to participate due to their
over-representation in this older population. Kaiser Permanente Colorado
membership includes approximately 56% women members whose ages are 65 or
older. Therefore, it was expected that at least 56% of the study population would be
women. Since participant accrual was completed by neighborhood, sampling was
stratified to assure that the proportion of men invited to participate reflected that of
the population as a whole.
The KPCO health plan does not routinely collect information on the race and
ethnicity of its members. However, we expected the racial and ethnic representation
of this managed care patient sample to be representative of insured racial and ethnic
populations in the state of Colorado, recognizing that this may not be representative
of the entire population of ethnic and racial minorities, which also includes
individuals without insurance. In addition, since participants were selected from
specific neighborhoods, census data on the neighborhood demographics were
reviewed, with 3 of the neighborhoods selected having a non-white (including
Latino) population greater than 75% (range 7.1% to 95.3%). Specific actions to
increase participation among minority patients included: (a) patient communication
materials that emphasized the importance of conducting research that involves
54


people of all racial and ethnic backgrounds (i.e., the need to develop quality care that
meets the needs of diverse racial and ethnic groups), (b) selection of racial and
ethnically diverse neighborhoods using statistical neighborhoods that are tied to
census blocks, using GIS, and (c) offer of help in reviewing consent forms and
instruments via the telephone.
All potential study participants were mailed an invitation to participate in the
study, an Informed Consent document that explained the project in detail, and a
HIPAA authorization for the use and disclosure of protected health information for
research. A postage-paid refusal postcard was included to allow participants to
indicate if they were not interested in being contacted further for the study. Copies
of both the consent and authorization forms were enclosed, and potential participants
were asked to sign one copy of each form if they elected to participate, return it in
the postage-paid envelope provided within 10 days, and retain the other copy for
their records. Once the signed consent and authorization were received, the first set
of surveys were mailed with instructions on how to complete and return them.
Surveys included questions on: 1) participant background (e.g., race/ethnicity,
income, healthcare utilization); 2) neighborhood environment; 3) neighborhood
cohesion; 4) self-efficacy for walking, driving, or using public transportation; 5)
physical activity, 6) loneliness, and 7) social support.
The direct phone number of the investigator was prominently displayed in all
mailings, encouraging the recipient to call with any questions, or if they needed
55


assistance filling out the forms. Follow-up telephone contact was made to the
potential participant in the event that neither the postcard nor the forms or surveys
were returned within a 10-day timeframe. Additional mailings were done, as needed,
in order to reach neighborhood targets.
Data Collection and Measures
Data were obtained using published GIS data available on all Denver
Statistical Neighborhoods (PitonFoundation, 2004), published business directories,
and primary observational data. The latter was collected using a modified version of
the Systematic Pedestrian and Cycling Environment Scan (SPACES) instrument, an
environmental audit form that provided a structured method for collecting detailed
data on factors related to neighborhood walkability (Pikora et al., 2002). Definitions
and pictures provided in the SPACES observer manual were utilized in order to
minimize subjectivity and maximize inter-rater reliability. The instrument was tested
in two ways: the proportion of segments for where at least three of the four raters
were in agreement (i.e., .75 agreement) and the calculation of kappa statistics for
each item. Intra-rater reliability was also analyzed by calculating perfect agreement
between two consecutive audits of the same segments and by the same observers,
conducted at least 7 days apart.
In addition, items from the Neighborhood Brief Observation Tool (NBOT)
(Caughy, OCampo, & Patterson, 2001) were used to access features of the
neighborhood related to social Capital, such as condition of the built environment
56


(e.g., upkeep of property and proportion of structures that were burned, boarded up
or abandoned), signs of territoriality and protection from crime (e.g., fences or other
border demarcations, neighborhood watch signs, personal security elements such as
bars on windows, security signs or guards, or dogs), public courtesies (e.g., public
trash cans, public phones, public benches, transit stops, parks or recreational spaces),
people and their activities, and non-residential land use (e.g., types of businesses and
services ob served). Average reliability of the instrument was 87%, based on percent
agreement among multiple raters (see Appendix A for adapted audit instrument).
Individual Demographic. Health and Fitness Data
Survey questions to confirm age, income, education, race/ethnicity,
functional mobility limitations, and hospital stays were combined with other mailed
surveys. Additional health data were obtained from the electronic medical records of
all members who had received an invitation to. participate in the study (data were de-
identified and combined by neighborhood for non participants to protect privacy).
The reasons for collecting non-participant data were to allow comparisons to be
made between participants and non-participants, as well as to increase the power to
detect hypothesized health differences between neighborhoods.
Data extracted were the most recent available, within a year prior to or six
months after their study enrollment date. For non-participants, the date of invitation
to enroll in the study was used. For participants, data dates were cross-checked with
self-reported tenure at their current address to assure that all data used were obtained
57


while the participant resided at their current address. Only one datum from one
participant was excluded due to its being collected prior to his moving to his current
address.
Indicators of Coronary Vascular Disease Risk
Health data collected from electronic medical records included the most
common risk factors used to diagnose metabolic syndrome, which increases the risk
of coronary heart disease and type 2 diabetes and is associated with obesity and
physical inactivity. Five out of seven risk factors for metabolic syndrome (AHA,
2005):
1. BMI (as a proxy for central obesity) >30 obese
2. Fasting blood triglycerides > 150 mg/dL
3. Blood HDL cholesterol (<40 mg/dL in men, <50 mg/dL in women)
4. Blood Pressure > 130/85 mmHg
5. Fasting Glucose >110 mg/dL
Other data collected from medical charts included resting pulse rate as an
indicator of level of conditioning, smoking status and the total number of diagnosed
comorbid conditions, associated with an increased risk for heart attack (i.e.,
hypertension, hypercholestemia, diabetes, and heart disease).
Chronic Disease Score (CDS)
A total chronic disease score (CDS) which reflects the total number of
comorbid conditions an individual has, was used as an overall indicator of health.
58


The CDS was constructed using a validated comorbidity measurement tool, the
RxRisk score (Fishman et al., 2003). The RxRisk score is a measure of comorbidity
that incorporates age, gender, health insurance benefit status and an RxRisk category
based on diagnoses derived from administrative pharmacy data. It was originally
developed and validated to identify chronic conditions and to predict cost of health
care, and subsequently revised to assess disease burden in certain populations.
Administrative pharmacy data were used to apply the RxRisk tool to our study
population in order to identify chronic conditions.
Individuals were scored a 1 if they met the higher risk criteria for each factor, they
were scored a 0 if they were within normal limits.
Individual Perceptual Data Regarding Neighborhood Environment
Individual perceptual data were collected using a mailed survey.
Self-Efficacv for Transport
Confidence level ratings (on a scale of 1-10) for walking, driving, or using
public transportation, adapted from self-efficacy assessments used in self-
management studies (Bandura, 1997; Williams, Rodin, Ryan, Grolnick, & Deci,
1998). In addition, multiple choice questions to assess preferences and functional
ability with regard to mobility (e.g., What mode of transportation do you most often
use to get your errands done (e.g., walk, drive myself, get a ride, etc.). If you drive
yourself, do any of the following apply to you: I (hive only during the daylight; I
drive only in good weather; I limit the distance I drive; I limit the routes I will take)
59


were adapted from interview questions asked in studies of older drivers (Fonda et al.,
2001).
The Neighborhood Environment Walkabilitv Scale (NEWS)
This measure was used to assess perceptions of neighborhood factors related
to walking and cycling. It consisted of 9 subscales that were pilot tested within two
San Diego neighborhoods (N=106). One-week test-retest reliability coefficients
ranged from .58 to .80. The scale also showed excellent validity with greater
walkability corresponding to higher residential density, land use mix, access to
resources, street connectivity, aesthetics, and traffic safety. Accelerometers verified
that residents in the more walkable neighborhoods engaged in approximately 52
more weekly minutes of moderate physical activity than those living in less walkable
neighborhoods, explained by greater walking for errands (P=1.04, SE=.50, p=.01)
(Saelens, Sallis, Black et al., 2003).
Neighborhood Cohesion Scale
The scale consisted of 5 items and is a subscale of a larger collective efficacy
questionnaire with questions on both cohesion and informal social control. The
larger questionnaire demonstrated significant (p<.01) positive associations with
friendship and kinship ties (i=0.49), organizational participation (r=0.45), and
neighborhood services (r=0.21). In this study only the cohesion items are included,
due to their close association with the social control subscale (r=.80, pc.OOl) and a
60


need to minimize the length of the survey. The cohesion subscale has a reliability of
.70 (Cagney, Browning, & Wen, 2005; Sampson et al., 1997).
Outcome Measures for Active Lifestyle and Social Isolation
Community Health Activities Model Program for Seniors (CHAMPS)
The CHAMPS is a self-report instrument designed to measure physical
activity in older adults, by assessing frequency per week of specific activities. The
CHAMPS also measures weekly energy expenditure (calories/week), calculated by
using the metabolic cost weights for each activity and adjusting for body weight. In
a study of older adults, 3-month stability coefficients for the frequency and energy
expenditure measures for control group participants were r=.57 and .84, respectively.
When compared with interviewer data and activity logs across four different groups,
the CHAMPS demonstrated good construct validity (p <.001) (Stewart et al., 1997).
A more recent study comparing sedentary older adults (N = 173) with active older
adults (N = 76) demonstrated a six-month stability (estimated on the physically
active control group who were assumed unlikely to change their activity levels)
ranging from 0.58 to 0.67, using intraclass correlation coefficients. All measures
were sensitive to change (p < 0.01), with small to moderate effect sizes (0.38-0.64).
Participants in the study were aged 65-90 yr (mean = 74, SD = 6); 64% were women,
and 9% were minorities (Stewart et al., 2001).
61


The Loneliness Scale
The Loneliness scale (de Jong Gierveld & Kamphuis, 1985) consists of 11-
items that assess perceptions of available companionship, support and help for older
adults. The scale has been used in the Longitudinal Aging Study Amsterdam
(LASA) and the Widowhood Adaptation Longitudinal study (WALS). Reliability
of the scale has been reported in the .80 -.90 range (Cronbachs alpha or rho). It also
is moderately correlated with the Center for Epidemiologic Studies Depression Scale
(CES-D) in total (r=.51) and with CES-D item 14 During the past week I felt
lonely, r=.54).
The Social Provisions Scale
The 24-item Social Provisions scales purpose was to assess the degree that
respondents social relationships provide social support in the areas of attachment,
social integration, reassurance of worth, reliable alliance, guidance and opportunity
for nurturance (Cutrona & Russell, 1987). When tested on a sample of 100 elderly
subjects, the scales internal consistency across all domains was >.70; test-retest
reliability ranged from .37 to .66; and total score correlations with life satisfaction,
loneliness, and depression ranged from .28 to .31 (p<.05). In addition, the scale
correlated with measures of social networks (i.e., number of relationships and
frequency of contact) and satisfaction with social relationships (Cutrona, Russell, &
Rose, 1984; Cutrona & Russell, 1987). Both subscale and total scale scores will be
used in this study.
62


Pilot of survey instruments
To assure feasibility of administering the proposed surveys, 10 older adults of
mixed gender and racial/ethnic background, were asked to complete the survey and
provide feedback, particularly on items not previously standardized. By using
qualitative methods such as open-ended questions, candid information was obtained
to reveal potential barriers to participation, and to identify the need for revisions to
the questions or survey instructions to assure accurate responses.
63


Table 4.1. Summary of Constructs and Measures
INDEPENDENT VARIABLES CONSTRUCTS METHOD OF COLLECTION
Level 1: Individual Factors
Participant Characteristics Demoeraohics - Mailed surveys Includes items on sex, marital status, household members, race, ethnicity, education, income
Health and Functional status -Electronic Medical Records -Questions on mailed survey
Participant Perceptions Mobility Self-efficacv -Walking -Driving -Use of Public Transport - Mailed survevs: Ouestions on mode of transport most used and confidence level for walking, driving, using public transport (5 items)
Safety -from Traffic & Crime - Mailed survevs: NEWS (14 items)
Services & Access - Mailed survevs: NEWS (7 items)
Neighborhood Cohesion - Mailed survevs: Cohesion Scale (5 items)
Outcomes
Active Lilestyle Activity Eneaeement - Mailed survevs: CHAMPS (33 items)
Social Isolation Loneliness - Mailed survevs: Loneliness Scale (11 items)
Social SuDDort - Mailed survevs: Social Provisions Scale (24 items)
64


"able 4.1 (Cont.). Summary of Constructs and Measures
INDEPENDENT VARIABLES CONSTRUCTS METHOD OF COLLECTION
Level 2: Neighborhood Factors
Structural Environment: Residential Density - Neighborhood facts, (Census data at piton.org)
Walkability Land Use mix - Neighborhood facts (piton.org) - Field audit
Street Connectivity - Neighborhood maps - Field audit
Green Soaces - Neighborhood maps - Field audit
Sidewalks - Field audit
Availability of Aesthetics - Field audit
Transportation TransDortation - Field audit: Count of bus or train stops
Social Environment Socio-Demographic's - Neighborhood facts (Census data, at piton.org)
Resources - Neighborhood facts (piton.org) - Field audit
Crime - Neighborhood stats (piton.org)
65


Data Management
As can be seen in Table 4.1, data for the project were collected from
several different sources: a) patient completed data on self-report questionnaires;
b) electronic medical record extraction; c) staff-entered data from GIS and field
audit data. Data were checked at the time of collection for missing values and
out-of-range responses so that errors could be corrected immediately via the
telephone. A system file was created using EXCEL for combining data sets.
Prior to the analyses, all data were again checked for missing and out-of-range
values. Protocols such as locked hard-copy fifes and password-protected
databases were employed to protect patient privacy in keeping with HIPAA
protocols. Preliminary analyses were conducted to assure that all variables met
the assumptions of the analyses to be performed.
Strategies Used to Overcome Implementation Barriers
Potential barriers to the study included 1) poor response rate to survey; 2)
failure to complete all questions due to respondent fatigue or confusion; 3)
inconsistency of published population and neighborhood data for specific
neighborhoods with regard to demographics, crime rates; timeframes, and map
accuracy; 4) privacy concerns with regard to use of individual residence
information versus aggregate data, in order to represent the true context of the
individual.(Jacquez, 1998)
66


Maximizing Survey Response Rate and Completion
To maximize response rates several techniques recommended by Dillman
(Dillman, 1991) were employed. A personalized invitational letter accompanied
surveys explaining their purpose and importance of the research. In addition, a
postage-paid postcard was provided to allow potential respondents to request a
phone call to answer questions or complete questionnaires via the telephone, or
refuse participation in the study. A follow-up phone call was made to non-
responders, followed by repeat mailing of the survey and a final phone-call to
attempt to collect the data over the telephone. Survey respondents were offered a
token thank-you gift for their efforts in returning the survey. All returned surveys
were checked for completeness and responses for missing information were
obtained via the telephone.
Maximizing Accuracy in Characterization of Neighborhoods
To overcome potential bias with regard to inaccurate, dated or inconsistent
neighborhood data, the eight-neighborhood sample was drawn from the 77
distinct City and County of Denver statistical neighborhoods. An advantage of
using only City and County of Denver neighborhoods is that the published census,
housing, economic and crime rate data for the year 2000 are consistently reported
for each neighborhood and are available through a user-friendly GIS website
developed jointly by the Piton Foundation and the Community Planning and
Development Agency of the City and County of Denver (PitonFoundation, 2004).
67


In addition, other sources were used to characterize specific
neighborhoods including telephone directories, and personal observation.
Assuring Participant Privacy and Avoiding Location Uncertainty-
Individual residences were mapped as points within neighborhood units so
that the research staff could collect observational data on the actual environmental
context of the individual using mapping and the SPACES audit instrument.
Observed data were then compared with perceptual data reported by respondents.
Privacy was protected by analyzing the data and displaying the results of the
analyses in aggregate form (Rushton, 2000).
68


CHAPTER 5.
RESULTS OF ANALYSES
Research Questions
The main research questions of this exploratory study are:
Question 1: What components, if any, of the environmental context enable or
inhibit activity engagement and social isolation in older adults?
Question 2: What are the potential causal pathways through which neighborhood
and individual factors influence these two outcomes?
Question 3: Do Objective Health outcomes cluster by neighborhood?
Analyses Used
To answer the research questions, this chapter will describe the statistical
tests used to analyze the data and present the results of these analyses. SPSS
version 11.0 was used (SPSS, Inc., Chicago, IL) to perform most of the analyses
in this report, with the exception of hierarchical regression modeling, which used
SAS version 9.1 (SAS Institute, 2003).
To answer research question 1, univariate analyses of factors related to
neighborhoods and factors related to individuals provided descriptive and
frequency data for the total sample, as well as grouped by neighborhood. Equality
of neighborhood means for continuous variables were assessed using ANOVA or
Kruskal Wallis tests, as appropriate. A Pearson Chi-Square test was used for
categorical data. The magnitude of effect size was calculated using Cohens d
69


(Cohen, 1988; Rosnow & Rosenthal, 1996). Post hoc analyses were performed to
identify sources of neighborhood differences by first comparing 95% confidence
intervals across neighborhoods to focus the analyses, and then performing either
the Dunnett T3 test for continuous data or the Fishers Exact test for categorical
data. A 2-level hierarchical regression modeling technique was used to test direct
associations of neighborhood variables with respondent-reported physical activity
and loneliness. To address the potential for collinearity among the many
predictors tested, separate regressions were used for each independent variable
tested. This required many individual analyses to be performed, which
admittedly increased the likelihood that some of the direct effects were significant
by chance. Due to the exploratory nature of this study, adjustments for multiple
analyses were not made. The very small neighborhood sample size (N=8) and the
fact that such statistical adjustments tend to be overly conservative and would
greatly reduce our already limited power, were weighed as important
considerations. It is understood that this decision greatly increased the chance of a
Type I error and was a trade-off between that and increasing the chance of a Type
II error. In all analyses, the alpha level used was the conventional two-sided .05.
However, given that adjustments for multiple analyses were not made, the reader
will need to interpret the p values presented with this limitation in mind.
To answer research question 2, mediation analyses were subsequently
conducted to test for intervening variables between the initial neighborhood
70


variables and outcome variables (Newsom, 2001). Possible mediator variables
were selected based upon the literature, previously hypothesized relationships,
and whether it made sense conceptually. The mediator was then tested
statistically by adding it to the hierarchical regression model. To test for
mediation a regression analysis to examine the direct association between the
independent variable (IV) and the dependent variable (DV) was conducted. If a
relationship was found then three separate regression analyses were performed to
examine the association between 1) the independent variable (TV) and mediating
variable; 2) the mediating variable and dependent variable (DV); and then to 3)
simultaneously include the IV and mediating variables as independent variables to
the dependent variable. A Sobel test was used to determine whether the indirect
effect (i.e., the reduction in magnitude between the IV and DV that occurred when
the mediating variable was included in the regression model) was statistically
significant. A variable was considered to be a full mediator if its inclusion in the
model resulted in the magnitude of the effect between the independent variable
and dependent variable being reduced by at least 10% and the p value changed
from <.05 to >.05 and if the Sobel t-test had a significance level <.05 (Preacher
& Hayes, 2004). A variable was considered to be a partial mediator if the above
criteria were met, except that in the case of partial mediation, a significant, albeit
reduced, direct effect remained.
71


Gender was also tested as a possible moderator of neighborhood variables
that may affect behavior differentially for women versus men (e.g., perceived
safety from crime). This test was conducted by adding an interaction term (sex x
neighborhood variable) to the main effects hierarchical regression model. While
other sociodemographic variables, such as age and income level, may also be
moderators, these would require more arbitrary distinctions (e.g., sorting
participants into older old versus younger old categories), and thus may be
more difficult to interpret. Since female gender has been shown to influence
decisions about where activity is performed, it is likely to be a factor in this study
(Eyler, 2003; Garcia Bengoechea, Spence, & McGannon, 2005; Krenichyn,
2005).
To answer research question 3, all of the methods described above were
used with health indices from both participants and non-participants electronic
medical records as the dependent variables and neighborhood characteristics as
the independent variables. The two datasets (participant and non-participant) were
compared and then combined to provide greater power to detect clustering of
health outcomes by neighborhood.
All data were cleaned and visually inspected to ensure they met
assumptions prior to analyses. Missing data for neighborhood audits were not
uncommon due to the observational nature of the data collection method. Missing
neighborhood dichotomous data were scored as not observed. Missing
72


neighborhood continuous data were ignored and not included in mean calculations
for that segment. Missing data for individual respondents were minimal, due to
extensive follow-up to assure survey completion, and were generally limited to
either refusal to answer a particular demographic item, such as income level, or
data that were missing from electronic medical records. Because missing self-
reported demographic data were so rare, they were in most cases ignored and the
remaining data were analyzed. Skipped answers for the standardized surveys were
scored according to the instructions for that specific survey. Total Ns were
reported for all analyses.
To aid in the presentation, analyses and discussion of this data this chapter
has been organized into five sections: 1) Characterizing Denver Neighborhoods;
2) Characterizing Study Participants; 3) the Built Environment and Activity
Engagement in Seniors; 4) Social Capital, Social Cohesion and Loneliness in
Seniors; and 5) The Relationship of Neighborhood Characteristics to Health.
73


Section 1: C.haracteri7inp Denver Neighborhoods
Figure 5.1.1. Conceptual Model: Neighborhood Characteristics.
This section analyzes the many variables used to describe and characterize
the eight Denver neighborhoods selected for the study. The literature describes many
factors that come into play when studying die link between neighborhoods and
health. Noise, deterioration of buildings and sidewalks, housing qualify, and
quantity and qualify of green spaces, and the variables of interest in this study,
physical activity and social isolation, have all been suggested as impacting perceived
and actual quality of life and health of older adults (Krause, 1993,1996; Takano et


sociodemographics, crime rates, land use, and quantity and types of destinations
interact with individual and group-level data is an important step to design of
relevant health interventions (Diez Roux, 2004).
Neighborhood data for this study included 1) reported data i.e., socio-
demographics, and violent crime rates, and 2) measured data, i.e., observational data
from the field audits. The latter contained the following constructs: walking, traffic
and permeability (existence and maintenance of sidewalks, traffic calming devices,
and street design); pedestrian safety from crime (surveillance), accidents (pathway
obstructions) and traffic (crossing aids and curb cuts); aesthetics (streetscape and
cleanliness); land use (presence of commercial and recreational destinations); and
social capital variables (public courtesies or amenities; incivilities or signs of
deterioration; territoriality, and vitality or people and activities observed).
Reported Data
Comparison of selected and non-selected neighborhoods
Using data reported from published sources described earlier, eight
neighborhoods in the City and County of Denver were selected from the 25
neighborhoods that had at least 80 independent living KPCO members ages 65 or
older. Neighborhoods were ranked according to 1) violent crime rate per 1000
persons (includes homicides, sexual assaults, aggravated assaults, and other
assaults); 2) average household income; 3) proportion of adult residents ages 65
75


years or older; and 4) walkability (determined using both population density and
street connectivity data), and one neighborhood was randomly selected from the top
and bottom five for each of the criteria (see Table 5.1.1).
Table 5.1.1. Denver Neighborhood Comparisons on Key Socio-Demographic Variables
Neighborhood Variables ; Neighborhood N- Mean Std. Deviation Std.Error Mean
Average Household Income All Denver 72 $55,645.79 25494.78 3004.59
Non-KP Neigh 25 $52,150.81 28184.40 5636.88
KP Neigh 47 $57,504.83 24054.46 3508.71
Selected 8 $48,000.03 13860.14 4900.30
% Residents 65 years or older All Denver 72 11.53 5.65 0.67
Non-KP Neigh 25 9.12 5.91 1.18
KP Neigh 47 12.81 5.12 0.75
Selected 8 12.61 7.69 2.72
Violent Crime Rate . All Denver 72 6.99 5.94 0.69
Non-KP Neigh 25 10.62 8.37 1.67
KP Neigh 47 5.06 2.62 .38
Selected 8 6.58 3.589 1.27
% Non-White - All Denver 72 46.31 28.58 3.37
Non-KP Neigh 25 52.98 29.76 5.95
KP Neigh 47 42.77 27.59 4.03
Selected 8 46.68 32.99 11.67
76


Table 5.1.2 shows that the 25 neighborhoods with at least 80 eligible KPCO
members were significantly different from the complete pool of Denver
neighborhoods on two variables: proportion of elderly and violent crime rates, with
the KPCO neighborhoods having a higher proportion of elderly residents and lower
violent crime rates.
Table 5.1.2. Independent Samples Test between KP Neighborhoods (i.e., at least 80
KP members who met study criteria) and Non-KP Neighborhoods____________
Neighborhood Variables Levene's Test for Equalityof Variances t-test for Equality of Means
F Sig. t df Sig. (2-tailed) : 95% Cl of the Difference
Lower Upper
Average Household Income Equal variances assumed .413 .523 .85 70 .400 -7258.06 17966.09
Equal variances not assumed .81 42.85 .424 -8037.58 18745.62
% Residents 65 years or older Equal variances assumed .050 .823 2.76 70 .007 1.03 6.37
Equal variances not assumed 2.65 43.43 .011 .88 6.52
Violent Crime Rate' Equal variances assumed 18.36 .000 -4.21 70 .000 -8.21 -2.93
Equal variances not assumed - -3.24 26.53 .003 -9.09 -2.04
% Non-White Equal variances assumed .328 568 -1.46 70 .150 -24.22 3.79
Equal variances not assumed -1.42 45.96 .162 -24.68 4.25
Table 5.1.3 illustrates how limiting the neighborhood selection pool to those
with at least 80 eligible KPCO members both narrowed the pool and somewhat
narrowed the range for each variable, although differences between those ranked
high and low for each variable were still quite large.


Table 5.1.3. Neighborhood Variation: Comparison of Denver Neighborhoods with
valanfA/i \Tai a/\/i a am/i 1 Average Household Income % Residents 65 years or older Violent Crime Rate % Non- White
All Denver Neighborhoods Low $12,000 1.55 0 5.6
High $163,000 28.74 38.77 95.3
Selected Neighborhoods Low $35,000 6.23 1.89 7.1
High $109,800 28.74 10.22 95.3
1 Selected from only those neighborhoods that had at least 80 Kaiser Permanente Members in 2004
who met the study criteria of 65 years or older and had seen their physician within the past year.
Measured Data
Field Audit: Inter-Rater Reliability Analysis
Two raters independently assessed a total of 26 segments (5% of total street
miles) on both weekdays and weekends, during the daytime only (See Appendix B
for sample neighborhood maps with selected segments). Raters used items from two
instruments that were adapted with the authors permission to capture aspects of the
environment that are relevant to older adults. Raters were randomly assigned to be
primary or secondary prior to observing each segment, and in cases where scores
did not match, the primary observers rating was used. This decision was based on
the categorical nature of much of the data (i.e., it was either observed or not), which
did not lend itself to being averaged for any specific segment. However, since the
data for all street segments within a neighborhood were ultimately combined into a
neighborhood score, and since both raters had the chance to be primary for an
equivalent number of segments, inconsistencies between raters should have been
reduced.
78


Nineteen pedestrian-related items from the SPACES (Pikora et al., 2002)
instrument were combined with 18 items from the NBOT (Caughy, O'Campo, &
Patterson, 2001), plus another 15 items thought to be relevant to senior pedestrians
(e.g., sidewalk width, curb cuts, timing of pedestrian walk signals, bus stops) for a
total of 52 items. Items were categorized into the domains of Walking Functionality,
Safety (Traffic and Personal), Aesthetics, Destinations, Social Capital, plus a
Subjective Assessment domain to capture the raters overall perceptions of the
neighborhood. Each of the domains contained sub-categories with several variables.
To assess inter-rater reliability, at least 2 randomly selected segments per
neighborhood were assessed independently by both raters at approximately the same
time for a total of 36 segments. More segments were selected in the first
neighborhood assessed, in order to capture inconsistencies early on and identify any
additional training needs. Because all variables were not observed within the selected
segments, inter-rater reliability could only be assessed for 39 of the 52 items (see
Table 5.1.4).
Interrater consistency of scaled items was assessed using Spearmanss rank
coefficient (p), which is appropriate for data that are not normally distributed
(Stemler, 2004). Inter-rater reliability was assessed using a kappa statistic which
provides a chance corrected measure of agreement (Portney & Watkins, 1993).
Since SPSS will not compute kappa for items where raters failed to use the same
range of scores (e.g., rating scale for an item was 1-3; rater 1 scored observations as
79


Full Text

PAGE 1

INDIVIDUAL AND NEIGHBORHOOD EFFECTS ON ACTIVE LIFESTYLES AND SOCIAL ISOLATION IN A SAMPLE OF COMMUNITY-DWELLING ELDERLY: A SOCIO-ECOLOGICAL STUDY by Diane Karen King B.S., State University of New York at Buffalo, 1981 M.B:A., State University of New York at Buffalo, 1983 M.S., Rush University, 1998 A thesis submitted to the University of Colorado at Denver in fulfillment of the requirements for the degree of Doctor of Philosophy Health and Behavioral Sciences 2006

PAGE 2

by Diane Karen King All rights reserved.

PAGE 3

This thesis for the Doctor of Philosophy degree by Diane Karen King has been approved

PAGE 4

King, Diane Karen (Ph.D., Health and Behavioral Sciences) Individual and Neighborhood Effects on Active Lifestyles and Social Isolation in a Sample of Community-dwelling Elderly: A Socio-ecological Study. Thesis directed by Professor Craig R. Janes ABSTRACT There is strong evidence that maintaining an active lifestyle into old age provides important physical and mental health benefits. Understanding the individual and neighborhood factors that facilitate or inhibit active lifestyles in this important population segment will contribute needed data toward the development of multi-level, multi-disciplinary interventions to prevent decline. The overall aim of this study was to investigate the affect of neighborhood contextual factors on activity engagement and social isolation in elderly, community-dwelling adults. Methods: Neighborhood and individual data were collected from adults, age 65 years and older, residing in distinct neighborhoods within. the city and county ofDenver, Colorado. Eight neighborhoods were selected to participate based upon the number of potential participants and variability with regard to neighborhood walkability, socio-economic status, proportion of adults over 65, and crime statistics. A list of patients age 65 or older was obtained from electronic medical records, geocoded, and inatched to the 8 selected neighborhoods using a Geographic Information System. A randomly selected sample (N=871) from the address-matched patient list were invited by letter to participate in the study. 190 seniors completed and returned the survey, providing data regarding perceptions of their neighborhood including access to resources, safety from traffic and crime, and social cohesion. They also reported on their weekly engagement in physical and social activities, loneliness and social support. Individual health data were extracted from electromc medical charts. Environmental data on both structural and social features of each neighborhood were collected using published data as well as an environmental audit tool adapted from the Systematic Pedestrian and Cycling Environment Scan (SPACES) instrument and the Neighborhood Brief Observational Tool (NBOT). Results: Neighborhood structural variables that maximized walking efficiency and mixed land-use significantly predicted greater walking for errands (p<.05). Social capital variables were indirectly associated with more frequent community engagement; more total lV

PAGE 5

and moderate-level P A and less loneliness, and their effects were mediated by perceptions of safety from crime and social cohesion (p<.05). Conclusions: Addressing social factors within neighborhood environments by enhancing perceptions of safety from crime and social cohesion may be just as important-as providing walkable communities with convenient destinations. This abstract accurately represents the content of the ........ .,, ......... ., .,. recommend its publication. Signed v

PAGE 6

DEDICATION I dedicate this dissertation to my husband Gene and my family for your incredible support and understanding, my cohort-member Bonnie, who kept me going, and my terrific committee members, for your patience and mentorship throughout this process.

PAGE 7

ACKNOWLEDGEMENT This dissertation was made possible with support from The Robert Wood Johnson Foundation, Active Living Research Program.

PAGE 8

CONTENTS Figures.......... . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv1 Tables............................................................................. XVIII CHAPTER 1. INTRODUCTION ... -...................................................... 1 The Health Problem. .. . . . . . . . . . .. . .. . .. .. . . . . . . ... 1 Conceptual Model and Study Framework......................... 3 Organization of the Dissertation............ . . . . . . . . . . . . 4 2. THEORETICAL BACKGROUND ................. ..................... 5 Theories of Aging ............................................. ........ -. . ... 5 Theories of Individual Behavior. . . . . . . . . . . . . . . . . . 6 Social Cognitive Theory .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .. .... 6 Human Occupation Perspective............................ 8 Theories of Social Cohesion ............... . . . . . . . . . . . .. 1 0 Social Identity and Integration.................... . .. .. .... 10 Social Capital........ . . . . .. . . . . . .. . . . . . . . . .. 12 Theories of Social Ecology. . . . . . . . . . . . . . . . . . . . . 16 Social-Ecological Perspective.............................. 16 Analyzing multi-level data, concepts & challenges...... 18 Vlll

PAGE 9

3. REVIEW OF THE LITERATURE ...................................... 22 Environmental Influence of Health Behaviors .................... 22 hnportance of Activity to the Health of Older Adults ........... 26 Social Support, Loneliness and Health of Older Adults ........ 30 Social Support ............................................... 30 Loneliness .................................................... 34 Culturally-Mediated Perspectives on Loneliness ........ 37 Contribution of This Stiidy ......................................... 40 Understanding How Context Mfects Lifestyles for Older Adults .. ................................... 40 4. RESEARCH DESIGN AND METHODS.............................. 41 Research Questions & Specific Aims.............................. 42 Measurable Objectives ......................... : ... ,.................. 43 Hypotheses .................. .-......................................... 43 Research Question 1: What components of the environmental context enable or inhibit physical and social activity in older adults? .................................................... ;...................... 43 Research 2: What are the Pathways through which Neighborhood and Individual Factors Influence Outcomes?..... 44 Research Question 3: Do Objective Health Outcomes Cluster by Neighborhood?......................................................... 45 Study Design........................................................... 45 lX -:::--

PAGE 10

Participant and Neighborhood Sampling Frame.................. 46 Participant Sampling Frame................................. 46 Geocoding ... . . . . . . . . . . . . . . . . . . . . . .. 4 7 Neighborhood Sampling Frame............................ 48 Neighborhood Selection .................................. .... 51 Participant Selection & Inclusion Criteria................ 53 Data Collection and Measures........................................... 56 Individual Demographics, Health & Fitness Data................ 57 Indicators of Coronary Vascular Disease Risk. . . . . .. 58 Chronic Disease Score....................................... 58 Individual Perceptions ofNeighborhood Environment.......... 59 Self-efficacy for Tmnsport..................... . . . . . . 59 The Neighborhood Environment Walkability Scale..... 60 Neighborhood Cohesion Scale............... . . . . . . ... 60 Outcome Measures for Active Lifestyle and Social Isolation... 61 Community Health Activities Model Program for Seniors (CHAMPS) .................................. 61 The Loneliness Scale........................................ 62 The Social Provisions Scale................................. 62 Pilot of Survey Instruments... . . . . . . . . . . . . . .... 63 X

PAGE 11

Participant and Neighborhood Sampling Frame ................. 46 Participant Sampling Frame................................. 46 Geocoding......................................... . . . . . .. 4 7 Neighborhood Sampling Frame................ . . . . . . 48 Neighborhood Selection........................ . . . . . ... 51 Participant Selection & Inclusion Criteria ............. 53 Data Collection and Measures........................................... 56 Individual Demographics, Health & Fitness Data.... . . . . . . 57 Indicators of Vascular Disease Risk. ..... ;.... 58 Chronic Disease Score ........................................ 59 Individual Perceptions of Neighborhood Environment.......... 59 Self-efficacy for Transport ................... :.. . . . . . . 59 The Neighborhood Environment Walkability Scale..... 60 Neighborhood Cohesion Scale.......... . . . . . . . . . .. 60 Outcome Measures for Active Lifestyle and Social Isolation... 61 . Community Health Activities Model Program for Seniors (CHAMPS).......................................... 61 The Loneliness Scale........................................ 62 The Social Provisions Scale.. . . . . . . . . . . . . . . ... 62 Pilot of Survey Instruments.................................. 63 X

PAGE 12

Data Management. . . . . . . . . . . . . . . . . . . . . . . . . . .. 66 Strategies Used to Overcome Implementation Barriers........................................................................... 66 Maximizing Survey Response Rate & Completion...... 67 Maximizing Accuracy in Characterizing Neighborhoods................................................ 67 Assuring Participant Privacy and A voiding Location Uncertainty ................ . . . . ..... 68 5. RESULTS OF ANALYSES .......................... 69 Research Questions.......................................... . . . . 69 Analyses Used......................................................... 69 Section 1 .. Characterizing Denver Neighborhoods .......... _..... 74 Reported Data......................................................... 75 Comparison of Selected and Non-Selected Neighborhoods................................................ 75 Measured Data ................................................. .... 78 Field Audit: Inter-Rater Reliability Analysis............. 78 Functionality: Walking...................................... 82 Functionality: Traffic... .. .. . .. . .. . .. .. . .. .. . .. . .. .. .. 84 Safety: Traffic................................................ 86 Safety: Personal............................................... 88 XI

PAGE 13

Aesthetics...................................................... 90 Destinations: Land-Use...................................... 93 Destinations: Facilities...................................... 95 Destinations: Recreation............ . . . . . . . . . . ... 96 Destinations: Public Courtesies........................... 98 Social Capital: Incivilities................................... 100 Social Capital: Territoriality................................ 102 Social Capital: Stability..................................... 105 Social Capital: Vitality... . . . . . . . . . . . . . . . . . .. 106 Subjective Assessment... . . . . . . . . . . . . . . . . ... 108 Discussion...... . . . . . . . . . . . . . . . . . . . . . . . . . ...... 110 Section 2. Characterizing Study Participants..................... 115 Pilot Study. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116 Main Study. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119 P ... Ra arttctpation te ..................................................... 119 P . Ch .. artic1pant aractenstics ........................................... 120 Sociodemographic Variables .............................. 120 Objective Health Variables ................................ 122 P . p arttc1pant erceptions ................................................ 126. Community Mobility......................................... 126 xii

PAGE 14

Neighborhood Environment for Walking................. 128 Neighborhood Social Cohesion............................ 130 Discussion... . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. 131 Section 3. The Built Environment and Activity Engagement in Seniors.................................... 135 Individual Level Outcomes.......................................... 136 Active Lifestyles: Weekly Activity Frequency........... 136 Active Lifestyles: Weekly Energy Expenditure.......... 138 Results of Multi-Level Hierarchical Modeling.................... 139 Research Question 1: Neighborhoods and Active Lifestyles: Does Context Matter?.................................. 139 Hypothesis 1: Summary ofResults................................. 153 Results of Mediation Analyses.................................... 153 Research Question 2: What Are the Potential Causal Pathways through which Neighborhood and Individual Factors Influence Outcomes?......................................... 153 Perceived Safety from Crime and Social Cohesion as Mediators of P A...... . . . . . . . . . . . . . . . . . . . . 164 Gender as a Moderating Variable.......................... . . . .. 166 Hypothesis 2: Summaey of Results .... .'........................... 167 Discussion.. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168 Section 4. Social Capital, Social Cohesion and Social Isolation in Seniors........................................... 172 xiii

PAGE 15

Individual Level Outcomes ...................................... ; .. ; 173 Social Isolation: Perceived Loneliness and Social Support... . . . . . . . . . . . . . . . . . . . . . . 173 Results of Multi-Level Hierarchical Modeling.................... 175 Research Question 1: Neighborhood Influence on Social Isolation: Does Context Matter?......................... 175 Results of Mediation Analyses...................................... 178 Research Question 2: What Are the Potential Causal Pathways through which Neighborhood and Individual Factors Influence Loneliness?........................................ 178 Gender as a Moderating Variable...... . . . . . . . . . . . . . .. 182 Hypothesis 2: Summary of Results................................ 182 Discussion............ . . . . . .. . . . . . . . . . . . . . . . . . . 183 Section 5. The Relationship ofNeighborhood to Health........ 186 Participant and Non-Participant Comparison..................... 187 Health Characteristics of Combined Sample..... . . . . . 188 Research Question 3: Do Objective Health Outcomes Cluster by Neighborhood?............................................. 192 Neighborhood Factors and Cardiovascular Disease Risk........ 194 Hypothesis 3: Summary ofResults................................ 196 Discussion... . . . .. .. . . . . . .. . . . . . .. .. .. . . .. . . . . .. 197 6. CONCLUSIONS AND FUTURE DIRECTIONS..................... 199 Answers to the Research Questions................................ 199 xiv

PAGE 16

Question 1: What Components of the Environmental Context Enable or Inhibit Physical and Social Activity in Older Adults?............................................................. 199 Question 2: What are the Potential Causal Pathways through which Neighborhood and Individual Factors Influence Outcomes?...................................................... 200 Question 3: Do objective health outcomes cluster by neighborhood? ........................................ 201 Limitations of the Study... . . . . . . . . . . . . . . . . . . . . . 202 Generalizability ........................................ ..... .. 202 Culture and Ethnicity...... . . . . . . . . .. . . . . . . . .. 203 Spatial Autocorrelation........................ . . . . . . .. 203 Contributions of the Study....... . . . . . . . . . . . . . . . . . . 203 Neighborhood Characterization............................ 204 Safety and Social Cohesion.. . . . . . . . . . . . . . . . 205 Revised Conceptual Model... . . . . . . . . . . . . . . . 206 Implications for Future Research and Intervention...... . . . . 207 GLOSSARY.............................................................................. 209 APPENDICES A: NEIGHBORHOOD AUDIT & DATA COLLECTION FORM.... 212 B: SAMPLE NEIGHBORHOOD MAPS.................................. 217 BffiLIOGRAPHY................................................................... .... 220 XV

PAGE 17

FIGURES Figure 1.1. Conceptual Model. ........................ . . . . . .. . . .. .. . . . . . .. .. 3 4.1. Conceptual Model. . . . .. . . . . . . . . . . . . .. . . .. .. . . .. . . .. .. 41 4.2. Geocoding results for active KPCO members ages 65 and older............................................................................. 48 4.3. Final Eight Selected Study Neighborhoods............................... 52 5.1.1. Conceptual Model: Neighborhood Characteristics...................... 74 5.2.1. Conceptual Model: Participant Characteristics........................... 115 5.3.1. Conceptual Model: Neighborhood Effects on PA........... ......... ... 135 5.3.2. Mediation effect: yard maintenance with social cohesion............... 157 5.3.3. Mediation effect: window bars with social cohesion ....... 158 5.3.4. Mediation effect: litter with social cohesion.............................. 159 5.3.5. Mediation effect: violent crime rate with social cohesion ........ 160 5.3.6. Mediation effect: yard maintenance with perceived safety.............. 162 5.3. 7. Mediation effect: window bars with perceived safety. . . . . . . . . . 163 5.3.8. Mediation effect: incivilities with perceived safety...................... 165 5.3.9. Mediation effect: window bars with perceived safety .. ."................ 166 5.3.10. Conceptual Model: Direct and Mediated Effects of Neighborhood Variables on Activity Engagement.. . . . . . . . . . . . . . . . . . . . 168 XVI

PAGE 18

5.4.1. Conceptual Model: Neighborhood Effects on Social Isolation......... 172 5.4.2. Mediation effect: window bars with perceived safety................... 180 5.4.3. Mediation effect: window bars with perceived social cohesion........ 181 5.4.4. Conceptual Model: Direct and Mediated Effects ofNeighborhood Variables on Social Isolation............................ . . . . . . . . . .. 182 6.'1. Revised Conceptual Model................................ . . . . . . . . .. 206 xvn

PAGE 19

TABLES Table 4.1. Summary of Constructs and Measures....................................... 64 5.1.1. Denver Neighborhood Socio-Demographic Comparison................. 76 5.1.2. Independent Samples Test between KP and Non-KP Neighborhoods... 77 5 .1.3. Neighborhood Variation: Comparison of Denver Neighborhoods with Selected Neighborhoods, highest and lowest on selection criteria .... :.. 78 5 .1.4. Inter-rater Reliability Results .......................................... ...... 81 5.1.5. The Built Environment for Walking: Sidewalk Functionality............ 82 5 .1.6. The Built. Environment for Walking: Traffic Calming. . . . . . . . . . .. 85 5.1.7. The Built Environment for Walking: Traffic Safety....................... 87 5.1.8. The Built Environment for Walking: Personal Safety..................... 89 5.1.9. The Built Environment for Walking: Aesthetics........................... 91 5.1.10. The Built Environment for Walking: Land Use............................ 94 5 .1.11. The Built Environment for Walking: Destination Types. . . . . . . . . 9 5 5 .1.12. The Built Environment for Walking: Parks and Recreation.............. 97 5.1.13. The Social Environment for Walking: Public Courtesies................. 99 5 .1.14. The Social Environment for Walking: Physical Incivilities. . . . . . .... 101 5.1.15. The Social Environment for Walking: Territoriality....................... 103 5.1.16. The Social Environment for Walking: Stability............................ 105 xviii

PAGE 20

5 The Social Environment for Walking: Vitality............................. 107 5.1.18. Subjective Walking Rating.................................................... 109 5.1.19. Summary ofNeighborhood Characteristics................................. 112 5.2.1. Participant Recruitment Results by Neigh.................................. 119 5.2.2. Sociodemographic Characteristics of Participants by Neigh............. 121 5.2.3. Participant Cardiac Risk Metrics & Chronic Disease Score by Neighborhood .................. :................. 123 5.2.4. Self-Reported Mobility by Neighborhood.................................... 127 5.2.5. Perceptions ofWalking Environment (NEWS Scale results)............ 129 5.2.6. Perceptions of Social Cohesion ....................... ;....................... 130 5.3.1. Frequency ofWeekly Activities (CHAMPS results)...................... 137 5.3.2. WeeklyPA (CHAMPS results)............................................... 138 5.3.3. Multilevel Analysis: Frequency of Walking for Errands.................. 140 5.3.4. Multilevel Analysis: Frequency ofHome-Based PA .......... .. .. . 142 5.3.5. Multilevel Analysis: Frequency of Community-Based Activity......... 144 5.3.6. Multilevel Analysis: Weekly calorie expenditure for total PA........... 146 5.3.7. Multilevel Analysis: Weekly calorie expenditure for Mod PA........ ... 148 5.3.8. Community-based Activity (Self Efficacy for Walking as Mediator)... 156 5.3.9. Community-based activity and yard maintenance (with social cohesion).......................................................... 157 5.3.10. Community-based activity and window bars (with social cohesion).... 158 XIX

PAGE 21

5.3 .11. Community-based activity and litter (with social cohesion).............. 159 5.3.12. Community-based activity and violent crime rate (with social cohesion).......................................................... 160 5.3.13. Community-based activity and yard maintenance (with perceived safety)......................................................... 161 5.3 .14. Community-based activity and window bars (with perceived safety)......................................................... 162 5.3.15. Total PA and incivilities (with perceived safety)........................... 164 5.3.16. Total PA and window bars{with perceived safety)........................ 165 5.4.1. Self-Reported Loneliness and Social Support (LASA and SPS results)....................................................... 174 5.4.2. Loneliness, Social Support, Social Cohesion and PA level............... 175 5.4.3. M:ultilevel Analysis: Loneliness.............................................. 176 Loneliness (with perceived safety from crime)............................. 179 5.4.5. Loneliness (with perceived social cohesion)................................ 181 5.5 .1. Characteristics of participants compared with non-participants for sex, age, CDS and CVD ................... ................ ................ 187 5.5.2. Combined Participant and Non-Participant Objective Health Metrics by Neighborhood ...... ......................................................... 189 5.5.3. Proportion ofParticipants and Non-Participants (combined) at higher MI risk, by Neighborhood..................................................... 191 5.5.4. Multilevel Analysis: Combined sample (N=781) BMI............. ... .... 193 5.5.5. Multilevel Analysis: Combined sample (N=781) CVD.. .. .. .. .. . . .. 195 XX

PAGE 22

The Health Problem CHAPTER I INTRODUCTION Maintaining an active lifestyle into old age, including engagement in meaningful activities such as walking or gardening, has known physical and mental health benefits. These include reduced risks of illness and depression, retention of functional fitness, and decreased mortality (Glass, de Leon, Marottoli, & Berkman, 1999; Tak.ano, Nakamura, & Watanabe, 2002). In addition, an understudied and equally important area of inquiry is the potential relationship between community based physical activity and social isolation .in the elderly. Research conducted on physical activity engagement and use of discretionary time among older adults has shown that as individuals age, the time spent engaged in outdoor and active leisure activities declines(Dallosso et al., 1988; Lee & King, 2003). While physiological realities of aging cannot be overlooked as contributing to the decline in activities that require greater energy expenditure (Shephard, 1997), ecological features of the surrounding neighborhoods as well as individual self efficacy with regard to driving or use of alternative modes of transportation may also contribute to an individual's.tendency to remain at home as opposed to "getting out." Longer life spans will likely result in longer retirements accompanied by decreased participation in what society deems "productive" or ''valuable" activities and roles (Glass et al., 1999), despite the potential to maintain health and 1

PAGE 23

independence into old age (Rakowski, 1997). The biological process of senescence, accompanied by an inactive lifestyle, often results in chronic illness and disability, leading to decreased activity, increased isolation, and loneliness (Khaw, 1997), thus, preserving ability to maintain an active lifestyle into old age may be one way to simultaneously maintain physical and mental health and diminish loneliness. Theories of aging that do not reflect the contextual physical and social environments in which behavior occur, limit our ability to predict behavior change across the lifespan (Rakowski, 1997). There is a need for theory ap.d models that include the interactions among environmental and individual factors that affect the behavior of the elderly. Current research has shown that as individuals age their "activity space" (Cromley & McLafferty, 2002) often shrinks to the locale of their home or immediate neighborhood (Herzog, Ofstedal, & Wheeler, 2002; Lee & King, 2003; Porter, 1994). Thus, more studies of individual and neighborhood factors that affect physical activity engagement in the elderly are needed to develop models that can inform multi-level, multi-disciplinary interventions to prevent decline in older adults. Diez-Roux (Diez-Roux, 1998) emphasizes the importance of developing a causation model that extends across multiple levels when attempting to study the inter-relationships among potentially important variables. The conceptual model presented below summarizes the hypothesized relationships between neighborhood and individual factors targeted in this study. 2

PAGE 24

Conceptual Model and Study Framework Neighborhood (Level ll Vital Statistics: Socio-demographlcs. Crime Rates Structural Factors: Walking Paths, Traffic, Land Use, Aesthetics Social Factors: Capital, Safety, Vitality Individual (Level 11 Active Lifestyle Weekly Activity Frequency Weekly Energy Expenditure Figure 1.1. Conceptual Model Social Isolation Perceived Loneliness Social Support This study provides amulti-level analysis of environmental and individual factors that affect active lifestyles and social isolation of adults, age 65 years and older. Particular attention was paid to structural variables in the neighborhood, such as presence and condition of sidewalks, walking paths, parks and traffic, as well as the existence of amenities that are desirable to older adults, such as banks, pharmacies, markets, health and recreation centers (Paul, 1993), and individual perceptions of neighborhood resources, pedestrian safety and social cohesion. Data on social isolation were also collected and analyzed, given its known association 3

PAGE 25

with health decline (Herzog et al., 2002; Seeman, 1996) and the potential role of an active life style in its prevention or diminution. Organization of the Dissertation This dissertation is organized to follow the author's path of exploration beginning with the theoretical for the study (Chapter 2) and a comprehensive review ofthe literature (Chapter 3). The research questions, specific aims, hypotheses, study design, selection of participants and neighborhoods, and data collection methods and measures are detailed next (Chapter 4). Results and their relevant discussion are presented in Chapter 5, which has been sub-divided into five sections in order to focus the reader on key areas components of the conceptual model: 1. Characterizing Denver Neighborhoods; 2. Characterizing Study Participants; 3. The Built Environment and Activity Engagement in Seniors; 4. Social Capital, Social Cohesion and Loneliness in Seniors; and 5. Neighborhood. Context and Health. The dissertation concludes with a summary of the key study findings, along with study limitations and areas for future research (Chapter 6). A glossary with definitions of key terms used is provided along with Appendices that contain maps, measures and copies ofstudy materials. 4

PAGE 26

CHAPTER2 THEORETICAL BACKGROUND Theories of Aging Traditional theories of aging and lifelong development are primarily stage based perspectives that begin at infancy, progress through childhood and then stop after adolescence (Chapman, 2000). Few acknowledge phenomena that predict behavior during old age; and those that do tend to describe more traditional views of aging, including detachment from society, reflection on the past, and preparation for death as the primary developmental "tasks,(Boeree, 1997). Current beliefs about aging suggest a need for a life course perspective that includes the of many adults to continue to lead productive and socially integrated lives into old age, while coping with the challenges of retirement, illness, and disability (Antonovsky, 1979; Mendes de Leon, Glass, & Berkman, 2003). This proposal draws on several theoretical perspectives to create a model that integrates neighborhood and individual factors that may impact the phenomenon of active aging and its positive health effects. These include theories of individual behavior, i.e., social cognitive theory and the human occupation perspective; theories of social cohesion, i.e., Durkheim's theory of social integration, and various perspectives on social capital; and theories of social ecology. 5

PAGE 27

Theories oflndividual Behavior Social Cognitive Theory Albert Bandura developed his theory of social development in the 1960's and 1970's, which later evolved into what is currently known as Social Cogriitive theory (SCT) (Bandura, 2001). SCT follows an agency perspective that assumes human beings act with intention and forethought, and are capable of adjusting behavior based on social and environmental feedback as well as on self-reflection. The theory is developmental in that behavior is adopted and revised through observational learning, self regulation, and reciprocal determinism (Bandura & Wood, 1989). Observational/earning, or modeling, involves attention to modeled behaviors resulting in actions that are similar to the behavior observed. Self-regulation is a response to the reaction of others with regard to the behavior, which may result in the behavior being repeated, modified, or stopped. Reciprocal determinism emphasizes that cognition, actions and the environment interact and may also impact activity engagement. Within the framework of Social Cognitive Theory, there are three primary cognitive processes: self-efficacy, outcome expectations, and goals. Self-efficacy, which is an indicator of an individual's confidence-in his or her own ability, related to how much control he or she believes to have over events and potential threats, and outcome expectations, which is a belief about the likelihood of success and/or the perceived consequences of failure when the specific behavior is attempted, are key 6

PAGE 28

components of a reciprocal process that produces goal-directed behavior (Bandura, 1997; Clark & Dodge, 1999). When individuals desire to achieve a behavioral goal they evaluate their own beliefs and assumptions as well as information and advice from external sources. They will then either attempt the behavior or not, based on the conclusion they reach regarding their likely success and/or the perceived consequence of failure. An important point to emphasize about self-efficacy is that it is situation and behavior-specific, as opposed to a trait or generalized state. Thus, it would be possible to have high self-efficacy regarding engaging in exercise within one's own house, and low self-efficacy regarding walking for exercise outside, in the neighborhood. For instance, if an older adult believes her neighborhood is unsafe and confirms this with external information, (e.g., broken glass in the garden or loitering teenagers on the comer), she might limit going outside. The reciprocal effect from her behavior may be a deterioration of her yard that engenders more littering, graffiti, and loitering around the vicinity of her house. Thus, individual characteristics, cognition, behavior, and environmental factors maintain a dynamic interaction that leads to multi-directional outcomes. Many older adults report that maintaining their independence and living their lives on their own terms are the most important determinants of their life satisfaction (Langlois et al., 1997; Letvak, 1997; Porter, 1994). Because self-monitoring of behavior is an important component of social cognitive theory (Bandura & Wood, 1989; Baranowski, Perry, & Parcel, 1997), interventions aimed at improving 7

PAGE 29

perceptions of autonomy and control often target increasing self-efficacy for a specific behavior or set of behaviors. Self-efficacy also contributes to autonomy and control by helping to shape aspirations and goals as well as predicting outcomes (Senecal, Nouwen, & White, 2000). Bandura and Locke (Bandura & Locke, 2003), in a review of nine large-scale meta-analyses across a range of activities, found consistent evidence that beliefs about self-efficacy are a significant factor in both motivation and performance, and that perceived self-efficacy is a strong and independent predictor of future performance (Bandura & Locke, 2003; McAuley, Jerome, Marquez, Elavsky, & Blissmer, 2003). For older adults in urban settings, the notion of pedestrian self-efficacy is as important to maintaining independence as driving in more suburban settings (Fonda; Wallace, & Herzog, 2001). In a study of older pedestrians in New Haven, Connecticut, a telephone survey found that 11.4% of residents aged 72 and older reported problems crossing the street (Langlois et al., 1997). Because of this connection between mobility and independence, understanding perceived self efficacy regarding mobility (walking, driving, and use of alternative modes of transportation) is included in this proposal. Human Occupation Perspective The Human Occupation perspective is the major tenet for the occupational therapy discipline and is derived from multiple of references including cognitive behavioral theories that emphasize learning and adaptation, developmental 8

PAGE 30

theories where roles and priorities change along a continuum, and socio-ecological systems theories that emphasize how human occupation is effected by individual skills, roles, values, interests as well as the environmental context in which activities are performed (Dutton, Levy, & Simon, 1993). These three theories combine to produce the key assumption that is the core of the Human Occupation perspective: human beings, throughout their-lifespan, have an innate desire for mastery of their environment that produces an "occupational" nature. Mastering one's environment requires the ability to continuously change and adapt, not only for survival, but also for the purpose of self-actualization (Hopkins & Smith, 1993). Having an occupational nature encompasses a wide range of purposeful behaviors that occupational therapists typically group into the broad domains of activities of daily living, leisure, and work (K.ielhofner, 1997; Kielhofner, Burke, & Igi, 1980). For older adults, activities of daily living include activities that are basic and instrumental, such as grooming, eating, taking medication, and driving. Leisure activities may include social activities with friends and family, group recreation, as well as solitary pastimes such as gardening, reading, or engagement in hobbies. Work may include barrie management activities such as shopping, banking, and household maintenance, as well as caring for grandchildren or spouses, educational activities, and vocational activities such as volunteer work and paid work. Self-identity is fashioned from the many occupational roles an individual assumes in order to perform a wide variety of activities. For example, one individual 9

PAGE 31

may simultaneously hold the varied roles of worker, parent, grandparent, spouse, and friend. As individuals age, some or all of these roles may be lost through retirement, children leaving home, and the death of spouses and friends. Thus, retention of meaningful roles and occupation is theorized as preserving self-efficacy and health (Bryant, Corbett, & Kutner, 2001). Because Human Occupation theory specifically targets activity engagement as healthful, it is an important component of an 'Active Lifestyles' model of aging. While Active Lifestyle, for the purpose of this study, is operationalized as total energy expenditure, questions about community-based activities (both active and sedentary) are included in the proposed measures, and will be examined to better understand the mix of activities performed by this population. Theories of Social Cohesion Social Identity and Integration Emile Durkheim, in 1897, discussed how individuals integrate into society in his study of suicide as a social phenomenon (Durkheim, 1897). He that healthy individuals formed attachments with groups and the strength of those attachments defined their own identity as individuals or as group members. Durkheim concluded that while suicide is an individual act, it was also a social act, resulting from disturbance to the social equilibrium. In his studies of suicide rates in different European countries as well as rates between different religious groups, Durkheim found that suicide was relatively stable within a society, and highly 10

PAGE 32

variable among societies. Suicide could be anti-social or "egoistic", when an individual has low attachment to society, or "altruistic", when an individual's group identity supersedes h{s or her own and death is seen as way to advance the group. In addition, he observed that suicides occurred in societies that were very disorganized and unregulated. These he called "anomie", and were akin to "egoistic" suicide, but the former differs in its emphasis on society's deficiency in regulating individuals as opposed to the individual's own low attachment. Durkheim' s theory emphasizes the importance of social integration and cohesion to individual mortality (Berkman, Glass, Brissette, & Seeman, 2000). McMillan and Chavis (McMillan & Chavis, 1986) define social cohesion as having four elements: 1) membership; 2) influence; 3) integration and fulfillment of needs; and 4) shared emotional connection or history. Theability to trust and reciprocity from neighbors has been shown to be a positive predictor of social cohesion (Macink.o & Starfield, 2001; Sampson, Raudenbush, & Earls, 1997). Having a sense of belonging, as well as being able to adapt and participate in life within a defined community, has known benefits to the health of older adults including improved prognosis for recovery from illness, improved immune function, and decreased mortality (Patrick & Wickizer, 1995; Seeman, 1996). Durkheim' s theory of social integration is reflected in studies of how communities maintain order and enforcement of social norms. For instance, Sampson and Raudenbush studied neighborhoods and violent crime and found that 11

PAGE 33

neighborhood norms of reciprocity, trust, and a willingness to intervene in order to prevent social disturbances, such as truancy or teenage loitering, were significantly associated with reduced violent crimes. The authors terlned the combined phenomenon of social cohesion and informal social control as "collective.efficacy", a concept that is related to, but not synonymous with Bandura's use of the same term to describe individual beliefs about the collective capability, shared knowledge and skills of a group to achieve desired outcomes (Bandura, 1997). Sampson and Raudenbush's measure of social cohesion will be used in this study (Sampson et al., 1997). Social Capital Society is not the sum of individuals just as population health status is not equal to the composition of indiVidual risk factors, but is determined by collective characteristics of commwrities and societies (Kawachi & Berkman, 2000). According to Macinko and Starfield (Macinko & Starfield, 2001) the social capital construct has a long history of use, beginning with Hanifan in 1920, who defined it as "good will, fellow.ship, sympathy and social intercourse among the individuals and families who make up a social unit." The works ofMarx, Durkheim, Bourdieu, and the more contemporary Coleman and Putnam, have engendered diverse perspectives on social capital, with definitions ranging from the individual to more structural/political concepts. For instance, social capital has been defined as resources that accrue to individuals via social connections, group level norms, 12

PAGE 34

cohesion and participation (Coleman, 1988) and also as the existence of orgairizations and resources within a community that can be accessed by its members. The question of whether a "tight" community with strong values, norms and sanctions also has strong social capital has engendered much discussion (Coleman, 1988; Putnam, 1995). From a Durkheimian perspective, attachment to a social group may be protective against despair and other negative health consequences (Durkheim, 1897). From a Marxist viewpoint, communities that are poorer in assets may experience demoralization, resentment of the society at large, and consequently lead to negative health outcomes (Kawachi & Berkman, 2000; Waitzkin, 1988). Bourdieu examined how possession of the "right" credentials, education, occupation, or taste, may dictate whether or not individuals have wealth, legitimacy or status within society, and can access the form of capital most valuable to them, i.e., economic, cultural, or social capital. His notion of"habitus," i.e., the typical customs, attitudes and behaviors practiced by groups or individuals, as a determination of their position within a larger society, may explain why marginalized groups who experience strong social cohesion often lack the resources enjoyed by those who are integrated into the mainstream. Bourdieu emphasized economic capital as the foundation of all types of capital (Bourdieu, 1984). Coleman, like Bourdieu, viewed social capital as membership in social groups, an especially important aspect of adolescent life, where approval from peers 13

PAGE 35

is of preeminent importance. However, his theory does not account for other types of capital that may be even more important to ensuring success in life versus mere acceptance (Coleman, 1988). Putnam views social capital as residing in communities, including not only a Durkheimian sense of belonging, but also a sense of trust, reciprocity, and cooperation, and is a proponent of activism to achieve these goals (Putnam, 1995). His view shares many ofthe elements of a true structural perspective, where social capital is an element of social cohesion that has to do with specific features of social structures that proyide resources for individuals and facilitate collective action (Bourdieu, 1984; Kawachi & Berkman, 2000; Putnam, 1995). These features include trust, available social organizations, norms and sanctions, and information channels. Structuralists consider social capital as ecological and external to individuals, as opposed to social networks that are often measured at the individual level (Kawachi & Berkman, 2000). From this perspective, social capital is a public good that all members of society can benefit from, although individual contributors may only reap a small part of the benefits themselves. The lack of agreement as to whether social capital resides in individuals, groups, or communities, presents. challenges for measuring the concept of social capital as well as its effect on health outcomes, and is a chief reason for criticism (Hawe & Shiell, 2000; Macinko & Starfield, 2001; Morrow, 1999). 14

PAGE 36

The mechanism that links social capital to health outcomes can be evaluated in several ways. First, compositional effects of social capital (e.g., attributes of the individuals who reside in a given community) indicate that individuals who are more socially isolated and less integrated, reside in areas where social capital is low. Second, contextual effects associated with the shared environment (e.g., at the state or neighborhood level) suggest that social capital influences health-related behaviors and access to services and amenities. Health-related behaviors may be affected due to promotion of rapid diffusion of health information, and increased likelihood that healthy behavioral norms are adopted. Access to services and amenities include transportation, recreation, neighborhood health resources, and existence of local groups who can lobby for services. Psychosocial processes affected include provision of support, trust and reciprocity. Third, social capital may influence policy on a broader scale, as it has been shown that states that have low levels of interpersonal trust are less likely to invest in human security and social safety net programs (Kawachi & Berkman, 2000). The definition of social capital is still evolving, and it is unclear at present as to the best way to operationalize it for practical application (Hawe & Shiell, 2000). Measurement of social capital is primarily of aggregate variables that are made up of individual responses to social surveys and more integral variables that involve direct social observation ofneighborhoods(e.g., number of establishments that still accept personal checks may be an indicator of trust). Thus, while conceptually interesting, 15

PAGE 37

further honing of the definition and its unique contribution to our understanding of its potential impact on health outcomes is warranted. For the purpose of this study, the aspects of social capital that will be examined include individual perceptions of access to resources, social cohesion and safety from crime, that may impact their activity levels, as well as individual perceptions of loneliness and social support, objective neighborhood factors -such as existence of resources, reported crime rate, and observed public courtesies (such as public benches and transportation) and incivilities (such as vacant lots, home security features, litter and graffiti). Theories of Social Ecology Social-Ecological Perspective The notion of "reciprocal" relationships between environmental factors and behavior is a key part of social cognitive theory 1986; Bandura, 1997), human occupation theory, and other theories of social ecology that delineate multiple levels of influence on individual behavior (Dzewaltowski, 1997; Saelens, Sallis, & Frank, 2003; Sallis & Owen, 1996). The impact of ecological factors on health is a key tenet of public health and was popularized in the 1800's. More recently, theorists and researchers in a variety of disciplines are acknowledging that there are multiple levels of influence on behavior. This is in contrast to earlier psychological and sociological perspectives that greatly emphasized individual agency over social/environmental influence. Psychologists, who made early contributions to the social-ecological perspective included Kurt Lewin, who in the 1930's theorized that 16

PAGE 38

perceptions of the external environment had an important influence on behavior, Roger Barker in the late 60's who believed environments_ had a direct influence on behavior, and Urie Bronfenbrenner in the late 70's who described how individual behavior and development are influenced by multiple .levels (Brown, 1998; Bull.& Shlay, 2005, In press). Brofenbrenner developed a multi-level model that emphasized interactive systems, described as microsystems (e.g., an individmil's immediate environment such as their family, home and neighborhood), mesosystems, that link individuals with other settings important to their day-to-day lives (e.g., school and work), exosystems, which affect individuals indirectly (e.g., community organizations, spouse or parent's workplace), and macro systems that create the larger cultural environment (e.g., policy, media, and norms) (Brofenbrenner, 1979; Brown, 1998). Sallis and Owen (Sallis & Owen, 1996) outline five common tenets of ecological models: 1) there are multiple dimensions that influence health behaviors; 2) these various dimensions interact with each other; 3) within the environment there are multiple levels of influence; 4) environmental factors directly influence behaviors; and 5) different environmental factors will influence specific behaviors differently, so ecological models should specify which factors influence which behaviors. A socio-ecological perspective was applied in this. study to evaluate both neighborhood and individual variables that niay influence individual levels of \. 17

PAGE 39

physical activity and loneliness. Multi-level modeling, a method that has been used in past research to better explain the multiple influences on individual behavior (Li, Fisher, Bauman et al., 2005), was used to increase understanding of the complex associations among neighborhoods, individuals and individual behavior. Analyzing Multi-Level Data; Concepts and Challenges Contextual analysis is the study of the effects of collective or group characteristics on individual level outcomes (Diez-Roux, 1998). Sometimes called "multi-level analysis" it measures the iinpact of ecological independent variables on individual dependent variables. Group-level variables (also known as ecological; macro-level; contextual; or aggregate) illustrate that individual risk for disease depends not just on individual factors (e.g., income; health) but also on community factors (resources; safety). For instance, some studies have shown that high levels of community unemployment are associated with increased individual stress levels regardless ofthe individual's actual employment status (Diez-Roux, 1998). There is a conceptual distinction between group-level and aggregate individual-level variables. Group level variables can be derived or integral in nature. Derived variables are analytical or aggregate in that they summarize composites of individual characteristics and are usually presented as means, percentages, medians, and other "distribution" variables. Derived variables often have analogues at both levels (e.g., individual income and mean neighborhood income) but measure very different constructs (i.e., the purchasing power of an individual versus the resources 18

PAGE 40

available to all individuals in that neighborhood). Derived variables may influence integral variables. Integral variables describe characteristics of the group, not derived from its individual members (e.g., regulations, availability of health care, political systems, population density). Integral variables have no analogues at the individual level (Diez-Roux, 1998). An illustration of these various level variables is as follows: the number of community organizations to which eaqh individual belongs (individual level); the percentage of persons in the community who belong to at least one organization (group level, derived); the number of organizations in the community (group level, integral). The analytical challenge of examining multi-level data and drawing the correct inferences is to avoid fallacies that arise when data collected at one level are used to draw conclusions about a different level (Diez-Roux, 1998). The often-cited Ecological or aggregate fallacy involves drawing inferences at the individual level based upon data collected at the group-level. For example: the finding that increased per capita income at the country level is associated with increased motor vehicle deaths, does not mean that individuals with higher income have more motor-vehicle deaths. It may mean that wealthier countries as a whole contain more drivers. The flip side is the Atomistic fallacy that involves drawing inferences at the group level based on data collected at the individual level. For example: even ifwithin countries, increased individual income is associated with decreased coronary heart disease (CHD) mortality, at the country level, increased per capita income may be 19

PAGE 41

associated with increased CHD death rates. Such inferential fallacies can be overcome if the data collected, and analyses conducted, match the level at which the inferences are to be drawn. Two other fallacies that support using a multi-level approach include the psychologistic (individualistic) fallacy and the sociologistic fallacy. The former assumes that social phenomena can be understood by looking at individual characteristics and the latter that assumptions about individuals can be drawn from studying social phenomena (Diez-Roux, 1998). Thus it is important to include relevant factors from multiple levels to avoid these latter two fallacies and account for confounders that may have an independent effect on the dependent variable and are not a part of the hypothesized "causal chain". Selection of the appropriate level data sources, measures and construct is an important process. While the level ofthe construct and the level of the data source do not have to be the same (e.g., surveying individuals about a group construct such as their perceptions of community cohesion), it is important to justify aggregating individual level data to perform a group-level analysis (Hofmann, 2004). In addition, defining the contextual unit and variables appropriately is another key challenge of designing socio-ecological studies (Diez-Roux, 1998). For instance, specifying the boundaries of a "community'' or "neighborhood" is often subjective. Most individuals belong to a variety of overlapping social and spatial contexts ( neighborhood, work, school, organizational affiliations, and virtual communities) whose effects may be difficult to separate. Another issue that may arise is that 20

PAGE 42

individual-level outcomes within groups may be correlated; even after group and individual-level variables are controlled for,. due to other shared contextual variables. Developing models that illustrate the linkages social structure and health outcomes are important to developing approaches to study these processes (Marmot, 2000). For older adults, social and spatial phenomena that exist within their immediate surroundings are important due to an assumed narrowing of activity space, in many cases. This study will be limited to the neighborhoods in which the selected individuals reside. The neighborhood boundaries used in this proposal are the Statistical Neighborhoods defined by the City and County of Denver, linked to Census Tracts. To address the fallacy, data were collected at both the neighborhood level and the individual level to avoid making inferences about individuals based on neighborhood demographics. A two-level hierarchical modeling approach was used to determine whether the participants' activity and social isolation outcomes were affected by the physical and social characteristics of their neighborhood. 21

PAGE 43

CHAPTER3 REVIEW OF THE LITERATURE Environmental Influence of Health Behaviors The assumption that environmental variables may affect behavior and health outcomes is the rationale for community-based public health programs, education, and inteiVentions (Patrick & Wickizer, 1995). Support for this assumption is found in studies that control for individual sociodemographic variables, access to health care, and other individual-level variables, yet disparities in health outcomes (De Bourdeaudhuij, Sallis, & Saelens, 2003; Saelens, Sallis, Black, & Chen, 2003; Saelens, Sallis, & Frank, 2003; Takano et al., 2002) are still present. In a review of studies that looked at the association between neighborhood built environmental variables and physical activity, significant differences in walking and cycling were found based upon street layout and the mix of residential and commercial space (Saelens, Sallis, & Frank, 2003). These findings were replicated by Saelens, et al. (Saelens, Sallis, Black et al., 2003) who obseiVed that neighborhoods designed around car use contributed to a more sedentary lifestyle. In their study of the relationship between physical activity and neighborhood walkability, they found that residents in neighborhoods with higher residential density, a mix of residential and commercialamenities, and gridlike street patterns recorded more weekly minutes of moderate level aCtivity than those residing in neighborhoods with long, winding streets and few retail establishments. However, few studies (Fisher, Li, Michael, & 22

PAGE 44

Cleveland, 2004; Li, Fisher, & Brownson, 2005; Michael, Green, & Farquhar, 2005) have focused on older adult activity within neighborhood contexts. A recent study of 582 community-dwelling seniors recruited from 56 Portland neighborhoods did find significant increases in levels of individual physical activity that were associated with neighborhood-level variables as well as neighborhood social cohesion after 'controlling for individual-level variables (Fisher et al., 2004). These findings were promising, and the importance of social cohesion to both activity engagement and social isolation were specifically explored withinthis study. When studying the link between neighborhoods and health, many factors come into play. Environmental toxins, noise, overcrowding, deterioration of buildings and sidewalks, and housing quality have all been suggested as impacting the health of older adults (Krause, 1996). Neighborhood deterioration has also bee:n linked to social isolation of older adults due to fear of crime and reduced neighborhood cohesion (Kr8:use, 1993). Alternatively, neighborhoods with walkable parks and tree-lined streets were found to increase longevity in a prospective study of Japanese elderly (Takano et al., 2002). Developing models that illustrate the relationships among neighborhood contextual factors such as land use, transportation, and senior serVices, and individual and group-level data, is an important step to developing relevant health interventions (Diez Roux, 2004). An example of differential access to services and resources for physical activity was illustrated in a study by Estabrooks, et al. (Estabrooks, Lee, & Gyurcsik, 2003). This 23

PAGE 45

study found that high SES neighborhoods had a significantly greater number of free resources for physical activity than did the low and medium SES neighborhoods. Despite the importance of the physical environment to active lifestyles, it is unknown whether the actual environment or individuals' perceptions about their environment have the greater influence on physical activity (Brownson et al., 2004). Measures of residents' perceptions of their neighborhood's walkability, safety, aesthetics, access to recreation facilities, and other amenities, have been shown to have good test-retest reliability, particularly with regard to specific elements of the built environment, such as estimating distances to destinations and presence of sidewalks. Determining the validity of those perceptions using Geographic Information Systems (GIS) and neighborhood audits have shown mixed results (Kirtland et al., 2003; Pikora et al., 2002). Kirtland, et al. (Kirtland et al., 2003) compared perceptions of supports for physical activity, such as existence and condition of sidewalks and public recreation facilities; street lighting; and overall aesthetics, with objective environmental measures using a sample of 1237 residents of Sumter County, South Carolina .. Self-reported physical activity frequency and duration, and perceptions of the immediate neighborhood (i.e., surroundings within a .5-mile radius or 1 0-minute walk), and the community (i.e., destinations within a 10mile radius or 20-minute drive), were included. Geographic Information Systems, a computer-based tool that allows the user to capture, store, retrieve, analyze and visually display database information that is "spatial" in nature (Cromley & 24

PAGE 46

McLafferty, 2002), was then used to create a map of objective supports and barriers to physical activity, and to confirm distances between respondent residences and specific locations. Overall agreement between perceptions and objective data for neighborhood items was highest for access to sidewalks, access to public recreation facilities, safety/crime, public spending, and streetlights. Perceptions of access to recreation facilities were highest for those reporting at least some physical activity as opposed to those categorized as iriactive. Community items showed the highest agreement for access to malls. Other community items, such as public pools, trails and parks showed low agreement, suggesting that perceptions may better match reality for more proximal destinations, and/or those that are accessed more frequently. Thus, inclusion of both perceptual and objective measures is recommended to adequately characterize features of neighborhoods that may affect active lifestyles (Brownson et al., 2004; Kirtland et al., 2003). Along with the increased interest in the importance of context to health, discussions of how best to measure, analyze and interpret socio-ecological effects are actively evolving. Oakes (Oakes, 2004) cautions researchers about making causal inferences with regard to neighborhood effects on health when observational data are used alone, without intervention or treatment. Oakes' reasoning is that inferences about neighborhood residents that are based on neighborhood factors are inherently confounded, since the neighborhood characteristics are not independent of the resident characteristics. Diez Roux (Diez Roux, 2004) disagrees that resident and 25

PAGE 47

characteristics are necessarily interdependent, though she stresses the need to directly measure objective and integral neighborhood characteristics that are independent of resident characteristics. She also emphasfzes the importance of not characterizing neighborhoods using group level demographics as proxies for other attributes, e.g., using neighborhood socio-economic status to indicate lack of neighborhood resources or high crime rates, that could be more directly measured (Diez Roux, 2004; Oakes, 2004). The present study evaluated the effect of both physical and social characteristics of neighborhoods on older adult physical activity and social isolation by characterizing neighborhoods using both independent, objective neighborhood variables such as population density, street lay out, and parks in addition to potentially interdependent variables such as violent crime rates, land use mix, and neighborhood aesthetics. Importance of Activity to the Health of Older Adults Studies of"successful" or "robust aging" dispute the traditional "Western" notion that disengagement and separation from society are "normal' and inevitable conditions of aging (Garfein & Herzog, 1995). In a study of well-elderly, participants in a group program designed around the Occupational Therapy (OT) principle that activities ofhurnan occupation (i.e., self-care, work, play and leisure activities) are beneficial to physical and mental health and well-being, did achieve more benefits in life satisfaction, health perceptions, and overall mental health than 26

PAGE 48

those who participated in either a no-treatment control group or a social activity group (Clarket al., 1997). The OT group targeted basic activities of daily living (ADLs) such as grooming and instrumental ADLs such as using transportation and shopping. The OT group also focused on preserving independence by providing skill building in the areas of safety, joint protection, energy conservation, exerCise and nutrition. Participants in the OT study were ethnically diverse, implying that focus on purposeful activity and maintenance ofindependence is valued cross-culturally. However, all participants, by nature of the program's design, needed to be motivated, healthy and mobile enough to travel to the program. Dallosso, et al. (Dallosso et al., 1988) used an activity inventory for older .adults in England to assess participation in four categories of activities: productive; indoor productive; leisure; and walking. A large sample (n=l042) of both men and women age 65 and older was interviewed. Only 53% performed outdoor activities, including gardening (88%), with a small percent performing car maintenance and house repairs. Leisure physical activities were least reported, including social walking (22.4%) and cycling (5.5%). Significantly more men reported participation in outdoor and leisure activities, including walking. Lee & King (Lee & King, 2003) investigated how differential use of discretionary time among older adults affects energy expenditure. They evaluated data from two studies designed to increase physical activity among older adults, to determine the impact on use of discretionary time. The Community Health 27

PAGE 49

Activities Model Program for Seniors (CHAMPS) questionnaire was used to assess weekly activities of varying metabolic equivalent (MET values). The findings were that the majority of activities performed were less than 3 METs, i.e., required less than 3 times the resting energy expenditure. Activity choices varied by gender with women spending more time engaged in social activities such as visiting, helping others, and attending meetings. Thus, activities selection in old age correspond to their physical demands, underscoring the importance of limiting environmental restrictions to promote more physically active lifestyles. Herzog, et al. (Herzog et al., 2002) categorized activities as productive or helping, i.e., activities that produce a good or a service; educational or intellectual, including activities such as reading, taking courses, or using computers, and leisure, including both formal pastimes such as classes, group travel and outings, and informal pastimes such as attending movies or concerts, walking, knitting or cooking. Everard, et al. (Everard, Lach, Fisher, & Baum, 2000) present the results of survey research conducted with a convenience sample of 244 members of an organization for older adults. All participants were community dwelling, aged 65 and older. The purpose of the survey was to study the association between engagement in activities and function. Male gender was associated with higher physical health scores. Performance of instrumental ADLs andhigh-demand leisure activities was positively associated with physical health. Maintenance of low28

PAGE 50

demand leisure activities was independently associated with better mental health. Interestingly, this was the only activity item that was significantly associated with increased mental health (p =.0001). Because this is a cross-sectional study it is unclear if better physical health promotes maintenance of more physically demanding activities, or if performance of these activities promotes better physical health. The finding that maintenance of low-demand leisure was associated with better mental health, also cross-sectional, is interesting since it implies that even sedentary activities may provide some important health benefits, particularly to older adults whose mobility may be impaired. The mechanism for this was not explored, but it is consistent with other studies that have found associations performance of social, sedentary activities and greater life expectancy (Glass et al., 1999; Sugisawa, Liu, 1994). This latter finding is iinportant for many older adults who are isolated or homebound due to chronic disease or disability, caregiving responsibilities for an afflicted spouse, or limitations with regard to transportation and financial resources, making external socialization and activities difficult. Emphasis on how solitary pursuits may be employed to defray loneliness need more attention (Rane-Szostak & Herth, 1995). For example, crafts, listening to music, gardening, and reading have been cited by some older adults as fulfilling activities that stave off loneliness. Such activities, as suggested by Csikszentmihalyi' s theory of optimal experience, provide a state "flow" (Csikszentmihalyi, 1991, 1997). Csikszentmihalyi defines "flow" as 29

PAGE 51

the times when individuals are fully absorbed in activities that cause them to lose all sense oftime and provide them with a feeling of satisfaction. He proposes that such "flow" experiences enhance feelings of control, mastery, and enjoyment and usually entail concentration, clear goals, immediate feedback, effortless involvement, an altered sense oftime, and a chance for completion. The present study focused on engagement in a wide variety of activities, particularly those that promoted greater energy expenditure, as a way to measure "active lifestyles" in older adults. While the CHAMPS was used to measure physical activity level, sedentary social and leisure activities were also. analyzed, since they may provide health benefits, regardless of MET level. Individual activity mix was quantitatively and qualitatively analyzed to determine the frequency of engagement in specific categories of activities and are included in the analysis. Social Support. Loneliness and Health of Older Adults Social Support Longitudinal studies have shown that people who are socially isolated have two to five times the risk of dying from all causes, after adjusting for age, sex, chronic diseases, use of alcohol and tobacco, self-rated health, and functional limitations (Penninx et al., 1997; Seeman, 1996). In addition, both human and animal studies have shown that presence of social support during stressful situations reduces blood pressure and enhances immune function (Berkman et al., 2000; Cacioppo et al., 1998; Seeman, 1996). Social support may be emotional, i.e., 30

PAGE 52

providing sympathy, caring and love, or instrumental, i.e., providing tangible help or assistance (Berlanan et al., 2000), and has been shown to speed recovery from heart attack and increase physical function and psychological adjustment after a stroke (Berlanan & Glass, 2000). Social networks, similarly, provide access to material resources, such as job opportunities, housing, health care referrals, and other benefits. Networks operate through micro psychosocial and behavioral processes such as social support, influence, engagement, attachment, and access to resources and material goods. They also operate through macro social forces such as labor markets, economic pressures, organizational relations, culture, social change, industrialization and urbanization (Berkman & Glass, 2000). Assessments of social networks on network structure and strength ofties. Structural factors include range or size (number of members); _density (extent of boundedness (degree to which networks are defined); and homogeneity; (extent to which individuals are similar). Strength of individual ties is measured by frequency of contact (number and type); multiplexity (number of transaction types); duration (how long individuals know one another); and reciprocity (extent that transactions are even). Social networks may help promote social participation, engagement in activities and a sense of identity, since they provide opportunity for hosting and attending social gatherings, group recreation, and rituals (Berkman et al., 2000). Older adults who live alone are becoming increasingly common (Letvak, 1997). Maintaining connections to family and community are often difficult for 31

PAGE 53

those who lack transportation, perform caregiving duties for an ailing spouse, or who ha:ve a physical or mental disability .. Social relationships may impact health in a variety of ways (Heaney & Israel, 1997; Letvak, 1997; Stewart, 1989), particularly for older adults who are at highest risk for isolation. In a prospective cohort study of 2812 community-dwelling adults age 65 and older, in-home and telephone interviews were conducted.over a twelve-year period to determine if there was a relationship between social ties, social support and cognitive function (Bassuk, Glass, & Berkman, 1999). Participants were assessed on cognitive performance, and a composite index of their social connections, activities, and emotional support, called the "social disengagement" index. After adjustment for age and baseline cognitive scores, higher social disengagement scores were significantly associated with a higher probability of cognitive decline at three, six, and twelve-year follow .up. Participants with no social ties had about twice the odds of cognitive decline as those with five or more social ties .. In addition, the social disengagement index predicted a higher mortality rate across all follow-ups. The relationship between disengagement and subsequent decline held for those with the best initial cognitive performance scores. The summary measure provided a strong and consistent predictor of cognitive decline that was not present with any of the individual scale items. The authors concluded that having multiple opportunities for social contact and activity may be more important than any particular type of social support. This 32

PAGE 54

may explain why many of the interventions used to improve social support in order to ameliorate loneliness in the elderly, have shown mixed success. Numerous studies show that social support reduces perceptions of loneliness (Chin-Sang & Allen, 1991; Fratiglioni, Wang, Ericsson, Maytan, & Winblad, 2000; Mullins, Elston, & Gutkowski, 1996; Penninx et al., 1997; Sugisawa et al., 1994; Tilvis, Pitkala, Jolkkonen, & Strandberg, 2000), however, these differ in type of support measured. For instance, a longitudinal study of older Swedish adults, Lennartsson & Silverstein (Lennartssort & Silverstein, 2001) investigated the relationship between engagement irt sociai, leisure and productive activities and mortality. Activities were looked at along two co11tinuums: solitary to social; and sedentary to active. In this way they sought to provide insight into whether activities that were social, physical, or neither social nor physical, had different effects. It was found that frequency of contact with children, grandchildren, other relatives and marital status were not significant predictors of mortality, so those activities were not included. In addition, when the impact of various activities and organizational participation were examined using multivariate analysis adjusted for gender, age, education, functional health, circulatory/heart problems and smoking, the only significant relationship that remained was that older men appeared to benefit from participating in solitary but active pursuits such as carpentry, gardening, or other hobbies. No benefits were found for social activities, group involvement, religious participation, or family contact for this study. The mechanisms at work were not 33

PAGE 55

investigated and no measures of social-psychological variables were used, so one can only hypothesize possible causes for the outcome. In addition, since survival was solely looked at as the outcome instead of overall health or quality of life, the impact of these various activities on physical and mental health is unlmown. On the other hand, a separate study of Japanese elders (Sugisawa et al., 1994) found that so'Cial participation (i.e., organizational attendance) was significantly related to decreased mortality. Similar to the Swedish study, marital status and number of social contacts showed little effect on mortality for this group. Also; in a Netherlands-based study of mortality in the elderly, a large social network was associated with less loneliness and lower mortality (Penninx et aL, 1997). Others have suggested that having a relationship that provides emotional support (i.e., closeness or intimacy) is the most important factor to reducing loneliness (Mullins et al., 1996; Russell, Cutrona, de la Mora,-& Wallace, 1997), although emotional support alone was not found to be a strong predictor of prevention of cognitive decline in the above-mentioned study on social disengagement (Bassuk et al., 1999). Loneliness Loneliness is defmed as a perceived lack or loss of a meaningful network of interpersonal contacts and/or companionship (Bergman-Evans, 1994; Penninx et al., 1997; Walker & Beauchene, 1991). Social isolation is not with loneliness (Mullins et al., 1996), but it may contribute to it by reducing the opportunities for social contact and activities. Since loneliness is a subjective state, 34

PAGE 56

the factors that contribute to it are many. They include loss of employment, income, mobility, hearing, vision, health, independence, enjoyed pastimes, relatives, friends, spouses, and pets (Bergman-Evans, 1994; Chin-Sang & Allen, 1991; Rane-Szostak & Herth, 1995; Russell et al., 1997). Loneliness has been implicated as a major contributor to declines in mental and physical health, self-care and nutritional inadequacy, and increased depression and suicide in older adults (Forbes, 1996; Khaw, 1997; Rane-Szostak & Herth, 1995;_ Russell et al., 1997; Sugisawa et al., 1994; Walker & Beauchene, 1991). Longitudinal studies have shown that people who are lonely are at greater risk of dementia leading to institutionalization, than those who have satisfying, albeit infrequent, contacts with friends/relatives (Fratiglioni et al., 2000; Russell et al., Tilvis et al., 2000). While theories of healthy aging indicate that loneliness and its accompanying health declines are not an inevitable component of aging (Bonneux, Barendregt, & Van der Maas, 1998; Garfein & Herzog, 1995; Khaw, 1997), the social and cultural factors that contribute to loneliness are becom1ng increasingly common. First, norms that foster geographically dispersed, nuclear families have made it more difficult for older adults to maintain the community ties and social participation that are important to sustained health (Chin-Sang & Allen, 1991; Forbes, 1996). Second, longer life expectancy for both men and women may result in longer retirements accompanied by decreased participation in what are perceived as "productive" or "valuable" activities and roles (Glass et al., 1999). 35

PAGE 57

Third, despite the potential to maintain health and independence into old age (Rakowski, 1997), the biological process of senescence often results in chronic illness and disability, leading to increased isolation and loneliness (Khaw, 1997). Fourth, reductions in health services that are reimbursable by Medicare and .decreased length ofhospital stays in favor ofhome health, have created a situation where many older adults are sent home alone to manage their own health, or the health of an ailing spouse, thus constraining access to services, social support and valued activities (Bergman-Evans, 1994; Forbes, 1996; Gustafson et al., 1998; McTavish et al., 1995). Studies disagree about the contribution of health, income, increased age, gender, ethnic status, rural/urban dwelling, existence of children, or education as risk factors for loneliness. A few have shown a tendency for more loneliness among people who perceive their health as poor (as opposed to their actual health status), people with disabilities, perceived poverty (as opposed to their actual income level), and female gender have been implicated as risk factors (Mullins et al., 1996). While these results seem to be inconsistent among studies, what is interesting to note, is that the existence of a few good friends seem to be more critical to avoiding loneliness then the existence of a spouse or children. Thus friends, coupled with a sense of utility and perceived control seem to be consistently important factors (Chin-Sang & Allen, 1991; Garfein & Herzog, 1995; Mullins et al., 1996; Porter, 1994; Sugisawa et al., 1994). 36

PAGE 58

Culturally-Mediated Perspectives on Loneliness The emphasis on youth, independence, and autonomy in the United States provides some distinct cross-cultural challenges for those whose cultural background and traditions esteem aging and integrate the elderly into the extended family unit. Preservation of traditional roles for the elderly varies and the impact on perceived health and well-being is dramatic. Individuals ofHispanic descent are diverse in ethnicity .and culture, yet extended families are still common among Hispanic groups in the United States. In many families Hispanic grandparents play a key role in helping to raise, or in some cases, raising, their grandchildren (Bagley, Angel, Dilworth-Anderson, Liu, & Schj.nke, 1995). In a qualitative study of 83 Latino elders (65 years and older), social support from families was reported as the most important source of well being, regardless of living arrangement (Beyene, Becker, & Mayen, 2002). Participants were. interviewed three times over 12-18 months. 62% of the participants had been in the United States for more than 20 years. While most appreciated the economic and health services available to seniors in the United States, many felt that respect for --elders was lacking, even among young Hispanics. All participants felt that ''feeling old" was a function oflosing their role or purpose in life, as opposed to chronology. Those with close relationships with their children and grandchildren said that they did not perceive themselves to be old, despite multiple chronic illnesses in all cases, and severe disability in some. The three greatest fears expressed by these Latino 37

PAGE 59

elders were outliving their usefulness and having to go to a "rest home", losing their autonomy and being lonely. Respect for elders and the important role of the elderly is also a cultural norm among Asian families. For instance, Chinese "nuclear families" in China and America often include one or two elderly grandparents who assist in child rearing. Adult children are considered the preferred source of social support to elderly parents, and are expected to provide care when they are ill. This tradition of family elder-care has also been noted in American families of Vietnamese, Cambodian, and Laotian descent (Bagley et al., 1995). Similarly, large extended families and clans are central to most Native American cultures, with customs of cooperation, shared resources and distinct roles. However, for some older Native Americans. who live in poverty with inadequate sanitation, lack of finances, and little access to health services, neglect and isolation accompanied by loneliness and despair, is similar to that experienced by socially isolated non: Indian elders (Mercer, 1996; Williams & Ellison, 1996). For African American families, large kin networks that are not necessarily defined by blood _relationship are usual (Bagley et al., 1995; Chin-Sang & Allen, 1991). African American elders play important roles with regard to childcare, wisdom and experience. In rural communities, even elders with dementia are deemed able to perform roles within the "kin system" (Bagley et al., 1995; Gaines, 1989). However, many African American elders, particularly in urban settings, find 38

PAGE 60

themselves increasingly alone and isolated from friends and relatives. In a qualitative study of thirty elderly Black women, church participation, attendance at senior citizens' centers, and affiliation with friends and other older adults were cited as the major coping strategies for loss of kin and loneliness (Chin-Sang & Allen, 1991 ). These women also emphasized that nurturing and helping others provided a sense of security, camaraderie, love, and belonging. This study did not specifically target any one ethnic or cultural group, thus limiting its ability to reveal cultural factors that affect active lifestyles and social isolation. It is recognized that this is a limitation of the study; given the varied cultural views of aging, and the unique challenges experienced by those whose traditional roles are altered by life in the U.S. It is also probable that ethnicity and culture may have an independent effect on active lifestyles and social isolation, apart from other neighborhood factors. In addition, it is also probable that in some neighborhoods, group ethnic or cultural identity affects neighborhood characteristics (Diez Roux, 2004). Analysis of self-reported ethnic and racial identity was performed and compared to neighborhood demographics from Census Tract data to better interpret the study results. However, possible cultural influences on perceptions and health behavior, as well as neighborhood characteristics, were limited by the small proportion of ethnic and racial minorities who participated in the study. 39

PAGE 61

Contribution of this Study Understanding How Context Affects Active Lifestyles for Older Adults The research discussed above (Garfein & Herzog, 1995; Rakowski, 1997) disputes the "traditional" notion that disengagement and separation from society are "normal" and inevitable conditions of aging. A primary goal of the present research was to understand how the specific physical and social content of an individual's activity space, and his or her perceptions of self-efficacy, neighborhood Gohesioil, and access to resources, were associated with his or her ability to be physically and socially active. This information has the potential to make an important contribution to a more socio-ecological model of aging, which includes maintaining active lifestyles as a desirable goal and a possible means to prevent social isolation. One innovation and strength of this study was the ability to access address information for a cohort of Kaiser Permanente HMO members over the age of 65, which was matched to participant's electronic medical records. This allowed for efficient random selection of a large population of older adults, and also permitted the use of GIS to integrate complex information, in order to better understand the relationship between the environment; individual attributes and perceptions, and broader indicators of health such as BMI and number of co-morbidities. In addition, the relationship between physical activity and social isolation was specifically examined, and its implications for intervention and prevention of the health declines associated with inactivity and loneliness are discussed. 40

PAGE 62

CHAPTER4 RESEARCH DESIGN & METHODS The purpose of this study was to determine if neighborhood factors affect active lifestyles and social isolation of adults, age 65 years and older, either directly, or indirectly via individual perceptions of self-efficacy with regard to mobility, safety, and access to desirable amenities (such as banks, pharmacies, markets, health and recreation centers) (Paul, 1993). The study also investigated whether a relationship exists between active lifestyles and social isolation. Neiahborhood (Level 2) Vital Statistics: Socio-demographics, Crime Rates Structural Factors: Walking Paths, Traffic, Land Use, Aesthetics Social Factors: Social Capital, Safety, Vitality Individual (Level 11 Characteristics: Socio-demograpbics, Health Perceptions: Self Efficacy, Pedestrian Safety, Crime Safety, Access to Resources, Social Cohesion Active Lifestyle Weekly Activity Frequency Weekly Energy Expenditure Figure 4.1. Conceptual Model 41 Social Isolation Perceived Loneliness Social Support

PAGE 63

Research Questions and Specific Aims The key research questions to be answered by the study are: Question 1: What components, if any, of the environmental context enable or inhibit active lifestyles and social isolation in older adults? Aim 1: Use GIS to select neighborhoods that vary with regard to walkability, crime rate, SES and proportion of adults ages 65+. Aim 2: Survey a sample of community dwelling older adult members of Kaiser Permanente Colorado (K.PCO) who reside in the selected neighborhoods, regarding their perceptions of their neighborhood's resources, safety, and social cohesion, and on the outcome variables of activity and social isolation. Aim 3: Use hierarchical regression modeling to analyZe the contextual effects on senior aCtivity engagement and social isolation. Question 2: What are the potential causal pathways through which neighborhood and individual factors influence these two outcomes? Aim 4: Conduct mediation analyses to test for intervening variables between the initial neighborhood variables and outcome variables Question 3: Do objective health outcomes cluster by neighborhood? Aim 5: Use GIS to map individual health data, obtained from KPCO electronic medical records and individual self-report (e.g., BMI), to reveal spatial patterns. 42

PAGE 64

Aim 6: Revise conceptual models and develop hypotheses for future studies. Measurable Objectives Measurable objectives of the study included: identifying neighborhood and individual variables associated with a more active lifestyle ( operationalized as greater frequency of specific types of activity and energy expended in physical activity) and less social isolation ( operationalized as lower perceived loneliness and greater perceived social support). The relationship between active lifestyle and social isolation was also explicitly examined. Hypotheses Research Question 1 : What components ofthe environmental context enable or inhibit physical and social activity in older adults? Hypothesis 1: Neighborhood structural and social factors, including those that contribute to: walkability; aesthetics; desirable destinations; opportunities for exercise; numerous public courtesies;few incivilities; less territoriality; high vitality; and high stability, as measured by walking audits and reported neighborhood data, will be associated with: 1.1. greater frequency of walking for errands; 1.2. greater frequency of community-based activity-engagement; 1.3. higher energy expenditure, and 1.4. less social isolation in adults ages 65 years and older. 43

PAGE 65

Research Question 2: What are the Pathways through which Neighborhood and Individual Factors Influence Outcomes? Hypothesis 2: Specific environmental factors will affect specific behaviors differently, and the associations may be direct,. indirect, or both. 2.1. higher frequency ofwalking for errands will be associated with: greater walkability features; numerous public courtesies; and numerous desirable destinations and will be partially mediated by perceived selfefficacy for walking, perceived pedestrian safety and perceived access to resources. 2.2. higher frequency of engagement in community-based activities will be associated with: greater walkability features; numerous public courtesies; numerous desirable destinations and fewer incivilities; less territoriality; higher vitality; and higher stability; and will be partially mediated by perceived self-efficacy for walking, perceived pedestrian safety and perceived social cohesion. 2.3. higher energy expended in physical activity will be associated with: walkability; desirable destinations; opportunities for exercise; high vita"!ity; and will be partially mediated by perceived self-efficacy for transport (walking versus driving) and access to resources. 2.4. lower perceived loneliness will be associated with: neighborhood aesthetics; desirable opportunities for exercise; less 44 j

PAGE 66

individual territoriality; greater neighborhood vitality and will be mediated by perceived self-efficacy for transport, safety from crime and social cohesion. Research Question 3: Do Objective Health Outcomes Cluster by Neighborhood? Hypothesis 3: Neighborhood factors conducive to more active lifestyles and lower perceived social isolation, are associated with better individual health indicators (i.e., lower BMI, fewer comorbidities, cardiov.ascular risk) in Kaiser Permanente members, ages 65 years and older. Study Design A cross-sectional survey and neighborhood observation design was conducted to understand the association of neighborhood contextual factors with older adult physical activity engagement and social isolation. regression models were used to analyze results. Methods included fmalizing, piloting, and refining the individual and environmental assessments, and use of GIS to select neighborhoods and geocode participant address information in order to identify residents of the neighborhoods of interest. Mailing of study invitations, consent forms and surveys to randomly selected neighborhood residents followed, and data collection activities concluded with environmental assessments of neighborhoods. 45

PAGE 67

Participant and Neighborhood Sampling Frame According to Bryk and Raudenbush (Bryk & Raudenbush, 1992), an optimal sample size for a 2-level model considers the relative cost of sampling level 1 units (i.e., individuals) versus level2 units (i.e., neighborhoods) and amount ofvariability at each level. Since the time and cost of sampling more neighborhoods is high compared with sampling more individuals per neighborhood, and variation within neighborhoods is expected to be relatively large, the number oflevel1 units will be substantially greater than the number of level 2 units. In addition, since the rule of thumb for survey research is to have at least 10 subjects for each variable measured, a Ievell sample size ofn=200 should be The smaller level2 sample size of eight neighborhoods will be selected to vary on walkability, crime rates, SES and proportion of residents aged 65 years and older, thus providing adequate power to detect whether neighborhood factors explains variability in individual activity and isolation. Participant Sampling Frame ARC GIS geocoding software was used to match an address list of 13,928 KPCO members, ages 65 and older, to Denver City and County streets. This list consisted of all KPCO members who met the following criteria: 1) Currently are active KPCO members 2) Age 65 or older as of July, 2004 3) Reside in Denver City or County 46

PAGE 68

4) Have a full street address with zipcode and phone number 5) Registered for an outpatient clinic visit within the past 12 months 6) Have NOT been admitted or discharged from a Skilled Nursing Facility within the past 12 months. The list, obtained from KPCO electronic medical records (EMR), was geocoded and overlaid onto Denver City and County street maps that delineated 77 distinct Denver statistical neighborhoods. Street map data were obtained from esri.com tigerline files. Geocoding Sixty-nine percent, or 9629 records, matched to street file coordinates. Of those addresses that matched, 9543 (68%) were identified as falling within the boundaries of a Denver city or county_ neighborhood. The others were within or on the border of Adams, Arapahoe, or Jefferson counties. Of the unmatched addresses, 124 had only a P.O. Box, with no street address provided. For the remaining 4175 records, two address matching services available on the Internet were used to attempt to detemtine if they were within the Denver county limits (getzips.com, 2005; mapquest.com, 2005). 3614 records appeared to contain a street address and/or zipcode that fell outside of the Denver county limits. The remaining 561 records appeared to have a Denver county zipcode, but may not have matched due to errors within the street address, such as misspellings or unrecognized formatting, or errors within the street database, such as missing address 47

PAGE 69

range information due to new residential developments. Given their relatively small number and the large amount of effort it would take to correct these addresses, no further attempt to match them was made. Thus, by eliminating the addresses that fell outside of the Denver county limits, the match rate for just Denver City and County addresses improved from 68% to 93.7% ; j i I ---Figure 4 2. Geocoding results for active KPCO members, ages 65 and older Neighborhood Sampling Frame Neighborhoods in the City and County of Denver were selected based on having at least 80 active independent Ilving KPCO members ages 65 or older plus variation in 1) violent crime rate per 1000 persons (includes homicides, sexual assaults, aggravated assaults, and other assaults); 2) average household income; 48

PAGE 70

3) proportion of adult residents age 65 years or older; and 4) walkability (determined using both population density and street connectivity data). Neighborhood data were obtained from the Piton Foundation (PitonFoundation, 2004) and web page for the Denver Department of Safety (CCDOS, 2005). Year 2000 Census data were used for three of the main criteria: average household income, proportion of adult residents 65 or older and population density. The latter was calculated by dividing the total population number for a particular neighborhood by lan2. High walkability neighborhoods. were selected based upon a high population density number plus street connectivity (which was dichotomized as high if street layouts were gridlike and low if street layouts were winding or mixed). 2003 Denver Department of Safety statistics (CCDOS, 2005) were used to determine neighborhood violent crime rates (calculated as the sum of total homicides, sexual assaults, violent assaults and other assaults per resident population from Census 2000). Violent crime rate was selected instead of total crime rate since property and other forms of non-violent crimes encompass such a broad range of infractions, and are less likely to be associated with perceptions of safety among older adults (Ferraro & LaGrange, 1992; Ziegler & Mitchell, 2003). In addition the measure of social cohesion used in this study showed an association between higher cohesion levels and lower violent crime rates (Sampson et al., 1997). 49

PAGE 71

Of the 77 Denver city and county neighborhoods, 5 had too small a resident population in the year 2000, and therefore lacked census data. The remaining 72 neighborhoods were sorted separately based on each ofthe 4 variables and a top stratum and bottom stratum were obtained for each criterion. This was to determine the potential range for each of the four strata, if all 72 neighborhoods were included .. Neighborhoods with less than 80 KPCO non-cohabitating (i.e., only one member per household was counted) members residing there were deleted and the remaining 47 neighborhoods were resorted using the same criteria. Again the top and bottom five neighborhoods were identified, and these were compared with the total neighborhood strata. While omitting neighborhoods with fewer than 80 KPCO members tended to eliminate neighborhoods that were highest in violent crime (none of the top five high crime neighborhoods appeared on the KPCO member list), lowest in average household income (only two of the five lowest income neighborhoods appeared on the K.PCO member list), lowest in walkability (only two of the five neighborhoods designated as least walkable appeared on the KPCO member list), and lowest in proportion of elderly residents (only one of the five neighborhoods with the lowest proportion of elderly residents was on the KPCO member list), the remaining range of neighborhoods was relatively broad and thought to provide adequate diversity on key characteristics. Thus, the decision to use a discrete sample was a trade-offbetween having a well-defined denominator with access to contact and health data and overall generalizability. 50

PAGE 72

Neighborhood Selection Random selection of the eight study neighborhoods was accomplished using a computerized research randomizer (Urbaniak & Pious, 2005) that utilizes a J avaScript random number generator, which produced eight sets of 5 unique random numbers, ranging from 1 to 50. These were assigned to each neighborhood within strata that matched the four main selection variables: 1) violent crime rate per 1000 persons; 2) average household income; 3) proportion of adult residents age 65 years or older; and 4) walkability. Neighborhoods were then re-sorted accordingly, resulting in a randomly ordered, ranked list within each stratum. The neighborhood assigned the top ranking for each stratum (using the random order) was-selected to participate in the study. If a neighborhood had already been selected, the next neighborhood on the list was chosen. Once all 8 neighborhoods were selected, the means of the selected neighborhoods were compared with those of the un-selected neighborhoods for each of the 4 neighborhood variables. In addition, 2000 census data on the proportion of non-whites living in these neighborhoods were included to assess whether selected neighborhoods were representative of Denver's ethnic/racial population. Comparisons were made using an independent samples t-test. In general, selected neighborhoods showed no significant differences from un-selected neighborhoods on each of the variables, with the exception of the more walkable neighborhoods. This 51

PAGE 73

is most likely due to the fact that a much larger proportion of Denver City and County neighborhoods had a grid-like street pattern, regardless of population density. The more suburban-like street layout, of windy streets and cui du sacs, were only observed in 17 neighborhoods. Since both higher population density and street layout are associated with greater walkability, the selected neighborhoods were drawn from those with the highest population density, so are not representative of neighborhoods with lower population density and grid-like street layout. This ls acceptable since it is important that both types of neighborhoods be represented for comparison purposes. Figure 4.3. illustrates the geographic location of the selected neighborhoods. Figure 4.3. Final eight selected study neighborhoods labeled by selection strata 52

PAGE 74

Participant Selection and Inclusion Criteria Kaiser Permanente Colorado (KPCO) was selected due to its cohort of more than 51,000 active patients, ages 65 or older, in the Denver/Boulder vicinity and accessible contact information and electronic health data. The fact that all patients have at least some level of health coverage controls for overly wide differences in perceived access to health care due to factors beyond the scope of their environment or transportation. The investigator's access to this list of individuals and access to a very rich electronic medical record (EMR) dataset, (subject to HIP AA regulations and agreements) were also Important considerations in selecting a KPCO sample. KPCO's Research Review Committee and Institutional Review Board approved the study design and the use of member database to contact potential participants . To achieve a study sample of approximately 25 participants per neighborhood (n=200) a list of patients, ages 65 and older, was drawn from KPCO's electronic medical records system, identifying all adult patients, ages 65 and older. Additional inclusion criteria were: completed a clinic visit within the 12 months prior to the date the list was drawn, were not residing in or discharged from a Senior Nursing Facility, were residing in a zip code that corresponded to a Denver City and County statistical neighborhood, and had been at that address for at least 6 months prior. Using GIS and street address data, members whose residences fell within the boundaries of one of the eight selected neighborhoods were identified. A subset of 50 individuals per neighborhood were randomly selected using random 53

PAGE 75

number tables. These 400 individuals were invited to participate in the study initially, with the goal more than a 50% (N=200 participants) return rate, which seemed to be a reasonable estimation for a targeted sample of elderly (Faler, 1988). Slightly more women than men were expected to participate due to their over-representation in this older population. Kaiser Permanente Colorado membership includes approximately 56% women members whose ages are 65 or older. Therefore, it was expected that at least 56% of the study population would be women. Since participant accrual was completed by neighborhood, sampling was stratified to assme that the proportion of men invited to participate reflected that of the population as a whole. The. KPCO health plan does not routinely collect information on the race arid ethnicity of its members. However, we expected the racial and ethnic representation of this managed care patient sample to be representative ofinsmed racial.and ethnic populations in the state of Colorado, recognizing that this may not be representative of the entire population of ethnic and racial minorities, which also includes individuals without insmance. In addition, since participants were selected from specific neighborhoods, census data on the neighborhood demographics were reviewed, with 3 of the neighborhoods selected having a non-white (including Latino) population greater than 75% (range 7.1% to 95.3%). Specific actions to increase participation among minority patients included: (a) patient communication materials that emphasized the importance of conducting research that involves 54

PAGE 76

people. of all racial and ethnic backgrounds (i.e., the need to develop quality care that meets the needs of diverse racial and ethnic groups), (b) selection of racial and ethnically diverse neighborhoods using statistical neighborhoods that are tied to census blocks, using GIS, and (c) offer of help in reviewing consent forms and instruments via the telephone. All potential study participants were mailed an invitation to participate in the study, an Informed Consent document that explained the project in detail, and a HIP AA authorization for the use and disclosure of protected health information for research. A postage-paid 'refusal' postcard was included to allow participants to indicate if they were not interested in being contacted further for the study. Copies of both the consent and authorization forms were enclosed, and potential participants were asked to sign one copy of each form if they elected to participate, return it in the postage-paid envelope provided within days, and retain the other copy for their Once the signed consent and authorization were received, the first set of surveys were mailed with instructions on how to complete and return them. Surveys included questions on: 1) background (e.g., race/ethnicity, income, healthcare utilization); 2) neighborhood environment; 3) neighborhood cohesion;. 4) self-efficacy for walking, driving, or using public transportation; 5) physical activity, 6) loneliness, and 7) social support. The direct phone number of the investigator was prominently displayed in all mailings,.encouraging the recipient to call with any questions, or if they needed 55

PAGE 77

assistance filling out the forms. Follow-up telephone contact was made to the potential participant in the event that neither the postcard nor the forms or surveys were returned within a 1 0-day timeframe. Additional mailings were done, as needed, in 'order to reach neighborhood targets. Data Collection and Measures Data were obtained using published GIS data available on all Denver Statistical Neighborhoods (PitonFoundation, 2004), published business directories, and primary observational data. The latter was collected using a modified version of the Systematic Pedestrian and Cycling Environment Scan (SPACES) instrument, an environmental audit form that provided astructured method for collecting detailed data on factors related to neighborhood walkability (Pikora et al., 2002). Definitions' and pictures provided in the SPACES observer manual were utilized in order to minimize subjectivity and maximize inter-rater reliability. The instrument was tested in two ways: the proportion of segments for where at least three of the four raters were in agreement (i.e., .75 agreement) and the calculation ofkappa statistics for each item. Intra-rater reliability was also analyzed by calculating perfect agreement between two consecutive audits of the same segments and by the same observers, conducted at least 7 days apart. In addition, items from the Neighborhood Brief Observation Tool (NBOT) (Caughy, O'Campo, & Patterson, 2001) were used to access features of the neighborhood related to social Capital, such as condition of the built environment 56

PAGE 78

(e.g., upkeep of property and proportion of structures that were burned, boarded up or abandoned), signs and protection from crime (e.g., fences or other border demarcations, neighborhood watch signs, personal security elements such as bars on windows, security signs or guards, or dogs), public courtesies (e.g., public trash cans, public phones, public benches, transit stops, parks or recreational spaces), people and their activities, and non-residential land use (e.g., types of businesses and services observed). Average reliability of the instrument was 87%, based on percent agreement among multiple raters (see Appendix A for adapted audit instrument). Individual Demographic. Health and Fitness Data Survey questions to confirm age, income, education, race/ethnicity, functional mobility limitations, and hospital stays were combined with other mailed surveys. Additional health data were obtained from the medical records of all members who had received an invitation to. participate in the study (data were de identified and combined by neighborhood for non participants to protect privacy). The reasons for collecting non-participant data were to allow comparisons to be made between participants and non-participants, as well as to increase the power to detect hypothesized health differences between neighborhoods. Data extracted were the most recent available, within a year prior to or six months after their study emollment date. For non-participants, the date of invitation to enroll in the study was used. For participants, data dates were cross-checked with self-reported tenure at their current address to assUre that all data used were obtained 57

PAGE 79

while the participant resided at their current address. Only one datum from one participant was excluded due to its being collected prior to his moving to his current address. Indicators of Coronary Vascular Disease Risk Health data collected from electronic medical records included the most common risk factors used to diagnose metabolic syndrome, which increases the risk of coronary heart disease and type 2 diabetes and is associated with obesity and physical inactivity. Five out of seven risk factors for metabolic syndrome (AHA, 2005): 1. BMl(as a proxy for central obesity) 2:30 obese 2. Fasting blood triglycerides 2: 150 mg/d.L 3. Blood HDL cholesterol (<40 mg/dL in men, <50 I.D.g/d.L in women) 4. Blood 2: 130/85 mmHg 5. Fasting Glucose 2: 110 mg/d.L Other data collected from medical charts included resting pulse rate as an indicator of level of conditioning, smoking status and the total number of comorbid conditions, associated with an increased risk for heart attack (i.e., hypertension, hypercholestemia, diabetes, and heart disease). Chronic Disease Score (CDS) A total chronic disease score (CDS) which reflects the total number of comorbid conditions an individual has, was used as an overall indicator of health. 58

PAGE 80

The CDS was constructed using a validated comorbidity measurement tool, the Rx.Risk score (Fishman et al., 2003). The Rx.Risk score is a measure of comorbidity that incorporates age, gender, insurance benefit status and an RxRisk category based on diagnoses derived from administrative pharmacy data. It was originally developed and validated to identify chronic conditions and to predict cost of health care, and subsequently revised to assess disease burden in certain populations. Administrative pharmacy data were used to apply the Rx.Risk tool to our study population in order to identify chronic conditions. Individuals were scored a 1 if they met the higher risk criteiia for each factor, they were scored a 0 if they were within normal limits. Individual Perceptual Data Regarding Neighborhood Environment Individual perceptual data were collected using a mailed survey. SelfEfficacy for Transport Confidence level ratings (on a scale of 1-1 0) for walking, driving, or using public transportation, adapted from self-efficacy assessments used in self management studies (Bandura, 1997; Williams, Rodin, Ryan, Grolnick, & Deci, 1998). In addition, multiple choice questions to assess preferences and functional ability with regard to mobility (e.g., "What mode of transportation do you most often use to get your errands done (e.g., walk, drive myself, get a ride, etc.)." "If you drive yourself, do any of the following apply to you: I drive only during the daylight; I drive only in good weather; I limit the distance I drive; I limit the ro1:1tes I will take") 59

PAGE 81

were adapted from interview questions asked in studies of older drivers (Fonda et al., 2001). The Neighborhood Environment Walkability Scale (NEWS) This measure was used to assess perceptions of neighborhood factors related to walking and cycling. It consisted of 9 subscales that were pilot tested within two San Diego neighborhoods (N=106). One-week test-retest reliability coefficients ranged from .58 to .80. The scale also showed excellent validity with greater walkability corresponding to higher residential density, land use mix, access to resources, street connectivity, aesthetics, and traffic safety. Accelerometers verified that residents in the more walkable neighborhoods engaged in approximately 52 more weekly minutes of moderate physical activity than those living in less walkable neighborhoods, explained by greater walking for errands SE=.50, p=.Ol) (Saelens, Sallis, Black et al., 2003). Neighborhood Cohesion Scale The scale consisted of 5 items and is a subscale of a larger collective efficacy questionnaire with questions on both cohesion and informal social control. The larger questionnaire demonstrated significant (p<.O 1) positive associations with friendship and kinship ties (r-0.49), organizational participation (r-0.45), and neighborhood services (r=0.21). In this study only the cohesion items are included, due to their close association with the social control subscale (r=.80, p<.OOl) and a 60

PAGE 82

:peed to minimize the length of the survey. The cohesion subscale has a reliability of .70 (Cagney, Browning, & Wen, 2005; Sampson et al., 1997). Outcome Measures for Active Lifestyle and Social Isolation Community Health Activities Model Program for Seniors (CHAMPS) The CHAMPS is a self-report instrument designed to measure physical activity in older adults, by assessing frequency per week of specific activities. The CHAMPS also measures weekly energy expenditure (calories/week), calculated by usirig the metabolic cost weights for each activity and adjusting for body weight. In a study of older adults, 3-month stability coefficients for the frequency and energy expenditure measures for control group participants were r=.57 and .84, respectively. When compared with interviewer data and activity logs across four different groups, the CHAMPS demonstrated good construct validity (p <.001) (Stewart et al., 1997). A more recent study comparing sedentary older adults (N = 173) with active older adults (N = 76) demonstrated a six-month stability (estimated on the physically active control group who were assumed unlikely to change their activity levels) ranging from 0.58 to 0.67, using intraclass correlation coefficients. All measures were sensitive to change (p 0.01), with small to moderate effect sizes (0.38-0.64). Participants in the study were aged 65-90 yr (mean= 74, SD = 6); 64% were women, and 9% were minorities (Stewart et al., 2001). 61

PAGE 83

The Loneliness Scale The Loneliness scale (de Jong Gierveld & Kamphuis, 1985) consists of 11-items that assess perceptions of available companionship, support and help for older adults. The scale has been used in the Longitudinal Aging Study Amsterdam (LASA) and the Widowhood Adaptation Longitudinal study (W ALS). Reliability of the scale has been reported in the .80-.90 range (Cronbach's alpha or rho). It also is moderately correlated with the Center for Epidemiologic Studies Depression Scale (CES-D) in total (r=.51) and with CES-D item 14 "During the past week I felt lonely," r-.54). The Social Provisions Scale The 24-item Social Provisions scale's purpose was to assess the degree that respondent's social relationships provide social support in the areas of attachment, social integration, reassurance of worth, reliable alliance, guidance and opportunity for nurturance (Cutrona & Russell, 1987). When tested on a sample of 100 elderly subjects, the scale's internal consistency across all domains was'>.70; test-retest reliability ranged from .37 to .66; and total score correlations with life satisfaction, loneliness, and depression ranged from .2.8 to .31 (p<.05). In addition, the scale correlated with measu,res of social networks (i.e., nuniber of relationships and frequency of contact) and satisfaction with social relationships (Cutrona, Russell, & Rose, 1984; Cutrona & Russell, 1987). Both subscale and total scale scores will be used in this study. 62

PAGE 84

Pilot of survey instruments To assure feasibility of administering the proposed surveys, 10 older adults of mixed gender and racial/ethnic background, were asked to complete the survey and provide feedback, particularly on items not previously standardized. By using qualitative methods such as openended questions, candid information was obtained to reveal potential barriers to participation, and to identify the need for revisions to the questions or survey instructions to assure accurate responses. 63

PAGE 85

Table 4.1. Summary of Constructs and Measures INDEPENDENT VARIABLES CONSTRUCTS METHOD OF COLLECTION Levell: Individual Factors Participant Characteristics Demographics Mailed surveys Includes items on sex, marital status, household members, race, ethnicity, education, income Health and Functional status -Electronic Medical Records -Questions on mailed survey Participant Perceptions Mailed Questions on -Walking mode of transpo_rt most used and -Driving confidence level for walking, -Use of Public Transport driving, using public transport (5 items) Safety Mailed NEWS -from Traffic & Crime (14 items) Services & Access Mailed survevs: NEWS (7 items)_ Neighborhood Cohesion Mailed Cohesion Scale (5 items) Outcomes Active Lifestyle Engagement Mailed surveys: CHAMPS (33 items) Social Isolation Loneliness Mailed : Loneliness Scale (II items) Social Support Mailed Social Provisions Scale (24 items) 64

PAGE 86

4.1 (Cont.). SumrnaryofConstructs and Measures INDEPENDENT CONSTRUCTS METHOD OF VARIABLES COLLECTION Levell: NeiRhborhood Factors Structural Residential Density Neighborhood facts, Environment: (Census data at piton.org) Land Use mix Neighborhood facts Walkability (piton.org) Field audit Street Connectivity Neighborhood maps Field audit Green Spaces -Neighborhood maps Field audit Sidewalks Field audit Aesthetics Field audit Availability of Transportation Transportation Field audit: Count of bus or train stops Social Environment Socio-Demommhics -Neighborhood facts (Census data, at piton.org) Resources Neighborhood facts (piton.org) Field audit Crime -Neighborhood stats (piton.org) 65

PAGE 87

Data Management As can be seen in Table 4.1, data for the project were collected from several different sources: a) patient completed data on self-report questionnaires; b) electronic medical record extraction; c) staff-entered data from GIS and field audit data. :bata were checked at the time of collection for missing values and out-of-range responses so that errors could be corrected immediately via the telephone. A system file was created using EXCEL for combining data sets. Prior to the analyses, all data were again checked for missing and out-of-range values. Protocols such as locked hard-copy and password-protected databases were employed to protect patient privacy in keeping with HIP AA protocols. Preliminary analyses were conducted to assure that all variables met the assumptions of the analyses to be performed. Strategies Used to Overcome Implementation Barriers Potential barriers to the study included 1) poor response rate to survey; 2) failure to complete all questions due to respondent fatigue or confusion; 3) inconsistency of published population and neighborhood data for specific neighborhoods with regard to demographics, crime rates; timeframes, and map accuracy; 4) privacy concerns with regard to use of individual residence information versus aggregate data, in order to represent the ''true" context of the individual.(Jacquez, 1998) 66

PAGE 88

Maximizing Survey Response Rate and Completion To maximize response rates several techniques recorrimended by Dillman (Dillman, 1991) were employed. A personalized invitational letter accompanied surveys explaining their purpose and importance of the research. In addition, a postage-paid postcard was provided to allow potential respondents to request a phone call to answer questions or complete questionnaires via the telephone, or refuse participation in the study. A follow-up phone call was made to nonresponders, followed by repeat mailing of the survey and a final phone-call to attempt to collect the data over the telephone. Survey respondents were offered a token thank-you gift for their efforts in returning the survey. All returned surveys. were checked for completeness and responses for missing information were obtained via the telephone. Maximizing Accuracy in Characterization of Neighborhoods To overcome potential bias with regard to inaccurate, or inconsistent neighborhood data, the eight-neighborhood sample was drawn from the 77 distinct City and County ofDenver statistical neighborhoods. An advantage of using only City and County of Denver neighborhoods is that the published census, housing, economic. and crime rate data for the year 2000 are consistently reported for each neighborhood and are available through a user-friendly GIS website developed jointly by the Piton Foundation and the Community Planning and Development Agency of the City and County of Denver (PitonFoundation, 2004). 67

PAGE 89

In addition, other sources were used to characterize specific neighborhoods including telephone directories, and personal observation. Assuring Participant Privacy and Avoiding Location UncertaintyIndividual residences were mapped as points within neighborhood units so that the research staff could collect observational data on the actual environmental context of the individual using mapping and the SPACES audit instrument. Observed data were then compared with perceptual data reported by respondents. Privacy was protected by analyzing the data and displaying the results of the analyses in aggregate form (Rushton, 2000). 68

PAGE 90

CHAPTERS. RESULTS OF ANALYSES Research Questions The main research questions of this exploratory study are: Question 1: What components, if any, of the environmental context enable or inhibit activity engagement and social isolation in older adults? Question 2: What are the potential causal pathways through which neighborhood . and individual factors influence these two outcomes? Question 3: Do Objective Health outcomes cluster by neighborhood? Analyses Used To answer the research questions, this chapter will describe the statistical tests used to analyze the data and present the results of these analyses. SPSS version 11.0 was used (SPSS, Inc., Chicago, IL) to perform most ofthe analyses in this report, with the ofhierarchical regression modeling, which used SAS version 9.1 (SAS Institute, 2003). To answer research question 1, univariate analyses of factors related to neighborhoods and factors related to individuals provided descriptive and frequency data for the total sample, as well as grouped by neighborhood. Equality of neighborhood means for continuous variables were assessed using ANOV A or Kruskal Wallis tests, as appropriate. A Pearson Chi-Square test was used for categorical data. The magnitude of effect size was calculated using Cohen's d 69

PAGE 91

(Cohen, 1988; Rosnow & Rosenthal, 1996). Post hoc analyses were performed to identify sources of neighborhood differences by first comparing 95% confidence intervals across neighborhoods to focus the analyses, and then perfonning either the Dunnett T3 test for continuous data or the Fisher's Exact test for categorical data. A 2-level hierarchical regression modeling technique was used to test direct associations of neighborhood variables with respondent-reported physical activity and loneliness. To address the potential for collinearity among the many predictors tested, separate regressions were used for each independent variable tested. This required many individual analyses to be performed, which admittedly increased the likelihood that some ofthe direct effects were significant by chance. Due to the exploratory nature of this study, adjustments for multiple analyses were not made. The very small neighborhood sample size (N=8) and the fact that statistical adjustments tend to be overly conservative and would greatly reduce our already limited power, were weighed as important considerations. It is understood that this decision greatly increased the chance of a Type I error and was a trade-off between that and increasing the chance of a Type II error. In all analyses, the alpha level used was the conventional two-sided .05. However, given that adjustments for multiple analyses were not made, the reader will need to interpret the p values presented with this limitation in mind. To answer research question 2, mediation analyses were subsequently conducted to test for intervening variables between the initial neighborhood 70

PAGE 92

variables and outcome variables (Newsom, 2001). Possible mediator variables were selected based upon the literature, previously hypothesized relationships, and whether it "made sense" conceptually. The mediator was then tested statistically by adding it to the hierarchical regression model. To test for mediation a regression analysis to examine the direct association between the independent variable (N) and the dependent variable (DV) was conducted. If a relationship was found then three separate regression analyses were performed to examine the association between 1) the independent variable (N) and mediating variable; 2) the mediating variable and dependent variable (DV); and then to 3) simultaneously include the N and mediating variables as independent variables to the dependent variable. A Sobel test was used to detennine whether the indirect effect (i.e., the reduction in magnitude between the N and DV that occurred when the mediating variable was included in the regression model) was statistically significant. A variable was considered to be a full mediator if its inclusion in the model resulted in the magnitude ofthe effect between the independent variable and dependent variable being reduced by at least 10% and the p value changed from <.05 to and if the Sobel t-test had a significance level <.05 (Preacher & Hayes, 2004). A variable was considered to be a partial mediator if the above criteria were met, except that in the case of partial mediation, a significant, albeit reduced, direct effect remained. 71

PAGE 93

Gender was also tested as a possible moderator of neighborhood variables that may affect behavior differentially for women versus men (e.g., perceived safety from crime). This test was conducted by adding an interaction term (sex x neighborhood variable) to the main effects hierarchical regression model. While other sociodemographic variables, such as age and income level, may also be moderators, these would require more arbitrary distinctions (e.g., sorting participants into "older old" versus "younger old" categories), and thus may be more difficult to interpret. Since female gender has been shown to influence decisions about activity is performed, it is likely to be a factor in this study (Eyler, 2003; Garcia Bengoechea, Spence, & McGannon, 2005; Krenichyn, 2005). To answer research question 3, all of the methods described above were used with health indices from both participants and non-participants electronic medical records as the dependent variables and neighborhood characteristics as the independent variables. The two datasets (participant and non-participant) were compared and then combined to provide greater power to detect clustering of health outcomes by neighborhood. All data were cleaned and visually inspected to ensure they met assumptions prior to analyses. Missing data for neighborhood audits were not uncommon due to the observational nature of the data collection method. Missing neighborhood dichotomous data were scored as "not observed". Missing 72

PAGE 94

neighborhood continuous data were ignored and not included in mean calculations for that segment. Missing data for individual respondents were minimal, due to extensive follow-up to assure survey completion, and were generally limited to either refusal to answer a particular demographic item, such as income level, or data that were missing from electronic medical records. Because missing self reported demographic data were so rare, they were in most .cases ignored and the remaining data were analyzed. Skipped answers for the standardized surveys were scored according to the instructions for that specific survey. Total N's were reported for all analyses. To aid in the presentation, analyses and discussion of this data this chapter has been organized into five sections: 1) Characterizing Denver Neighborhoods; 2) Characterizing Study Participants; 3) the Built Environment and Activity Engagement in Seniors; 4) Social Capital, Social Cohesion and Loneliness in Seniors; and 5) The Relationship of Neighborhood Characteristics to Health. 73

PAGE 95

Section 1: Characterizing Denver Neighborhoods Neiqhborfloocl CLeel 21 Figure 5.1.1. Conceptual Model: Neighborhood Characteristics. This section analyzes the many variables used to describe and characterize Denver neighborhoods selected for the study. The literature describes many ... . .. factors that coDl.einto play when studying the link between neighborhoods and d,eteriomtion and sidewalks, houSiDg quality,. and. quantity and quality of green spaces, and the variables _of interest in this study, and social isolation, have all bJl suggested as impacting perceived 8rui,;u:tua\ quality of life and health of older adults (Knlw;e, 1993, 1996; Takano et .. . . . aL,-2o62j: De"Vel()ping tbatrepresent how factOrs such as ... ;

PAGE 96

sociodemographics, crime rates, land use, and quantity and types of destinations interact with individual and group-level data is an important step to design of relevant health interventions (Diez Roux, 2004). Neighborhood data for this study included 1) reported data i.e., sociodemographics, and violent crime rates, and 2) measured data, i.e., observational data from the field audits. The latter contained the following constructs: walking, traffic and permeability (existence and maintenance of traffic calming devices, and street design); pedestrian safety from crime (surveillance), accidents (pathway obstructions) and traffic (crossing aids and curb cuts); aesthetics (streetscape and cleanliness); land use (presence of commercial and recreational destinations); and social capital variables (public courtesies or amenities; incivilities or signs of deterioration; territoriality; and vitality or people and activities observed). Reported Data Comparison of selected and non-selected neighborhoods Using data reported from published sources described earlier, eight neighborhoods in the City and County of Denver were selected from the 25 . . neighborhoods that had at least 80 living K.PCO membeJ:s ages 65 or older. Neighborhoods were ranked aceording to 1) violent crime rate per 1000 penions (includes homicides, sexual assaults, aggravated assaults, and other 2) average household income; 3) prqportion of adult residents ages 65 75

PAGE 97

years or older; and 4) walkability (determined using both population density and street connectivity' data), and one neighborhood was randomly selected from the top and bottom five for each of the criteria (see Table 5.1.1). Table 5.1.1. Denver Neighborhood Comparisons on Key Soci;:_,_-'_-,_.. >1 -Mean-Average :_, :Jneome : ,-: -> Non-KP Neigh 25 $52,150.81 28184.40 5636.88 -::: : ;; !,_ ;-,';_. ;.. !; .. .-' Selected 8 $48,000.03 13860.14 4900.30 VIolent Crinle Rate . +-. > .: .. -. Selected 8 6.58 3.589 1.27 % Noll"Wbite .; ... 3:::-.:..3'!-'7::---_, ,, . '< Selected 8 46.68 32.99 11.67 76

PAGE 98

Table 5.1.2 shows that the 25 neighborhoods with at least 80 eligible KPCO members were significantly different from the complete pool of Denver neighborhoods on two variables: proportion of elderly and violent crime rates, with. the KPCO neighborhoods having a higher proportion of elderly residents and lower violent crime rates. Table 5.1.2. Independent Samples Test between KP Neighborhoods (i.e., at least 80 KP members who met study criteria) and Non-KP Neighborhoods . c.':. /:: 1' : .' . . Equalicy-of .. -o-:1 : . ,, ' .. ..... '.' 1: -,F ,Sig. t df .. :. 'Sig. '95% CI .413 .523 .85 70 2.:tailed) of the Difference Lower Upper .400 -7258.06 17966.09 fEqual variances !Household ___ -+----+---4-....,.., ___ l--___ -+......,'"""'"-+----..:..-+----l fEqual variances not .81 42.85 .. .. assumed llo ResidentS. .. 0 Equal variances or ._ .. ,. Equal variances not 2.65 43.43 >\.. assumed .050 .823 2.76 70 .000 -4.21 70 variances RS.te > ., . : )Equal variances not -3.24 26.53 :. .: .568 > variances ;;)!;> .... '" ,._ 'EiaS;.;:;lsu=m:;;;:ed:.=....,. ...;--"""'--"--+-1 _....;... ___ .!J3qual variances not -1.42 45.96 . . ;328 -1.46 70 Table 'how furuting the neighborhood-selection . :--::. : --. ._ : .. : at'least8o eligible kfico members both narrowed the poolandsoniewhat . -. -. narrowed the range for each variable, although differences between -those bi@t and lowfor each variable were still quite large.

PAGE 99

Table 5.1.3. Neighborhood Variation: Comparison ofDenverNeipborhoods with Selected Neighborhoods, highest and lowest on Selection Criteria Average Household % Residents 65 Violent %Non-Income years or older Crime Rate White All Denver Neighborhoods Low $12,000 1.55 0 5.6 High $163,000 28.74 38.77 95.3 Selected Neighborhoods Low $35,000 6.23 1.89 7.1 High $109,800 28.74 10.22 95.3 Selected from only those neighborhoods that had at least 80 Kaiser Permanente Members m 2004 who met the study criteria of 65 years or older and had seen their physician within the past year. Measured Data Field Audit: Inter-Rater Reliability Analysis Two raters independently assessed a total of26 segments (5% of total street mlles) on both weekdays and weekends, dUring the daytime only (See Appendix B for sample neighborhood maps with selected segments). Raters used items from two instruments that were adapted with the authors' permission to capture aspects of the environment that are relevant to older adults. Raters were randomly assigned to be "primary'' or "secondary'' prior to obserVing each segment, and in cases where scores did not match, the primary observer's rating was used. This decision was based on the categorical nature of much of the data (i.e., it was either observed or not), which did not lend itself to being averaged for any specific segment. However, since the data for all street segments within a neighborhood were ultimately combined into a neighborhood score, and since both raters had the chance to be "primary'' for an equivalent number of segments, incon$istencies between raters should have been reduced. 78

PAGE 100

Nineteen pedestrian-related items from the SPACES (Pikora et al., 2002) instrument were combined with 18 items from the NBOT (Caughy, O'Campo, & Patterson, 2001 ), plus another 15 items thought to be relevant to senior pedestrians (e.g., sidewalk width, curb cuts, timing of pedestrian walk signals, bus stops) for a total of 52 items. Items were categorized into the domains of Walking Functionality; Safety (Traffic and Aesthetics, Destinations, Social Capital, plus a Subjective Assessment domain to capture the rater's overall perceptions of the neighborhood. Each of the domains contained sub-categories with several variables. To assess inter-rater reliability, at least 2 randomly selected segments per neighborhood were assessed independently by both raters at approximately the same time for a total of 36 segments. More seginents were selected in the first neighborhood assessed, in order to capture inconsistencies early on and identify any additional training needs. Because all variables were not observed within the selected segments, inter-rater reliability could only be assessed for 39 of the 52 items (see Table 5.1.4). Interrater consistency of scaled items was assessed using Spearmans' s rank coefficient (p ), which is appropriate for data that are not normally distributed (Stemler, 2004). Inter-rater reliability was assessed using a kappa statistic which provides a chance corrected measure of agreement (Portney & Watkins, 1993). Since SPSS will not compute kappa for items where raters failed to use the same range of scores (e.g., rating scale for an item was 1-3; 1 scored observations as 79

PAGE 101

lor 2, while rater 2 used the whole scale) in those cases, an extra observation was created usingthe missing category, but weighting it only :001 so as not to unduly influence the kappa statistic (observations from the original set are weighted 1) (UCLA, 2005). Following the methods presented by Pikora, et al. (Pikora et al., 2002) Spearman correlations between raters that exceeded 70 were considered to have a high level of consistency. A kappa score of >0. 75 was considered excellent agreement, 0.40-0.75 fair to good agreement, and <0.40 poor agreement (Aday, 1989; Fleis, 1981). Twenty-four of the 39 items had high inter-rater consistency, p>.70; 15 items also had excellent agreement (K >.75); 18 had good agreement (K = .40 to .75) and 6. had poor agreement of(K <.40). Items with poor inter-rater agreement tended to require more judgment ( e;g., assessment of whether or not a pedestrian could be seen from most homes) or were observed less frequently (e.g., security guards on premises). These items were still.included in preliminary analyses usi.p.g the primary rater's scores, however, their weak agreement is considered in interpretation of the results. 80

PAGE 102

Table 5 .1.4. Inter-rater Reliability Results Inter-rater reliability Inter-rater Factors Total items consistency 1 Kappa Statistics2 High Low Excellent FairGood WALKING FUNCTIONALITY Walking (Path type, maintenance, 5 5 3 2 continuity, width, gradient) Traffic (# lanes, traffic calming devices) 2 2 1 1 SAFETY Safety from Traffic (buffer zones, 2 2 1 1 curb cuts) Safety from Crime (front porches, 2 2 1 surveillance) AESTHETICS Streetscape (shade, front yard 6 4 2 1 3 maintenance, vacant lot maintenance, graffiti, litter, park maintenance) DESTINATIONS Land Use (Housing & Destination 6 3 3 2 2 Density) Facilities (street lights, 2 2 2 public transportation stop) SOCIAL CAPITAL Physical Incivilities (building condition. 2 1 I I 1 boarded up buildings) Territoriality (fences, security bars, 6 3 3 2 3 security signs, dogs, security guards, neigh watch signs) Stability (for sale and for rent signs) 2 1 1 1 1 SUBJECTIVE ASSESSMENT Summary Opinion of the NeighborhOod 4 1 3 I 3 (attractiveness, physical challenges, safety from traffic & crime) TOTALS 39 24 15 15 18 I Spearman correlations between raters > .70, considered to have a high level of consistency. 2 Kappa score >0.75 excellent agreement, 0.40-0.75 fair to good agreement, and <0.40 poor agreement 81 Poor 1 2 2 I 6

PAGE 103

Functionality: Walking Data collected during field audit observations are reported based on the constructs described earlier. Large between-neighborhood differences were found for the measures of sidewalk availability (-j(7)= 31.757, p<.001); sidewalk continuity (:(7)= 166.396, p<.001); sidewalk width (:(7)= 105.938, p<.OOI); sidewalk gradient or slope (-j(7)= 48.200, p<.001); and sidewalk maintenance (-(7)= 48.065. p<.001), which is not surprising given that neighborhoods were selected so as to maximize their differences. Tbl 515 Th B 'ltE a e e Ul t fi Walki nvuonmen or ng: S'd alk F t' l't1 1 ew unc 10na 1 y Walking Functionality Availability Continuity Width Gradient Maintenance Neigh N 40 40 39 39 40 1 Mean 2.85 1.45 1.82 2.13 2.13 Std. Dev. .427 .714 .389 .864 .607 Neigh N 35 35 34 35 35 2 Mean 2.97 2.97 2.26 2.20 2.06 Std. Dev. .169 .169 .448 .833 .684 Neigh N 34 34 27 34 29 3 Mean 2.53 2.82 1.30 2.82 2.21 Std. Dev. .825 .387 .609 .387 .675 Neigh N 39 40 30 40 30 4 Mean 2.38 2.45 1.63 2.63 2.40 Std. Dev. .877 .749 .669 .705 .675 Neigh N 38 39 38 39 38 5 Mean 2.87 3.00 2.00 2.72 2.11 Std. Dev. .414 .000 .329 .456 .388 Neigh N 26 26 24 26 22 6 Mean 2.77 2.81 1.63 2.31 1.77 Std. Dev. .587 .491 .576 .884 .152 Neigh N 40 39 40 39 40 7 Mean 3.00 2.74 1.18 2.13 2.00 Std. Dev. .000 .595 .385 .833 .320 Neigh N 32 32 26 31 26 8 Mean 2.59 1.25 2.00 2.97 2.81 Std. Dev. .798 .568 .000 .180 .402 Total N 284 285 258 283 260 Mean 2.75 2.43 1.73 2.48 2.17 Std. Dev. .616 .831 .575 .750 .619 df 7 7 7 7 7 p UnadJusted, umvanate Kruskal Wallis, "'p<.05, *"'p<.Ol Between Group Differences 82

PAGE 104

Post hoc comparisons of the means showed significant differences in sidewalk availability between Neighborhood 7 (selected for its low proportion of elderly residents), which had the highest sidewalk availability on average, and was significantly greater than Neighborhood 3 {high violent crime) and 4 (low violent crime), which had the lowest availability of sidewalks {p<.05). Neighborhood 2 (selected Tor its high walkability), had the next highest sidewalk availability and was also significantly greater than Neighborhood 4 (low crime) {p<.01). Neighborhoods 5 (selected for its high SES) and i had sidewalks with direct, continuous routes, rating significantly better than the lowest scoring neighborhoods: 1 (selected for its low walkability), 4 and 8 {p<.01), that tended to have routes that were winding and disjointed. Neighborhoods 1 (low walkability) and 8 (selected for its high proportion of elderly residents) scored significantly worse than all others, including Neighborhood 4 (p<.001). Neighborhood 2 (high walkability), had the widest sidewalks, significantly wider than all other neighborhoods except 5 (high SES) {p<.05). Neighborhood 7 (low elderly) had the narrowest sidewalks, scoring significantly lower than all neighborhoods except neighborhood 3 {p<.05). Neighborhoods 8 (high elderly), 3 {high crime) and 5 (high SES) had the lowest or flattest gradient for walking and were significantly less hilly than Neighborhoods 1 and 7 {p<.05). Neighborhoods 8 (high elderly) and 3 (high crime) were also less hilly than Neighborhood 2 {high walkability) (p<.Ol). 83

PAGE 105

Neighborhood 8 (high elderly), had the best maintained sidewalks, better than all neighborhoods except Neighborhood 4 (low crime) (p<.Ol). Neighborhood 6 (selected for its low SES) scored the lowest on for sidewalk maintenance, but was only significantly worse than Neighborhood 8 (high elderly) (p<.OOl). Summary. No single neighborhood had high scores across all sidewalk functionality variables. selected for its high walkability characteristics (grid-like street pattern, direct routes, mixed land-use) predictably scored high on sidewalk availability, continuity and width. However, Neighborhood 8, with its high proportion of elderly residents, scored best for flatter gradient and well-maintained sidewalks, two very important features for elderly pedestrians. Functionality: Traffic -Differences in the presence of Traffic Calming measures such as posted speed limits under 30 mph, fewer lanes, and more traffic calming structural features like all-ways stop signs, were observed across neighborhoods, but significant differences were only present for posted speed limit Ci!C7)= 19.868,p<.Ol) and presence of traffic calming features (i!(7)= 28.053,p<.001), all large in magnitude. The number of traffic lanes were not significantly different between neighborhoods Ci!C7)= 13.084,p>.05). 84

PAGE 106

Table 5.1.6. The Built Environment for Walking: Traffic Calming1 Traffic Calming Neighborhood speed limit lanes traffic calming Neigh N 16 40 40 1 Mean 2.50 2.95 2.00 Std. Dev. .816 .316 .816 Neigh N 8 35 35 2 Mean 2.13 2.83 2.60 Std. Dev. .354 .453 .553 Neigh N 11 34 34 3 Mean 2.09 2.76 2.03 Std. Dev. .831 .496 .627 Neigh N 16 40 40 4 Mean 2.75 2.77 2.20 Std. Dev. .577 .620 .608 Neigh N 21 39 39 5 Mean 2.81 2.90 2.26 Std. Dev. .402 .307 .751 Neigh N 11 26 26 6 Mean 2.64 2.65 2.12 Std. Dev. .505 .745 .711 Neigh N 10 40 40 7 Mean 2.60 2.88 2.13 Std. Dev. .843 .463 .757 Neigh N 14 32 31 8 Mean 2.86 3.00 2.00 Std. Dev. .363 .000 1.125 Total N 107 286 285 Mean 2.60 2.85 2.17 Std. Dev. .642 .468 .769 df 7 7 7 p * UnadJusted, umvanate Kruskal Walhs, *p<.05, **p<.Ol Between Group Differences Post hoc .comparisons showed that Neighborhoods 8 and 5 (hlgh elderly and high SES, respectively) had the lowest posted speed limits, significantly lower than Neighborhood 2 (high walkability) (p<.05). Neighborhood 2 had the greatest .number of observed traffic calming features, significantly better than Neighborhoods 1 (low walkability) and 3 (high crime) (p<.OS). 85

PAGE 107

Summary. Neighborhoods 2 (high walkability), 5 (high SES) and 8 (high elderly) scored best for presence of traffic calming features such as lower speed limits and numerous stop signs. Safety: Traffic Neighborhood features that enhance pedestrian safety from traffic showed large and significant differences by neighborhood for presence of cross walks (T(7)= 34.865,p<.001), crossing aids, such as medians (:j(7)= 31.159,p<.001), width of the buffer zone between the sidewalk and street (y}(7)= 127.047,p<.001), and curb cuts that connect both sides of the street ci(7)= 55.583,p<.001). 86

PAGE 108

Table 5 .1. 7: The Built. Environment for Walking: Traffic Safety1 Traflk Safi!ty Cross Crossing Buffers Cw-b Cuts Neighborhood Walks Aids (Parkways) Neigh N 40 40 39 38 1 Mean .OS .93 1.23 1.59 Std. Dev. .267 .267 .617 .719 Neigh N 36 36 34 35 2 Mean .56 .42 2.47 2.86 Std. Dev. .504 .500 .896 .355 Neigh N 38 38 17 31 3 Mean .Hi .74 1.30 1.76 Std. Dev. .370 .446 .724 .815 Neigh N 40 40 30 30 4 Mean .17 .72 1.33 2.100 Std. Dev. .452 .452 .758 .712 Ntjgh N 39 39 38 36 5 Mean .21 .19 2.82 1.99 Std. De"1. .409 .409 .563 .858 Neigh N 26 26 24 25 6 Mean .23 .11 1.58 2.280 Std. Dev. .4:30 .430 .929 .792 Neigh N 40 40 40 40 7 Mean .10 .85 1.13 1.850 Std. Dev. .304 .362 .463 .769 Neigh N 33 33 26 14 8 Mean .15 .19 1.00 2.39 Std. Dev. .364 .415 .000 .8315 Total N 292 2P2 258 249 Mean .22 .75 1.154 2.064 Std. Dev. .412 .432 .928 .833 ) ;;;/'c ".:;:-::;::; [;: 1: '' l';;t;<: .< ,, 1: :.'.1: ;; ... ; ;.; < : .. t ..... ;..->.:. 1_:, . .;; ... 0: ; 1 Unadjusted, univariate Kruskal Wallis or Pearson's Chi-Square; *p<.05, **p<.O 1 Presence of crosswalks differed between neighborhoods with Neighborhood 2 '(high walkability), having significantly more cross walks than Neighborhoods 1 87

PAGE 109

(low walkability), 3, 5, 7, and 8 (p<.OS). Presence of medians and signal lights with pedestrian walk signals (i.e., "crossing aids") also differed between neighborhoods with Neighborhood 1, having the most of these features proportionatelysignificantly better than Neighborhoods 2, 3 and 4 (p<.OS). Neighborhood 2 had significantly fewer observed crossing aids than all the other neighborhoods {p<.05). Neighborhoods 2 and 5 had significantly wider buffer zones than all the other neighborhoods {p<.01 ). Neighborhood 2 had a greater proportion of curbcuts connecting both sides of the street then every other neighborhood (p<.OOl) except for Neighborhood 8 (high proportion of elderly). Neighborhood 6 had more curbcuts than Neighborhood 1 {p<.05). Summary. Neighborhood 2 (high walkability) bad more crosswalks and curb cuts, and, along with Neighborhood 5 (high SES), wider buffer zones to protect pedestrians from street traffic. Safety: Personal Presence of street lights and good surveillance (the that a pedestrian would be seen from a house or yard) are considered important aspects of safety while walking, however there were no significant differences between neighborhoods on these two variables. The presence of sidewalk obstructions differed greatly between neighborhoods fi(7)= 17.7 41, p<.05). 88

PAGE 110

Table 5.1.8: The Built Environment for Walking: Personal Safety1 Personal Safety !Neighborhood Street Lights Surveillance Path Obstructions Neigh N 40 40 40 1 Mean .90 2.70 2.55 Std. Dev. .304 .516 .552 Neigh N 36 35 35 2 Mean .97 2.49 2.66 Std:Dev. .167 .742 .482 Neigh N 38 33 28 3 Mean .92 2.42 2.32 Std. Dev. .273 .867 .772 Neigh N 40 39 29 "4 Mean .88 2.28 2.24 Std. Dev. .335 .759 .830 Neigh N 39 38 37 5 Mean .85 2.55 2.65 Std. Dev. .366 .760 .538 Neigh N 26 25 25 6 Mean .88 2.44 2.36 Std. Dev. .326 .870 .757 Neigh N 40 38 40 7 Mean .. 95 2.39 2.10 Std. Dev. .221 .755 .810 Neigh N 33 31 26 8 Mean .97 2.45 2.58 Std. Dev. .174 .810 .703 Total N 292 279 260 Mean .91 2.47 2.43 Std. Dev. .. 280 .757 .703 df 7 7 7 P UnadJusted, uruvanate Kruskal Walbs or Pearson Chi-Square; Between Group Differences Post hoc comparisons indicated that the fewest sidewalk obstructions were observed for Neighborhoods 2 and 5, significantly better than Neighborhood 7, which had the most observed obstructions (p<.05). Summary. Neighborhoods 2 (high walkability) and 5 (high SES) had few sidewalk obstructions; Neighborhood 7, with its low proportion of elderly residents, had many. 89

PAGE 111

Aesthetics Neighborhood aesthetics for walking included streetscape variables such as shade and yard maintenance, as well as cleanliness, and parks. Large differences across Neighborhoods were observed for quantity of trees shading the path (x2(7)= 94.736,p<.001); front yard maintenance (-/(7)= 98.364,p<.001); quantity oflitter (-/(7)= 141.925,p<.001); graffiti (-/(7)= 140.944,p<.001); presence of parks (x2(7)= 22.821,p<.Ol); and upkeep of parks (-/(6)= 14.115,p<.Ol). 90

PAGE 112

bl 5 1 9 Th B "1 E Ta e ... e mt nvrronment w 1ki or a ng: A th I es etlcs AESTIIETICS Tree Tree Yard Litter Graffiti Parks Park quantity size Maintenance Maintenance INei2bborboods Neigh N 40 38 38 40 39 39 7 1 Mean 2.17 2.95 2.55 3.55 3.79 .21 2.71 Std. Dev. .501 .226 .645 .504 .409 .409 .. 756 Neigh N 35 35 35 35 35 35 5 2 Mean 2.86 3.00 2.23 2.80 2.49 .14 1.80 Std. Dev. .355 .000 .731 .677 .981 .355 1.095 Neigh N 34 29 33 35 33 35 4 3 Mean 1.91 3.00 1.70 2.49 3.82 .14 2.50 Std. Dev. .514 .000 .883 .981 .392 .355 1.000 Neigh N 40 36 39 38 40 40 4 4 Mean 2.58 3.00 2.64 3.53 3.78 .10 3.00 Std. Dev. .636 .000 668 .506 .423 .304 .000 Neigh N 38 33 39 39 39 37 11 5 Mean 2.58 2.94 2.44 3.59 3.85 .32 3.00 Std. Dev. .599 .242 .641 ;498 .366 .475 .000 Neigh N 26 22 23 26 25 25 6. 6 Mean 2.04 2.82 1.52 2.04 3.04 .28 2.33 Std. Dev. .599 .501 .846 1.076 .735 .458 .516 Neigh N 40 31 38 40 40 40 -7 Mean 1.85 3.00 1.29 2.35 2.57 .00 Std. Dev. .533 .000 .565 .949 1.107 .000 -Neigh N 32 31 32 32 33 33 2 8 Mean 2.66 2.97 2.72 3.88 3.94 .06 3.00 Std. Dev. .545 .180 .523 .554 .242 .242 .000 Total N 285 255 277 285 284. 284 39 Mean 2.33 2.96 2.17 3.06 3.42 .15 2.64 Std. Dev. .643 .205 .852 .960 .868 .359 .707 df 1 1 7 1 7 7 6 p ** ** UnadJusted, uruvanate Kruskal Wallts or Pearson Chi-Square, *p<.OS; *p<.Ol 91

PAGE 113

Post hoc comparisons found quantity of shade was greatest for Neighborhoods 2 and 8 which were significantly greater than Neighborhoods 1, 3, 6 and 7 (p>.05). Neighborhoods 4 and 5 also had significantly more shade than Neighborhoods 3, 6 and 7 (p<.05). Neighborhoods 8 (high elderly) and 4 (low crime) had the best maintained yards, on average, and were significantly better than Neighborhoods 3 (high crime), 6 (low income) and 7 (low elderly) that had the lowest average yard maintenance ratings (p<.OOl). Neighborhoods 3, 6, and 7 were also significantly lower than Neighborhoods 1 (low walkability) and 5 (high income) (p<.Ol). Cleanliness measures included quantity of litter and graffiti observed as well as park maintenance. The least amount of litter was observed in Neighborhood 8 (high elderly}, which was cleaner than all neighborhoods except Neighborhood 5 (high income) (p<.05). Neighborhoo-d 6 (low income) had the most litter and was signifi_cantly worse than Neighborhoods 1, 4, 5, and 8 (p<.001). Neighborhoods 8 and 5 also had significantly less graffiti than Neighborhoods 2, 6 and 7 (p<.OOl). Neighborhoods 2 and 7, had the most graffiti and were significantly worse than all neighborhoods except for 6 (p<.001). Park maintenance was rated only when a park was observed within a neighborhood, thus no post hoc analyses were conducted for this variable due to too few cases of an.observable park within some neighborhoods. As mentioned above, presence of parks differed significantiy between neighborhoods. All neighborhoods except Neighborhood 7 had at least one park. 92

PAGE 114

Neighborhood 5, which had a very large park that could be observed from many of the audited segments, was significantly greater than Neighborhoods 4 and 8 (p<.05). Summary. Neighborhoods differed with regard to aesthetic features, but Neighborhood 8 (high elderly) scored best with regard to shade, well maintained yards, and very little litter. Neighborhood 5 (high SES) also scored high on cleanliness of both its streets and its nicely maintained, centrally located park. Neighborhoods 3 (high crime), 6 (low SES), and 7 (low elderly) scored poorly on most of these same features. Destinations: Land-Use Land Use differed between neighborhoods for both housing density Cr.!(?)= II7.72l,p<.OOI) and non-residential uses (i'(7)= 29.599,p<.OOI). Proportion of land with?: l/3rd used for non-residential purposes versus that which was mostly dedicated to residences was a large and significant difference (i(7)= I6.053,p<.05). 93

PAGE 115

Table 5 .1.1 0. The Built Environment for Walking: Land Use 1 DESTINATION: Land Use housing destination >or= l/3rd N eigbborboods density density non residential Neigh N 39 40 40 I Mean 1.23 1.18 .05 Std. Dev. .627 .446 .221 Neigh N 31 36 36 2 Mean 1.68 1.86 .31 Std. Dev. .541 .723 .467 _Neigh N 25 35 35 3 Mean 1.12 1.80 .31 Std. Dev. .332 .901 .471 Neigh N 36 40 40 4 Mean 1.39 1.48 .23 Std. Dev. .803 ./84 .423 Neigh N 36 38 38 5 Mean 1.28 1.50 .18 Std. Dev. .659 .. 726 .393 Neigh N 24 26 26 6 Mean 1.67 1.81 .31 Std. Dev. .868 .694 .471 Neigh N 35 40 40 7 Mean 1.03 1.43 .15 Std. Dev. .169 .675 .362 Neigh N 32 32 32 8 Mean 2.44 1.44. .09 -Std. Dev. .878 .564 .296 Total N 258 287 287 Mean 1.59 1.54 .20 Std. Dev. .866 .727 .400 df 7 7 7 .JJ... ** ** ,, UnadJusted, umvanate Kruskal Walbs or Pearson Chi-Square, **p<.Oi Post hoc comparisons indicated that Neighborhood 2, with its many apartment complexes, had significantly higher residential density compared with all other neighborhoods except Neighborhood 8 that also had a lot of multi-occupancy housmg (p<.001). Neighborhood 8 also had significantly higher residential density than most neighborhoods (except Neighborhoods 2 and 6) (p>.05). Neighborhoods 1 and 7, with their many single family homes, had the lowest residential density. 94

PAGE 116

Non-residential land-use density post hoc comparisons indicated that Neighborhood 1, with its more traditional suburban design, had the least proportion of mixed land-use segments, significantly lower than Neighborhoods 2, 3 and 6 (p<.05), where more than 30% of audited segments were non-residential. Destinations: Facilities There were significant differences across neighborhoods for shops (l(7)= 45.987, p<.OOl); services ciC7)= 40.707,p<.001) and restaurants (-(7)= 77.652,p<.001). T bl 5 1 11 Th B "It E t fi W 1kin D f f n Types1 a e .. e Ul nvtronmen or a lg: es ma 10 DESTINATION: Facilities IN eighborhood Sho_ps Services Restaurants Neigh N 40 40 40 1 Mean .03. .18 .08 Std. Dev. .158 .385 .267 Neigh N 36 36 36 2 Mean .31 .78 .89 Std. Dev. .786 1.072 .515 Neigh N 38 38 38 3 Mean .08 .76 .32 Std. Dev. .273 .490 .471 Neigh N 40 40 40 4 Mean .50 .52 .25 Std. Dev. 1.754 .960 .439 Neigh N 39 39 39 5 Mean .IS .36 .23 Std. Dev. .432 .584 .427 Neigh N 26 26 26 6 Mean .23 .50 .35 Std. Dev. .652 .648 .485 Neigh N 40 40 40 7 Mean .10 .48 .48 Std. Dev. .496 .847 .506 Neigh N 33 33 33 8 Mean .06 .33 .09 Std. Dev. .242 .816 .292 Total N 292 292 292 Mean .18 .49 .33 Std. Dev. .789 .775 .500 df 7 7 7 p Pearson Chi-Square, *p<.OS, p<.Ol 95

PAGE 117

Post hoc comparisons indicated that Neighborhood 2, which was within the heart of the city, and Neighborhood 6, which was bisected by a major road that had numerous strip malls, had significantly more shops than Neighborhood 1 (p<.0.5). Neighborhoods 2 and 3 (which had a large number of churches), had significantly more services than Neighborhoods 1, 5 and 8 (p<.05). Neighborhood 2 had significantly more restaurants than all other neighborhoods (p<.05), although Neighborhood 7 (a highly diverse neighborhood with a large number of Mexican and Asian eating establishments) had a significantly greater number than Neighborhoods 1 and 8 (p<.001). Summary. 2 (high walkability) and 8 (high elderly) had the highest residential density and Neighborhood 2 (high walkability), with its central urban location, also hadthe greatest number of shops, restaurants and services. Destinations: Recreation Similar to the above-described facilities, features of the built environment that provided opportunities for exercise varied by neighborhood. Specifically parks (-(7)= 22.821,p<.Ol); walking or biking trails ci(7)= 20.22l,p<.Ol); lakes (l(7)= 40.556,p<.001); or tennis (..j(7)= 22.108,p<.Ol). Availability of a gyin or public gardens did not vary significantly between neighborhoods. Overall opportunities for exercise, which was a sum-score of all available within-neighborhood options, did show a large and significant difference across neighborhoods ("l(7)= 22.47l,p<.Ol). 96

PAGE 118

T bl 5 1 12 Th B "lt E t W 1kin P ks dR a e .. e U1 nvrronmen or a tg: ar an f ecrea IOn DESTINATION: Recreation Gym Park trails lake tennis Other oppfor courts recreation exercise !Neighborhood summed Neigh N 41 39 40 40 40 41 41 I Mean .00 .21 .13 .00 .15 .17 .63 Std. Dev. .000 .409 .335 .000 .362 .381 1.356 Neigh N 37 35 36 36 36 37 37 2 Mean .00 .14 .03 .00 .03 .08 .27 Std. Dev. .000 .355 .167 .000 .167 .277 .652 Neigh N 39 35 38 38 38 39 39 3 Mean .00 .14 .03 .00 .03 .10 .31 Std. Dev. .000 .355 162 .000 .. 162 .307 .893 Neigh N 41 40 40 40 40 41 41 4 Mean .05 .10 .10 .00 .00 .02 .27 Std. Dev. .312 .304 .304 .000 .000 .156 .742 Neigh N 40 37 39 39 39 40 40 5 Mean .03 .32 .21 .15 .10 .05 .83 Std. Dev. .158 .475 .409 .366 .307 .221 1.318 Neigh N 27 25 26 26 26 27 27 6 Mean .00 .28 .12 .19 .00 .26 .81 !Std. Dev. .000 .458 .326 .402 .000 .447 1.442 Neigh N 41 40 40 40 40 41 41 7 Mean .00 .00 .00 .00 .00 .00 .00 Std. Dev. .000 .000. .000 .000 :ooo .000 .000 Neigh N 34 34 34 34 34 34 34 8 Mean .03 .06 .00 .00 .00 .06 .15 Std. Dev. .171 .239 .000 .000 .000 .239 ;500 Total N 300 285 293 293 293 300 300 Mean .01 .15 .08 .04 .04 .09 .40 Std. Dev. .141 .359 .264 .190 .199 .282 .995 df 7 7 7 7 7 7 7 p ** ** ** ** ** ** UnadJusted, uruvanate Kruskal Wall1s or Pearson Chi-Square, *p<.OS, **p<.Ol. Post hoc comparisons indicated Neighborhoods 5 and 6 had access to a park with a large lake with a walking path around it, which was significantly different from all other neighborhoods. For total opportunities to exercise, Neighborhood 5 scored highest, but this was only statistically significant when compared with Neighborhood 7 (p<.05) where observable public recreational facilities were scarce, a large difference. 97

PAGE 119

Summary. Neighborhood 5 (high SES), selected for its high average household income, had the greatest number of recreational resources. Neighborhood 7 (low elderly) was at the opposite end of the continuum with regard to opportunities for exercise. Destinations: Public Courtesies Public courtesies include conveniences that contribute to resident quality of life. Courtesies that did not significantly differ across neighborhoods included street lights, public phones and senior services. Courtesies with significant between neighborhood differences included mail boxes for posting letters, a moderate difference (-(7)= 27.34l,p<.001); and large, significant differences for public benches (-(7)= 63.122,p<.001); public trash receptacles (-(7)= 21.933,p<.Ol); newspaper dispensers {i(7)= 51.331,p<.OOI) and public transportation stops (-(7)= 30.202,p<.001). Condition of public transportation stops also varied significantly (j(7)= 21.542,p<.Ol), a large difference. 98

PAGE 120

Table 5.1.13. The Social Environment for Walking: Public Courtesies1 DESTINATION: PUBLICCOURTESIES mailbox public trashcan senior news RTDStop RTDStop SUM bench services dispensers Maint SCORE Neighborhood Neigh N 40 40 40 40 40 40 II 41 I Mean .OS .00 .05 .00 .00 .28 1.91 1.24 Std. Dev. .221 .000 .221 .000 .000 .452 .539 .70 Neigh N 36 36 36 36 36 37 14 37 2 Mean .25 .14 .II .00 .31 .43 2.14 2.27 Std. Dev. .439 .351 .319 .000 .467 .502 .770 1.09 Neigh N 38 38 38 38 38 39 12 39 3 Mean .00 .03 .05 .00 .03 .31 1.58 1.33 Std. Dev. .000 .162 .226 .000 .162 .468 .515 .89 Neigh N 40 40 40 40 40 41 9 41 4 Mean .05 .10 .12 .00 .00 .22 1.89 1.49 Std. Dev. .221 .304 .335 .000 .000 ..419 .782 1.05 Neigh N 39 39 39 39 39 40 4 40 5 .Mean .05 .21 .05 .00 .03 .10 2.25 1.28 Std. Dev. .223 .409 .223 .000 .160 .304 .500 1.04 Neigh N 26 26 26 26 26 27 10 27 6 Mean .00 .08 -.08 .00 .04 .37 2.20 1.70 Std. Dev. .000 .272 272 .000 .196 .492 .632 1.03 Neigh N 40 40 40 .40 40 41 3 41 7 Mean .03 .00 .03 .00 .00 .07 2.00 1.15 Std. Dev. .158 .000 .158 .000 .000 .264 .000 .62 Neigh N 34 34 34 34 34 34 16 34 8 Mean .15 .50 .29 .03 .21 .47 2.69 2.79 Std. Dev. .359 .508 .462 .171 .410 .507 .479 2.01 Total N 293 293 293 293 293 299 79300 Mean .07 .13 .10 .00 .07 .27 2.11 1.62 Std. Dev. .26 .33 .29 .06 .26 .45 .68 1.22 df 7 7 7 7 7 7 7 7 p ** .. ** Unadjusted, umvanate Kruskal Walhs or Pearson Chi-Square, p<.OS, p<.Ol. 99

PAGE 121

Post hoc comparisons found that Neighborhoods 2 (high wal.kable) and 8 (high elderly) had significantly more public transportation stops observed than Neighborhoods 4 (low crime) and 5 (high income) (p<.05). Neighborhood 6 (low income) also had more public transportation stops than Neighborhood 5 (p<.05). In addition, Neighborhoods 2 and 8 provided significantly more total public courtesies than all other neighborhoods with the exception of Neighborhood 6 (p<.05). Neighborhood 7 (low elderly) provided the fewest total public courtesies. Summary. Neighborhoods 2 (high walkability) and 8 (high elderly) had the greatest number of public courtesies, and Neighborhood 7 (low elderly), the fewest. Social Capital: Incivilities Physical incivilities include variables that contribute to degradation of a neighborhood. Large and significant between-neighborhood differences were observed for all physical incivility variables including condition of buildings (-(7)= buildings (:i(7)= 23.993,p<.Ol); existence ofnuisance properties (e.g., bars, checks cashed sites) (-(7)= 38.218,p<.001); condition of front yards (-(7)= 89.746,p<.OOI); condition of parks (-(6)= 13.884,p<.05); graffiti (:l(7)= 59.315,p<:oot) and litter (:i(7)= 66.353,p<.001). Significant differences remained between neighborhoods when incivilities were summed cl(7)= 96.877, p<.OOl). 100

PAGE 122

T bl 5 1 14 Th S 1 E a e .. e oct a fi w 1kin Ph a1 In '1" I nvtronment or a g: lYSlC ctvt ttles SOCIAL CAPITAL: Physical Incivilities Building Decay Nuisance Yard Park Litter Graffiti SUM Neighborhood Maint. Properties Maint. Maint. SCORE Neigh N 38 38 40 38 7 40 39 40 1 Mean .00 .00 .00 .08 .14 .00 .00 .10 Dev. .000 .000 .000 .273 .378 .000 .000 .379 Neigh N 35 35 36 35 5 35 35 36 2 Mean .03 .00 .44 .17 .60 .11 .29 1.11 Std. Dev. .169 .000 .504 .382 .548 .323 .458 1.141 Neigh N 33 34 38 33 4 35 33 38 3 Mean .06 .03 .39 .25 .29 .00 .89 Std. Dev. .242 .171 .393 .496 .500 .458 .000 1.247 Neigh N 38 39 40 39 4 38 40 40 4 Mean .03 .03 .27 .10 .00 .00 .00 .43 Std. Dev. .162 .160 .452 .307 .000 .000 .000 1.010 Neigh N 37 37 39 39 11 39 39 39 5 Mean .03 .03 .03 .08 .00 .00 .00 .15 !Std. Dev. .164 .164 .160 .270 .000 .000 .000 .432 Neigh N 25 25 26 23 6 26 25 26 6 Mean .20 .08 .31 .70 .oo .so .08 1.77 Std. Dev. .408 .277 .471 .470 .000 .510 .277 1.478 Neigh N 39 39 40 38 40 40 40 7 Mean .15 .18 .15 .68 -.33 .30 1.75 Std. Dev. .366 .389 .362 .47i -.474 .464 1.391 Neigh N 32 33 33 32 2 32 33 33 8 Mean .03 .00 .12 .03 .00 .03 .00 .21 Std. Dev. .177 .000 .331 .177 .000 .177 .000 .740 Total N 277 280 292 277 39 285 284 292 Mean .06 .04 .18 .26 .13 .14 .08 .77 Std. Dev .240 .203 .386 .439 .339 .352 .279 1.202 df 7 7 7 7 6 7 7 7 p ** Unadjusted, umvanate K.ruskal Walhs or Pearson Chi-Square, *p<.OS, **p<.Ol Post hoc comparisons found Neighborhoods 6 (low income) and 7 (high ethnic minority, low elderly) had significantly higher levels of many physical incivilities including worse condition of yards and more litter, than Neighborhoods 1, 2, 5 and 8 (p<.01). Significantly more boarded up buildings were observed in Neighborhood 7 than Neighborhoods 1, 2 and 8 (p<.05). More graffiti was obseryed for Neighborhoods 2 and 7 than Neighborhoods 1, 3, 4, 5 and 8 (p<.01). A greater 101

PAGE 123

number of nuisance properties and were obsexved in Neighborhoods 2 than Neighborhoods 1, 3, 5, 7 and 8 (p<.01). Nuisance properties were relatively high in Neighborhood 6 as well, significantly more than Neighborhoods 1 and 5 (p<.01). The total number of incivilities were greatest for Neighborhoods 6 and 7, significantly more than Neighborhoods 1, 4, 5 and 8 (p<.01). Neighborhoods 1 (low walkable) and 5 (high SES) had the fewest physical incivilities across all variables. Summary. Observed incivilities such as boarded up buildings, litter, graffiti, and deteriorated yard condition were highest in Neighborhoods 6 (low SES) and 7 (low elderly) and lowest in Neighborhood 1 (low and 5 (high SES). Social Capital: Territoriality Between neighborhood differences were noted for all variables related to how vigilant or territorial a neighborhood might seem to pedestrians. High fences or hedges (F(7,272)=6.179, .001); security bars on the windows (:(7)= 96.031, p<.001); security warning signs posted (:(7)= 16.100,p<.05); Security guards or attendants (-f(7)= 70.390,p<.001); visible dogs in the yard (-f(7)= 20.677,p<.Ol); neighborhood watch signs posted (:(7)= 19.740,p<.01); all with large and significant differences. When Territoriality indicators were summed, the total indicated large and significant between-neighborhood difference (F(7,273)=6.224, p<.001). 102

PAGE 124

Table 5.1.15. The Social Environment for Walking: Territoriality' SOCIAL CAPITAL: TerritorialiCy High Window Security Security Dog Neigh SUM Neighborhood Borders Bars Signs Guards on Premise Watch SCORE (fences) Sign Neigh N 40 39 38 40 38 40 40 1 Mean 1.53 1.26 1.37 1.05 1.11 .38 1.29 Std. Dev. .877 .442 .633 .221 .311 .490 .313 Neigh N 35 35 35 35 35 36 35 2 Mean 1.94 2.20 1.63 1.17 1.03 .11 1.59 Std. Dev. 1.083 .901 .843 .707 .169 .319 .361 Neigh N 33 33 32 33 33 38 33 3 Mean 1.79 2.18 1.50 1.00 1.03 .16 1.49 Std. Dev. .927 .950 .672 .000 .174 .37 .301 Neigh N 39 39 39 39 39 40 39 4 Mean 2.00 1.15 1.41 1.00 1.08 .28 1.33 Dev. 1.051 .489 751 .000 .270 .452 .316 Neigh N 37 37 37 37 37 39 37 5 Mean 1.65 1.11 1.65 1.08 1.22 .44 1.34 Std. Dev. 978 .393 .676 .277 .417 .502 .274 Neigh N 25 25 25 25 25 26 25 6 Mean 2.20 2.00 1.32 1.04 1.08 .38 1.53 Std. Dev. .764 .866 .476 .200 .277 .496 .215 Neigh N 40 40 40 40 40 40 40 7 Mean 2.73 1.95 1.15 1.00 1.33 .15 1.63 Std. Dev. 905 .846 .362 000 .572 .362 .242 Neigh N 31 31 31 32 31 33 32 8 Mean 1.61 1.23 1.48 2.25 1.06 .24 1.59 Dev. 1.086 .560 .769 1.481 .250 .435 .557 Total N 280 279 277 281 278 292 281 Mean 1.93 1.61 1.44 1.19 1.12 .26 1.47 Std. Dev 1.03 .83 .6876 .69 .35 .44 .3565 df 7 7 7 7 7 7 7 p Unadjusted, umvanate ANOVA, Kruskal Wallis or Pearson Chi-Square, *p<.OS, **p<.Ol 103

PAGE 125

Post hoc comparisons indicated that Neighborhood 7 had proportionately more high property borders (e.g., fences, hedges) than all other neighborhoods except 6 (p<.05) .. Neighborhood 6 had significantly more high property borders than Neighborhood 1 (p<.05). Neighborhoods 2, 3, 6 and 7 had the highest proportion of barred windows, significantly greater than Neighborhoods 1, 4, 5 and 8 (p<.OS). Differences in posting security signs were only significant between Neighborhood 5, which posted a lot, and Neighborhood 7, which did not post signs (p<.01). Security guards were most common in Neighborhood s,significantly greater than all other neighborhoods (p<.05). The post hoc comparisons showed no significant differences for the presence of dogs. Overall, Neighborhoods 2 and 7 had the greatest number of features that would indicate higher territoriality; significantly more than Neighborhoods 1, 4 and 5 (p<.001). Yet the number of neighborhood watch signs observed was highest for Neighborhood 5, significantly more than both Neighborhoods 2 and 7 (p<.01). Summary. Neighborhoods 2 (high walkability), 6 {low SES) and 7 (low elderly) had the greatest number of territoriality signs (high fences, barred windows). Neighborhood 5. (high SES) had few signs of territoriality, although it did have the highest number of observed neighborhood watch signs. 104

PAGE 126

Social Capital: Stability Neighborhood stability, resident transience, appeared to have large and significant between-neighborhood variance. This was measured by observed numbers of"for sale" signs (:(7)= 20.513,p<.01) and "for rent" signs (x2(7)= 73.413, p<.OO 1) with fewer signs representing greater neighborhood stability. Table 5 .1.16. The Social Environment for Walking: Stability1 SOCIAL CAPITAL: Stability Neighborhoods For Sale Signs" For Rent Neigh N 39 39 1 Mean 2.64 3.00 Std. Dev. .486 .000 Neigh N 35 35 2 Mean 2.80 2.17 Std. Dev. .406 .618 Neigh N 34 34 3 Mean 2.68 2.79 Std. Dev. .535 .479 Neigh N 40 40 4 Mean 2.78 2.70 Std. Dev. .423 .564 Neigh N. 38 38 5 Mean 2.63 2.84 Std. Dev. .489 .437 Neigh N 26 26 6 Mean 2.46 2.69 Std. Dev. .582 .471 Neigh N 40 40 7 Mean 2.43 2.85 Std. Dev. .675 .362 Neigh N 32 32 8 Mean 2.84 2.97 Std. Dev. .515 .177 Total N 284 284 Mean 2.66 2.76 Std. Dev .531 .491 df 7 7 p ** ** I' ** Unadjusted, umvanate Kruskal Wallis, p<.OS, p<.Ol 2Higher score denotes fewer signs/greater stability. 105

PAGE 127

Post hoc comparisons found no significant differences between specific neighborhoods with regard to numbers of"for sale" signs. Neighborhood 2 (high walkable) had the greatest number of residences for rent compared with all other neighborhoods (p<.OS) and Neighborhoods 1 (low walkable) and 5 (low income) had the least number of residences for rent. Summary. Neighborhood 1, with its suburban location and single family homes, and Neighborhood S, with its expensive real estate, had few observed rental properties. Neighborhood 2, located _downtown Denver, had many rental properties, which may indicate a more transient neighborhood. Social Capital: Vitality Large and significant between-neighborhood differences for people outside (F(7,284)= 3.592,p<.01); and activities (:(7)= 34.863,p<.001) were observed. Neighborhoods with no people observed by auditors (:(7)= 17.514,p<.05); and no activities observed also had large and significant between-neighborhood variance (-j(7)= 22.008,p<.01). Neighborhoods did not vary significantly for observed sedentary activities like sitting, parking or standing around. 106

PAGE 128

Table 5.1.17. The Social Environment for Walking: Vitality1 SOCIAL CAPITAL: Vitality Total people No activities Sitting, Total activities Neighborhood No people observed observed parking, observed observed standing Neigh N 40 40 40 40 40 I Mean .05 1.65 .03 .23 1.30 Std. Dev. .221 .834 .158 .423 :823 Neigh N 36. 36 36 36 36 2 Mean .00 1.53 .00 .36 1.61 Std. Dev. .000 .696 .000 .487 .688 Neigh N 38 38 38 38 38 3 Mean .24 1.08 .:Zl .37 .68 Std. Dev. .431 .997 .413 .489 .842 Neigh N 40 40 40 40 40 4 Mean .:zo 1.25 .20 .13 .98 Std. Dev. .405 .870 .405 .335 .620 Neigh N 39 39 39 39 39 5 Mean .05 1.87 .05 .21 u:z Std. Dev. .223 .978 .223 .409 1.335 Neigh N 26 26 26 26 26 6 Mean .12 1.54 .08 .27 1.12 Std. Dev. .326 .811 .272 .452 .816 Neigh N 40 40 40 40 40 7 Mean .10 1.80 .03 .32 1.13 Std. Dev. .304 .911 .158 .474 .853 Neigh "N 33 33 33 33 33 8 Mean .09 1.70 .09 .36 1.33 Std. Dev. .292 .918 .292 .489 .854 Total N 292 292 292. 292 292 Mean .11 1.55 .09 .28 1.25 Std. Dev. .309 .912 .280 .448 .935 elf 7 7 7 7 7 p ** ** ** UnadJusted, umvanate ANOVA, Kruskal Wallis or Pearson s Chi-Square, *p<.OS, *p<.Ol Post hoc comparisons found Neighborhoods 3 (high crime) had significantly fewer people observed than Neighborhoods 1 and 5 (p<.05). Neighborhoods 3 and 4 also had significantly fewer physical activities observed than Neighborhoods 1, 2 and 7 (p<.05). Neighborhoods 5 and 7 had the_most people observed and were significantly higher than Neighborhood.3 (p<.05). Neighborhood 2 (high walkable with high non-residential density) and Neighborhood 5 (high income with a large public park) had the most people observed performmg physical activities (e.g., 107

PAGE 129

walking, running, biking) and were significantly higher than Neighborhoods 3 and 4 (p<.OS). Neighborhood 1 also had more physically active people observed than Neighborhood 3 (p<.OS). Summary. Neighborhoods 2 (high walkability) and 5 (high SES) were lively, vital neighborhoods with numerous people out and about, many of whom were engaged in physical activities such as dog walking, jogging or biking. Neighborhood 3 (high crime) had very few people observed engaged in outside physical activity. Subjective Assessment Raters provided an overall summary opinion for each segment walked. Large and significant between-neighborhood differences were found for perceived safety from traffic (-(7)= 16.962, p<.OS); perceived safety from crime (-(7)= 38.033, p<.OOl); how aesthetically pleasing the walk seemed (X2(7)= 113.497, p<.OOl); and how physically difficult the walk seemed (F(7,279)=11.018, p<.OOl. 108

PAGE 130

Table 5.1.18. Subjective Walking Rating1 SUBJECTIVE ASSESSMENT Safe from Safe from Aesthetically Physically easy Neighborhood Traffic Crime Pleasin_g_ to walk Neigh N 39 40 40 40 1 Mean 2.69 2.80 2.00 2.10 Std. Dev. .614 .405 .453 .672 Neigh N 35 35 35 35 2 Mean 2.51 2.57 2.34 2.66 Std. Dev. .507 .502 .482 .539 Neigh N 35 34 36 35 3 Mean 2.23 2.62 1.69 2.06 Std. Dev. .877 .551 .525 .765 Neigh N 40 40 40 40 4 Mean 2.12 2.83 2.38 2.28 Std. Dev. .822 .385 .705 .599 Neigh N 39 39 39 39 5 Mean_ 2.44 2.77 2.46 2.54 Std. Dev. .680 .427 .555 .555 Neigh N 26 26 26 26 6 Mean 2.27 2.31 1.65 1.85 Std. Dev. .778 .618 .689 .732 Neigh N 40 40 40 39 7 Mean_ 2.22 2.40 1.20 1.90 Std. Dev. .768 .632 .405 .598 Neigh N 33 33 33 33 8 Mean 2.52 2.91 2.52 2.79 Std. Dev. .667 .292 .667 .485 Total N 287 287 289 287 Mean 2.38 2.66 2.03 2.28 Std. Dev. .737 .516 .716 .693 df 7 7 7 7 p Unadjusted, umvanate ANOVA or Kruskal Walhs; *p<.05, **p<.Ol Post hoc comparisons found Neighborhood 1 (low walkable, suburban) having the highest perceived safety from traffic, differing significantly only from Neighborhood 4 (low crime, suburban) {p<.05). Neighborhood 8 (high elderly) had the greatest perceived safety from crime, significantly better than Neighborhoods 2 (high walkable, urban), 6 (low income) and 7 (low elderly) {p<.05). Neighborhoods 109

PAGE 131

1 and 4 were also perceived as relatively safe, significantly better than Neighborhoods 6 and 7 (p<.05) which were perceived as the least secure neighborhoods. Neighborhoods 5 (high income) and 8 (high elderly) were perceived as the most aesthetically pleasing to walk in, significantly more pleasing than Neighborhoods.!, 3, 6 and 7 (p<.01). Neighborhood 7 was considered less pleasing to walk in than all other neighborhoods except 6 (p<.01). Neighborhood 8 was the easiest physically to walk in and was significantly less challenging than Neighborhoods 1, 3, 4, 6 and 7 (p<.Ol). Neighborhood 2 was also relatively easy to walk in compared with Neighborhoods .1, 3, 6 and 7 (p<.Ol). Neighborhoods 6 and 7 were the most physically challenging for walking, significantly more difficult than 2, 5 and 8 (p<.Ol). Summary. Neighborhoods 5 (high SES) and 8 (high elderly) were rated as pleasing, safe and free of physical challenges for walkers. Neighborhood 2 (high walkability) was also considered "easy" to walk in, but was not perceived to be as safe from crime. Neighborhoods 6 (low SES) and 7 (low elderly) were rated least favorably with regard to safety, aesthetics and physical challenges. Discussion This section was the first step to answering Research Question 1 : What components, if any, of the environmental context enable or inhibit physical and social activity in older adults? By examining the relationships among a large number 110

PAGE 132

. of reported and measured variables, the above analyses provides a basis for characterization of the neighborhoods. Since neighborhoods were selected to maximize variation in walkability, violent crime rate, SES and proportion of elderly, it was not surprising that there were large and significant differences between neighborhoods on most variables, although post hoc analyses indicated that these differences tended to be limited to few neighborhoods that were at the extremes for specific characteristics. In addition, since many of the measured varfables were based on field audit observations of randomly selected street segments that represented 5% of street miles in each neighborhood, and since observations were limited to. daylight hours which could impact observation of more dynamic features, such as people and activities observed, failure to observe a specific variable cannot be construed to mean that the variable did not exist anywhere in the neighborhood. However, assuming that the street segments observed were representative of the overall neighborhood, and considering that both weekday and weekend visits were made to all neighborhoods, relative comparisons could be made (see Table 5.1.19). 111

PAGE 133

Table 5.1: l9. Sliinmary ofNeiKhb.orhood Characteristics N. 1 Sdty, I::Aesihedcs-1 Des1inan8m I _Sodal" .. L .. mr 'waikiiag. siahmty. 1 vtianiy. Suhjeedve Assess. 1 I + T:raflic +M&Dy, +FeW--+I +Some +ffidt Low waik dgnals,wi. people smi;r -o:i1imiile6 +Few : (tramc) -Hillyw1th ._ .. -. . ( ' wizldll1g. . . tennit > non-little shade: comitmous signs or "Few retail .-Poor routes_ cuibcuts facilities High2Wa1k .:r::. sicleWaiiB ..... I I fU:.&orn. 10t of pUblic -Some I I I -Low saretY. N rOuB sileet, p-aff:di, rec-reatDn' inciVilitks& '(c:rilne)3 I +Ma:DY +So;me I +Med.: u.;.;.._ ... c -; -Sidewalks : &dli:aes 1iJ.Jic nJne ' ' p nor.. umfoJ::mly shade; recma&:n mcivilities & I I I .-Naimw S_lgr!S :m,aD:d:amed ,sigUS of:. people aestlehcs sidewalks-. yams . 4_ I +WeD.-_ +WeD.-: +Some +Med. LO C maintainied. maintained. fadi.iieB of. sate'f;y w .ilidewa]hj yu& .+Sonll/ p -'.. -Non.. <:.Little actny 1 + ComiJJsJi!d an aset. to enh3ncil.lltt.he 3J:ld cap.ital .C:On!;idered:a lia,hiiity._ Wiih regarcho .socicil capitai aspects :oi:the:n:eigbborh()9d.'

PAGE 134

.... .... w 5. L 19:.( Stlinlliaiy ofNeighbothood .. . . Nmp 5 6 Lo:w SES 7 LOw Dderly : ... 8 IJigh Eiderir. I mr .-tlni..;..r +:w&. li.clewallti ... comin:u.ow :..Poarly maintained sidewalks to ;.steep; comiriuity +Wen-" +Low .eel iimi. f&rfivm .. +Some tra1&c c:aln'ling 8ncl cl:om.ic cumeu'6r . _.: .. tidic obstructiOnS limi .. .. r :m.an;y .i\esihe:ti.Cs t Deiha'tions +vexY.litae Jiror : cram :a; +WeRpark m.mitiined. y .. . ,.Alot 'of ijttie . shade; poariy. mailltained y..lrds, of gr.afi"Jti . Ji-i6!Jiiar. orcm.Sir +ManY puh1i.: ftCJI!&iian .opporluni .... facilities fr.lany faci&.iiM' .. . reclll!llliion . .. ctrlu.ni&i. -Few public: rec:reati:>IW + ;iiuh1iC recmah opporbiiie. Few.faeilities Sociai 1 S .ml)mty I Vitilijy Capital +Few in.ciViliDe. J tem.i)lialDy : n:u::ivilities -Few courtesies . ;I mal1lf . mciVilil:ies .:aiul sp_ o'f . +l'tfa#y -sonie tltiP. +Mecl +Me4 pebpie; mw:Ji' +Some. peOple +l'tlaDy pc1iOpJe. :i-SOJn!! peoPle; some aA:iMty. Assess. &. +!a:syto Walk safety (c:rime);bw aestl:Etic:s; -:nat easy to Waik. _Low aesthet!cs . .. -noteasyio walk +ffid..# (trafli.c & : aeshd.cB EUytoMdk + Gomidemd \ a set. wiilf:regard u. enluliicn)g th3 :W31k8bility aiul .of tm !Eigbbo:diriod. Considef.e. d a liability; oftlie ncigll.horh,ood.

PAGE 135

. Observations were limited to thos e that could be made within specific neighborhood boWidaries. Thus, they do not accoWit for the presence of a variable in an adjacent neighborhood that may still be of close proximity to some residents. Despite these limitations, important distinctions could be made among the study neighborhoods whose relationship with the outcome variables of active lifestyles and social isolation will be further investigated in parts 3 and 4 of this chapter. The next section will characterize study participants. 114

PAGE 136

Section 2: Participants Figure 5.2.1. Conceptual Model: Participant Characteristics. .. P.espite Plt}'!licafepyironment.to quality ()flife and . . . ::_ --:. : : .. -.-: . ,_,__ ,:,.:' :, ...... : . : -: ..

PAGE 137

of self-efficacy for transport (i.e., walking, driving, or taking public transportation to accomplish errands), personal safety, neighborhood services, and neighborhood cohesion, collected on the main study participants. Individual outcome data were measured using self-reported weekly physical activity and self-reported loneliness and social support. Pilot Study To assure feasibility of administering the proposed surveys, six women and four men, all white, ages ranging from 68 to 88 (mean age= 78), completed the survey. Half had an average household income of $30,000 or more, and all had completed high school, with 40% completing a 2-year college program or technical school and 30% completing a bachelors or graduate degree. Each "tester" was asked to complete the two-part survey on their own, and track their time for each part. The PI then met individually with the tester, and reviewed all the survey questions, asking open-ended questions, particularly with regard to any missed answers. Tester feedback focused particularly on survey format, ease of completion, time burden, clarity of instructions, understa11ding of questions and scales, and appropriateness of content. Most testers found the survey to be straightforward and said it was easy to complete. None thought it was overly burdensome or time consuming. Completion times for part 1 ranged from 5 to 30 minutes and for part 2 ranged from 9 to 30 minutes, with total average completion time for both sections combined equal to 116

PAGE 138

approximately 43 minutes (median completion time, 40.5 minutes). All but one tester were able to complete most sections of the survey without any missed answers. The one section that consistently yielded missed answers for many of the testers was the CHAMPS, which requires the respondent to fill in both frequency of activity performance (i.e., the number of times per week they do the activity) and duration (i.e., the total amount oftime per week they do it). Many respondents skipped the frequency-section, answering only the duration section (which asked them to check a box corresponding to the total amount of time per week). In addition, when questioned further, some respondents who completed the duration section said their answer reflected how long they did the activity per time, as opposed to calculating the total amount of time spent per week. Based on this feedback, some format changes were made to the CHAMPS. The frequency question was boxed and highlighted for each activity. Arrows pointed next to the duration question, which is slightly below the frequency question. Finally, the duration question now asks "each time you do this activity, how long do you usually spend?" with choices offered in 1 hour increments. This will require the investigator to calculate the weekly duration instead of the respondents having to do the niath. In addition, three items were added to the CHAMPS questionnaire to -...... ascertain time spent on instrumental activities in the domains of: 1) meal preparation; 2) shopping for groceries and other essentials; 3) banking and managing 117

PAGE 139

of finances. These additional questions will not be included in calculating energy expenditure, given their low MET value, but will be used in an analysis of overall activity mix in the leisure, work, and activities of daily living domains, and its potential relationship with active lifestyles and isolation. Finally, the CHAMPS does not ask seniors about the time they spend watching television, which could be significant in some cases, and has implications for amount of time spent in sedentary activity. I therefore added it as an item, but to avoid making the survey even longer, I also deleted the seldom selected activity: "shoot pool or billiards". If someone does indeed spend weekly time shooting pool they have the option to write it into the "Other Activity'' item. Since the "shoot pool" item is not includedin the overall scoring of the CHAMPS it made sense to eliminate it. One suggestion from two respondents was to add "marriage" as one of the possible reasons for moving to a neighborhood. Another: respondent had a lot of difficulty understanding the ''strongly disagree" to "strongly agree" anchors, preferring more of a yes or no, or true or false choice. Since most others indicated no problem with the anchors the :PI did not change them, but did expand the instructions and provide examples to :fuit:her clarify those sections of the survey. 118

PAGE 140

Main Study Participation Rate Participation rate in the main study varied along neighborhood liries, ranging from 47% to 12% with an overall rate of23%. Recruitment targets of25 respondents per neighborhood were met or exceeded in five ofthe eight neighborhoods. For the harder-to-reach neighborhoods, repeat mailings and follow-up calls, as well as three additional recruitment mailings were sent to residents, but goals fell short with N=22, 17, and 16. Recruitment was cut-off after mailing lists were exhausted for the three harder-to-reach neighborhoods, resulting in a total sample of 190 returned surveys. Table 5.2.1. Participant Recruitment Results, by Neighborhood Neighborhood Total lneligible Not Interested Unable Surveys Surveys Partie. invited Post Phone to mailed returned Rate card reach Neiehl 60 2 18 6 6 28 27 .47 Neieh1 122 3 44 10 35 30 26 .22 Neigh3 143 4 24 34 56 25 16 .12 Neieb4 79 4 22 9 14 30 25 .33 NeiehS 60 1 13 5 16 25 25 .42 Neigh6 130 10 37 27 31 25 22 .18 Neieh7 130 7 40 33 31 19 17 .14 Neieh8 151 1 49 16 52 33 32 .21 "TOT 875 32 247 140 241 215 190 .23 119

PAGE 141

Participant Characteristics Sociodemographic variables Participant characteristics were 43% male, 15% non-white, 44% with income levels below $30,000, and 8% with less than a high school education. The mean age of respondents was 74 years (ranging from 65 to 99 years) and the average amount of time residing at their current address was 21 yeats (ranging from 1 to 64 years). Neighborhoods varied significantly on age of respondents (a large difference), (F(7,182)=3.161,p<.Ol); proportion ofrespondents who lived alone {i(7)=15;314, p<.05); proportion of respondents who reported a race/ethnicity other than white (i(7)=93.185,p<.01); proportion ofrespondents who reported their average annual household income was less than $30K {i(7)=25.240, p<.01); and years of residence at the current address {i(7)=48.98, <.01), (also a large difference). 120

PAGE 142

Table 5.2.2. Sociodemographic Characteristics of participants by Neighborhood 1 Neigh Neigh Neigh Neigh Neigh Neigh Neigh Neigh Total 1 l 3 4 5 6 7 8 N=l90 N=27 N=l6 N=16 N=l5 N=l5 N=ll N=17 N=32 (unless specifiec!}_ Sex (%Male) 48.2 46.2 43.8 52.0 44.0 31.8 29.4 43.8 43.0 Race/Ethnic (% Non-White) 0.0 12.0 "93.8 4.0 8.0 22.7 5.9 3.1 15.0** Income (N=I86) -(%<30K) 26.9 50.0 43.8 20.0 39,0 77.3 70.6 35.5 44.0** Education (%
PAGE 143

Objective Health Variables Summary statistics for health indicators that were _obtained from participant medical records are presented in Table 5.2.3, below. Complete data on all the indicators of interest were available for 87% participants. However, only 57% of participants had a blood glucose test during the year prior or six months after their enrollment in the study, thus interpretation of those data were limited. A significant overall between-neighborhood difference was found for systolic blood pressure only (F(7,182)=2.575, p<.05). Review of neighborhood confidence intervals confmned Neighborhood 3 respondents had higher blood pressure, on average, than Neighborhood 5, however, this difference was not statistically significant (p>.05). Neighborhood patterns of health metrics will be explored further in section 5 of this chapter, using a combined sample of participants and non-participant data. 122

PAGE 144

Table 5.2.3. Participant Cardiac Risk Metrics & Chronic Disease Score, by Neighborhood 1 METRIC Neigh Neigh Neigh Neigh Neigh Neigh Neigh Neigh TOTAL 1 2 3 4 5 6 7 8 BMI: N 25 25 16 17 25 19 15 32 174 Mean 28.17 28.19 28.72 26.09 25.26 27.53 28.42 28.43 27.62 SD (3.69) _(_6.28) (5.91) (4.10) (3.92) (4.99) (5.67) (4.63) (4.95) BP: N 27 26 16 25 25 22 17 32 190 Mean 134/76 133/75 138174 133173 118/70 135177 131174 125172 130/74 (SD sys, (18.50, (20.01, (22.28, (22.03, (17.10, (19.25, (19.44, (17.12, (19.88, SD dia) 11.03) 11.57) 9.54) 1 0.13) 14.21) 11.05) 10.43) 9.43) 11.04) HDL: N 23 23 12 21 21 17 15 29 161 Mean 56.83 61.30 64.08 54.71 59.67 54.76 58.73 55.38 57.80 SD (13:12) (22.62) (14.63) (9.95) (21.42) (11.82) (17.43) (15.35) (16.40) TRIGLYC: N 23 17 10 21 20 18 14 29 151 Mean 135.35 183.59 137.5 153.90 151.3 180.58 200.78 151.38 159.85 SD (71.39) (108.72) (34.38) (70.64) (81.69) (86.37) (114.25) (70.9) (82.93) .PULSERT: N 27 26 16 25 25 22 17 32 190 Mean 73.07 71.77 69.56 80.68 78.44 79.59 73.65 79.56 76.21 .. SD (11.18) (12.20) (10.22) (17.07) (27.61) (11.74) (13.19) (13.02) (15.87) -BGLU: N 18 11 9 15 9 15 11 20 108 Mean 106 105.91 98.56 98.53 98.22 108.2 112.09 114.40 106.17 SD (26.78) (22.41) (11.32) (11.2) (9.88) (21.99) (13.27) (38.18) (23.83) CVD N 27 26 16 25 25 22 17 32 190 Mean 1.22 1.38 1.44 1.08 1.16 1.18 1.82 1.09 1.26 SD (1.15) (1.36) (.96) (.99) (.89) (1.22) (.95) (1.09) (1.10) TotalCDSJ N 27 26 16 25 25 22 17: 32 190 Mean 2.15 3.15 2.44 2.20 2.36 2.36 3.59 2.28 2.52 SD (1.41) (2.34) (1.59) (1.47) (1.73) _(1.99) (1.46) (1.59) (1.76) UnadJusted, umvanate ANOVA, *p<.05, p<.Ol Between Group Differences. 2CVD Score indicates number of chronic conditions that increase risk for heart attack (scored 1-4). 3Total CDS is the total number of co-morbid chronic diseases estimated by pharmacy data in the electronic medical chart. 123

PAGE 145

Confidence interval differences in proportion of non-white respondents existed between Neighborhood 1, which had the lowest proportion of non-white respqndents and Neighborhoods 3 and 6, respectively, which had the highest (p<.01). Between-neighborhood differences in years living at the current address existed across the board with Neighborhoods 2 and 8 having, on average, the shortest-tenured residents (approximately 11 years); significantly less than Neighborhoods 1, 3 and 7 (p<.05). This is not surprising given that Neighborhood 2 also had the highest proportion of rental property observed, and Neighborhood 8 contained a large residential community that was limited to adults over the age of 55, while Neighborhoods 1, 3 and 7 had mostly houses. Neighborhood 7 respondents had, on average, the longest-tenured residents (approximately 42 years); significantly greater than Neighborhoods 2, 4, 5, 6, and 8 (p<.05). Examination of confidence intervals indicated that differences existed between respondents from Neighborhoods 1 arid 6 on living alone and having an average household income of< 30K and post hoc analysis revealed that both these differences were significant (p<.01). Significant respondent income differences also existed between Neighborhoods 4 and 6 (p<.01). Neighborhood 6 had the highest proportion of respondents who both lived alone and had an income lower than $30K. Differences in the confidence intervals for age existed between Neighborhoods 1 and 6, but the post hoc test was not significant (p>.OS). 124

PAGE 146

Because, as previously mentioned, neighborhoods were selected to reflect diversity using Census demographic data, these differences among respondents were expected, although demographics of respondents were not always congruent with their neighborhood's demographics, for some variables. While quantitative comparisons were difficult for some dimensions, due to how the census data were combined into categories and reported as proportions, some qualitative inferences could be made. Specifically, the proportion of non-white participants was somewhat lower than the proportion of non-white residents in most neighborhoods, and for Neighborhood 6 (proportion of non-white residents =76.5%; Proportion of non-white participants= 22.7%) _and Neighborhood 7 (non-white residents= 83.1 %; non-white participants= 5.9%) the differences appeared to be quite large. Similarly, for average household income, participant income levels appeared to be fairly comparable to neighborhood income levels, with the exception ofNeighborhood 5, where average household income was approximately $110,000 for the neighborhood overall, yet nearly 40% of study participants living in Neighborhood 5 had household income levels below $30,000. The average age of residents was not available by neighborhood; instead, proportion of residents ages 65 and older was reported. Comparing these proportions to the average ages of participants, two of the neighborhoods with the fewest elderly residents (Neighborhood 6 with 9.3% elderly; Neighborhood 7 with 6.3% elderly) had participants whose average ages exceeded that of the overall study sample 125

PAGE 147

(average ages were 76.9 and 76.5 years old respectively). Such inconsistencies may be relevant and will be discussed more in sections 3 and 4. Participant Perceptions Community Mobility Participant responses to survey questions with regard to having a physical impairment that limited their ability to perform activities outside of their home, preferences for walking, driving, or using public transportation for transport; and perceived self-efficacy for walking, using public transportation (RTD), or driving to get to places they need to go for errands or appointments, are summarized below (Table 5.2.4). Large and significant differences were detected across neighborhoods for all self-reported mobility preferences and self-efficacy ratings, with functional limitations at (i(7)= 21.325,p<.Ol); preference for walking or using public transportation at (;l(7)=21.325,p<.Ol); frequency of walking to do errands (x.2(7)= 23.114,p<.Ol); limiting their driving (i(7)=20.145,p
PAGE 148

Table 5.2.4. Self-Reported Mobility. Percentages/Means (SD) compared across Neighborhoods (Neigh) 1 Neigh Neigh Neigh Neigh Neigh Neigh Neigh Neigh Total 1 2 -3 4 5 6 7 8 N=27 N=26 N=16 N=25 N=25 N=22 N=17 N=32 N=190 Functional Impairment limits-ability to perform activities outside of the home) % 11.11 26.92 50.0 40.0 16.0 36.36 47.06 25.0 30.0 Preference for walking or using RTD for transport ** % 0.0 23.08 0.0 4.0 4.0 0.0 0.0 6.25 5.26 Limit their own driving ** % 25.93 46.15 56.25 64.0 40.0 72.73 58.82 28.13 46.84 Self-Efficacy: Walking for Transport ** Mean 8.67 7.96 6.5 7.44 9.0 6.91 4.88 9.0 7.78 (SD) (2.69) (3.23) (3.08) (3.33) (2.58) (3.01) (3.81) (2.11) (3.15) Self-Efficacy: Drivin2 for Trans JOrt ** Mean 9.41 7.23 8.13 8.32 8.72 6.55 6.0 9.56 8.16 (SO) (1.65) _(4.12) (2.65) (3.43) (3.19) (4.29) (4.19) (.98) (3.33) Self-Efficacy: Usin2 Public Transportation (RTD ** Mean 7.67 6.81 4.88 5.32 7.80 6.45 4.59 6.97 6.49 (SD) (3.08) (3.36) (3.14) (3.53) (3.33)_ (3.54)_ (3.76) (3.06) (3.45) Weeki' Frequency ofwalkin2 for errands 1 ** Mean3 .13 .78 .11 .59 .07 .46 .259 .321 .35 (SD) (.46) (.96) (.43)(.80) (.35) (.87) (.65) (.64) .71 I .. Unadjusted umvanate statistics, usmg ANOVA, Kruskal Walbs or Pearson Chi-Square, *p<.05, **p<.Ol 2 Walk to do errands is self-reported times/week, frequency score (CHAMPS) 3 Data reported in table were normalized using a square root transformation 127

PAGE 149

Significant differences for functional impairments were observed between Neighborhoods 1 and 3, with 3 having a much higher proportion of respondents reporting they were limited in their functional ability outside of their home (p<.Ol). Neighborhood 2 significantly preferred to walk or use public transportation versus driving as compared with Neighborhoods 1 and 6 (p<.05). Participants in Neighborhood 2 also indicated that they more frequently walked to do errands than those living in Neighborhoods 1 and 5, a significant difference (p<.05). Self-efficacy for walking was lowest in Neighborhood 7, which significantly differed from Neighborhoods 1, 5, and 8 (p<.05). Post hoc comparisons found no significant differences for self-efficacy for driving or using public transportation. Summary. Functional mobility appeared to be high in Neighborhood 2 (high walk) where participants indicated a preference walking. While a high proportion of Neighborhood 3 (high crime) participants reported physical impairments that limited their activity outside of their home, participants residing in Neighborhood 7 Oow elderly) reported the lowest self-efficacy for walking. Neighborhood Environment for Walking Participant responses to survey questions regarding their access to resources (e.g., shopping and public transportation) within their neighborhood, and their perception of their safety from traffic and crime while walking, are summarized 128

PAGE 150

below (Table 5.2.5). Large and significant differences were detected across neighborhoods for access to resources (F(7, 182)=11.944, p<.001); safety from traffic (F(7, 182)=4.768, p<.001); and safety from crime (i(7)=69.368,p<.001). Table 5.2.5. Perceptions of Neighborhood Environment for Walking (NEWS Scale 1 results)1 Neigh Neigh Neigh Neigh Neigh Neigh Neigh Neigh Total df Sig 1 2 3 4 5 6 7 8 N=27 N=26 N=16 N=25 N=25 N=ll N=l7 N=32 N=l90 Access to Resources Mean 2.78 3.27 2.79 3.03 2.83 2.59 2.09 3.00 2.84 7 .ooo Std. Dev .4644 .4997 .5216 .5281 .4226 .3800 .5469 .3744 .5480 Safety from Traffic Mean 3.02 2.90 2.56 2.83 3.07 2.45 2.46 2.94 2.82 7 .000 ... Std. Dev .4817 .4213 .5553 .5851 .5354 .6530 .6672 .4601 .5743 from Crime Mean 3.35 3.02 2.54 3.42 3.53 2.66 2.52 3.40 3.12 7 .000 ... Std. Dev .5007 .4952 .5725 .3604 .4194 .6009 .6741 .3409 .6098 umvanate ANOVA or Kruskal Walhs, *p<.OS, *p<.Ol between-group differences. Access to Resources between specific neighborhoods was quite variable. Neighborhood 2 had the highest perceived access to resources-significantly higher than Neighborhoods 5, 6, and 7 (p<.05). Neighborhood 8 also had relatively high perceived access, compared with Neighborhoods 6 and 7 (p<.05). Neighborhood 7 had the lowest perceived access and in addition to being lower than Neighborhoods 2 and 8, it was also significantly lower than Neighborhoods 1, 3, 4 and 5 (p<.05). Between-neighborhood post hoc comparisons for perceptions of pedestrian safety from traffic found that Neighborhood 6 had significantly worse perceptions of safety from traffic than Neighborhoods 1 and 5 (p<.05). 129

PAGE 151

Neighborhood 5 had the highest mean perception of being safe from crime and was statistically higher than Neighborhoods 2, 3, 6 and 7 (p<.Ol). Neighborhood 7 had the lowest perceptions of being safe from crime. Summary. Perceptions of neighborhood environment for walking were highest for Neighborhoods 2 (high walkability), 8 (high elderly), and 5 (high SES) and lowest for Neighborhoods 6 Oow income) and 7 Oow elderly). Neighborhood Social Cohesion There were large and significant between-neighborhood differences for mean perceptions of Social Cohesion (i(7)=56.625, p<.OOI). Table 5.2.6. Self-Reported Perceptions of Social Cohesion (Cohesion Scale results)1 Neigh Neigh Neigh Neigh Neigh Neigh Neigh Neigh Total 1 2. 3 4 5 6 7 8 N=27 N=l6 N=16 N=l5 N=l5 N=ll N=17 N=32 N=190 COHESION Mean 3.96 3.31 1.84 3.93 4.04 3.37 3.05 4.08 3.65 Std. Dev .6973 .5885 .4745 .8080 .7234 .7863 .9125 .4948 .8055 Unadjusted Kruskal Wallis; *p<.05, *p<.Ol Between Group Differences hoc analyses revealed lowest reported Social Cohesion for df Sig_ 7 .ooo Neighborhoods 3 and ?(high crime and low proportion of elderly residents, respectively), and significantly lower than Neighborhoods 1, 5 and 8 (p<.05). The highest reported Social Cohesion was found in Neighborhoods 5 and 8 (high income and high proportion of elderly residents, respectively). 130

PAGE 152

Summary. High social cohesion was perceived in Neighborhoods 5 (high SES) and 8 (high elderly)--low cohesion in Neighborhoods 3 (high crime) and 7 (low elderly). Discussion This section presented detailed data on study participation rate, participant socio-demographic characteristics, and responses to mailed surveys with regard to their perception of their own ability to be mobile within their neighborhood and the neighborhood environment for walking. Perceived social cohesion was also reported here. Analyses revealed large magnitudes ofbetween-neighborhood differences for nearly all dimensions -measured, which were also statistically significant. Beginning with participation rates, the neighborhoods that required the least amount of effort to recruit the targeted number (Neighborhoods 1, 4, and 5) had a higher proportion of non-hispanic whites in both the neighborhood overall as well as among the participants and a higher average household income "for the neighborhoods overall, although not necessarily for the study participants. A much greater effort was required to recruit participants in Neighborhoods 3, 6 and 7 which had a much higher proportion of minorities among neighborhood residents, although those who participated from Neighborhood? were mostly white, and also tended to have lower average household incomes, for both the neighborhood and participants. The implications are two-fold. Our recruitment method, which was largely accomplished 131

PAGE 153

by mail, with telephone follow-up was efficient in recruiting a white sample with greater resources. Health may also have been an issue with a greater proportion of participants from the more challenging to recruit neighborhoods reporting that they had a health limitation that made it difficult for them to perform activities outside of their home. Specific health indicators, including number ofco-morbidities for both non-participants as well as participants will be discussed in detail in Section 5. Participant perceptions of their own mobility also varied greatly by neighborhood. Predictably, participants in Neighborhood 2, an urban neighborhood that was selected for its walkability features, including gridded street design, short blocks, and mixed land-use, indicated that they walked or used public transportation at a much higher rate, and also walked to do errands more frequently than participants in the other neighborhoods. Also predictably, participants in Neighborhood 1, a suburban neighborhood that was selected for its winding, non continuous streets and single occupancy houses, reported that they mostly drove for transportation and limited their driving at a much lower rate than participants in most of the other neighborhoods. Neighborhood 1 participants also seldom walked to do errands. Participant perceptions of their neighborhood environment for walking varied in a similar way, with participants in the highly walkable.neighborhood (Neighborhood 2) having the highest perceived access to resources, although neighborhoods with lowest perceived access seemed to be those with lower income 132

PAGE 154

and higher minority populations (Neighborhoods 6 and 7). Perceptions of safety from traffic and crime seemed to vary similarly, with the mostly white, higher income neighborhoods that also reported higher perceptions of safety. Neighborhood 3, which was the selected for having a high violent crime rate reflected this same pattern (i.e., participants reported low perceived safety from crime), although Neighborhood 7 was even lower. Social Cohesion, defined as reciprocity, trust, shared behavioral norms, and common values among neighbors (Sampson et al., 1997) was one of the most interesting variables in this study, and wiil be looked at in more depth in Section 4 of this chapter. Perceived social cohesion was lowest for participants from Neighborhoods 3 and 7. Possible reasons for these poor perceptions include 1) participants in these two neighborhoods reported more physical impairments, so may have been less likely to interact with neighbors; 2) participants in Neighborhood 7 were mostly non-Hispanic whites who had lived in their homes for decades, the neighborhood's demographics were now largely Asian and Hispanic; thus language and cultural barriers may have existed; and 3) Neighborhoods 3 and 7 both had lower proportions of elderly residents. On the other hand, Neighborhood 8, selected for its large proportion of elderly residents, scored very high on perceived social cohesion and relatively high on access to resources and safety from traffic and crime, despite its windy, non continuous streets and lack of mixed land-use. Because most of the study participants 133

PAGE 155

lived in an active living community for adults over the age of 55, many amenities, such as grocery store delivery, free shopping shuttles, centralized community center, and visible security services, were available. In addition, a large cohort of seniors may have provided built in social support that was less easily found in neighborhoods that were populated with young families. As mentioned previously, trust and reciprocity from neighbors has been shown to be. a positive predictor of social cohesion.(Macinko & Starfield, 2001; Sampson et al., 1997) Having a sense of belonging, as well as being able to adapt and participate in life within a defined community, may be more closely associated with having shared demographics than a walkable physical environment, although Neighborhood 8 also had many mobility enhancing aspects such as well-maintained sidewalks, shade, and excellent asthetics (Patrick & Wickizer, 1995; Seeman, 1996). Thus, social cohesion may interact with both environmental and behavioral variables, serving as a marker or proxy for communitybelongingness as well as activity engagement. Its role will be explored further in Sections 3 and 4. 134

PAGE 156

Section 3: The Built Environment and Activity Engagement in Seniors Neighborhood (Level 2) Vital Statistics: Socio-deniographics, Crime Rates Structural Fact:Or5: Walking Paths, Traffic, Land Use, Aesthetics Social Fad;ors: Social Capital, Safety, Vitality Weekly Activity Frequency Weekly Energy Expenditure Figure 5.3.1. Conceptual Model: Neighborhood on PA . ... : and he,alth aciults may illclude. 1Q96) .. earlier,,. ::=; ii;,. =<: ::::.. . :;:.: := . . . 'usiorl"ofbpth i:lldi.Viawu}chai8cteristics' . !Ua-1 of

PAGE 157

outcomes, clustered within neighborhoods, are first presented, followed by two-level hierarchical regression models that were used to analyze direct effects of neighborhood variables on weekly activity engagement. Individual perceptions including perceived access to resources, perceived self-efficacy for walking, driving or using public transportation, perceived safety from crime, and perceived social cohesion were entered into the models as possible mediators. In addition, sex x neighborhood variable was added as an interaction term in the models, to determine if selected neighborhood characteristics affected activity levels differentially for males and females. Individual Level Outcomes Active Lifestyles: Weekly Activity Frequency Large and significant differences were observed between neighborhoods for self-reported frequency of walking for errands {i(7)=23.114,p<.Ol), frequency of home-based physical activity {i(7)=14.686,p<.04), frequency of community-based activities (F(7,182)= 2.719, p<.OS) as well as frequency of sedentary activities (F(7,182)= 2.14l,p<.05), but not for frequency of total physical activity or frequency of moderate intensity physical activity. 136

PAGE 158

Table 5.3.1. Frequency ofWeekly Activities (CHAMPS Results)1 Neigh Neigh Neigh Neigh Neigh Neigh Neigh Total df Sig 1 l 3 4 5 6 7 8 N=l7 N=26 N=16 N=25 N=25 N=22 N=17 N=32 IN=t9o ,requency of Weekly Physical Activity (Total) Mea11 21.13 14.48 13.69 18.36 21.70 15.93 22.09 17.56 18.19 7 .426 Std. Dev. 16.41 10.01 11.21 9.43 16.03 18.76 20.37 11.94 14.50 !Frequency of Weekly Physical Activity (Moderate Level) Mean 10.52 4.58 4.81 6.46 9.76 8.23 9.85 7.98 7.84 7 .515 Std. Dev. 9.63 4.73 5.64 7.14 10.14 17.20 20.29 8.13 10.99 Frequency of Walking for Errands Mean .13 .78 .11 .59 .07 .46 .26 .32 .35 7 .002 Std. De' .46 .96 .43 .80 .35 .87 .60 .64 .71 ** of Home-Based Physical Activity Mean 7.70 4.35 4.50 5.44 6.76 .4.27 7.41 3.59 5.43 7 .04 Std. Dev. 5.82 2.99 3.48 4.9J 5.80 3.84 6.50 3.09 4.82 ,requency of Weekly Community-Based Activity Mean 13.24 13.94 10.41 15.24 15.46 9.68 14.32 17.45 14.05 7 .01 Std. Dev. 7.14 7.91 7.52 6.83 6.39 6.02 7.13 10.10 7.86 Frequency of Weekly Sedentary Activity Mean 17.06 20.54 14.41 22.31 19.20 17.45 14.41 19.70 18.54 7 .04 Std. Dev. 7.71 10.05 6.59 11.66 8.89 8.33 5.79 8.51 8.99 UnadJusted, umvanate ANOV A or Kruskal Wallis; **p<.Ol Between Group D1fferences 2 Data reported in table were normalized using a square root transformation Post hoc analyses showed participants in Neighborhood"2, selected for high walkability, walked to do errands more than six times as frequently(p<.05) as participants in Neighborhood 5 (high SES). Participants in Neighborhood 8 (high elderly) had the greatest frequency of engagement in community-based activities, and were significantly higher than Neighborhood 6, which had the lowest frequency (p<.05). Neighborhood 1, selected for being least walkable, had the greatest 137

PAGE 159

frequency of home-based PA, however post hoc analysis found that that the difference only approached significance compared to the Neighborhood with the lowest mean score for home-based PA, Neighborhood 8 (p=.054). Post hoc analysis revealed no significant differences between individual neighborhoods for frequency of engagement in sedentary activity. Active Lifestyles: Weekly Energy Expenditure Differences in participant mean P A as measured by total and moderate level weekly calorie expenditure and kcallkg/hour (p<.05), were significant. Table 5.3.2. Weekly Self-Reported Physical Activity (CHAMPS results)1 Neigh Neigh Neigh Neigh Neigh Neigh Neigh Neigh Total df Sig 1 2 3 4 s 6 7 8 N=27 N=26 N=16 N=2S N=2S N=12 N=17 N=32 N=190 Calories Expended in TOTAL Physical Activity Mean B036.76 3922.47 B405.04 14499.50 2894.12 3541.71 7 .01<.l Std. Dev. 1909.45 3581.44 2977.03 2232.95 2905.79 Calories Expended in MODERATE-LEVEL Physical Activity Mean 1266.10 1863.61 1651.51 1443.03 7 .004 Std. Dev. 1249.34 2659.81 1696.8(] 1975.94 ** TOTAL Physical Activity Mean 72.97 39.09 42.99 43.61 261.28 1.41 47.38 51.75 50.89 7 .034 Std. Dev. 45.16 25.99 35.49 36.05 44.82 35.53 40.78 34.94 38.71 ikcallkglbrMODERATE-LEVEL Physical Activity Mean 46.58 16.43 19.46 20.81 33.72 21.20 18.23 28.70 26.60 7 .007 Std. Dev. 37.10 16.65 27.44 27.17 35.80 27.23 26.79 27.99 30.18 ** 11 UnadJusted, umvanate ANOVA or Kruskal Wallis, *p<.05, *p<.01 Between Group Differences 138

PAGE 160

Post hoc analyses found significant differences for only moderate level P A, with Neighborhood 1 significantly exceeding Neighborhood 2 (p<.OS). No significant differences were found for Total P A for the post hoc analyses. Summary. Between-neighborhood differences in type of activity and level of activity intensity were found. Greater frequency of walking for erran.ds occurred in neighborhood 2, selected for its typical urban design and its high walkability, and greater frequency of home-based PA occurred in neighborhood 1, selected for its typical suburban design and its low walkability. Contrary to expectations, this latter neighborhood's residents also expended the greatest amount of energy in moderate-intensity P A. Greater frequency of communitybased activity occurred in neighborhood 8, selected for having the highest proportion of elderly residents. To better understand specific neighborhood variables_that predict these outcomes, the hierarchical modeling results will be presented next. Results of Multi-Level Hierarchical Modeling Research Question 1: Neighborhoods and Active Lifestyles: Does Context Matter? Hypothesized relationshipsbetween physical activity and neighborhood variables were tested using a 2-levet hierarchical regression model. For all analyses, variables related to sidewalk functionality (i.e., "Walking"), safety from traffic and crime (i.e., "Safety"), aesthetics, destinations, vitality, and social capital variables 139

PAGE 161

were tested one at a time, in separate models, adjusting for individual level covariates. Results are reported in tables 5.3.3-5.3.7 below. T bl 5 3 3 Multil l Anal 0 t a e . eve ys1s-ucome: F requencyo f W alkin fi Errands 1 tg or Variable Parameter SE p Estimate MEASURED DATA WALK FUNCTIONALITY Intercept 3.1823 1.5518 .0794 Continuity .2486 .2869 .3874 Intercept 2.7930 1.8199 .1687 Path width .4501 .5603 .4229 Intercept 42803 2.1225 .0836 Path maintenance -.2503 .6808 .7136 SAFETY Intercept 3.4179 1.5390 .0618 Traffic buffer .1646 .3111 .5974 Intercept 12782 1.5604 .4397 Curb cuts 1.0104 .3272 .0023 Intercept 5.2270 1.4957 .0101 Peel. signals -2.7296 .7715 .0005 Intercept 2.5403 1.3838 .1090 Walking: Cross-walks 2.8559 .7032 <.0001 AESTHETICS Intercept 4.4764 1.7403 .0369 Litter -.3228 .3986 .4191 Intercept 5.6132 1.6986 .0130 Graffiti2 -.7451 .3342 .0271 Intercept 3.8029 1.7363 .0647 Yard maintenance -.03844 .3951 .9227 DESTINATIONS Intercept 3.2108 1.4518 .. 0626 Proportion non-residential 2.7652 1.7263 .1110 Intercept 3.1167 1.4203 .0706 Retail .04784 .01883 .0440 Intercept 3.1852 1.4616 .0721 Services .05774 .03866 .1859 Intercept 3.9167 1.4558 .0360 Opportunities to Exercise -.05097 .04758 .3253 Intercept 3.3043 1.4837 .0612 Facilities/Courtesies .3841 .3751 .3072 140

PAGE 162

Table 5.3.3 (Cont.). Multilevel Analysis-Outcome: Frequency of Walking for Errands 1 Variable Parameter SE Estimate MEASURED DATA VITALITY Intercept 4.3863 1.9015 People around -.4371 .7834 Intercept 3.5582 1.6706 Activities .1093 .6053 SOCIAL CAPITAL Intercept 3.1981 1.4393 Nuisance Properties .05197 .02733 Intercept 3.5749 1.4373 Inc:ivilities3 .3463 .2849 Intercept 3.0087 1.5749 Window Bars .4164 .4057 Intercept 4.1334 1.4938 Neighborhood Watch -1.8334 1.5759 REPORTED DATA Intercept 2.8645 1.4635 Population Density .00016 .00007 Intercept 4.4639 1.5282 Average Household Income -.00002 .000008 Intercept 3.9301 1.5006 % Residents 65 years + -.01377 .02693 Intercept 3.5017 1.4947 Violent Crime Rate .03413 .05852 Intercept 3.8409 1.4668 %Non-White -.00315 .. 00654 I, Covanates. adjusted for age, sex, tncome level, and chrontc dtsease score. 2 A higher rating indicates a neighborhood that is free of litter and graffiti. 3 A higher rating indicates a greater number of incivilities observed. 141 p .0544 .5776 .0707 .8569 .0680 .1060 .0418 .2258 .0977 .3062 .0278 .2463 .0981 .0768 .0207 .0739 .0396 .6274 .0576 .5810 .0397 .6465

PAGE 163

T b1 5 3 4 M 1fl 1 An 1 0 t a e .. u 1 eve a ys1s-u come: F requencyo fH ornease Variable Parameter SE p Estimate MEASURED DATA WALKING FUNCTIONALITY Intercept 13.2098 5.1591 .0375 conti.nuity .3350 .9414 .7224 Intercept 16.8842 5.9436 .0250 Path width -1.5036 1.7573 .3880 Intercept 19.0014 6.5189 .0225 Path Maintenance -2.2401 1.9103 .2426 SAFETY Intercept 13.4063 5.0982 .0339 Traffic buffer .2751 .9856 .7805 Intercept 19.2364 5.2942 .0084 Curb cuts -2.7061 1.1547 .0202 Intercept 9.5425 5.4843 .1254 Ped. signals 5.7744 3.7363 .1241 Intercept 14.8406 4.8601 .0185 Cross-walks -4.5826 3.9565 .2484 AESTHETICS Intercept 13.9454 5.7894 .0469 Litter2 -.02392 1.3150 .9855 Intercept 14.1765 6.1139 .0535 Gramte -.1075 1.4031 .9390 Intercept 15.0652 5.6848 .0329 Yard Maintenance -.04792 1.2254 .6963 DESTINATIONS Intercept 14.8147 4.8686 .0188 Proportion -6.5470 6.0601 .2815 non-residential Intercept 14.0978 4.8478 .0270 Retail -.02775 .08596 .7578 Intercept 14.3459 4.9171 .0267 Services -.06229 .1422 .6766 Intercept 12.8046 4.8179 .0376 Opportunities to Exercise .1905 .1392 .2201 Intercept 17.0106 4.7084 .0086 Facilities/Courtesies -2.4613 .6672 .0003 142

PAGE 164

Table 53 4 (Cont.). Multilevel Analys1s-Outcome: Frequency ofHome-Based PA1 Variable Parameter SE I P Estimate VITALITY Intercept People around Intercept Activities SOCIAL CAPITAL Intercept Nuisance Properties Intercept Incivilities3 Intercept Window Bars Intercept Neighborhood Watch Intercept Population Density Intercept Average Household Income Intercept % Residents 65 years + Intercept Violent-Crime Rate Intercept %Non-White MEASURED DATA 9.4471 6.0525 2.7389 2.3295 12.6991 5.4797 .8330 L8701 14.1729 4.8600 -.0421 .1085 13.9138 4.8113 -.1119 .9906 15.4301 5.2313 -.9982 1.3262 12.4986 4.9592 5.0676 5.0456 REPORTED DATA 17.7334 4.9584 -.00017 .0003 12.6303 5.1864 .00002 .00003 15.2660 4.9310 -.0841 .0766 14.4520 4.9520 -.0659 .1841 14.2046 4.8632 -.0050 .0204 1Covanates. adjusted for age, sex, mcome level, and chronic d1sease score. 2 A higher rating indicates a neighborhood that is free of litter and graffiti. 3 A higher rating indicates a greater number of incivilities observed. 143 .1625 .2413 .0536 .6566 .0268 .7112 .0233 .9102 .0214 .4527 .0398 .3166 .0249 .5792 .0508 .5088 .0212 .3143 .0267 .7325 .0266 .8142

PAGE 165

Table 5.3.5. Multilevel Analysis-Outcome: Frequency of Community-Based Activity1 Variable Parameter SE p Estimate MEASURED DATA WALKING FUNCTIONALITY Intercept 26.8367 8.2771 .0142 continuity -1.5958 1.3848 .2508 Intercept 17.8739 9.6739 .1071 Path width 2.8910 2.7561 .2957 Intercept 8.6642 9.0311 .3704 Path Maintenance 6.2591 1.9616 .0017 SAFETY Intercept 23.7490 8.3710 .1071 Traffic buffer .00834 1.6064 .9959 Intercept 20.7950 9.5670 .0663 Curbeots 1.4194 2.5851 .5837 Intercept 23.1101 9.4657 .0047 Ped. signals .8907 7.1236 .9006 Intercept 23.8737 8.0492 .0209 Cross-walks -.4916 7.0589 .9446 AESTHETICS Intercept 11.5985 8.4171 .2106 Litte.-2 4.2832 1.2481 .0008 Intercept 18.9941 9.7806 .0933 Graffiti2 1.7597 2.1326 .4104 Intercept 14.7377 8.7837 .1373 Yard Maintenance 3.3466 1.5094 .0279 DESTINATIONS Intercept 25.8501 7.8937 .0136 Proportion -14.8444 8.8085 .0937 non-residential Intercept 23.6204 7.9653 .0251 Retail .01856 .1401 .8989 Intercept 23.9522 8.0815 .0252 Services -0.0219 .2326 .9173 Intercept 23.3548 7.9871 .0265 Opportunities to Exercise .08555 .2598 .7532 Intercept 21.5410 8.0673 .0320 Facilities/Courtesies 2.1693 1.8060 .2313 144

PAGE 166

Table 5.3.5 (Cont.). Multilevel Analysis-Outcome: Frequency of 1 Community-Based Activity' Variable Parameter SE Estimate MEASURED DATA VITALITY Intercept 16.1658 9.7907 People around 4.7758 3.6651 Intercept 18.8495 8.8245 Activities 3.4121 2.7843 SOCIAL CAPITAL Intercept 24.3805 7.9758 Nuisan(!e Properties -.08729 .1735 Intercept 23.9139 7.8074 Incivilities3 -2.2187 1.3356 Intercept 29.7484 8.1819 Window Bars -3.7478 1.6955 Intercept 23.7952 8.2268 Neighborhood Watch -0.1175 8.9060 REPORTED DATA Intercept 23.4470 8.1777 Population Density .0001 .0005 Intercept 22.8489 8.5646 Average Household Income 00002 .00006 Intercept 20.9708 8.0250 % Residents 65 years + .1931 .1126 Intercept 28.1418 7.7512 Violent Crime Rate -.6046 .1952 Intercept 25.6716 7.8544 %Non-White -.05223 .02813 Covanates. adjusted for age, sex, mcome level, and chrontc drsease score. 2 A higher rating indicates a neighborhood that is free of litter and graffiti. 3 A higher rating indicates a greater number of incivilities observed. 145 p .1427 .1943 .0700 .2221 .0223 .6328 .0182 .0994 .0083 .0284 .0232 .9895 .0285 .8017 .0371 .7397 .0399 .1372 .0110 .0212 .0171 .1128 r

PAGE 167

-Table 5.3.6. Multilevel Analysis-Outcome: Weekly calorie expenditure for total PA 1 Variable Parameter SE p Estimate MEASURED DATA WALKING FUNCTIONALITY Intercept 12538 3063.99 .0046 Continuity -700.65 433.55 .1079 Intercept 11190 3678.11 .0188 Path width -37.3355 1011.74 .9706 Intercept 9951.01 4039.56 .0432 Path Maintenance 498.58 1152.33 .6658 SAFETY Intercept 11531 3157.28 .0082 Traffic buffer -226.14 531.21 .6708 Intercept 15471 3143.22 .0017 Curb cutS -1608.82 564.29 .0049 Intercept 8827.80 3090.21 .0245 Pedestrian signals 4553.67 1492.87 .0026 Intercept 13017 2926.72 .0030 Cross-walks -4249.67 1487.13 .0048 AESTHETICS Intercept 8696.54 3396.29 .0375 utter 1012.15 610.11 .0989 Intercept 8031.11 3421.83 .0513 Graffiti2 1240.37 597.52 .0394 Intercept 9505.80 3447.11 .0282 Yard Maintenance 649.58 642.31 .3133 DESTINATIONS Intercept 16441 3184.16 .0013 Proportion non-residential -3183.48 977.32 .0014 Intercept 12048 2979.61 .0068 Retail -67.9704 36.4968 .1119 Intercept 12240 3021.49 .0067 Services -102.37 64.1887 .1618 Intercept 9834.20 3002.00 .0136 Tennis Courts 12143 4163.38 .0040 Intercept 11799 3087.29 .0065 Facilities/Courtesies -550.11 643.69 .3939 146

PAGE 168

Table 5.3.6 (Cont.). Multilevel Analysis-Outcome: Weekly calorie expenditure for total P A 1 Variable Parameter SE Estimate MEASURED DATA VITALITY Intercept 9066.39 3638.55 People around 1356.91 1271.93 Intercept 10498 3371.91 Activities 427.31 1028.19 SOCIAL CAPITAL Intercept 12284 2985.41 Nuisance Properties -90.1506 45.8583 Intercept 11351 2955.57 Incivilities3 -920.18 425.31 Intercept 13563 3086.36 Window Bars -1248.51 581.56 Intercept 10411 3032.01 Neighborhood Watch 4349.18 2421.99 REPORTED DATA Intercept 12698 3047.35 Population Density -.2254 .1360 Intercept 9528.20 3168.75 Average Household Income .03004 .01555 Intercept 10709 3084.98 % Residents 65 years + 39.3126 44.0511 Intercept 12366 3037.66 Violent Crime Rate -133.55 89.2284 Intercept 11510 3030.98 %Non-White -6.6641 11.2777 Covanates. adJusted for age, sex, mcome level, and chrome dtsease score. 2 A higher rating indicates a neighborhood that is free of litter and graffiti. 3 A higher rating indicates a greater number of incivilities observed. 147 p .0415 .2875 .0170 .6782 .0063 .0969 .0064 .0319 .0032 .0332 .0109 .0743 .0059 .1485 .0238 .1015 .0133 .4065 .0066 .1851 .0090 .5761

PAGE 169

T bl 5 3 7 Multil 1 Anal 0 t a e . eve JYSlS_ ucome: w k1 a1 ee yc oneexpen diture for Mod P A 1 Variable Parameter SE p Estimate MEASURED DATA WALKING FUNCTIONALITY Intercept 9298.73 2346.05 .0054 Continuity -672.09 348.50 .0554 Intercept 7558.94 2873.59 .0339 Path width 151.94 856.29 .8594 Intercept 6981.97 3231.41 .0675 Path Maintenance 383.91 984.48 .6970 SAFETY Intercept 8251.73 2429.61 .0115-Traffic buffer -211.83 450.69 .6389 Intercept 10898 2526.47 .0035 Curb cuts -1233.32 542.95 .0243 Intercept 6045.89 2351.14 .0369 Pedestrian signals 3884.80 1135.83 .0008 Intercept 9227.71 2261.32 .0047 Cross-walks -3473.34 1332.46 .0099 AESTHETICS Intercept 5825.12 2634.06 .0627 Litte? 869.45 513.14 .0920 Intercept 5290.23 2682.47 .0892 Graffiti 1028.59 510.85 .0456 Intercept 6405.22 2663.10 .0471 Yard Maintenance 607.44 532.15 .2552 DESTINATIONS Intercept 9057.63 2215.61 .0046 Proportion non-residential -5948.18 1807.34 .0012 Intercept 8712.00 2271.98 .0086 Retail -62.4471 29.6637 .0799 Intercept 8999.30 2298.58 .0078 Services -101.18 50.5353 .0921 Intercept 6913.94 2296.17 .0196 Tennis Courts 10292 3561.73 .0044 Intercept 8253.38 2665.83 .0101 Facilities/Courtesies -356.40 564.74 .5288 148

PAGE 170

Table 5.3.7 (Cont). Multilevel Analysis-Outcome: Weekly calorie expenditure forModPA1 Variable Parameter SE Estimate MEASURED DATA VITALITY Intercept 6156.21 2867.62 People around 1130.13 1079.09 Intercept 7315.59 2611.17 Activities 394.08 868.98 SOCIAL CAPITAL Intercept 8918.54 2274.41 Nuisance Properties -83.4974 36.9949 Intercept 8133.37 2262.63 Incivilities3 -753.38 363.41 Intercept 9908.63 2382.26 Window Bars -1075.04 488.49 Intercept 7187.83 2307.23 Neighborhood Watch 4254.81 1899.62 REPORTED DATA Intercept 9177.08 2332.30 Population Density -.2057. .1126 Intercept 6625.60 2438.25 Average Household Income .02445 .01360 Intercept 7425.10 2363.65 % Residents 65 years + 38.5666 36.8529 Intercept 8832.55 2335.14 Violent Crime Rate -112.43 75.9050 Intercept 8199.39 2322.89 %Non-White -6.4485 9.4481 Covanates. adjusted for age, sex, mcome level, and chrome d1sease score. 2 A higher rating indicates a neighborhood that is free oflitter and graffiti. 3 A higher rating indicates a greater number of incivilities observed. p .0689 .2964. .0265 .6508 .0078 .0648 .0088 .0396 .0042 .0291 .0170 .0264 .0077 .1175 .0348 .1223 .0200 .3356 .0092 .1891 .0124 .5204 Hypothesis 1.1. Neighborhood Characteristics will be Associated with Greater Frequency of Walking for Errands. Neighborhood characteristics consistent with Denver's urban neighborhood design, specifically those variables that supported safer walking with regard tQ safety from traffic, such as presence of curb cuts and crosswalks, fewer traffic signals, and greater number of retail destinations, showed 149

PAGE 171

modest(r = .23 to .28), significant {p<:01) associations with frequency of walking for errands, in the hypothesized directions. More walking for errands was also associated with worse neighborhood ratings for graffiti, (r = -.21, p=.0035), however, it is likely that this relationship is an artifact of urban versus suburban dwelling as opposed to a direct relationship between graffiti and walking for errands. Reported measures based on. census and crime statistics were not significantly related to walking for errands with the exception of population density which approached significance {p=.07) Hypothesis 1.2. Neighborhood Characteristics Will Be Associated With Greater Frequency of Home-Based P A. Few neighborhood characteristics were related to frequency of performing home-based PA, such as housework and yard work. Lack of curb cuts and lack of public courtesies predicted more home based P A. However, these associations may be an artifact of suburban versus urban dwelling, since home-based PA was strongly associated with gardening {r=.73, p=.OOO) and heavy yard work (r=.645, light house work {r=.72, p=.OOO) and heavy house work (r= .581, p=.OOO). Thus having a house with a yard as opposed to an apartment in the heart of the city may increase frequency of home-based P A, although the reported statistic of population density was not significantly related to this outcome. 150

PAGE 172

Hypothesis 1.3. Neighborhood Characteristics Will Be Associated With Greater Frequency of Communi tv-Based Activity Engagement. Neighborhood characteristics that supported safer walking with regard to personal safety from falling, such as path maintenance; pleasant aesthetics, such as yard maintenance and less litter, also showed modest (r-.22 to .25), significant (p<.Ol) associations with frequency of engaging in community-based activities, in the hypothesized directions. The negative correlation between community-based activities and presence of window security bars on residences (r=-.20, p=.005) may have social implications with regard to territoriality and the need to protect one's home from intruders. This interpretation is further supported, as there was a significant inverse relationship between neighborhood violent crime rates and community-based activity engagement. Thus, crime rates and window bars will be explored further during the discussion of social cohesion, in Section 4. Hypothesis L4. Neighborhood Characteristics Will Be Associated With Greater Weekly Calories Expended in Total and Moderate-Level PA. Neighborhood characteristics significantly associated with more calories expended in total and moderate level physical activities included walking variables such as route continuity, curbcuts, cross walks and traffic signals with pedestrian walk/don't walk signs; land use variables, such as proportion of non-residential land use and recreational variables, specifically tennis courts; and social capital variables such as graffiti, window bars, and neighborhood watch signs. However, many of these 151

PAGE 173

associations were not in the hypothesized direction. Variables that promoted greater walking for errands such as path continuity, curb cuts, cross-walks, and greater proportion of non-residential land-use were negatively associated with weekly calories expended in both total and moderate-level PA, although the magnitude of these effects was modest (r = -.19 to -.23,p<.Ol). Other variables associated with more urban neighborhoods, such as greater presence of graffiti, total incivilities and window bars, were also negatively associated with calories expended in physical activities; while neighborhood watch signs showed significant positive association with calories expended in PA (r =,19,p=.0096). While initially puzzling, these patterns seem to be indicative of differences between urban and suburban environments. This interpretation is strengthened by a stronger positive association between the presence of tennis courts, which were more often observed in more suburban neighborhoods, and moderate-level PA (r =.2788,p<.0001). In addition, since greater frequency of home-based PA was strongly associated with both calories expended in total PA (r =.572,p<.001) and in moderate PA (r=.4l,p<.001), once again having a house and yard may present more opportunities for older adults to engage in physical activity. Presence of traffic lights with pedestrian walk/don't walk signals also showed a modest, positive association with calories expended (r=.20,p<.Ol), however, such signals tended to be located at busy intersections in Denver's more suburban neighborhoods to control traffic on multi-lane major roads. These "crossing aids" seemed to favor drivers versus since the timing of 152

PAGE 174

walk signals did not provide ample time for the auditors to cross multi-lane streets, let alone older adults. Hypothesis 1: Summary ofResults Hypothesis 1 was partially confirmed, with modest but statistically significant associations found between the outcome variable, Active Lifestyles and some Structural as well as some Social factors at the Neighborhood level. However different neighborhood characteristics supported different forms of activity, and not all of these relationships supported the initial hypotheses. Variables that supported more efficient walking (curbcuts and crosswalks) and places to walk to (density of retail), were associated with more walking for errands. A few variables associated with less urban environments were associated with more frequent engagement in home-based P A, however this may be more tied to having a house and yard than to features that support walking or social capital variables. Safety and aesthetic variables were associated with more frequent community engagement as well as more total and moderate-level P A. Results of Mediation Analyses Research Question 2: What Are the Potential Causal Pathways through which Neighborhood and Individual Factors Influence Outcomes? Developing models that represent how neighborhood factors such as land use, transportation, and senior services may be directly associated with individual outcomes or may operate indirectly, through individual perceptions, to affect 153

PAGE 175

outcomes. Understanding these relationships is an important step to devel.oping relevant health interventions (Diez Roux, 2004). While the data collected in this study are cross-sectional, the use ofmulti-lev:el analysis enables identification of statistical predictors of current behavior. To do this, mediation analyses were conducted, particularly with regard to the possible relationships between those neighborhood characteristics that appeared to be directly related to the outcomes variables and individual perceptions of their neighborhood. Possible mediator variables were selected based upon the literature, previously hypothesized relationships, and whether a logical conceptual rationale was present. Using the methods described at the beginningofChapter 5 (p.70), three additional regression analyses were performed once a direct relationship was identified between the Independent and Dependent variables, to. see if including the proposed mediator in the model eliminated or significantly reduced the strength of the relationship between the independent variable and dependent variables. Hypothesis 2.1. The Relationship Between Neighborhood Characteristics and Higher Frequency of Walking for Errands Will Be Mediated by Perceived Self-efficacy for Walking. Safety from Traffic and Crime. and Access to Resources. Individual perceptions of self-efficacy for walking, pedestrian safety from traffic, pedestrian safety from crime, and access to resources were hypothesized as possible mediators between the neighborhood variables (i.e., curbcuts, cross-walks and density of retail), and the outcome variable (i.e., walking for errands). The individual 154

PAGE 176

perception variables were moderately correlated with the neighborhood variables (r = .29 to .35,p<.0001), but only perceived access to resources was correlated with walking for errands (r=.21, p=.003). When entered into the regression model, perceived access to resources did not significantly mediate the relationships between the neighborhood variables and walking for errands. Hypothesis 2.2. The Relationship Between Neighborhood Characteristics and Higher Frequency of Engagement in Comrnunitv-Based Activities Will be Mediated by Perceived Self-Efficacy for Walking, Safety from Crime and Social Cohesion. Neighborhood variables that appear to have a direct association with weekly frequency of engagement in community-based activity, included sidewalk maintenance, quantity of litter, yard upkeep, and presence of window security bars. These variables may be related to individual perceptions on the degree of neighborhood deterioration, which from a social capital viewpoint may be symbolic of other factors such as crime levels, p edestrian vulnerability, and environmental threats. Thus, mediators tested included perceived self-efficacy for walking, perceived pedestrian safety from crime, and perceived social cohesion." All three of these mediators were correlated with each other (r = .36 to .6l,p<.0001) and were modestly correlated with frequency of engagement in community-based activities (r =-.23 to .28,p<.001). The three potential mediators were also modestly to strongly correlated with the neighborhood variables (r= .19 to .58,p<.Ol). Each mediator was entered separately into the hierarchical regression model as described above. 155

PAGE 177

Self-efficacy for Walking for Errands did not mediate the relationships bf!tween the dependent variable, frequency of community-based activity, and either path maintenance or litter. Based on the regression analyses it appeared.to mediate the dependent variable's relationships with yard maintenance and window bars. However, the Sobel test for yard maintenance only approached significance (p=.059) and was not significant for window bars, indicating that the direct between these neighborhood variables and the outcome Walking for Errands was not significantly mediated by Table 5.3.8. Community-based Activity (Self Efficacy for Walking as Mediator)1 Variables Parameter SE p Estimate First Analysis Intercept 14.7377 8.7837 .1373 Yard Maintenance 3.3466 1.5094 .0279 Second Analysis Intercept 3.3997 9.2391 .7238 Yard Maintenance 2.4259 1.5781 .1261 Self Efficacy for Walking .7258 .2085 .0006 Sobel t= 1.8862 .059 First Analysis Intercept 29.7484 8.1819 .0083 Window Bars -3.7478 1.6955 .0284 Second Analysis Intercept 14.4263 9.0601 .1553 Window Bars -2.9907 1.6918 .0789 Self Efficacy for Watkin .7321 .2071 .0005 Sobel t=-1.3971 .1624 I, Covanates. adjusted for age, sex, mcome level, and chroruc disease score Perceived social cohesion did not mediate the relationship between path maintenance and the dependent variable frequency of community-based activity. However, it mediated the relationships between yard maintenance and the dependent 156

PAGE 178

variable (Table 5.3.9, figure 5.3.2) and between window bars and the dependent variable (Table 5.3.10, figure 5.3.3). Table 5.3.9. Community-based activity and yard maintenance 1 (with social cohesion) Variables Parameter .SE p Estimate First Analysis Intercept 14.7377 8.7837 .1373 Walking: yard maint 3.3466 1.5094 .0279 Second Analysis Intercept 8.4323 9.0613 .3830 Walking: yard maint 2.0273 1.5707 .1985 Perceived Social Cohesion 1.8396 .8026 .0231 Sobel t=l.l452 .025 1Covariates: adjusted for age, sex, income level, and chronic disease score R=.22 DV: IV: Freq of Community-Yard Maintenance Based Activity -.03 (p=) Mediating Variable: ::Y Social Cohesion DV: IV: Freq of Community-Yard Maintenance J3 = 2.03 Based Activity (NS) Figure 5.3.2. Mediation effect: yard maintenance with social cohesion. 157

PAGE 179

Table 5.3.10. Community-based activity and window bars I (with social cohesion)' Variables Parameter SE p Estimate First Analysis Intercept 29.7484 8.1819 .0083 Territor: window bars -3.7478 1.6955 .0284 Second Analysis Intercept 17.5198 9.8289 .1179 Territor: window bars -2.1084 1.8255 .2497 Perceived Social Cohesion 1.8029 .8233 .0299 Sobel t=-2.6493 .008 1, Covanates. adJusted for age, sex, mcome level, and chrome dtsease score R=-.20 DV: IV: Freq of CommunityWindow Bars P=-3.75 Based Activity (p =.03 Mediating Variable: ::Y Social Cohesion DV: IV: Freq of CommunityWindow Bars p = -2.11 Based Activity NS Figure 5.3.3. Mediation effect: window bars with social cohesion. 158

PAGE 180

Perceived social cohesion also partially mediated the relationship between litter and the dependent variable (Table 5.3.11, figure 5.3.4). Table 5.3.11. Community-based activity and litter (with social cohesio Variables Parameter SE p Estimate First Analysis Intercept 11.5985 8.4171 .2106 Aesthetics: litter 4.2832 1.2481 .0008 Second Analysis Intercept 6.9295 8.8226 .4580 Aesthetics: litter 2.9335 1.4815 .0493 Perceived Social Cohesion 1.6295 .8062 .0448 Sobel t=2.4368 .015 l, Covanates. adJusted for age, sex, mcome level, and chrome disease score R=.25 DV: IV: Freq of CommunityLitter 13 =4.28 Based Activity (p <.000 Mediating Variable: :Y S-ICOOnlon "' DV: IV: Freq of CommunityLitter 13 = 2.93 Based Activity (p =.049 Figure 5.3.4. Mediation effect: litter with social cohesion. 159

PAGE 181

Perceived social cohesion mediated the relationship between neighborhood violent crime rate and community-based activity engagement (Table 5.3.12, Figure 5.3.5). Table 5.3.12. Community-based activity and Violent Crime Rate (with social cohesion)1 Variables Parameter SE p Estiinate First Analysis Intercept 28.1418 7.7512 .0110 Reported: Violent Crime -.6046 .1952 .0212 Second Analysis Intercept 17.5919 9.1369 .1025 Reported: Violent Crime -.4102 .2217 .1137 Perceived Social Cohesion 1.7317 .8101 .0340 Sobel t=-2.422 .0154 .. Covanates. adJUSted for age, sex, mcome level, and chrome disease score R=-.23 DV: IV: Freq of CommunityViolent Crime Rate P=-.60 Based Activity =.02 (p ) Mediating Variable: y DV: IV: Freq of CommunityViolent Crime Rate p = -.41 Based Activity -.11 (p= Figure 5.3.5: Mediation effect: violent crime rate with social cohesion. 160

PAGE 182

Perceived pedestrian safety from crime did not mediate the relationships between independent variables path maintenance, litter or violent crime rate, and the dependent variable frequency of coi:nmunity-based activity. It did mediate the relationship between yard maintenance and the dependent variable (Table 5.3.13, Figure 5.3.6). Table 5.3.13. Community-based activity and yard maintenance (with perceived safety) 1 Variables Parameter SE p Estimate Fint Analysis Intercept 14.7377 8.7837 .1373 Walking: yard maint 3.3466 1.5094 .0279 Seeond Analysis Intercept 10.8299 8.9671 .2664 Walking: yard maint 1.9821 1.6663 .2359 Perceived Safety fr-4Jm Crime.-2.0554 1.1805 .0835 Sobel t=2.3706 .0177 1Covariates: adjusted for age, sex, income level, and chronic: disease score 161

PAGE 183

R=.22 DV: IV: Freq of CommunityYard Maintenance P=3.35 Based Activity (p =.03 Mediating Variable: :Y "' DV: IV: Freq of CommunityYard Maintenance P= 1.98 Based Activity (NS) Figure 5.3.6. Mediation effect: yard maintenance with perceived safety. It al&o mediated the relationship between window bars and the dependent variable (Table 5.3.14, Figure 5.3.7), although the magnitude ofthe relationship between perceived safety from crime and community-based activity also decreased, which may imply another factor interacting with both of these variables. Table 5.3.14. Conimunity-based activity and window bars (with percei ved safety) 1 Variables Parameter SE p Estimate First Analysis Intercept 29.7484 8.1819 .0083 Territor: window bars -3.7478 1.6955 .0284 Second Analysis Intercept 20.0111 9.8213 .0810 Territor: window bars -2.3242 1.8136 .2017Perceived Safety from Crime 2.0383 1.1637 .0816 Sobel t=-2.2510 .024 I, Covanates. adjusted for age, sex, mcome level, and chrome disease score 162

PAGE 184

R= -.20 DV: IV: Freq of CommunityWindow Bars Based Activity (p =.03 Mediating Variable: :/ Safe from cnme "" DV: IV: Freq of CommunityWindow Bars Based Activity NS ( ) Figure 5.3.7. Mediation effect: window bars with perceived safety. Hypothesis 2.3. The Relationship Between Neighborhood Characteristics and Higher Energy Expended in Physical Activity will be Mediated by Perceived Self-Efficacy for Transport and Access to Resources. Independent vanables were similar between weekly calories expended in total and moderate-level physical activity, and mediator analyses conducted showed no significant differences in effects, so only total physical activity analyses are presented here. Because of the inverse relationships between neighborhood variables conducive for walking and greater calories expended in P A, self-efficacy for driving was looked at as a possible mediator between the independent variable pedestrian signals and both dependent 163

PAGE 185

variables (energy expended in total and moderate level PA). Self-efficacy for driving did not mediate the relationships. Perceived access to resources also did not mediate the relationship between neighborhood variables and P A. Perceived safety from crime and social cohesion as mediators of P A Since quantity of incivilities and window bars were also inversely associated with energy expended in P A, perceived safety from crime and social cohesion were tested as potential mediators. Social cohesion was not a mediator between either independent variable and P A, but perceived safety from crime was a mediator between both independent variables and the dependent variables of total and moderate PA. The results for total PA are shown below (Tables 5.3.15-16, Figure 5.3.8-9), but the same analyses using moderate PA yielded similar results. Table 5.3.15. Total PA and incivilities (with perceived safety) Mediation Analysis: Weekly Parameter SE p calorie expenditure for total Estimate PA (with Perceived Safety from Crime)Variables First Analysis Intercept 11351 2955.57 .0064 Inc:iviUties -920.18 425.31 .0319 Second Analysis Intercept 8153.24 3354.53 .0454 loc:ivilities -482.00 509.67 .3456 Perceived Safety from Crime 865.74 442.41 .0520 Sobel t=-2.1026 .0276 1Covariates: adjusted for age, sex, income level, and chronic disease score 164

PAGE 186

R= -.22 IV: Incivilities 13 =-920.18 (p =.03 Mediating Variable: ,'?/ Safe from Crime / IV: Incivilities 13 =-482.00 NS DV: Total PA DV: Total PA Figure 5.3.8. Mediation effect: incivilities with perceived safety. Table 5.3.16. Total PA and window bars (with perceived safety) Mediation Analysis: Weekly Parameter SE p calorie expenditure for total Estimate PA (with Perceived Safety from Crime)Variables ''First Analysis Intercept 13563 .36 .0032 Window Bars -1248.51 581.56 .0332 Second Analysis Intercept 9166.70 3769.45 .0453 Window Bars -628.35 709.32 .3769 Perceived Safety from Crime 870.63 446.01 .0526 Sobel t=-2.289 .022 I, Covanates. adJusted for age, sex, mcome level, and chrome disease score 165

PAGE 187

R = -.18 IV: DV: Window Bars ll = -1248.51 Total PA = (p .03) Mediating Variable: :/ IV: DV: Window Bars ll=-628.35 Total PA NS ( ) Figure 5.3.9. Mediation effect: window bars with perceived safety. Gender as a moderator variable Gender differences in perceived environmental correlates of physical activity have been examined in limited studies (Bengoechea, 2005; Dunton, Jamner, & Cooper, 2003; Eyler, 2003). Women in particular were more likely to perceive their neighborhood as Unsafe and less likely to perceive easy access to shopping within walking distance (Bengoechea, 2005). Gender was therefore considered as a possible moderator between independent variables that were associated with perceived access to resources (i.e., density of retail), safety from crime and social cohesion (i.e., window bars, incivilities, litter) and outcome variables (i.e., weekly frequency of 166

PAGE 188

walking for errands, community-based activity, and calories expended in total and moderate P A). No significant associations were found between gender and any of the outcome variables, however, thus gender was not a significant moderator. Other possible moderators of activity, such as older age and health require more complicated analyses, given their continuous nature, and power may be a greater issue, as larger sample sizes (N>200) are needed to have reasonable power to detect moderator effects when one of the variables is continuous (Kenny, 2004). Dichotomizing these variables would require some arbitrary decisions to be made regarding young old and older old, or healthy and unhealthy. Future explorations, however, should consider these and other possible moderators. Hypothesis 2: Summary of Results Hypothesis 2 was partially confirmed, with different neighborhood variables affecting different behaviors, and interactions betWeen individual perceptions and neighborhood factors were a factor for some types of activity. Contrary to expectations, perceptions of the neighborhood safety, resources or self-efficacy of walking did not mediate the relationship between neighborhood variables that promoted efficient walking and individual frequency of walking for errands. However, the relationships between aesthetic and safety-related variables on community-based activity were mediated by perceived social cohesion and 167

PAGE 189

perceived safety from crime. Perceived safety from crime also mediated the inverse relationships of incivilities and window bars on total and moderate P A. Neighborhood Clevel 2) Structural Factors: Sidewalk Functionality, Land Use, Aesthetics Social Factors: Social Capital (incivilities, window bars) Perceptions: Crime Sal'ety ) Social Cohesion Community-Based Activity Energy expended in P A Figure 5.3.1 0. Conceptual Model: Direct and Mediated Effects of Neighborhood Variables on Activity Engagement. Discussion Saelens, et al.(Saelens, Sallis, Black et al., 2003) observed that neighborhoods designed aroWid car use contribute to a more sedentary lifestyle and that residents in neighborhoods with higher residential density, a mix of residential and commercial amenities, and gridlike street patterns recorded more weekly minutes of moderate level activity than those residing in neighborhoods with long, winding streets and few retail establishments. These findings, along with those of 168

PAGE 190

other recent studies that focused primarily on the built environment for walking or bicycling, were partially supported by study for the outcome of frequency of walking for errands. Elderly study participants who lived in an urban neighborhood with greater retail density, and features that promoted efficient walking, such as curb cuts, cross walks, and few signal lights, walked for errands at a significantly higher rate than those who lived in neighborhoods with less continuous routes and few retail Furthermore, these direct associations between the built environment for walking variables and frequency of walking for errands were not mediated by any of the individual perceptual variables tested. Different environmental variables were associated with frequency of engagement in community based activities, which encompassed any activity that was performed outside of the home including social activities such as going to the movies and recreational activities such as biking or golf. Neighborhood variables that were important here were more related to neighborhood upkeep and safety than to walkability or density of resources. Better yard maintenance, less litter, and fewer window bars were associated with higher frequency of community-based activities. The neighborhood whose participants scored highest also had the higllest proportion of elderly residents, with a large active-living senior community that was both well maintained and secure. Direct effects of yard maintenance and window bars were mediated by perceived social cohesion (which also mediated the direct effect of neighborhood violent crime rate) and perceived safety from crime, which 169

PAGE 191

underscores the importance of the perceived social environment to facilitating community-based activity. Thus observable neighborhood characteristics such as yard upkeep and window bars may serve as proxies for perceived social cohesion and safety from crime, which in turn facilitate greater community engagement. Despite the above relationships structural variables that promote walking for errands, and aesthetic and safety variables that were associated with community perceptions and activity-level, weekly energy expenditure for participants departed from these patterns. Calories expended in both total and moderate level PA were related to some of the same variables that were associated with more community-based activity, i.e., few window bars and fewer incivilities. However, greater calories expended in P A were negatively related to variables that facilitated greater walking for errands, contrary to expectations. In addition, variables that promoted higher levels of energy expenditure were typical of a more suburban environment, such as traffic lights, tennis courts, and neighborhood watch signs. When considered independently, these relationships seem spurious; however, when looked at in the context of suburban living, they may imply that having a house, a yard and nearby tennis courts supports engagement in the types of activities that produce greater energy expenditure than walking for errands. The relationships between incivilities and window bars were mediated by perceived safety from crime, which also supports the notion that suburban living may be more desirable to older adults who may feel more vulnerable to crime than younger adults. Lee & King (Lee 170

PAGE 192

. & King, 2003) found in general, activities selected in old age tended to be less physically demanding, thus living in an environment that demands more routine house and yard upkeep may actually promote higher activity levels for some seniors. Whether such environments are also socially desirable or more isolating will be discussed in the next section. 171

PAGE 193

Section 4: Social Capital, Social Cohesion and Social Isolation in Seniors Neighborhood Clevel 2) Vatal statistics: Socio-demographics, Crime Rates Structural Factors: Walking Paths, Traffic, Land Use, Aesthetics Social Factors: Social Capitai,Safety, Vitality Social Isolation Perceived Loneliness Social Support Figure 5.4.1. Conceptual Model: Neighborhood Effects on Social Isolation This sectiqn eompletes the exploration qf research questions. i and 2 by what variables, if any, are with social isolation, :a.tid the through which indiVidllal factors .. ; 'i1ifluerlce the outcome vSriables ofperceived)oneliness and social support. As discussed in Chapter 2 social capital has been defined as resources that accrue to in.diyiduals via social connections, group ievt;:1 norms, cohesion and participation (Coletnall11988) and .abo as the existence ofcuganizations and re$ources within a that can be accessed \ly its mi?ID.Pers .. Putnam s structuial perspective .':

PAGE 194

(Putnam, 1995), which views social capital as features and social structures that facilitate social cohesion and collective action (Bourdieu, 1984; Kawachi & Berkman, 2000; Putnam, 1995), will be emphasized in this section. These features include trust, shared norms and values, and available social organizations. In other studies, measurement of social capital has primarily consisted of aggregate variables that are made up of individual responses to social surveys and more integral variables that involve direct social observation of neighborhoods (e.g., number of establishments that still accept personal checks may be an indicator of trust). For the purpose of this study, the aspects of social capital that will be examined include individual perceptions of safety from crime, social cohesion, social support and loneliness, and observed neighborhood variables with emphasis on those variables that conceptually relate to social capital: density of resources, public courtesies, aesthetics, incivilities and territoriality. Individual Level Outcomes Social Isolation: Perceived Loneliness and Social Support Analyses of Variances found no significant between-neighborhood differences for perceived Loneliness or Social Support. 173

PAGE 195

Table 5.4.1. Self-Reported Lonelinl;)ss and Social Support (LASA and SPS results)l LONBLINESS Nli&bJ. NqM Nli&hJ Nlif:h' Nti&Jl5 Nlilb' Nli&h7 Nli&hJ Tcrtal elf lie N 27 215 115 25 :14 22 17 32 189 1 .08 I Uem 330 4.27 .5.015 3.28 2.46 4.23 4 .59 3.2.5 3.69 Std.DN. 2.771 3.118 3.021 25.68.5 2.484 3 .54.5 3.001 2.771 2.963. SOCIAL S'OP.PORT .a\:ti:ulmunt 1 .156 N 27 26 115 2.5 :15 2:1 17 32 190 Mem 3.41 3.25 3.27 3.51 3.60 3.28 3.18 3.48 3.39 Std.Dev. .5639 .6708 .54315 .51512 .5448 .15422 .5514 .44151 .5724 ldlpd.an 1 .:119 N :17 215 16 25 25 22 17 32 190 Uem 3.15 3.15 3.08 3.2P 3.40 3.15 3.00 330 3.21 Std.Dev. .15134 .5.5215 .3734 .415.5P .4330 .5382 .78.515 .37215 .5249 lbuiUIUUe cd"WoDh 1 .591 N 27 215 16 2.5 :1.5 2:1 17 32 190 Uem 33.5 3.41 3.27 3.30 3.47 3.23 3.2.5 3.2.50 3.32 Std.Dev. .4.560.5 .5145 .4784 .4449 .5368 .43.58 .51560 .4066 .47.51 Bdia1lle Allilnce 1 .08 N 27 215 115 25 25 22 17 32 190 Mem 3.53 3.4P 3.30 3.152 3.75 3.515 3.51 3.15:1 3.515 Std.Dev. .4458 .43:1P .4492 .44.515 .4018 .44P4 .4879 .37.515 .438.5 1 .0155 N 27 26 115 2.5 25 22 17 32 190 Me in 3.40 3.32 3.17 3.50 3.69 3.24 3.41 3.51 3.42 Std.Dev. .5247 .5318 .6565 .51537 .4748 .6049 .6611 .5019 .5655 -Opp.far Naduluue 1 .130 N 27 :115 115 2.5 25 22 17 32 190 Mem 2.!14 :il.78 :il.!l:il :il.89 3.15 :il.95 2.81 :ii.!IO :il.93 Std.Dev. .6563 .15014 .5680 .1809 .1360 .151521 .7133 .54:il0 .6551 S1>cial SUJIIari Tcrtd 1 .182 N 27. 215 16 25 25 22 17 32 190 Mem 3.29 3.23 3.17 3.3.5 3.51 3.23 3.20 3.34 3.30 Std.Dev. .4480 .4534 .41815 .4372 .43115 .4439 .5124 .310215 .4318 llJlwljust.td. UlliVuilte .ANOVA ar Kiu.Wal W.WS; "P<.OS, ololp<.01 Between Group Differences

PAGE 196

Loneliness and Social Support showed a strong negative correlation with each other (r= -.62,p>.0001) and were also moderately correlated with perceived Neighborhood Cohesion (r = -.34 and .34 respectively, p<.OOOl ); with higher social cohesion associated with less loneliness and greater social support. Loneliness, So'?ial Support and Neighborhood Cohesion were somewhat correlated to. engagement in all activities, including social and sedentary activities, {p<.05), but these associations were strongest with frequency of engagement in community-based activities (p<.Ol). T bl 54 2 L 1' a e one mess, s 1 s OCla upport, S lC h OCla 0 es10n, an Lonely Social Social TotPA Support Cohesion cal exp Lonely Pearson I -.62 -.34 -.16 Correlation N 189 189 189 189 Social Pearson -.62 1 .34 .12 Support Correlation NS N 189 190 190 190 Social Pearson -.34 .34 1 .19 Cohesion Correlation N 189 190 190 190 Correlation ts stgmficant at the 0.05 level (2-tailed). Correlation is significant at the 0.01 level (2-tailed). ModPA cal exp -.12 189 .07 NS "190 .21 190 dPAl Homebased -.17 189 .20 190 .07 NS 190 Results of Multi-Level Hierarchical Modeling Research Question 1: Neighborhood Influence on Social Isolation: Does Context Matter? eve CommALL based activities -.30 -.18 189 189 .23 .23 190 190 .26 :18 190 190 To answer the second part of research question 1 regarding what components of the environmental context are potentially protective against social isolation, hypothesized relationships between loneliness (which, due to its strong correlation 175

PAGE 197

with social support, will be considered the proxy variable for social isolation in these analyses) and neighborhood variables were tested using separate 2-level hierarchical regression models for each neighborhood variable. Table 5.4.3. Multilevel Analysis: Loneliness1 Variable Parameter SE p Estimate AESTHETICS Intercept -3.5754 3.1253 .2902 Litter -.8644 .4687 .0669 Intercept -4.5531 3.2784 .2075 Graffitiz -.5974 .5624 .2896 Intercept -4.2418 3.1803 .2240 Yard Maintenance -0.6923 .4981 .1664 DESTINATIONS Intercept -6.6314 2.8545 .0531 Prop. non-residential 2.8581 2.6717 .2862. Intercept -6.5664 2.8712 .0622" Retail .02005 .03684 .6059 Intercept -6.8356 2.8818 .0554 Services .05964 .05797 .3433 Intercept -5.7573 2.8588 .0907 Opportunities to Exercise -.08466 .06351 .2309 Intercept -5.7939 2.8134 .0784 Trails -6.2%09 3.3145 .0622 VITALITY Intercept -3.5153 3.2753 .3187 People around -1.5756 .9727 .1071 -4.5645 3.1137 .1861 -1.0499 .7740 .1768 176

PAGE 198

Table 5.4.3 (Cont.). Multilevel Analysis: Loneliness Variable Parameter SE p Estimate SOCIAL CAPITAL Intercept -6.7416 2.8412 .0553 Nuisance Properties .05870 .04002 .1928 Intercept -6.0830 2.8362 .0691 Incivilities3 .4886 .4025 .2265 Intercept -7.9545 2.8531 .0270 Window Bars 1.2608 .4768 .0089 Intercept -4.9428 2.8152 .1226 Neighborhood Watch -4.5151 1.8780 .0173 REPORTED DATA Intercept -6.8524 2.9052 .0564 Population Density .00013 .0001 .3176 Intercept -5.6549 3.0708 .1151 Average Household Income -.00001 .00002 .5906 Intercept -5.9128 2.9238. .0896 % Residents 65 years + -.0227 .0364 .5557 Intercept -6.8131 2.8186 .0523 Violent Crime Rate .1336 .0644 .0833 Intercept -6.2843 2.8085 .0666 %Non-White .0147 .0079 .1119 I, Covanates. adjusted for age, sex, mcome level, and chrome disease score. 2 A higher rating indicates a neighborhood that is free of litter and graffiti. 3 A higher rating indicates a greater number of incivilities observed. A few of the neighborhood variables that appeared to have a direct effect on loneliness were also important_ factors in another dependent frequency of community-based activity engagement, discussed in the previous section. Frequency of community-based activity and loneliness were moderately correlated (r= -.30, p<.Ol) and it also makes sense that individuals who more frequently engage in activity outside of their homes may also experience less loneliness. Important mediators between relevant independent variables and community-based activity engagement were perceived safety from crime and social cohesion. These two 177

PAGE 199

variables will be similarly evaluated here, as potential mediators between the relevant independent variables window bars and neighborhood watch signs and the dependent variable loneliness. Hvoothesis 1.4: Neighborhood Structural and Social Factors will be Associated with Less Social Isolation in Adults Ages 65 and Older. Hypothesis 1.4 predicted the association of various neighborhood characteristics and social isolation variables (i.e., perceived loneliness). However, the analyses failed to find associations with walkability, destination, public courtesy or stability variables and a weak association with only one aesthetics variable, i.e., litter (p=.0669), one recreation variable, i.e., presence of trails (p=.066), and one vitality variable, i.e., presence of people (p=.1 07). Two social capital indicators, specifically fewer window bars and greater presence of neighborhood watch signs, were modestly and significantly associated with less loneliness (p<.05). Results of Mediation Analyses Research Question 2: What Are the Potential Causal Pathways Through Which Neighborhood and Individual Factors Influence Loneliness? To answer research question 2, mediation analyses were conducted, particularly with regard to the possible interaction between the two social capital variables that appeared to have a direct effect on loneliness and two mediator variables that were important in earlier analyses: perceived safety from crime and 178

PAGE 200

perceived social cohesion. The hypothesized mediators of self-efficacy for walking, using public transportation, or driving were not associated with perceived loneliness, so were not entered into the mo<;lel. Hypothesis 2.4. The Association of Neighborhood Characteristics with Perceived Loneliness will be Mediated by Perceived Safety from Crime and Social Cohesion. The association between presence of window bars and more loneliness is mediated by perceived safety from crime (Table 5.4.4, Figure 5.4.2). The inverse relationship between having a neighborhood watch and loneliness also appeared to be mediated by perceived safety from crime, since the magnitude of effect decreased by more than 10% and the p value increased to >.05 when perceived safety from crime was entered into the model. However the Sobel test was not significant. Table 5.4.4. Loneliness (with perceived safety from crime) Variables Parameter SE p Estimate First Analysis Intercept -7.9545 2.8531 .0270. Window Bars .4768 .0089 Second Analysis Intercept -2.1360 3.4428 .5546 Window Bars .4571 .5435 .4014 Perceived Safetyfrom Crime -1.1892 .4106 .0043 Sobel t= 3.0195 .0025 First Analysis Intercept -4.9428 2.8152 .1226 Neighborhood Watch -4.5151 1.8780 .0173 Second Analysis Intercept -.6064 3.0276 .8469 Neighborhood Watch -2.8538 1.8923 .1334 Perceived Safety from Crime -1.2218 .3642 .0010 Sobel t= -1.3766 .1686 .. Covanates: adJusted for age, sex, mcome level, and chrome disease score 179

PAGE 201

R=.24 DV: IV: Window Bars 13 = 1.26 Loneliness (p =.008 Mediating Variable: / Safe from Crime / "' IV: DV: Window Bars 13 = .46 Loneliness NS ( ) Figure 5.4.2. Mediation effect: window bars with perceived safety. The association between window bars and more loneliness was also mediated by perceived social cohesion. Again the direct effect between neighborhood watch and loneliness appeared to be mediated by social cohesion, although the Sobel test only approached significance at (p=.0566). 180

PAGE 202

T bl 54 5 L 1' ( 'th a e .. one mess Wl perceive d SOCia CO eSIOn Variables Parameter SE p Estimate First Analysis Intercept -7.9545 I 2.8531 .0270 Window Bars 1.2608 I .4768 .0089 Second Analy_sis Intercept -1.6446 3.4253 .6458 Window Bars .4246 .5352 .4287 Perceived Social Cohesion -.9288 .2937 .0019 Sobel t= 3.4814 .0005 First Analysis Intercept -4.9428 2.8152 .1226 Neighborhood Watch -4.5151 1.8780 .0173 Second Analysis Intercept -.3737 3.0307 .9053 Neighborhood Watch -2.3323 1.9272 .2279 Perceived Social Cohesion -.9378 .2694 .0006 Sobel t= -1.9058 .0566 I, Covanates: adJusted for age, sex, mcome level, and chrome disease score R=.24 DV: IV: Window Bars p == 1.26 Loneliness (p =.008 Mediating Variable: Social Cohesion ( IV: DV: Window Bars P= .42 Loneliness NS Figure 5.4.3. Mediation effect: window bars with perceived social cohesion. 181

PAGE 203

Gender as a moderating variable Despite the literature regarding differential experiences of environmental variables among men and women (Bengoechea, 2005; Dunton et al., 2003; Eyler, 2003), gender was not significantly associated with loneliness. Older age was positively correlated with more loneliness (r = .27,p<.0001) and will be considered in future explorations, given the complexities mentioned in the previous chapter. Hypothesis 2: Summary ofResults Hypothesis 2 was partially confirmed, with specific neighborhood variables having an indirect effect on loneliness, via perceptions of the safety and cohesiveness of the neighborhood. Direct effects of the neighborhood characteristics measured were significantly reduced when these mediators were entered into the model. Perceived Louelluess Figure 5.4.4. Conceptual Model: Direct and Mediated Effects of Neighborhood Variables on Social Isolation 182

PAGE 204

Discussion Significant between-neighborhood differences in individual social isolation outcomes were not found, however a few environmental variables do appear to be associated with individual perceptions of crime and social cohesion, both of which were associated with differences in loneliness. The relationship between loneliness, social cohesion and social capital have been discussed by many theorists (Coleman, 1988; Durkheim, 1897; Putnam, 1995). Durkheim and McMillan and Chavis (Durkheim, 1897; McMillan & Chavis, 1986) discussed the inverse relationship between group membership and depair. Sampson and Raudenbush have demonstrated how trust and reciprocity from neighbors predicts social cohesion (Sampson et al., 1997). Thus, the relationship that this study found between the presence of window bars and both perceived safety from crime and perceived social cohesion is interesting and supports Putnam's views regarding how structural features within communities may engender a sense of trust, reciprocity and (Putnam, 1995). The presence ofwindow bars implies an "every person for themselves" approach to security, particularly when compared with neighborhood watch activities, that suggest collective action. Neighborhood watch activities benefit all households within a neighborhood, as it implies neighbors are keeping "eyes on the street". Window bars only benefit the individual household that has them, and could even direct crime to other homes that have not taken these 183

PAGE 205

precautions. The fact that neighborhood watch signs variable was not mediated by perceptions of safety from crime or social cohesion suggest that"it has its own direct effect on perceived loneliness. Since it is likely that more cohesive neighborhoods will get together and establish a neighborhood watch versus the other way around (i.e., neighborhood watch results in higher cohesion) its importance to perceived loneliness is not clear, althoughperhaps, that once established, neighborhood watch activities may work to maintain both a higher sense of cohesion and lower perceived loneliness among residents. The results of these analyses suggest that the overall importance ofbuilt environmental variables with loneliness is less then the importance of variables that relate to the social environment, i.e., safety and cohesion. While it was surprising that presence of services in the neighborhood were not associated with less loneliness, neighborhoods with the lowest average loneliness among respondents were also those that both study participants and field auditors rated as well groomed and safe (i.e., Neighborhood 1, low walkable, suburban; Neighborhood 4, low crime; Neighborhood 5, high income; Neighborhood 8, high proportion of elderly), although the between-neighborhood difference for loneliness only approached statistical significance (p=.08). Further study of environmental features that might address loneliness and increase social cohesion will be important, given the association of loneliness with 184

PAGE 206

increased morbidity and mortality, dementia and institutionalization (Berkman & Glass, 2000; McMillan & Chavis, 1986; Rabin, 2000; Rane-Szostak: & Herth, 1995). Structural components that facilitate community-based activity as well as cohesion may help to reduce loneliness, however, there may be other more important factors, such as a sense of community belongingness and usefulness, that the current study design was unable to illuminate. The above analyses along with sections 1-3 of this chapter discussed the multi-level influences on individual behavior in order to answer research questions 1 and 2. The next and final section of this chapter is focused on research question 3: Do objective health outcomes cluster by neighborhood? The results presented. use data from a larger sample than the participant sample in order to illustrate the importance of "place" and health and highlight patterns for future study. 185

PAGE 207

Section 5: The Relationship ofNeighborhood Characteristics to Health Recent studies of the interactions between environment and objective health have primarily focused on the relationships between environments that promote physical activity and obesity (Frank, Andresen, & Schmid, 2004; Frank, Schmid, Sallis, Chapman, & Saelens, 2005; Lopez-Zetina, Lee, & Friis, 2005). Several of these studies have found associations between land-use, time spent commuting in cars, time spent walking, and body mass index (BMI). One Atlanta-based study of over 10,000 respondents to a travel survey found that land-use mix (measured using GIS) had the strongest aSsociation with obesity; and that each additional hour spent in a car per day was associated with a 6% increase and each additional kilometer walked per day was associated with a 4.8% reduction in the likelihood of obesity (Frank et al., 2004). In this section, previously presented (see Section 2, above) objective health data from participating Kaiser Pennanente HMO members' electronic medical records were compared with eligible non-participants and then combined and clustered by for the purpose of answering research question 3: Do objective health outcomes cluster by neighborhood? The indicators analyzed included smoking .status, blood glucose, HDL cholesterol, Triglycerides, BMI, blood pressure, pulse rate, whether the participant 186

PAGE 208

was diagnosed with diabetes, hypertension, hypercholestemia, and/or heart disease (i.e., CVD score), and total estimated nwnber of co-morbidities (i.e., chronic disease score or CDS) for the purpose of answering research question 3: Do objective health outcomes cluster by neighborhood? Participant and Non-Participant Comparison Because of the large sample size, t tests for independent samples were conducted to compare study participants to non-participants on sex, age and objective health data. Significant differences between participants and nonparticipants were noted for sex (p<.05), with a greater proportion of males in the participant group. Significantly higher blood glucose levels and significantly higher pulse rates were observed for non-participants (p<.Ol), although the mean pulse rates for both groups were within normal limits (Table 5.5.1.). Table 5.5.1. Characteristics of participants compared with non-participants for sex, age, Chronic Disease Score (CDS) and Cardiovascular Disease Risk (CVD) Levene's N #female ##male %male Test t test for equality of means equality ofvar Sig. t df Sig. (2-tailed) !sex participants 190 108 82 43.1 non-participants 681 424 257 37.7 --1.337 297.17 .182 SE Mean Std. Dev. Mean !Are participants 190 74.21 5.806 .421 non-participants 681 73.86 6.450 .247 .105 .668 869 .504 participants 190 2.52 1.760 .128. non-participants I 681 1 2.75 I 1.906 I .073 187

PAGE 209

Table 5.5.1 (Cont.). Characteristics ofpariicipants compared with non-participants for sex, age, Chronic Disease Score (CDS) and Cardiovascular Disease Risk (CVD)1 Levene's t test for equality of means Test equality ## ofvar non-## % Sig. t df Sig. N smokers smokers smokers _{_2-tailed) Smokinl! Status participants 175 160 15 7.9 npn-participants 608 535 73 10.7 -1.232 339.93 .219 -N Mean Std. Dev. SE Mean BMI partie. 174 27.6170 4.94711 .375 non-part. 607 27.3825 5.61891 .228 .220 .498 779 .619 systolic partie. 190 130.23 19.877 1.442 bp non-part. 681 130.85 18.131 .695 .214 -.410 869 .682 diastolic partie. 190 73.75 11.036 .801 bp non-part. 681 73.87 10.844 .416 .576 -.137 869 .891 pulse partie. 190 76.21 15.867 1.151 ** rate non-part. 681 79.97 14.710 .564 .558 -3.065 869 .002 BGlu partie. 108 106.17 23.83 2.30 ** non-part. 279 112.9 35.78 2.14 .008 -2.145 290.13 .033 CVD partie. 190 1.26 1.100 .080 score3 non-part. 681 1.43 1.190 .046 .01 -1.786 322.87 .075 UnadJusted umvanate statistics, usmg Levene s test for equality ofvanances and t-test for mdependent samples, *p<.OS, **p<.Ol. 2 Total CDS is the total number of co-morbid chronic diseases estimated by pharmacy data in the electronic medical chart. 3 CVD Score indicates number of chronic conditions that increase risk for heart attack (scored 1-4). Health Characteristics of Combined Sample When participant and non-participant objective health data were combined to better understand neighborhood differences, more differences were detected. Using ANOVA large differences for the combined sample were detected for Systolic (F(7, 863)=4.348, p<.001) Blood Pressure, and moderate differences for Diastolic Blood Pressure (F(7, 863)=3.344, p<.01); triglycerides (F(7,601)=2.360, p<.05); BMI (F(7, 188

PAGE 210

773)=2.889, p<.Ol); and Chronic Disease Score (CDS) (F(7,863)=2.193, p<.05) (see Table 5.5.2). Table 5.5.2. Combined Participant and Non-Participant Objective Health Metrics by Neighborhood1 METRIC: Neigh Neigh Neigh Neigh Neigh Neigh Neigh Neigh TOTAL 1 2 3 4 5 6 7 8 BMI ** N 53 115 141 62 54 109 100 147 781 Mean 28.57 26.72 28.81 26.90 25.84 27.57 27.56 26.89 27.43 SD (6.24) (5.19) (5.58) (6.11) (4.45) (5.33) (5.24) (5.36) {5.47) BP ** N 60 120 142 79 60 130 129 151 871" Mean 133/75 131/75 136/76 130/72 ll3170 133174 132/75 126/72 .131/74 (SD sys, (18.63, (19.72, (19.87, (19.58, (16.92, (17.71, (16.69, (16.90, (18.52, SD dia) 10.92) 10.50) 11.36) 9.93) 11.22) l0.96J 10.34) 10.88) BPMeds N 60 120 142 79 60 130 129 151 871 # onMeds 24 43 65 35 27 58 63 60 375 % 40% 35.8% 45.8% 44.3% 45.0%. 44.6% 48.8% 39.7% 43.0% HDL N 48 78 103 55 45 .94 103 124 650 Mean 59.38 60.08 58.23 57.93 57.98 54.99 54.77 59.00 57.62 .. SD (17.40) (13.5) (15.77) (18.2) (15.74) (15.33) (17.42) (16.41) TRIGLYC N 48 69 94 52 39 91 96 120 609 Mean. 152.46 147.81 134.65 168.56 152.44 167.76 167.25 144.16 153.54 SD (75.96) (79.87) (59.20) (82.91) (68.87) (87.85) (78.82) (70.82) _(_76.28) PULSERT N 60 120 142 79 60 130 129 151 871 Mean 78.65 77.31 79.10 80.20 76.65 81.00 79.02 79.81 79.15 SD (12.49) (14.33) (14.73). (15.84) (20.48). (14.80) (14.22) (14.77) (15.04) BGLU N 35 36 51 36 18 62 68 81 387 Mean l 03.11 109.19 104.8 105.0 110.0 ll5.18 117.57 113.37 111.02 SD (22.38) (27.68) (27.85] (16.93) _(35.63j_ (41.14) (32.99) Total CDS 1 N 60 120 142 79 60 130 129 151 871 Mean 2.38 2.66 2.99 2.73 2.22 2.95 2.84 2.44 2.70 SD (1.66) 11.96) (1.93) (1.67) (1.90) (1.95) (1.90) (1.88) I ** Unadusted, umvanate ANOVA, *p<.OS, p<.Ol Between Group Differences 2 Total CDS is the total number of co-morbid chronic diseases estimated by pharmacy data in the electronic medical chart. 189

PAGE 211

Subsequent post hoc analyses found significant differences for Neighborhood 3 (selected for its high violent crime rate, and high proportion of ethnic minorities), which scored worse than 5 (high income) and 8 (high proportion of elderly) for a of health indicators including BMI, Systolic and Diastolic blood pressure (p<.05), and CVD score (i.e., score of 1-4 representing total number of diagnoses including diabetes, hypertension, hypercholestemia, and heart disease) (p<.001). there appeared to be a difference in Chronic Disease score between Neighborhood 3 and5 based on the confidence intervals, but the post hoc test was not significant. Percent of individuals at higher risk for heart attack (MI) were compared for each neighborhood. Individual residents were scored a 1 if they smoked or if they were obese (i.e., had a BMI 2:30). As mentioned above, individuals with diagnosed diabetes, hypertension, hypercholestemia, and/or heart disease received a cumulative score of 1-4 (indicated in Table 5.5.3. as CVD score) depending on how many of these diagnoses were indicated in their medical record. the Pearson Chi Square, neighborhoods were found to vary significantly on BMI (x2(7)=23.528, p<.01); and CVD score (:i(7)=26.696,p<.001) (see Table 5.5.3). 190

PAGE 212

Table 5.5.3. Proportion of Participants and Non-Participants (combined) at higher MI k b N 'ghb h d1 ns >Y e1 or oo METRIC: standards for Neigh Neigh Neigh Neigh Neigh Neigh Neigh Neigh healthy adults 1 1 3 4 5 6 ., 8 %Smokers N 60 120 142 79 60 130 129 151 % 11.7% 10.8% 9.15% 12.6% 13.3% 6.15% 9.3% 4.6% BMI: >30 N 53 115 141 62 54 109 100 147 % 32% 19% 41% 21% 19% 27% 25% 25% CVD Score1 N 60 120 142 79 60 130 129 151 Mean 1.2 1.28 1.75 1.34 1.10 1.38 1.57 1.23 SD (1.13) (1.12) (1.14) (1.19) (.95) (1.19) (1.22) ( 1.18) Percent represents proportion of sample that exceeds the standards for each nsk factor. Pearson Chi-Square, significant at "'p<.OS, *"'p<.Ol TOT 871 10.1% ** 781 27% ** 871 1.39 (1.171_ 2 CVD Score indicates number of chronic conditions that increase risk for heart attack (scored 1-4 ). Post hoc analyses showed that Neighborhood 3 had a significantly greater CVD score than Neighborhoods 2, 5, and 8 (p<.01). Neighborhood 3 also had significantly greater obesity than Neighborhoods 2, 4, 5 and 8 (p<.01). Neighborhood 5 had the least number ofMI risk factors, significantly lower than Neighborhood 3 (p<.01). Summary. Residents in Neighborhood 3 (high crime) had greater number of:MI risk factors and poorer overall health than residents in Neighborhoods 5 (high SES) and 8 (high elderly). 191

PAGE 213

Research Question 3: Do Objective Health Outcomes Cluster by Neighborhood? It is apparent, from the above 3.nalyses, that BMI and blood pressure differed by neighborhood, however, interpretation of these patterns is complicated, particularly with regard to hypertension, which is variable and difficult to interpret based upon 1 medical chart reading. CVD score also varied by neighborhood and post hoc analyses did indicate significant differences between specific neighborhoods. Total CDS and triglycerides also appeared to vary by neighborhood, however few pairwise comparisons were significant when post hoc analyses were performed. Since BMI tends to be a much more stable measure over time and is also associated with many of the other cardiac risk metrics, that metric, along with CDV score, will be looked at next in greater detail. Hvoothesis 3. Neighborhood Factors Conducive to More Active Lifestyles and Lower Perceived Social Isolation are Associated with Lower BMI and Fewer CVD Risk Factors. To test whether neighborhood variables predicted BMI, the mea8ured variables that were used in the earlier analysis of neighborhood characteristics associated with different types of activity engagement and loneliness were entered into separate 2-level hierarchical regression models, with BMI as the dependent variable. 192

PAGE 214

T b1 55 4 Mul "1 1 Anal C b. ed 1 (N 781) BMI1 a e .. . tt eve ys1s: om m sampJe = Variable Parameter SE p Estimate WALKING FUNCTIONALITY Intercept 38.6817 2.4864 <.0001 continuity .05178 .4136 .9004 Intercept 41.5352 2.5144 <.0001 Path width -1.5745 .6166 .0109 Intercept 40.2494 3.1124 <.0001 Path Maintenance -.6342 .8967 .4796 SAFETY Intercept 39.2805 2.3953 <.0001 Traffic buffer -.3626 .4083 .3749 Intercept 42.2330 2.5177 <.0001 Curb cuts -1.5339 .4948 .0020 Intercept 37.0595 2.7529 <.0001 Ped. signals 2.3470 1.9738 .2348 Intercept 39.3223 2.3489 <.0001 Cross-walks -2.7143 1.8819 .1496 AESTHETICS Intercept 39.7655 2.6194 <.0001 Litter -.4456 .5493 .4175 Intercept 37.9773 2.8088 <.0001 Graffiti2 .3190 .6044 .5978 Intercept 40.0027 2.5960 <.0001 Yard Maintenance -.5687 .5486 .3002 193

PAGE 215

Table 5.5.4 (Cont.). Multilevel Analysis: Combined sample (N=781) B Variable Parameter SE p Estimate DESTINATIONS Intercept 38.7439 2.4209 <.0009 Proportion non-residential .2049 3.1852 .9487 Intercept 38.9475 2.3353 <.0001 Retail -.03485 .03804 .3599 Intercept 38.9093 2.3688 <.0001 Services -.02171 .06805 .7498 Intercept 38.7964 2.3402 <.0001 Opportunities to Exercise -.00589 .06232 .9248 Intercept 40.2995 2.3476 <.0001 Facilities/Courtesies -1.0304 .3901 .0084 VITALITY Intercept 40.6565 2.7119 <.0001 People around -1.2403 .9510 .1925 Intercept 39.9985 2.4210 <.0001 Activities -1.0625 .6406 .0976 SOCIAL CAPITAL Intercept 38.9247 2.3500 <.0001 Nuisance Properties -.02360 .04897 .6299 Intercept 38.7170 2.3594 <.0001 Incivillties3 .09597 .4544 .8328 Intercept 38.2239 2.5717 <.0001 Window Bars .3093 .5908 .6008 Intercept 38.8185 2.4049 <.0001 Neighborhood Watch -.1494 2.4530 .9515 l Covanates: adJusted for age, sex, and chrome disease score. 2 A higher rating indicates a neighborhood that is free of litter and graffiti. 3 A higher rating indicates a greater number of incivilities observed. Neighborhood Factors and Cardiovascular Disease Risk To continue our test of hypothesis 3, these same neighborhood variables were entered into separate 2-level hierarchical regression models, with CVD score as the dependent variable. 194

PAGE 216

Table 5.5.5. Multilevel Analysis: Combined sample (N=781) CVD Variable Parameter SE p Estimate WALKING FUNCTIONALITY Intercept .-.5610 .6279 .4013 continuity .02716 .1021 .7902 Intercept -.0337 .6764 .9586 Path width -.2407 .1800 .1816 Intercept -.4082 .7792 .6166 Patb Maintenance -.04105 .2224 .8536 SAFETY Intercept -.3699 .6131 .5653 Traffic buffer -.08375 .1091 .4428 -.6771 .7423 .3920 Curb cuts .07716 .2031 .7042 Intercept -.3237 .7217 .6673 Peel. signals -.2493 .5144 .6644 Intercept -.5231 .6039 .4149 Cross-walks .07821 .5803 .8928 AESTHETICS Intercept .1363 .6127 .8303 Litterl -.2753 .09111 .0016 Intercept -.5241 .7196 .4900 Graffiti2 .0077 .1594 .9612 Intercept .2132 .6064 .7355 Yard Maintenance -.3117 .08966 .0003 DESTINATIONS Intercept -.7694 .5912 .2343 Proportion non-residential 1.5013 .5641 .0079 Intercept -.4614 .5923 .4615 Retail -.00679 .01075 .5280 Intercept -.4740 .6001 .4555 Services -.00491 .01835 .7890 Intercept -.4508 .5893 .4693 Opportunities to Esercille -.01652 .01545 .2852 Intercept -.5658 .6053 .3810 Facilities/Courtesies .0583 .1350 .6660 195

PAGE 217

Table 5.5.5 (Cont.}. Multilevel Analysis: Combined samp_le _ili_=781) C Variable Parameter SE p Estimate VITALITY Intercept .2668 .6776 .7055 People around -.4877 .%223 .0286 Intercept -.08638 .6124 .8918 Activities -.3211 .1497 .0323 SOCIAL CAPITAL Intercept -.4906 .5940 .4361 Nuisance Properties -.00246 .01337 .8538 Intercept -.6145 .5872 .3301 .1698 .08967 .0587 Intercept -.9296 .6202 .1776 Window Bars .2412 .1149 .0361 Intercept -.4413 .6111 .4939 Neigbborbood Watcb .6387 .7049 t Covanates. adJusted for age, sex, and chrome dtsease score. 2A higher rating indicates a neighborhood that is free of litter and graffiti. 3 A higher rating indicates a greater number of incivilities observed .. Hvoothesis 3: Summary of Results Hypothesis 3 was partially supported. In the above analyses that looked at BMI as the health indicator, significant associations for a few variables that support safe walking (i.e., sidewalk width and curbcuts) and neighborhood facilities/courtesies (e.g., public transportation stops, public benches, newspaper dispensers, mailboxes) were associated with BMI in the hypothesized directions, however variables that promoted walking for errands, such as cross walks and density of retail, were not significantly related to BMI. On the other hand, some of the same neighborhood characteristics that promoted commwrity-based activity and less loneliness (i.e., litter, yard maintenance, and window bars) did predict fewer CVD-related chronic conditions. Also, neighborhood vitality variables (i.e., people 196

PAGE 218

observed outside and people observed engaged in activities) and a higher proportion of residential land use, which was one of the predictors of greater energy expended in weekly P A, were related to fewer CVD-related conditions (lower CVD score). Because these analyses were performed on the larger sample, the above analyses were limited to information available in the record. Therefore, it was not possible to control for income or race, or to perform mediation. analyses using perceptions or other self-reported behaviors as potential Discussion An advantage of GIS is its ability to provide visual evidence that can be used to-identify potential problem areas and generate testable hypotheses. However, the reader should be cautious of drawing conclusions from such maps that, in and of themselves, do not provide statisticil evidence as to why such patterns exist or whether they are indicative of actual problem areas. Studies have shown that individual risk is not limited to individual factors (e.g., income; health) but may also depend on community factors (resources; safety). The previously mentioned studies that found associations between high levels of community unemployment and increased individual stress levels, regaz:dless of the individual's actual .employment status are just one example of this phenomenon (Diez-Roux, 1998) Thus it is possible that neighborhood level factors may be associated with physiological health, 197

PAGE 219

regardless of other individual attributes, such as personal resources and access to healthcare. All of the individual data in this study were from adults who were insured with Kaiser Permanente, yet between-neighborhood differences did emerge for so:q1e objective health metrics, i.e., BMI and number of CDV risk factors. Unfortunately, sociodemographic data were limited for this particular analysis, particularly regarding potentially relevant data such as income and race/ethnicity. It is notable that the two neighborhoods with the worst health picture were also neighborhoods with the highest proportion of racial and ethi:ric minorities and had the highest violent crime rate and lowest proportion of elderly, respectively. It is unknown ifthe individuals were themselves racial or ethnic minorities or whether they perceived that they had high access to resources and services or not. 198

PAGE 220

CHAPTER6 CONCLUSIONS AND FUTURE DIRECTIONS The purpose of this study was to explore components of the neighborhood environment that may facilitate or inhibit physical and social activity in older adults. One hundred and ninety medically insured, community-dwelling adults, aged 65 years and older and residing in one of 8 diverse Denver neighborhoods, participated in this cross-sectional study. This chapter summarizes the study results for each of the three research questions. Strengths and limitations of the study are presented and future directions for research and possible intervention are discussed. Answers to the Research Questions Question 1: What Components of the Environmental Context Enable or Inhibit Physical and Social Activity in Older Adults? Neighborhoods were selected to vary on crime rates, population demographics and walkability as defined by street layout and land use heterogeneity. No one neighborhood provided the ideal environment to support all types of activity. Instead, specific features of neighborhoods were associated with specific forms of activity. Neighborhoods with greater number of features with the traditional urban design (grid-like street layout, crosswalks and curbcuts, high density of retail) predicted greater frequency of walking for errands. However, walking for errands was not a primary form of activity for this elderly population and 199

PAGE 221

the highest levels ofPA were actually reported in the most suburban, least walkable neighborhood. These residents reported infrequent walking for errands but high frequency ofhome-based forms ofPA, such as housework, gardening and heavy yard work. Higher participation in community-based activity, including non-active social activities such as visiting friends, going to the movies, and attending meetings, as well as daily living activities su.ch as going to the grocery store or bank, was associated with neighborhoods that were aesthetically pleasing and had recreational resources and numerous public courtesies, such as public benches and public transportation stops. Engagement in activities, particularly community-based activity, was inversely related to loneliness, which was used as a proxy measure for social isolation. However, neighborhood variables that predicted community-based activity, such as less litter, fewer window bars and more neighborhood watch signs, were only modestly associated with less loneliness, which had a strong association with perceived social cohesion. Question 2: What are the Potential Causal Pathways through which Neighborhood and Individual Factors Influence Outcomes? The possible indirect relationships of neighborhood variables operating through individual perceptions were explored to help fine-tune the proposed model and provide insight into potential areas for intervention. While direct effects of the 200

PAGE 222

built environment on frequency of walking for errands did not appear to be mediated by ariy of the individual perceptions measured, levels of community-based activities and total energy expended in P A were both mediated by perceptions of pedestrian safety from crime and perceived social cohesion of the neighborhood. These same 2 variables also mediated the relationship between identified neighborhood variables and perceived loneliness. Question 3: Do objective health outcomes cluster by neighborhood? GIS .was used to map 871 KPCO members who qualified for the study and resided in the 8 study neighborhoods .. By combining non-participants with participants to create the larger sample, clustering of individuals with poorer health metrics could be visually identified and then tested using quantitative analytical methods. In this manner, between-neighborhood differences were found for BMI and CVD score, adjusting for age, sex and chronic disease score, however few of the observed neighborhood variables predicted these health outcomes. While a greater proportion ofindividuals with more risk factors for MI resided in neighborhoods with higher violent crime rates, more incivilities, lower neighborhood SES, lower proportion of elderly residents, and greater proportion of non-white residents, our ability to interpret these patterns is limited for a number of reasons. First, only a subset of the neighborhood population were mapped and potentially relevant indiviclual sociodemographic data individual race and income) were UI)known 201

PAGE 223

for 78% of the mapped sample, thus it is difficult to generalize these data to the neighborhood as a whole. Second, given the chronic nature of the health outcomes analyzed and the fact that many-of these conditions take years to develop, knowing the length of exposure to a particular neighborhood environment as well as information on changes within the neighborhood over time, is relevant to understanding the relationship of environmental features to health. On the other hand, from an intervention perspective, identifying geographic areas where specific health issues aie more prevalent can provide useful, cost effective information for screening, education, prevention, and delivery of services that target the greater needs of a specific group or neighborhood. Limitations of the Study Generalizabilitv To enhance consistency of sociodemographic and crime data used to select neighborhoods, the neighborhood sampling frame was restricted to Denver City and County's statistical neighborhoods. This eliminated the more remote suburban and rural communities in the Denver vicinity. To maximize variability, neighborhoods_ were selected to represent the extremes on walkability, violent crime rate, proportion of elderly residents, and average household income. However, another trade-off was made in favor ofrec:ruitment efficiency by limiting the sampling frame to adults who were insured by KPCO. Thus not only were neighborhoods restricted to Denver City 202

PAGE 224

and County, they were also required to have at least 80 residents who were active KPCO members, resulting in the exclusion of a few neighborhoods including those with the highest crime and lowest average household incomes. Generalizability is therefore narrowed based on the range of neighborhoods sampled. Culture and ethnicity Given the varied cultural views of aging and the known influence of culture and ethnicity on roles, values and behavioral norms, an important limitation of this study is its inability to study the ethnic and cultural factors that may have had an effect on active lifestyles, social isolation, and health, apart from other neighborhood factors. Similarly, it is unknown to what extent group ethnic or cultural identity affected neighborhood characteristics (Diez Roux, 2004). This information would be particularly useful, given that the two neighborhoods with the greatest proportion of non-white residents were also neighborhoods whose elderly KPCO members had the greatest number of CVD risk factors, low activity levels and higher loneliness. Although study participants were asked to self-identify their racial and/or ethnic background, analyses were limited due to the small-proportion of ethnic and racial minorities who participated in the survey. Spatial Autocorrelation Non-contiguous neighborhoods were selected to maximize between-group differences and minimize interaction among the selected neighborhoods. However, this sampling and data-collection method did not account for environmental features 203

PAGE 225

located in adjacent neighborhoods, which may have been as important to participant behavior. Collecting data within buffer zones that surround individual residences is a common method for defining relevant activity space. This method may be useful for studying older adults whose activity may be smaller relative to yom1ger adults. However, even this method does not take into accoWlt older adults who work or volm1teer outside of their neighborhood, a common scenario. The method used to characterize neighborhoods in this study included a walking audit of the participant's residential block as well as a random selection of neighborhood blocks to attempt to characterize the unique features of that neighborhood for comparison purposes. An assumption was made that the blocks where participants resided would have similar features to adjacent blocks, even if they are outside of the neighborhood bom1daries, and thusshould not contaminate the data collected. This assumption, however, may not be true, particularly for destinations such as retail, restaurants and services whose access and use may be less dependent on proximity to the immediate home environment. Contributions of the Study Neighborhood Characterization A chief goal of this study was to characterize neighborhoods by sampling a small, randomly selected percent of street miles (i.e;, 5%) for each neighborhood. The method allowed for quick and efficient data collection given limited time and ) 204

PAGE 226

resources, a common constraint in many studies. Based on the data collected, along with reported census data, it was possible to characterize each neighborhood on the built environment and social capital variables of interest and explore the relationship of these neighborhood variables with physical activity, loneliness and objective health indicators. The specific findings of this study, i.e., that built environment variables that support walking may be less relevant to the total P A levels of older adults as well as to their community engagement and social integration, have important implications for designing communities that will promote mental and physical health throughout the lifespan. These findings agree with other studies that demonstrated that older adults tend to favor indoor activity with fewer energy demands (Dallosso et al., 1988; Lee & King, 2003). While environmental variables were important to increasing frequency of walking for errands, other factors such as living in an aesthetically pleasing, safe neighborhood, with some opportunities for recreation, seemed to be of greater importance to overall activity levels and less loneliness. Safety and Social Cohesion The fact that perceived safety from crime and social cohesion substantially mediated the direct effects of aesthetic and social capital variables on community based activity engagement and loneliness emphasizes the importance of perceived safety to older adults. Thus, a neighborhood environment that is clean, safe and predictable, and fosters trust and reciprocity among neighbors, may be at least as 205

PAGE 227

important as the physical environment to promoting activity and reducing loneliness for this population. Revised Concemtual Model Neighborhood Structural Factors: Walkability Land Use Figure 6.1. Revised Conceptual Model. Incivilities Window Bars Neighborhood watch Perceptions: Crime Safety Social Cohesion Loneliness Structural and social factors are important to enhancing activity engagement and preventing social isolation in seniors. The revised conceptual model illustrates how different types of factors support different outcomes. Neighborhood aesthetics plus symbols of neighborhood decline and self-protection (e.g., incivilities and window bars) are indirectly related to outcomes, working via perceptions of safety and social cohesion. These perceptual variables are multi-directional in that they 206

PAGE 228

may also influence the level of social capital within a neighborhood by influencing pride, norms or organization of a neighborhood watch group. Implications for Future Research and Intervention The findings of this study suggest several areas for future research: 1. Direct measurement of objective neighborhood characteristics may be more important for understanding transport and mobility constraints than commUnity engagement and social isolation, which may be more effectively measured via resident surveys of perceptions of safety and cohesion. 2. More study of the trade-offs between l.irban environments that promote efficient walking for transport and pleasant, "safe" environments that community engagement and social cohesion, and whether planned communities that combine these features are successful in increasing overall PA. 3. Study of the role that culture and ethnicity. play in shaping neighborhoods with regard to enhancing or reducing social cohesion among neighbors. 4. Study the relationships among collective efficacy, neighborhood activism and perceived social cohesion. 5. Further study ofhow suburban environments may increase PA levels in older adults by promoting more walking for pleasure, home upkeep and yardwork. 207

PAGE 229

6. Further study of how seniors living in more rural communities compare with metropolitan area seniors, with regard to activity engagement and social isolation. 7. Longitudinal study of how built and social environmental variables effect objective health outcomes throughout the lifespan. 8. Experimental studies of planned interventions as well as natural expep.ments, to study the effects of changes to the built environment over time, e.g., change in automobile use after the establishment of a mixed-use transportation hub in a neighborhood previously zoned for residential land use only. Increasing physical activity and preventing are critical to preserving the health and independence of older adults. Addressing social factors within neighborhood environments by enhancing aesthetics and perceptions of safety from crime and social cohesion may be just as important as providing walkable communities with convenient destinations for seniors. Interventions that promote the interaction of neighbors to increase perceptions of social cohesion and safety may be one way to promote greater community engagement and reduce loneliness. 208

PAGE 230

GLOSSARY Activity Space: The area where a person spends time including residence, neighborhood, workplace, retail space, recreational space, and the pathways traveled to and from these sites (Cromley & McLafferty, 2002). Collective Efficacy: An individual's beliefs about the collective capability, shared lmowledge and skills of a group to achieve desired group outcomes (Bandura, 1997). connectivity: Ease of travel between two points directly related to design of streets and modes of available transportation (Saelens, Sallis, & Frank, 2003). GIS: Geographic Information System (GIS) is a computer-based tool that allows the user to capture, store, retrieve, analyze and visually display database information that is "spatial" in nature, using a geographic presentation, such as a map (Cromley & McLafferty, 2002). Land Use Mix: Level of integration within a given area of different types of uses for space (e.g., residential, retail, commercial, public) (Saelens, Sallis, & Frank, 2003). Loneliness: A subjective experience defined as a perceived lack or loss of a meaningful network of interpersonal contacts and/or companionship (Bergman Evans, 1994; Penninx et al., 1997; Walker & Beauchene, 1991). Loneliness is NOT synonymous with social isolation, which is a more objective concept (see definition below). 209

PAGE 231

MET value: 1 equivalent value is defined as the ratio for the body's working metabolic rate of a given activity divided by the resting metabolic rate of that individual (1 Kcal/kg body weight/hour) (Lee & King, 2003). Thus, 1.0 MET is equivalent to the body at rest, and 3.3 METs is equivalent to walking at a moderate pace, on level ground (Ainsworth, 2002). Neigp.borhood Effect: When attributes of neighborhoods affect health and/or social outcomes (Diez Roux, 2004; Oakes, 2004). Residential density: Number of dwellings per unit ofland area acre) (Saelens, Sallis, & Frank, 2003). Self-Efficacy: An individual's beliefs about his or her capabilities to organize and execute the required courses of action to produce specific behaviors or desired outcomes (Bandura, 1994, 1997). Social Cohesion: Reciprocity, trust, shared behavioral norms, and common v:;tlues among neighbors (Sampson.et al., 1997). Social Engagement: Also termed "activity engagement" or "occupational engagement". Engagement in activities performed within the context of the social environment. Such activities may be deliberately chosen in order to reach valued goals or may be activities required by social norms such a$ hygiene, housework, and other instrumental activities of daily living. (Herzog et al., 2002; Hopkins & Smith, 1993) Social engagement has, in some works, been defined as maintaining social ties 210

PAGE 232

(Bassuk et al., 1999), but for the purpose of this study, we term that Social Integration. Social Integration: Having strong ties and sense of membership or belongingness to a specific group or community (Berkman & Glass, 2000; Durkheim, 1897). Social Isolation: Individuals or groups who lack social contact due to separation or marginalization from their community (Berkman & Glass, 2000). Statistical Neighborhood: The City and County of Denver, in collaboration with the Denver Regional Council of Governance (DRCOG) and local neighborhood governments, has designated 77 geographic areas as statistical neighborhoods. Each statistical neighborhood is attached to one or more census tracts, as defmed by the U.S. Census Bureau. Census tracts consist of between 4,000 and 8000 persons with relatively homogenous social and economic characteristics and have the advantage of remaining generally stable from census to census. Thus, data on statistical neighborhoods can be used to compare geographic areas over time (DenverGovemment, 2002). 211

PAGE 233

APPENDIX A NEIGHBORHOOD AUDIT AND. DATA COLLECTION FORM 212

PAGE 234

Auditor 10: Date: --Iime : amlpm Neigh.:--------Study ID(s): ------bata EntrV Date: Rec II: BUILDING CHARACTERISTICS 1. Type of Residential housing In ( all) H: single family detached dwelling 04 RH: rowhouscs I townhomcs 03 MO: multiple occupancy (2 units) 01 A: > 7 units 01 2. MAIN/rredomlnant housing in ,iewlsecmenl'l II: single family detached dwelling 04 Rll: rowhouses /ro"nhomes 03 MO: multiple occupancy (2 units) O, A: > 7 units o; BUILDING/LOT CONDITION J. ProJiol1ion w/ wiudowslburntdlboarded up or None 04 Less than 03 o 112 ol More than half 01 4. GR,\FJTTI amount? None D4 Alittlc D3 A moderate amount Dl Alot D1 s. 0\erall condition of most buildings E'cellcnt/Well kept 03 Fair condition 01 Poor/deteriorated condition 01 Not Applicable (no bldg) 00 6. Proportion with front yard/grounds? None 01 Less rhan o, 113'd ro 112 o; than half 04 7. Proportion with front Jawonundscape/planterslgardens? None Dr Less than D, 1/J'J to 1/2 D; More than half D4 8. LITTER amount? None D, Alirrle D1 A moderate amount D, Alot D; 9. 0\'erall conditions of grounds/\'ards? (Wcll-maintalncd=trlm, clean, Excellent(> 7Wo well-maintained) 03 Fair (S0-74% well maintained) 01 Poor (7S% ofbldgs 03 -. IS. Proportion of buildings with attendants or security guards? None 01 Less than ltl"' Dt Jl)oltO 1/2 03 than half 04 16. "Nelchborhood Watch" signs? No D1 Yes D: llome Sccurltv 17. Proportion ofbulldinJ:S(homes& commcrtial) with security barslerates? None 0, Less than Ol to 112 0! More than half 01 213 Proportion with securit\' signs? None 01 l.m than Jl)'d 03 lilrd ro 112 ol More than half D 1 19 Propol1ion or buildines 1\ith ,islblc dogs? None 0, less than 1/Jol 03 1/)rd to 112 ol More than half 01 NEIGHBORHOOD VITALITY 10. People around: Are there people outside? ( all) None 0 Workers D <12y 0 Teens 0 Adults 0 Seniors 0 ll. What arc people doing? ( all) None obsm ed 0 Stand/SitiPatkcd 0 Walking 0 Running 0 Cycling 0 Playing D GardcninsfY ardwork 0 Otherwork 0 Other social 0 NEIGHBORHOOD STABILITY 12. For Sale. signs? ) or more 1 None 23. For R,nr signs? 3 or more 1-2 None NEIGHBORHOOD DESTINATIONS o, o. o; 24. Propo11ion of in \iew/Sfllmentthal are non-residential? NonC Less than D3 1/)N 10 112 ol More than half 01

PAGE 235

lS. Non-Residential Destinations tally Keep tally of the following categories an the line provided to the left. Add up total tally and write itlH:tween brackeis. ( denotes senior friendly) 1._: _( ) An Gallerie&IMuseums* 2. __ : _( ) Appliance sales, rentals repairs 3. __ ( ) Auto repair/body shops/garages 4. __ ._( ) Auto sales/rentals S. __ ( ) Bakery* 6. __ ( ) Bank* 7. __ ( ) Barbcrslbcauiy parlors* S. __ ( ) Bars 9. __ ( ) Business Services (copies) I 0. __ ( ) Check cashing services II. __ ( ) Clothing stores 12. __ ( ) Church/Religious ccntcrsW IJ __ ( ) Coffee shops -k 14. __ ( ) Convenience Store (7) IS. __ ( ) Crafts/fabric/yam store 'k 16. __ ( ) Criminal Justice (court/jail) 17. __ ( ) Day carclnurscr)'schools IS. __ ( ) Drug stores/pharmacies* 19. __ ( ) Drych:aningltailoring* 20. __ ( ) Electronics store 21. __ ( ) Employment services 22. __ ( ) Fast food/toke out 23. __ ( ) Fire station* 24 __ ( ) Florist* 25. __ ( ) Funeral Home/Mortuary 26. __ ( ) Furniture store (new) 27. __ ( ) Furniture store (used) 28. __ ( ) Gas Station 29. __ ( ) Giflic:1rd shop* 30. __ ( ) Green gram/deli* 31. __ ( ) Grocery storclsupennarket* 32 __ ( ) Hardware/Home repairs 33. __ ( ) Health clinics* 34. __ ( ) Hospital* 35. __ ( ) laundromats 36. __ ( ) library (Public);!( 37. __ ( ) liquor stores 38. __ ( ) Manufacturing __ ( ) Movie House/Cinema 40. __ ( ) Parking lot (commercial) 41. __ ( ) Pawn shops 42 __ ( ) Police Station* 43. __ ( ) Post office* 44. __ ( ) Professional offices (MD, DDS, Lawyers, Accountants)* 45 __ ( ) Real Estate sales office 46. __ ( ) Restaurants* 4i. __ ( ) Rec/Fimcss Center/Gym* 45. __ ( ) Schools: Collcgcs/Univ. 49 __ ( ) Schools: Parochial/Private SO __ ( ) Schools: Public -elem 51 __ ( ) Schools: Public -jr.high 52 __ ( ) Schools: Public :-high school 53. __ ( ) Second hand :Stores 54 __ ( ) Sex entertainment shops 55. __ ( ) Senior CcntcrW 56. __ ( ) Shoe Repair* 57. __ ( ) Social Service organizations 58. __ ( ) Tauoo parlors. 59 __ ( ) Travel agents 60._. _( ) Utilities (gas, water, electric) 61 __ ( ) Video/ pool/ bowling 62. __ ( ) Yoga or Pilatcs studio 63. __ ( ) Other: 26. Parking at Destinations? ( all) No reserved D 1 On-street spots 02 Lot or garage adjacent 03 Lot or garase across street 04 '1.7 Number Parklnl! Soaces (e.g in destinations app;oxl: 1 21 > YorN 0 so so (I) _ill I)) (4) (6) shops services other 28. Bike parkine facilities? None Dt Bike Rack or locker D2 l9. Public Amenities ( all) None Stn:et LighiS Mail boll (for posting a lcucr) Public benches 0 0 0 0 Public phones .that works 0 Tnish cans (non-residential) D Walking patbsltr:lils 0 Signage al stores offering sr services D Signage at stores offering sr discounts 0 Other indicators? 0 Parks 30. Park None Dt Yes (private, residents only) 02 Yes (public) D3 Jl. Park Amenities ( all) walking lnlils D exercise circuit 0 lake with walking path around it 0 tennis c.ouns 0 other recreation (describe) 0 Community Garden 0 32. Cooditioo of park Poor/deteriorated D1 Fair condition D2 Exccl!cntlwcll kept 03 Pablic. Transportation 33. Is there.a bus or slop? None Do Bus or train 01 214 34. Condition or bus/train stop? Poor (no bench or shade plus not clean) D1 Fair (bench 21: shade, not both) D2 Excellent (clean wlbeneh & shade) D3 PEDESTRIAN FRIENDLINESS 35. Continuity of walk Winding succts, disjointed path D1 Long blocks, mostly direct routes D2 Short blocks, grid, direct routes D3 FUNCTIONALITY Sidewalk 36. Type of sidewalk None 01 I nlcrmillcnl 0 2 Continuous Dl 37. cond!Uon of sidewalk Poor (broken, cracked, weedy) 01 Fair (some bumps, crocks, etc) 01 Excellent (\'cry few 'problem areas) 03 Width 38. Sidewalk width less thilll 3fcct D 1 3 feet but less than s Dz >S' 03 Location 39. Sidewalk location Nexllo the road D 1 Within 3 fcct'from the curb 02 More than 3 .feet from the curb D3 Obstacles 40. ( all) Total obstiuction (need 10 step off sidewalk or into sticct) D1 Pimial 9bstruc1ion (need to walk around it) 02 None 03 Curbcuts 41. Curb cu(S I) black end I None 01 Unconnected with other cuts or sidewalks (forces you irito the street) 01 Connects both sides of street/walkway 03 2) block end 2 None 01 Uncorincc"d with other cuts or sidewalks (forces you into the street) ol Connects both sides of streel/walkway 03

PAGE 236

Slope 4l. Slope Steep slop.! 01 Moderate slope D, Flat or Gcnrlc D3 Shade or lrccs shadln.: paths None 01 Less than I per lor D, I OJ 44. A,eroge heighr ofrrccs lhc 'pot h) Snull (head high) D 1 (between head & roonop) o, Large (higher than roofiops) 03 SAFETY FROM TRAFFIC .S5. l'ostcc.l speed limil _____ 46. '1'1pc uf Su-ccr Two.w.ay street Oncw:ly street .S1. Number oflancs on rnatl >-l lanes 3-J lar\t::S 1 lanes -lS. control dc,lces Dr D. D; None 01 2way srop (or stop at .... inlcrscct) 02 Traffic light D3 stop or Roundabout D.s 49. Intersection Allis (" all) None 01 Cross walk or Island D3 Pcdl!$.lri::m signal D.a seconds between w:tlkh.lon'l walk .50. Dri\cr heh.a\"ior Did not obscr\'c 00 Reckless, ron light or slop sign Dr Sped or Failed to yield ro pedestrians D. \l'cllbeha\'Cd dri\'CfS D; WALK SUMMARY Safe from Traffic 51. llow safe from Ira me did you feel "alklng this sq:ntenr? Unsafe (busy street, high speeds. narrov or missing sidc\\'alks. difficult crossing) D1 Cautious (some ofrhc above issues) D, Vety safc(no issues) D3 Safe from Crime Sl. llow secure did )'OU feel wnlklng !his Jegment"! Threatened (scary dogs/people, or dcscnod area) Dr. W;uy, rc: surroundings, people D, Exlrcmcl)' secure 01 Aesthics 53. Ho,, aur:acthe would you r.alt ll1is segment ror walking Poor/Mostly unauracti\c 01 Fair/some nice sections 02 Exccllcn!Ncry anracti\'C D, Phvsical Challenges 54. !low physically would you r.llc Chis segment for wulklng? Very diflieulr Dr Moderately difficult D, Easy D3 215

PAGE 237

N .... 0'1 DATA COLLECTION Auditor:------------Date: Time: ___ am/pm 0 Street Segment: (Orig. ___ / to End / ____ R/L DATA ENTRY: Date: Rec # itt. itz. Q3. H RH MO A VL H RII MO A VL If RH MO A VL H RH MO A VL H RH MO A YL H RH MO A VL Biz Biz Biz Biz Biz Biz NonRes"'" Non Res"'" NonRes0111 NonRes016 NonRes0111 NonRes011o OW YIN OW YIN OW Y/N OW YIN OW YIN OW YIN Condition: E I rIP Condition: E I rIP Condition: E IF I P Condition: E IF I P Condition: E IF I P Condition: E IF I P Porch/Border usn low Porch/Border nlg 1 luw Porch/Border "S 1 low Porcb/Oordcr us 1 low Porch/Border low Porch/Border usn low Secur: B/ S/ Dog! A Secur: B/ S/ Dog/ A Sccur: B/ S/ Dog/ A Secur: B/ S/ Dog/ A Secur: ll/ S/ Dog/ A Secur: B/ S/ Dog/ A Surveill: G /Part/ P Surveill: G /Part/ P Surveill: G /Part/ P Surveill: G /Part/ P Surveill: G /Part/ P Surveill: G /Part/ P Cement/ Plantings Cement/ Plantings Cement/ Plantings Cement/ Plantings Cement/ l'lantings Cement/ Plantings Tree: s M L # Tree: s M 1. # __ Tree: s M L # Tree: s M L # Tree: s M L # Tree: s M L # Pkg: St D L G andct Pkg: St D L G audct Pkg: St D L .G uudet l'kg: St D L G 811dct ;;uPkg: St D L det Pkg: St D L G dct #spots __ m #spots __ !(II __ #spots __ !IJ #spots __ R --#spots __ g --#spots __ ---Bb rack: YIN &b rack: YIN Bb rack: YIN rack: Y IN Bb rack: YIN YIN FYDIGrounds Y I N FYD/Grounds Y I N FYDfGrounds Y IN FYDIGrounds Y IN F\'DIGrounds Y IN FYDIGrounds Y IN FYD Cond: E I F I P FYD Cond: E IF I P FYD Cond: E I F I P FYD Cond: E IF I P FYD Cond: E IF I P FYD Cond: ElF I P Signaac: For Sale (II ) For Rent (II ) Neigh Watch Sign: 0 Yes 0 No Graf: None Some A lot Lit: None Some A lot Park uu Pvt Trnils Ex Circuit Lake wlnath Tennis Other rcc ) People: None Workers <12y Adults Comonunl!Y_Gnrdcn NIA Pnrk Condition: E I r I P None Sl:lndiSiVrnrkcd Walk Run nikc l'lav Gorden Oth work Oth Social Amenities: None Slrccl Lls News dispenser Mail Dox Pub Bench broKe l ;;: Bus or Train Y I N Pub Ph works Public Trash Cans Walking Paths Sr Svccs (signs) Other Amenities: Condition: E/ Fl P Street 1-wo>' 2-w"y #Lanes 1-2 1 3 1 >4 Trnffic Posted Speed ___ Driver Beh NIA R:m Stor> Sr>cd !':ailed to yield to peds Wcll-llehaved Continuity of Wnlk: Winding disjointed I L:h1llsks. Speed Control None I 2-wny(or"T-intcrscction") el A!l-Wnys!l>(orRoundabout) rnostlr t.li[Ccl ruul!t:i I grid Crossing Aids None Crosswalk Median Walk II seconds to "Don't Walk" Sidewalk None lntcnniltcnt Continuous Width <)'wide )-5' wide >S' wide Loc Abuts Rd :;:3 from curb > 3 from curb Obstructions TOiall Part/ None Curb Cuts I None UIC I doesn' t connect wl om comer\ Connects ll None UIC Connects Slo1e steen nu>d eentlc-nat Sidewalk Condition E/ F/ P Walk Summnrv: Safe Traff: Unl C:wt I SMe Safe Crime: Thr 1 W:ory 1 Secure Aesthetics: E F I'. Challenges: F.nsy Mod Diff IJII .,..,..,, ..... ,,_. __ kll#llo"'' I .,,,,;,_, ._ ...,_., 'II ,,.. - II IJI,, : . , '-' -111 . ... ,,,,, o o : ... t 1--# '1 .. ,,.,1 .&..1:-1:1.-..-. .. .,.l oOoooo l n o'\ ..,,,, '' .. -t...l ................ ..,. 1 ... .... 1 ''"'" ,nd Cmulilifttt: F.xccllcnt wcll kcptfl:'air/l'nur-dctcrioratl-d. 1urt'f1 : 31"\:a l::argc enough for a (:.hove hc:ul) lluw (hcluw Srcnrity: llars/SiJ;ns (n..:is,h Sur, rlll:mcc Ge1o d (could nh.J l!lgs Liller 1-3 per 1\ lot >3 per lnl l'nrk (publiclpri\';;IIC) indic:11c an)' n.:crc.:nion or s;ocial O()(lOI1 uniliL-s (.: g community l1l'CIItll \ x Ac;1i\'itk-,. (whal (lC''fliC dning't). ltTn stcl cnndilion: E (clean. bench & shade) IF {bench or 1 (no bench; IIC'I shade) Sid''''"'lk Cundiliun: (min. problem ;m."'.1sY Fair (S\11\Jt: humps, cra..:h)tlt)(Jr (bmkcn. weedy). \Valk Summotn : Sar ... fnun Trilrlic: (bll:')'. pnor sid'-=\V'.J!ks, hard lu crC'ssV (slltnl! I Safe. Safl' rnun Crhn11.: 11trcah:ncd ppl or tlugs; tl..::-;crtcdJ I I Seem\: Acslhcllc-s : J-:a:;;y/:'t-Tudcr.&lc1y hard/J)iflicult

PAGE 238

APPENDIXB NEIGHBORHOOD AUDIT MAPS 217

PAGE 239

Neighborhood 1: Low Walk .. .c:: .. .......... ........... 0 700 Map Created By DaraHlcks Projection: liTM Zone 13 Datum: NAD 1983 Revised 4/18/05 1,400 2,800 4,200 218 5,600 N W+E s

PAGE 240

ca ID Neighborhood 2: High Walk A Colfax .., 1:1: 1:: ca 1:: .3 e 0 (.) 1 h 1:: ca ca "2 e ca QJ .r: en 1:: 1:: 12th (1:1 1:: 0 1:: 11th LO "E .c ;:; I! "' ca 1:: (!) u QJ "0 Cl 0 10th F 1 G R Cl Cl "C 1:: ca 0 M 9th 0 1:: 0 I!! 1:: 0 8th w 1:: :::J 0 I!: 7th B c .... ===.. ............ ............. Feet 0 500 1,000 2,000 3,000 219 Map Created By DaraHicks Projection: UIM Zone 13 Datum: NAD 1983 Revised 4/18/05 4,000 I i

PAGE 241

BffiLIOGRAPHY Aday, L. (1989). Designing and conducting health surveys: a comprehensive guide. San Francisco: Jossey-Bass. AHA. (2005). Metabolic Syndrome. Retrieved, from the World Wide Web: www.americanheart.org/presenter .jhtml ?identifier=4 7 56 Ainsworth, B. (2002). The Compendium of Physical Activities Tracking Guide. Prevention Research Center, Norman J. Arnold School of Public Health, University of South Carolina. Retrieved April 17, 2004, from the World Wide Web: Antonovsky, A. (1979). Health, Stress and Coping. San Francisco: Jossey-Bass, Inc. Bagley, S. P., Angel, R., Dilworth-Anderson, P., Liu, W., & Schinke, S. (1995). Adaptive health behaviors among ethnic minorities. Health Psycho/, 1 4(7), 632-'640. Bandura, A. (1986). Social foundations ofthought and action: A social cognitive theory. Englewood Cliffs, NJ, USA: Prentice-Hall, Inc. Bandura, A. (1994). Self-Efficacy. In V. S. Ramachaudran (Ed.), Encyclopedia of Human Behavior (Vol. 4, pp. 71-81). New York: Academic Press. Bandura, A. (1997). Self Efficacy: The Exercise of Control. New York: WH Freeman and Company. Bandura, A. (2001). Social cognitive theory: an agentic perspective. Annu Rev Psycho/; 52, 1-26. Bandura, A., & Locke, E. A. (2003). Negative self-efficacy and goal effects revisited. J Appl Psycho/, 88(1), 87-99. 220

PAGE 242

Bandura, A., & Wood, R. (1989). Effect of perceived controllability and performance standards on self-regulation of complex decision making. J Pers Soc Psycho/,. 56(5), 805-814. Baronowski, T., Perry, C. L., & Parcel, G. S. -(1997). How individuals, environments, and health behavior interact. -In F. M. L. Karen Glanz, Barbara K. Rimer (Ed.), Health Behavior and Health Education: Thepry, Research, and Practice (seconded., pp. 153-178). San Francisco: Jossey-Bass, Inc. Bassuk, S. Glass, T. A., & Berkman, L. F. (1999). Social disengagement and incident cognitive decline in community-dwelling elderly persons. Ann Intern Med, 131(3), 165-173. Bengoechea, E. G., Spence, J.C., McGannon, K.R. (2005). Gender differences in perceived environmental correlates of physica/activity. Intematiomll J oumal of Behavioral Nutrition and Physical Activity. Retrieved Dember 4, 2005, from the World Wide Web: http://www.ijbnpa.org/content/pdf/1479-5868-212.pdf Bergman-Evans, B. F. (1994). Alzheimer's and related disorders: loneliness, depression, and social support of spousal caregivers. J Gerontal Nurs, 20(3), 6-16. Berkman, L. F., & Glass, T. (2000). Social Integration, Social Networks, Social Support, and Health. In I. Kawachi (Ed.), Social Epidemiology (pp. 137-173). New York: Oxford University Press, Inc. Berkman, L. F., Glass, T., Brissette, 1., & Seeman, T. E. (2000). From social integration to health: Durkheim in the new millennium. Soc Sci Med, 51(6), 843-857. Beyene, Y., Becker, G., & Mayen, N. (2002). Perception of aging and sense of well being among Latino elderly. J Cross Cult Gerontal, 17(2), 155-172. 221

PAGE 243

Boeree, C. G. (1997). Erik Erikson: 1902-1994. Personality Theories. Retrieved, from the World Wide Web: http://www.ship.edu/-cgboeree/erikson.htrnl Bonneux, L., Barendregt, J. J., & Vander Maas, P. J. (1998). The expiry date of man: a synthesis of evolutionary biology and public health. J Epidemiol Community Health, 52(10), 619-623. Bourdieu, P. (1984). Distinction: A Social Critique of the Judgement of Taste. Cambridge, Massachusetts: Harvard University Press. Brofenbrenner, U. (1979). The ecology of human development: Experiments by nature and design. Cambridge, MA: Harvard Uiliversity Press. Brown, K. M. (1998, 1999). Ecological Models of Health Behavior and Health Promotion. University of South Florida, Community and Family Health. Retrieved, from the World Wide Web: http://www.med.usf.edu/-kmbrown/Ecological Models Overview.htm. Brownson, R. C., Chang, l J., Eyler, A. A., Ainsworth, B. E., Kirtland, K. A., Saelens, B. E., & Sallis, J. F. (2004). Measuring the Environment for Friendliness Toward Physical Activity: A Comparison of the Reliability of 3 Questionnaires. Am J Public Health, 94(3), 473-483. Bryant, L. L., Corbett, K. K., & Kutner, J. S. (2001). In their own words: a model of healthy aging. Soc Sci Med, 53(7), 927-941. Bryk, A. S., & Raudenbush, S. W. (1992). Hierarchical Linear Models: Applications and Data Analysis Methods. Newbury Park, CA: Sage Publications, Inc. Bull, S. S., & Shlay, J. C. (2005). Promoting "Dual Protection" from pregnancy and sexually transmitted disease--a social ecological approach. Health Promotion Practice, 6, 72-80. 222

PAGE 244

Bull, S. S., & Shlay, J. C. (In press). Promoting "Dual Protection" from pregnancy and sexually transmitted disease--a social ecological approach. Health Promotion Practice. Cacioppo, J. T., Poehlmann, K. M., Kiecolt-Glaser, J. K., Malarkey, W. B., Burleson, M. H., Berntson, G. G., & Glaser, R. (1998). Cellular immune responses to acute stress in female caregivers of dementia patients and matched controls. Health Psycho/, 17(2), 182-189. Cagney, K. A., Browning, C. R., & Wen, M. (2005). Racial disparities in self-rated health at older ages: what difference does the neighborhood make? J Gerontal B Psycho! Sci Soc Sci, 60(4), 8181-190. Caughy, M. 0., O'Campo, P. J., & Patterson, J. (2001). A brief observational measure for urban neighborhoods. Health Place, 7(3), 225-236. CCDOS. (2005). Department of Safety Web page. The City arid County ofDenver, Department of Safety. Retrieved, from the World Wide Web: http:/ /www.denvergov .org/Safety/ Chapman, L. (2000). Freud's Theory of Development: Tutorial. Child Development Wesite. Retrieved, from the World Wide Web: http://www.idealist.com/childrenlfreud.html Chin-Sang, V., & Allen, K. R. (1991). Leisure and the older black woman. J Gerontal Nurs, 17(1), 30-34. Clark, F., Azen, S. P., Zemke, R., Jackson, J., Carlson, M., Mandel, D., Hay, J., Josephson, K., Cherry, B., Hessel, C., Palmer, J., & Lipson, L. (1997). Occupational therapy for independent-living older adults. A randomized controlled trial. Jama, 278(16), 1321-1326. Clark, N. M., & Dodge, J. A. (1999). Exploring self-efficacy as a predictor of disease management. Health Educ Behav, 26(1), 72-89. 223

PAGE 245

Cohen, J. (1988). Statistical power analysis for the behavioral sciences (2nd ed.). Hillsdale, NJ.: Lawrence Earlbaurn Associates. Coleman, J. (1988). Social capital in the creation ofhurnan capital. American Journal of Sociology, 94(S94-S 120). Cromley, E. K., & McLafferty, S. L. (2002). GIS and Public Health. New York: The Guilford Press. Csikszentmihalyi, M. (1991). Flow: The Psychology of Optimal Experience. New York: Harper Perennial. Csikszentmihalyi, M. (1997). Finding Flow: The Psychology of Engagement with Everyday Life. Cutrona, C. E., Russell, D., & Rose, J. (1984). Psychosocial factors and physical and me_ntal health among the elderly. Unpublished manuscript. Cutrona, C. U., & Russell, D. (1987). Social Provisions Scale. Retrieved, from the World Wide Web: http://www.sph.unc.eduliprc/longscan/Measures%20Manual/ Ages5to 11 /m3 8. 12M . Dallosso, H. M., Morgan, K., Bassey, E. J., Ebrahim;S. B., Fentem, P. H., & Arie, T. H. (1988). Levels of customary physical activity among the old and the very. old living at home. J Epidemiol Community Health, 42(2), 121-127. De Bourdeaudhuij, 1., Sallis, J. F., & Saelens, B. E. (2003). Environmental correlates of physical activity in a sample of Belgian adults. Am J Health Promot, 18( 1 ), 83-92. de Jong Gierveld, J., & Kamphuis, F. H. (1985). The development of a Raschtype loneliness scale. Applied Psychological Measurement, 9, 289-299. 224

PAGE 246

DenverGovernment. (2002). Statistical Neighborhoods Compared to Registered Neighborhood Organizations (Vol. 2004): Denver City Conncil. Diez Roux, A. V. (2004). Estimating neighborhood health effects: the challenges of causal inference in a complex world. Social Science & Medicine, 58, 19531960. Diez-Roux, A. V. (1998). Bringing context back into epidemiology: .variables and fallacies in multilevel analysis. Am J Public Health, 88(2), 216-222. Dillman, D. A. (1991). The Design and Administration of Mail Surveys. Annual Review of Sociology, 17,225-249. Dunton, G. F., Jamner, M. S., & Cooper, D. M. (2003). Assessing the perceived environment among minimally active adolescent girls: validity and relations to physical activity outcomes. Am J Health Pro mot, 18(1 ), 70-73. Durkheim, E. (1897). Suicide (G. Simpson, Trans.). New York: Press. Dutton, R., Levy, L. L., & Simon, C. J. (1993). Frames of reference in occupational therapy: introduction. In H. D. Smith (Ed.), Willard and Spackman's Occupational Therapy (8th ed., pp. 62-65). Philadelphia: J.B. Lippincott Company. Dzewaltowski, D. A. (1997). The ecology of physical activity and sport: Merging science and practice. Journal of Applied Sport Psychology, 9, 254-276. Estabrooks, P. A., Lee, R. E., & Gyurcsik, N.C. (2003). Resources for physical activity participation: does availability and accessibility differ by neighborhood socioeconomic status? Ann Behav Med, 25(2), 100-104. Everard, K. M., Lach; H. W., Fisher, E. B., & Baum, M. C. (2000). Relationship of activity and social support to the functional health.of older adults. J Gerontol B Psycho/ Sci Soc Sci, 55(4), S208-212 .. 225

PAGE 247

Eyler, A. A. (2003). Personal, social, and environmental correlates of physical activity in rural Midwestern white women. Am J Prev Med, 25(3 Supp11 ), 86-92. Ferraro, K. F., & LaGrange, R. L. (1992). Are older people most afraid of crime? Reconsidering age differences in fear of victimization. J Gerontol, 47(5), S233-244. Fisher, K. J., Li, F., Michael, Y._, & Cleveland, M. (2004). Neighborhood-level influences on physical activity among older adults: a multilevel analysis. J Aging Phys Act, 12(1), 45-63. Fishman, P. A., Goodman, M. J., Hornbrook, M. C., Meenan, R. T., Bachman, D. J., & O'Keeffe Rosetti, M. C. (2003). Risk adjustment using automated ambulatory pharmacy data: the RxRisk model. Med Care, 41(1), 84-99. Fleis, J. (1981). Statistical methods for rates and proportions. New York: Wiley. Foler, J., Floyd J. (1988). Survey Research Methods (Vol. 1). Newbury Park, CA: Sage Publications, Inc. Fonda, S. J., Wallace, R. B., & Herzog, A. R. (2001). Changes in driving patterns and worsening depressive symptoms among older adults. J Gerontol B Psycho/ Sci Soc Sci, 56(6), S343-351. Forbes, A. (1996). Caring for older people. Loneliness. B.M.J., 313(7053), 352-354. Frank, L. D., Andresen, M.A., & Schmid, T. L. (2004). Obesity relationships with community design, physical activity, and time spent in cars. Am J Prev Med, 27(2), 87-96. Frank, L. D., Schmid, T. L., Sallis, J. F., Chapman, J., & Saelens, B. E. (2005). Linking objectively measured physical activity with objectively measured urban form: findings from SMARTRAQ. Am J Prev Med, 28(2 Suppl 2), 117-125. 226

PAGE 248

Fratiglioni, L., Wang, H. X., Ericsson, K., Maytan, M., & Winblad, B. (2000). Influence of social network on occurrence of dementia: a community-based longitudinal study. Lancet, 355(9212), 1315-1319. Gaines, A. (1989). Alzheimer's disease in the context of Black southern culture. Health Matrix, 6, 33-38. Garcia Bengoechea, E., Spence, J. C., & McGannon, K. R. (2005). Gender differences in perceived environmental correlates of physical activity./nt J Behav Nutr Phys Act, 2, 12. Garfein,"A. J., & Herzog, A. R. (1995). Robust aging among the young-old, old-old, and oldest-old. J Gerontol B Psycho[ Sci Soc Sci, 50(2), S77-87. getzips.com. (200S). ZipExpress: The Zip Code and Area Code Authority. Insight Software Solutions. Retrieved, from the World Wide Web: www.getzips.com Glass, T.A., de Leon, C. M., Marottoli, R. A., & Berkman, L. F. (1999). Population based study of social and productive activities as predictors of survival --among elderly Americans. B.M.J., 319(1208), 478-483. Gustafson, D. H., McTavish, F., Hawkins, R., Pingree, S., Arora, N., Mendenhall, J., & Simmons, G. E. (1998). Computer support for elderly women with breast cancer. Jama, 280(15),.1305. Hawe, P., & Shiell, A. (2000). Social capital and health promotion: a review. Soc Sci Med, 51(6), Heaney, C. A., & Israel, B. A. (1997). Social Networks and Social Support. In B. K. Rimer (Ed.), Health Behavior and Health Education: Theory Research and Practice (2nd ed., pp. 179-205). San Francisco: Jossey-Bass. Herzog, A. R., Ofstedal, M. B., & Wheeler, L. M. (2002). Social engagement and its relationship to health. Clin Geriatr Med, j 8(3), 593-609, ix. 227

PAGE 249

Hofmann, D. A. (2004). Issues in Multilevel Research: Theory Development, Measurement and Analysis. InS. Rogelberg (Ed.), Handbook of Research Methods in Industrial and Organizational Psychology. Oxford: Blackwell. Hopkins, H. L., & Smith, H. D. (1993). Scope of occupational therapy. In H. D. Smith (Ed.), Willard & Spackman's Occupational Therapy (8th ed., pp. 3-4). Philadelphia: J.B. Lippincott .. Jacquez, G. (1998). GIS as an enabling technology. In M. Loytonen (Ed:), GIS and Health (pp. 17-28). Philadelphia: Taylor and Francis, Inc. Kawachi, I., & Berkman, L. F. (2000). Social Cohesion, Social Capital, and Health. In I. Kawachi (Ed.), Social Epidemiology (pp. 174-190). New York: Oxford University Press, Inc. Kb.aw, K. T. (1997). Healthy aging. B.MJ., 315(7115), 1090-1_096. Kielhofner, G. (1997). Conceptual Foundations of Occupational Therapy (2nd ed.). Philadelphia: F :A. Davis Company. Kielhofner, G., Burke, J.P., & Igi, C. H. (1980). A model ofhurnan occupation, part 4. Assessment and intervention. American Journal of Occupational Therapy, 34(12), 777-788. Kirtland, K. A., Porter, D. E., Addy, C. L., Neet, M. J., Williams, J.E., Sharpe, P. A., Neff, L. J., Kimsey, C. D., Jr., & Ainsworth, B. E. (2003). Environmental measures of physical activity supports: perception versus reality. Am J Prev Med, 24(4), 323-331. Krause, N. (1993). Neighborhood deterioration and social isolation in later life. International Journal of Aging and Human Development, 36, 9-36. Krause, N. (1996). Neighborhood deteriorationand self-rated health in later life. Psychology and Aging, 11, 342-352. 228

PAGE 250

Krenichyn, K. (2005). 'The only place to go and be in the city': women talk about exercise, being outdoors, and the meanings of a large urban park. Health Place. Langlois, J. A., Keyl, P.M., Guralnik, J. M., Foley, D. J., Marottoli, R. A., & Wallace, R. B. ( 1997). Characteristics of older pedestrians who have difficulty crossing the street. Am J Public Health, 87(3), 393-397. Lee, R. E., & King, A. C. (2003). Discretionary time among older_adults: how do physical activity promotion interventions affect sedentary and active behaviors? Ann Behav Med, 25(2), 112-119. C., & Silverstein, M. (2001): Does eQ.gagement with life enhance survival of elderly people in Sweden? The role of social and leisure activities. J Gerontal B Psycho/ Sci Soc Sci, 56(6), S335-342. Letvak, S. (1997). Relational experiences of elderly women living alone in rural communities: !l phenomenologic inquiry. J NY State Nurses Assoc. 28(2), 20-25. Li, F., Fisher, J., & Brownson, R. C. (2005). A multilevel analysis of change in neighborhood walking activity in older adults. J Aging Phys Act, 13(2), 145159. . Li, F., Fisher,K. J., Bauman, A., Ory, M.G., Chodzko-Zajko, W., Harmer, P., Bosworth, M., & Cleveland, M. (2005). Neighborhood influences on physical activity in middle-aged and older adults: a multilevel perspective. J Aging Phys Act, 13(1 ), 87-114. Lopez-Zetina, J., Lee, H., & Friis, R. (2005). The link between obesity and the built environment. Evidence from an ecological analysis of obesity and vehicle. miles oftravel in California. Health Place. Macinko, J., & Starfield, B. (2001). The utility of social capital in research on health determinants. Milbank Q, 79(3), 387-427, IV. 229

PAGE 251

mapquest.com. (2005). Map Generator Program. Retrieved, from the World Wide Web: http://www.mapguest.com Mannot, M. (2000). Multilevel Approaches to Understanding Social Determinants. In I. Kawachi (Ed.), Social Epidemiology (pp .. 349-367). New York: Oxford University Press, Inc. McAuley, E., Jerome, G. J., Marquez, D. X., Elavsky, S., & Blissmer, B. (2003). Exercise self-efficacy in older adults: affective, and behavioral influences. Ann Behav Med, 25(1), 1-7. McMillan, D. W., & Chavis, D. M. (1986). Sense of community: a definition and theory. Journal o[Community Psychology, 14,6-23. McTavish, F. M., Gustafson, D. H., Owens, B. H., Hawkins, R. P., Pingree, S., Wise, M., Taylor, J. 0., & Apantaku, F. M. (1995). CHESS (Comprehensive Health Enhancement Support System): an interactive cQmputer system for women with breast cancer piloted with an underserved population. J Ambulatory Care Manage, 18(3), 35-41. Mendes deLeon, C. F., Glass, T. A., & Berkman, L..F. (2003). Social engagement and in a community population of older adults: the New Haven EPESE. Am J Epidemiol, 157(7), 633-642. Mercer, S. 0. (1996). Navajo elderly people in a reservation nursing home: admission predictors and culture care practices. Soc Work, 41(2), 181-189. Michael, Y. L., Green, M. K., & Farquhar, S. A. (2005). Neighborhood design and active aging. Health Place. Morrow, V. (1999). Conceptualising social capital in relation to the well-being of children and young people: a critical review, The Sociological Review (pp. 744,;765). Oxford, UK: Blackwell Publishers. 230

PAGE 252

Mullins, L. C., Elston, C. H., & Gutkowski, S.M. (1996). Social determinants of loneliness among older Americans. Genet Soc Gen Psycho! Monogr, 122(4), 453-473. Newsom. (2001). Testing Mediation with Regression Analysis. Retrieved, from the World Wide Web: www.upa.pdx.edu/I0Nnewsom/usp534/ho rnediation.doc Oakes, J. M. (2004). The (mis)estirnation of neighborhood effects: causal inference for a practicable social epidemiology. Social Science & Medicine, 58, 19291952. Patrick, D. L., & Wickizer, T. M. (1995). Community and Health. fu D. C. Walsh (Ed.), Society and Health (pp. 46-92). New York: Oxford University Press, Inc. Paul, M. (1993). A little help, a lot of independence. Retirement community advance.s the cause of aging with dignity. Health Prog, 74(4), 43-45. Penninx, B. W., van Tilburg, T., Kriegsman, D. M., Deeg, D. J., Boeke, A. J., & van Eijk, J. T. (1997). Effects of social support and personal coping resources on mortality in older age: the Longitudinal Aging Study Amsterdam. Am J Epidemiol, 146(6), 510-519. Pikora, T. J., Bull, F. C., Jamrozik, K., Knuirnan, M., Giles-Corti, B.,& Donovan, R. J. (2002). Developing a reliable audit instrument to measure the physical environment for physical activity. Am J Prev Med, 23(3), 187-194. PitonFoundation. (2004). Neighborhood Facts. A joint project of The Piton Foundation and the Community Planning and Development Agency of the City and County of Denver. Retrieved, fi:orn the World Wide Web: http:/ /www.piton.org Porter, E. J. (1994). Older widows' experience of living alone at horne. Image J Nurs Sch, 26(1), 19-24. 231

PAGE 253

Portney, L., & Watkins, M. (1993). Foundations of clinical research: applications to practice. Norwalk, CT: Appleton Y Lange. Preacher, K. J., & Hayes, A. F. (2004). SPSS and SAS procedures for estimating indirect effects in simple mediation models. Behavior Research Methods, Instruments, & Computers, 3 6( 4 ), 717-731. Putnam, R. D. (1995). Bowling alone: America's declining social capital. Journal of Democracy, 6(1), 65-78. Rabin, B. S. (2000). Social networks and dementia. Lancet, 356(9223), 76-77. Rakowski, W. (1997). Health behavior in the elderly. In D. Gochman (Ed.), Handbook of Health Behavior Research III: Demography, Development, and Diversity. New York: Plenum Press. Rane-Szostak, D., & Herth, K. A. (1995). A new perspective on loneliness in later life. Issues Ment Health Nurs, 16(6), 583-592. Rosnow, R. L., & Rosenthal, R. (1996). Computing contrasts, effect sizes, and countemulls on other people's published data: General procedures for research consumers. Pyschological Methods, 1, 331-340. Rushton, G. (2000). Guest Editorial: GIS to improve public health. Transactions in GIS, 4, 1-4. Russell, D. W., Cutrona, C. E., de la Mora, A., & Wallace, R. B. (1997). Loneliness and nursing home admission among rural older adults. Psycho/Aging, 12( 4), 574-589. Saelens, B. E., Sallis, J. F., Black, J. B., & Chen, D. (2003). Neighborhood-based differences in physical activity: an environment scale evaluation. Am J Public Health, 93(9), 1552-1558. 232

PAGE 254

Saelens, B. E., Sallis, J. F., & Frank, L. D. (2003). Environmental correlates of walking and cycling: findings from the transportation, urban design, and planning literatures. Ann Behav Med, 25(2), 80-91. Sallis, J. F., & Owen, N. (1996). Ecological models. In B. K. Rimer (Ed.), Health Behavior and Health Education: Theory, Research, and Practice (second ed., pp. 403-424). San Francisco: Jossey-Bass. Sampson, R. J., Raudenbush, W., & Earls, F. (1997). Neighborhoods and violent crime: a multilevel study of collective efficacy. Science, 277(5328), 918-924. SAS Institute, I. (2003). SAS OnlineDoc 9.1.tJnpublished manuscript, Cary, NC. Seeman, T. E. (1996). Social ties and health: the benefits of social integration. Ann Epidemiol, 6( 5), 442-451. Senecal, C., Nouwen, A., & White, D. (2000). Motivation and dietary self-care in adults with diabetes: are self-efficacy and autonomous self-regulation complementary or competing constructs? Health Psycho!, 19(5), 452-457. Shephard, R. J. (1997). Aging, Physical Activity and Health. Champaign, IL: Human Kinetics Publishers. Stemler, S. E. (2004). A comparison of consensus, consistency, and measurement approaches to estimating interrater reliability. Practical Assessment, Research & Evaluation. Retrieved September 5, 2005, from the World Wide Web: http:/ IP AREonline.net/getvn.asp?v=9&n=4 Stewart, A. L., Mills, K. M., King, A. C., Haskell, W. L., Gillis, D., & Ritter, P. L. (2001). CHAMPS physical activity questionnaire for older adults: outcomes for interventions. Med Sci Sports Exerc, 3 3(1), 1126-1141. Stewart, A. L., Mills, K. M., Sepsis, P. G., King, A. C., McLellan, B. Y., Roitz, K., & Ritter, P. L. (1997). Evaluation of CHAMPS, a physical activity promotion program for older adults. Ann Behav Med, 19(4), 35l-361. 233

PAGE 255

Stewart, M. J. (1989). Social support: diverse theoretical perspectives. Soc Sci Med, 28(12), 1275-1282. Sugisawa, H., Liang, J., & Liu, X. (1994). Social networks, social support, and. mortality among older people in Japan. J Gerontal, 49(1 ), S3-13. Takano, T., Nakamura, K., & Watanabe, M. (2002). Urban residential environments and senior citizens' longevity in megacity areas: the importance ofwalkable green spaces. J Epidemiol Community Health, 56(12), 913-918. Tilvis, R. S., Pitkala, K. H., Jolkkonen, J., & Strandberg, T. E. (2000). Social networks and dementia. Lancet, 356(9223), 77-78. UCLA. (2005). SPSS FAQ Kappa statistic for variables with unequal score ranges. Academic Technology Services. Retrieved September 5, 2005, from the World Wide Web: http://www.ats.ucla.edu/stat/spss/fag/kappa.htm Urbaniak,. G. C., & Pious, S. (2005). Research Randomizer [website]. Retrieved; from the World Wide Web: www.randomizer.org Waitzkin, H. (1988). Health. Care, Social Contradictions, and the Dilemmas of Reform, The Second Sickness (pp. 3-43). New York: Free Press. Walker, D., & Beauchene, R. E. (1991). The relationship ofloneliness, social isolation, and physical health to dietary adequacy of independently living elderly. JAm Diet Assoc, 91(3), 300-304. Williams, E. E., & Ellison, F. (1996). Culturally informed social work practice with American Indian clients: guidelines for non-Indian social workers. Soc Work, 41(2), 147-151. Williams, G. C., Rodin, G. C., Ryan, R. M., Grolnick, W. S., & Deci, E. L. (1998). Autonomous regulation and long-term medication adherence in adult outpatients. Health Psycho/, 17(3), 269-276. 234

PAGE 256

Ziegler, R., & Mitchell, D. B. (2003). Aging and fear of crime: an experimental approach to an apparent paradox. Exp Aging Res, 29(2), 173-187. 235