UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Walk the Talk : the association between older adult mobility and the built environment Chudyk, Anna M. 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2017_february_chudyk_annamaria.pdf [ 3.88MB ]
Metadata
JSON: 24-1.0339983.json
JSON-LD: 24-1.0339983-ld.json
RDF/XML (Pretty): 24-1.0339983-rdf.xml
RDF/JSON: 24-1.0339983-rdf.json
Turtle: 24-1.0339983-turtle.txt
N-Triples: 24-1.0339983-rdf-ntriples.txt
Original Record: 24-1.0339983-source.json
Full Text
24-1.0339983-fulltext.txt
Citation
24-1.0339983.ris

Full Text

 WALK THE TALK: THE ASSOCIATION BETWEEN OLDER ADULT MOBILITY AND THE BUILT ENVIRONMENT by  Anna M. Chudyk  B.H.Sc., The University of Western Ontario, 2006 M.Sc., The University of Western Ontario, 2008  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY in THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES   (Experimental Medicine)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  November 2016  © Anna M. Chudyk, 2016     iiAbstract Mobility is vital to healthy aging. The built environment is central to mobility as it is the setting where mobility occurs. Older adults of low socioeconomic status (SES) may be especially reliant on built environments that support non-motorized mobility. Despite this, older adults of low SES are underrepresented in research.  Therefore, this dissertation applies quantitative methods to describe the mobility (capacity and enacted function) and investigate the association between the built environment and non-motorized mobility (walking for transportation and physical activity) of older adults of low income. The studies within this dissertation are set in Metro Vancouver; the last three utilize data from a cross-sectional study of 161 older adults of low income.   The first study investigates the reliability of a novel approach (virtual audits) for measuring the built environment. I found that virtual audits reliably measure macroscale (neighbourhood-level) built environment features that promote older adult walking, but may be inappropriate for measuring fine-grain details of the microscale (street-level) built environment.  The second study provides an in-depth description of the mobility of older adults living on low income. I noted that participants generally had the capacity to be mobile and made a high proportion of walking trips, although these did not together serve to meet physical activity guidelines for most.     iiiThe third study analyzes travel diary data to identify destinations that older adults most commonly visit in a week (i.e., grocery stores, malls, restaurants/cafes); I also found that each 10-point increase in Street Smart Walk Score (measure of the built environment) was associated with a 20% increase in walking for transportation (IRR = 1.20, 95% CI = 1.12, 1.29).  The fourth study analyzes accelerometry and self-report data to investigate associations between the built environment and physical activity and walking for transportation. Odds of any walking for transportation were 1.45 (95% CI = 1.18, 1.78) times greater for each 10-point increase in Street Smart Walk Score. There were no other built environment-mobility associations.   Taken collectively, these studies fill methodological gaps in the literature and provide data on an understudied population that may be especially reliant on built environments that support walking.         ivPreface This dissertation is an original intellectual product of the author, Anna M. Chudyk. Chapters 3-6 of this dissertation are expansions of stand-alone manuscripts for publication in the peer-reviewed academic literature. Two have already been published (Chapters 3 and 5) and a third has been accepted for publication (Chapter 4); a fourth has been submitted for publication and is currently undergoing peer-review (Chapter 6). As first author, I led each of these chapters. I provide the details of my contributions, and those of my collaborators, for each publication below. UBC’s Clinical Research Ethics Board (certificate: H10-02913) approved the work presented in Chapters 4-6.   Chapter 3: A version of this material was published as Chudyk, A. M., Winters, M., Gorman, E., McKay, H. A., & Ashe, M. C. (2014). Agreement between virtual and in-the-field environmental audits of assisted living sites. Journal of Aging and Physical Activity, 22(3), 414-420. doi: 10.1123/japa.2013-0047.  As lead author of this publication I was responsible for all major areas of concept development and study planning, data collection, analysis and presentation of findings, and manuscript writing. Ashe and Winters were supervisory authors and guided all aspects of the research, including contributing to concept development, design, approach, presentation and edits to the manuscript. Gorman was involved in the project from concept development to study planning, data collection, and provided manuscript edits. McKay was involved in concept development and contributed to manuscript edits. My overall contribution: 90%.     vChapter 4: A version of this material has been accepted for publication (July 13, 2016) as Chudyk, A.M, Sims-Gould, J., Ashe, M.C., Winters, M., and McKay, H.A. (2017). Walk the Talk: Characterizing mobility in older adults living on low income. Canadian Journal on Aging; 36(2).  As lead author of this publication I was responsible for all major areas of concept development, data collection, analysis and presentation of findings, and manuscript writing. McKay and Sims-Gould were the supervisory authors and guided all aspects of the research, including contributing to concept development, design, approach, presentation and edits to the manuscript. Ashe and Winters were involved in concept development and contributed to manuscript edits. My overall contribution: 90%.   Chapter 5: A version of this material was published as Chudyk, A. M., Winters, M., Moniruzzaman, M., Ashe, M. C., Sims-Gould, J., & McKay, H. A. (2015). Destinations matter: The association between where older adults live and their travel behavior. Journal of Transport & Health, 2(1), 50-57. doi: 10.1016/j.jth.2014.09.008.  As lead author of this publication I was responsible for all major areas of concept development, data collection, analysis and presentation of findings, and manuscript writing. McKay and Winters were the supervisory authors and guided all aspects of the research, including contributing to concept development, design, approach, presentation and edits to the manuscript. Moniruzzaman, Ashe, and Sims-Gould were involved in concept development and contributed to manuscript edits. My overall contribution: 90%.    viChapter 6: A version of this material will be submitted for publication as Chudyk, A.M, McKay, H.A., Winters, M., Sims-Gould, J., and Ashe, M.C. Can older adults’ neighbourhood walkability promote a walk to the shops?   As lead author of this publication I was responsible for all major areas of concept development, data collection, analysis and presentation of findings, and manuscript writing. McKay and Ashe were the supervisory authors and guided all aspects of the research, including contributing to concept development, design, approach, presentation of findings and edits to the manuscript. Winters and Sims-Gould were involved in concept development and contributed to manuscript edits. My overall contribution: 90%.      viiTable of contents Abstract .......................................................................................................................................... ii Preface ........................................................................................................................................... iv Table of contents ......................................................................................................................... vii List of tables................................................................................................................................ xvi List of figures .............................................................................................................................. xix List of abbreviations ....................................................................................................................xx Glossary ..................................................................................................................................... xxii Acknowledgements ....................................................................................................................xxv Dedication ................................................................................................................................ xxvii Chapter  1: Introduction, literature review, rationale, objectives, and hypotheses ................1 1.1 Introduction ........................................................................................................................ 1  The Canadian aging population .................................................................................. 1  Mobility as a key determinant of healthy and active aging ........................................ 2  The link between the built environment and older adult mobility .............................. 3  Mobility is influenced by the person-environment interaction ................................... 6 1.1.4.1 The Press-Competence Model ............................................................................. 6 1.1.4.2 Framework of person-level variables that influence older adult mobility ........... 9  SES and older adult mobility .................................................................................... 11  Summary ................................................................................................................... 12  Dissertation aims and outline .................................................................................... 13 1.2 Literature review .............................................................................................................. 15  How mobility (travel behaviour and physical activity) is measured ........................ 15   viii1.2.1.1 Measureable dimensions of travel behaviour .................................................... 15 1.2.1.2 Approaches used to measure travel behaviour ................................................... 16 1.2.1.2.1 Travel diaries .............................................................................................. 17 1.2.1.3 Measureable dimensions of physical activity .................................................... 18 1.2.1.4 Approaches used to measure physical activity .................................................. 21 1.2.1.4.1 Self-report questionnaires ........................................................................... 21 1.2.1.4.2 Accelerometry ............................................................................................. 22  How the built environment is measured ................................................................... 24 1.2.2.1 Measureable dimensions of the built environment ............................................ 24 1.2.2.2 Approaches used to measure the built environment .......................................... 27 1.2.2.2.1 Objective approach: GIS- based measures ................................................. 27 1.2.2.2.2 Self-report approach: Perceived measures .................................................. 28 1.2.2.2.3 Observational approach: Audits .................................................................. 29  Existing literature on the association between the built environment and older adult physical activity and walking for transportation ................................................................... 31 1.2.3.1 Recent systematic reviews ................................................................................. 31 1.2.3.2 Recent realist review .......................................................................................... 35 1.2.3.2.1 Overview of relevant studies published since the most recent systematic reviews ..................................................................................................................... 37 1.2.3.3 Studies of the association between the built environment and physical activity (as measured by accelerometry) ........................................................................................ 38 1.2.3.3.1 Associations between composite indices of walkability and MVPA ......... 43   ix1.2.3.3.2 Associations between objective measures of the built environment and MVPA ..................................................................................................................... 45 1.2.3.3.3 Associations between perceived measures of the built environment and MVPA ..................................................................................................................... 45 1.2.3.3.4 Associations between composite indices of walkability and light physical activity ..................................................................................................................... 45 1.2.3.3.5 Associations between objective and perceived measures of the built environment and total physical activity volume ........................................................... 46 1.2.3.3.6 Summary of findings................................................................................... 47 1.2.3.3.7 Studies of the association between the built environment and walking for transportation (as measured by self-report) .................................................................. 48 1.2.3.3.8 Associations between composite indices of walkability and duration of walking for transportation ............................................................................................. 57 1.2.3.3.9 Associations between audit measures of the built environment and duration of walking for transportation ........................................................................................ 58 1.2.3.3.10 Associations between perceived measures of the built environment and duration of walking for transportation .......................................................................... 58 1.2.3.3.11 Associations between the built environment and frequency of walking for transportation ................................................................................................................ 59 1.2.3.3.12 Summary of findings................................................................................. 60 1.2.3.4 Overall summary of the findings and direction for future research ................... 61  Aim, rationale, and objectives .................................................................................. 62   x1.2.4.1 Study 1. Agreement between virtual and in-the-field environmental audits of Assisted Living sites ......................................................................................................... 62 1.2.4.2 Study 2. Walk the Talk: Characterizing mobility of older adults living on low income ............................................................................................................................ 65 1.2.4.3 Study 3. Destinations matter: The association between where older adults live and their travel behaviour ................................................................................................. 66 1.2.4.4 Study 4. Can older adults’ neighbourhood walkability promote a walk to the shops ............................................................................................................................ 68 Chapter  2: Methods ....................................................................................................................70 2.1 Walk the Talk overview ................................................................................................... 70  Design of Walk the Talk: Engaging community stakeholders ................................. 70  BC Housing as key community partner .................................................................... 71  Integrated knowledge translation: approach to collaboration with BC Housing ...... 71  The Shelter Aid for Elderly Renters (SAFER) program ........................................... 72  The income of SAFER recipients relative to markers of low SES in Canada .......... 73 2.2 Study design ..................................................................................................................... 74  Setting ....................................................................................................................... 74  Inclusion and exclusion criteria ................................................................................ 75  Pilot study ................................................................................................................. 76  Sampling approach for main study ........................................................................... 76  Recruitment ............................................................................................................... 77 2.2.5.1 Recruitment target .............................................................................................. 77 2.2.5.2 Recruitment process overview ........................................................................... 78   xi2.2.5.3 Follow-up recruitment phone calls .................................................................... 79  Informed consent process ......................................................................................... 80  Incentives .................................................................................................................. 81  Data collection .......................................................................................................... 81 2.2.8.1 Data collection overview ................................................................................... 81 2.2.8.2 Measures of mobility (travel behaviour and physical activity) ......................... 82 2.2.8.2.1 Travel behaviour – travel diaries ................................................................ 82 2.2.8.2.2 Physical activity – accelerometry ............................................................... 83 2.2.8.2.3 Physical activity – self-report questionnaire ............................................... 84 2.2.8.3 Measures of person and environment-level characteristics of participants ....... 84 2.2.8.4 Environment-level variables .............................................................................. 85 2.2.8.4.1 Built environment – GIS-based measure .................................................... 85 2.2.8.4.2 Built environment – perceived measure ..................................................... 88 2.2.8.4.3 Social environment ..................................................................................... 89 2.2.8.5 Person-level variables ........................................................................................ 90 2.2.8.5.1 Sociodemographic information ................................................................... 90 2.2.8.5.2 Cognitive domain ........................................................................................ 91 2.2.8.5.3 Physical domain .......................................................................................... 91 2.2.8.5.4 Psychosocial domain ................................................................................... 92  Data entry .................................................................................................................. 93  Statistical analysis ................................................................................................... 94 2.2.10.1 Absolute agreement ......................................................................................... 94 2.2.10.2 Linear regression .............................................................................................. 95   xii2.2.10.3 Logistic regression ........................................................................................... 96 2.2.10.4 Poisson regression ............................................................................................ 98 2.2.10.5 Negative Binomial regression .......................................................................... 99 Chapter  3: Agreement between virtual and in-the-field environmental audits ..................101 3.1 Introduction .................................................................................................................... 101 3.2 Methods.......................................................................................................................... 105  Sampling of segments ............................................................................................. 106  Audits ...................................................................................................................... 106 3.3 Statistical analysis .......................................................................................................... 107 3.4 Results ............................................................................................................................ 108 3.5 Discussion ...................................................................................................................... 115 3.6 Conclusions .................................................................................................................... 118 Chapter  4: Walk the Talk: Characterizing mobility in older adults living on low income119 4.1 Introduction .................................................................................................................... 119 4.2 Methods.......................................................................................................................... 123  Measures of person-level characteristics ................................................................ 125  Measures of environment-level characteristics ....................................................... 126 4.2.2.1 Social environment domain ............................................................................. 126 4.2.2.2 Built environment domain ............................................................................... 127  Measures of participants’ mobility ......................................................................... 128 4.2.3.1 Physical activity ............................................................................................... 128 4.2.3.2 Travel behaviour .............................................................................................. 129 4.3 Statistical analysis .......................................................................................................... 129   xiii4.4 Results ............................................................................................................................ 130  Person and environment-level characteristics of participants ................................. 132 4.4.1.1 Sociodemographic information ........................................................................ 132 4.4.1.2 Cognitive domain ............................................................................................. 136 4.4.1.3 Physical domain ............................................................................................... 136 4.4.1.4 Psychosocial domain ........................................................................................ 136 4.4.1.5 Social environment domain ............................................................................. 137 4.4.1.6 Built environment domain ............................................................................... 138  Mobility................................................................................................................... 139 4.4.2.1 Physical activity ............................................................................................... 140 4.4.2.2 Travel behaviour .............................................................................................. 141 4.5 Discussion ...................................................................................................................... 142 4.6 Conclusions .................................................................................................................... 147 Chapter  5: Destinations matter: The association between where older adults live and their travel behaviour .........................................................................................................................149 5.1 Introduction .................................................................................................................... 149 5.2 Methods.......................................................................................................................... 151  Study measures ....................................................................................................... 151 5.2.1.1 Independent variables ...................................................................................... 152 5.2.1.2 Travel behaviour .............................................................................................. 152 5.2.1.3 Destinations...................................................................................................... 153 5.3 Statistical analysis .......................................................................................................... 154 5.4 Results ............................................................................................................................ 156   xiv5.5 Discussion ...................................................................................................................... 163 5.6 Conclusions .................................................................................................................... 167 Chapter  6: Can older adults’ neighbourhood walkability promote a walk to the shops ...168 6.1 Introduction .................................................................................................................... 168 6.2 Methods.......................................................................................................................... 171  Study measures ....................................................................................................... 171 6.2.1.1 Independent variables ...................................................................................... 171 6.2.1.2 Built environment domain ............................................................................... 172 6.2.1.3 Social environment domain ............................................................................. 172 6.2.1.4 Physical domain ............................................................................................... 172 6.2.1.5 Psychosocial domain ........................................................................................ 173 6.2.1.6 Sociodemographic factors ................................................................................ 173  Outcome measures .................................................................................................. 173 6.2.2.1 Physical activity ............................................................................................... 173 6.2.2.2 Self-reported walking for transportation .......................................................... 174 6.3 Statistical analysis .......................................................................................................... 175 6.4 Results ............................................................................................................................ 178 6.5 Discussion ...................................................................................................................... 193 6.6 Conclusions .................................................................................................................... 198 Chapter  7: Integrated discussion .............................................................................................200 7.1 Overview of findings and their contribution to my field of research ............................ 200 7.2 Points to ponder and future directions: .......................................................................... 204  Recruiting older adults of low SES ......................................................................... 204   xv Defining low SES ................................................................................................... 206  The person-environment fit..................................................................................... 208  Measuring the built environment ............................................................................ 209  Measuring walking for transportation ..................................................................... 210  The association between physical activity and walking for transportation ............ 212 7.3 Strengths and limitations................................................................................................ 213  Strengths ................................................................................................................. 213  Limitations .............................................................................................................. 214 7.4 Final conclusions ........................................................................................................... 216 Bibliography ...............................................................................................................................219  : Recruitment mail out package and consent form .......................................... 257  : On-site data collection booklet ......................................................................... 268  : Take-home data collection booklet .................................................................. 308  : Additional data for Chapter 3 ......................................................................... 345  : Additional data for Chapter 4 .......................................................................... 349  : Additional data for Chapter 5 .......................................................................... 355  : Additional data for Chapter 6 ......................................................................... 360      xviList of tables Table 1.1. Measureable dimensions of travel behaviour .............................................................. 16 Table 1.2. Measureable dimensions of physical activity .............................................................. 19 Table 1.3. Measureable dimensions of the built environment ...................................................... 25 Table 1.4. Measureable dimensions of urban design features that influence walking and cycling....................................................................................................................................................... 26 Table 1.5. Studies that investigated the association between the built environment and older adult physical activity (as measured by accelerometry) ........................................................................ 39 Table 1.6. Studies that investigated the association between the built environment and older adult walking for transportation (as measured by self-report). .............................................................. 49 Table 2.1. Eligibility criteria for the Shelter Aid for Senior Renters (SAFER) program ............. 73 Table 2.2. 2011 Low income cut-offs by size of family unit and community .............................. 74 Table 2.3. Name and location of data collection sites .................................................................. 81 Table 2.4. Comparison of Walk Score and Street Smart Walk Score methodology .................... 87 Table 2.5. Example contingency table of binary ratings by two observers .................................. 94 Table 3.1. Prevalence of select built environment features related to older adults’ walking at four Assisted Living sites ................................................................................................................... 112 Table 3.2. Comparison of virtual vs. in-the-field audits by rater ................................................ 114 Table 4.1. Select person-level, environment-level and mobility measures used in the study .... 124 Table 4.2. Descriptive statistics for select sociodemographic characteristics ............................ 132 Table 4.3. Descriptive statistics for select measures by domain of mobility .............................. 134 Table 4.4. Distribution of participants (n) across strata of Walk Score ...................................... 139 Table 4.5. Descriptive statistics for select mobility outcomes .................................................... 140   xviiTable 5.1. Descriptive statistics for select sociodemographic and mobility characteristics and bivariate association between each variable and number of walking for transportation trips (average/day)............................................................................................................................... 158 Table 5.2. Estimates from negative binomial regression analyses for number of walking for transportation trips (average/day) ............................................................................................... 162 Table 6.1. Descriptive statistics for select characteristics, by gender ......................................... 179 Table 6.2. Physical activity and walking for transportation outcomes, by gender ..................... 182 Table 6.3. Estimates from linear regression analyses for physical activity volume [total activity counts (TAC, number/day) and steps (number/day)] ................................................................. 184 Table 6.4. Estimates from linear regression analyses for physical activity intensity [light physical activity (minutes/day) and moderate-to-vigorous physical activity (MVPA, minutes/day)] ..... 186 Table 6.5. Estimates from logistic regression analyses for making any walking for transportation trip/week (vs. none) .................................................................................................................... 189 Table 6.6. Estimates from regression analyses for frequency (ntrips/week) and duration (hours/week) of walking for transportation ................................................................................. 191 Table D.1. Comparison of interrater agreement by audit approach ............................................ 346 Table D.2. Prevalence of select built environment features related to older adults’ walking at four Assisted Living sites ................................................................................................................... 347 Table E.1. Built environment domain variables: Correlations ................................................... 350 Table E.2. Social environment variables: Correlations .............................................................. 351 Table E.3. Physical domain variables: Correlations ................................................................... 352 Table E.4. Psychosocial domain variables: Correlations ............................................................ 353 Table E.5. Sociodemographic variables: Correlations ................................................................ 354   xviiiTable F.1. Select sociodemographic and mobility characteristics and bivariate association between each variable and number of walking trips (average/day) ............................................ 356 Table F.2. Estimates from negative binomial regression analyses for number of walking trips (average/day)............................................................................................................................... 357 Table F.3. Estimates from negative binomial regression analyses for number of walking for transportation trips (average/day) with a city fixed effect .......................................................... 358 Table F.4. Estimates from Negative Binomial regression analyses for number of walking for transportation trips (average/day) with a cluster robust variance ............................................... 359 Table G.1. Select characteristics of participants with and without valid accelerometry data .... 361 Table G.2. Estimates from linear regression analyses for physical activity volume [total activity counts (TAC, number/day) and steps (number/day)] and intensity [light physical activity (minutes/day) and moderate-to-vigorous physical activity (MVPA, minutes/day)] using a 10-hour valid day criterion ............................................................................................................... 363      xixList of figures Figure 1.1. Comparison of a car-oriented suburban vs. pedestrian oriented urban built environment. Map data: Google. .................................................................................................... 5 Figure 1.2. The Press-Competence Model. Reprinted with permission of the American Psychological Association from Lawton and Nahemow (1973). ................................................... 7 Figure 2.1. Map of study and surrounding area. Base map source: iMapBC. Copyright Province of British Columbia. All rights reserved. Reproduced/adapted with permission of the Province of British Columbia. .......................................................................................................................... 75 Figure 4.1. Flow of participants into the study ........................................................................... 131 Figure 5.1. Number of participants that made > 1 trip/week to most common destinations (n=150) ........................................................................................................................................ 160       xxList of abbreviations AL: Assisted living ASCQ: Ambulatory Self-Confidence Questionnaire BMI: Body mass index CHAMPS: Community Healthy Activities Model Program for Seniors CHHM: Centre for Hip Health and Mobility CI: Confidence interval CPMD: Counts per minute per day EQ-VAS: European Quality of Life-5 Dimensions visual analogue scale FCI: Functional Comorbidity Index GIS: Geographic Information Systems GPS: Global Positioning Systems IPAQ: International Physical Activity Questionnaire IRR: Incidence rate ratio LICO: Low income cut-off MCI: Mild cognitive impairment MoCA: Montreal Cognitive Assessment MVPA: Moderate-to-vigorous physical activity NAICS: North American Industry Classification System NEWS-A: New England Walkability Scale Abbreviated OPAL: Older People's Active Living questionnaire OR: Odds ratio PSS: Perceived Stress Scale   xxiRCT: Randomised controlled trial SAFER: Shelter Aid for Elderly Renters SC-3PT: 3-item measure of social interaction SC-5PT: 5-item measure of social cohesion and trust SD: Standard deviation SES: Socioeconomic status SNQLs: Senior Neighbourhood Quality of Life Study  SPPB: Short Physical Performance Battery SWEAT-R: Seniors’ Walking Environmental Assessment Tool – Revised TAC: Total activity counts   xxiiGlossary 5 Ds: system used to classify the measureable components (dimensions) of the built environment (Ewing & Cervero, 2010; Ewing, Meakins, Bjarnson, & Hilton, 2011). 5D’s stand for: density (compactness of an area), diversity (extent to which different land use types are integrated spatially), design (street network characteristics and character of space between buildings), destination accessibility (number of destinations reachable within a defined travel time), and distance to transit (extent of transit service in an area). Accelerometer: small device that measures physical activity by integrating data on accelerations caused by body movement, over a user-defined sampling period, to obtain a summed value of activity counts (Esliger, Copeland, Barnes, & Tremblay, 2005).  Active aging: the process of optimizing opportunities for health, participation, and security in order to enhance quality of life as people age (World Health Organization, 2002). Built environment: human-made infrastructure that comprises land use (e.g., density and diversity of commercial, residential, and industrial destinations), transportation systems (e.g., sidewalks and street networks), and urban design (e.g., quality of sidewalks and streets) (S. L. Handy, Boarnet, Ewing, & Killingsworth, 2002).  Geographic information systems (GIS): computer systems that integrate, analyze, and map geographically referenced information (Butler, Ambs, Reedy, & Bowles, 2011). Global positioning system (GPS): an electronic instrument that continuously measures individuals’ position in time and space (Shoval et al., 2010). Light physical activity: physical activity falling within the cut-point range of 100-1951 counts/minute (Freedson, Melanson, & Sirard, 1998). Can also be further divided into “low   xxiiilight” (100-1040 counts/minute) and “high light” (1041-1951 counts/minutes) sub-categories (Copeland & Esliger, 2009). Macroscale features of the built environment: features that exist at the neighbourhood-level and reflect spatial structure, such as land use mix and street connectivity (Susan L. Handy, 1993). Microscale features of the built environment: features that exist at the street-level and reflect the character of the space between buildings, such as sidewalk and road characteristics (Susan L. Handy, 1993). Moderate-to-vigorous physical activity (MVPA): physical activity that falls above 1951 counts/minute (Freedson, Melanson, & Sirard, 1998). Mobility: the ability of individuals to move themselves within community environments (Webber, Porter, & Menec, 2010). Mobility refers to both functional capability (e.g., underlying capacity to be mobile, like lower extremity function) as well as enacted function in the real world (e.g., actual behaviour, such as walking) (Glass, 1998). Mobility can be non-motorized (e.g., walking, biking) and/or motorized (e.g., travel by car, scooter). Mode: means by which travel takes place (e.g., car, walking, transit). Physical activity: any bodily movement produced by skeletal muscles that results in energy expenditure (Caspersen, Powell, & Christenson, 1985). Typically classified into: i) leisure time physical activity (recreational physical activity, such as jogging or weight lifting); ii) occupational physical activity (physical activity obtained during work); iii) transportation-related physical activity (physical activity for the purpose of travel, such as walking to the bus stop or cycling to the store); and iv) household physical activity (domestic physical activity, such as chores or gardening) (Sun, Norman, & While, 2013).   xxivPress-competence model: a model of stress and adaptation. Posits that a person’s behaviour (e.g., mobility) results from the interaction between the individual’s capacity (referred to as individual competence) and the supports and pressures present in his/her environment (referred to as environmental press) (Lawton & Nahemow, 1973).  Psychosocial: pertaining to thought and feelings.  Social environment: encompasses interpersonal relationships (e.g., social networks, social support), social inequalities (e.g., socioeconomic position, racial discrimination), and neighbourhood characteristics (e.g., social cohesion, neighbourhood physical and social disorder) (McNeill, Kreuter, & Subramanian, 2006). Sedentary behaviour: activities that occur while lying down or sitting and fall within the cut-point range of 0-99 counts/minute (C. E. Matthews et al., 2008). Segment: section of road between two intersections (T. J. Pikora et al., 2002). Travel behaviour: the how (e.g., walking, biking) and why of people’s movement through space. Travel diary: self-report instrument that measures an individual’s movements over a defined period of time (Behrens & Masaoe, 2009). Trip: movement from one street address to another (S. L. Handy et al., 2002), or more simply, one-way travel between two destinations. Walkability: the extent to which the built environment supports walking; commonly operationalized using composite indices of built environment features, such as the Walk Score or Walkability Index.    xxvAcknowledgements I would like to thank all of the people and organizations that made this dissertation possible. Firstly, I would like to thank my supervisor and mentor, Heather McKay. I look up to you as a leader, researcher, organizer and mover of the masses. Your whole-hearted support and guidance has given me the room to grow as a researcher and to be unafraid to pursue ideas, ask questions, and express my opinions. I am lucky to have been your student. I am equally lucky for the guidance of my other committee members, whom I now mention. To Maureen Ashe, for being so attentive and always making time to answer questions, provide guidance and/or feedback, as well as your physical activity expertise. To Joanie Sims-Gould, for being an endless well of support and providing me with an environmental gerontology perspective. And to Meghan Winters, for the methodological guidance throughout my dissertation (“I admire your brain”), GIS support, and built environment expertise. I could not have asked for a better committee of strong, female role models. Thank you to BC Housing, and especially Margaret McNeil and Rebecca Seeger, for your collaboration on the project and to the recipients of the Shelter Aid for Elderly Renters rental subsidy, without you this project would literally not have been possible. To the Walk the Talk frontline staff for all of your hard work during data collection – Suman Ajit Auluck, Catherine Craven, Paul Drexler, Kaitlyn Gutteridge, and Lutetia Wallis-Mayer – as well as to those working behind the scenes (Thea Franke, Lynsey Hamilton, Sarah Lusina-Furst, Md. Moniruzzaman and Josh Van Loon). I could not have asked for a harder working and more dedicated group. Finally, I’d also like to thank Penny Brasher for her statistical guidance throughout my dissertation and Erin Gorman for passing on her knowledge of accelerometry.   xxviVanier Canada Graduate Scholarship, UBC Four-Year Doctoral Fellowship and Center for Hip Health and Mobility provided financial support for my doctoral work. Financial support specific to the Walk the Talk project was provided by a Canadian Institutes of Health Research Emerging Team Grant (grant number 108607, PI: McKay, HA). Thank you, thank you, thank you. Finally, I’d like to thank my family and friends for their ongoing encouragement and support throughout my studies. To my parents, for always asking about my research, taking care of my dog when I was in the thick of study planning and data collection, and always believing in me. To my sister and Cassie for always being there for me. To my “gang” of Vancouver friends for the stress relief and connections that helped keep me balanced. And finally, to my sweet Heather and our baby girl Kate - for all that you are, all that you do, and all that you help me to become. With you in my life, I don’t have reason to worry too much about anything. Our love is the cake. Life is the icing with which to make our dreams come true.    xxviiDedication To: MySelf. My family. Happiness. Growth. Perseverance. And to all who benefit from this work.     1Chapter  1: Introduction, literature review, rationale, objectives, and hypotheses  1.1 Introduction  The Canadian aging population The world’s population is aging at a rapid rate. In 2010, individuals aged > 65 years accounted for eight percent of the world’s population; by 2050, this figure is expected to double (United Nations Department of Economic and Social Affairs, n.d.-a). While the population of most developed and developing countries is projected to age over the coming decades, the Canadian population is projected to age more rapidly than most other Organisation for Economic Co-operation and Development countries (United Nations Department of Economic and Social Affairs, n.d.-a). In Canada, the percentage of individuals aged > 65 years is expected to rise from 14% in 2010, to 25% by 2050 (United Nations Department of Economic and Social Affairs, n.d.-a); this is largely a result of the aging of the baby boomer generation (those born between 1947 and 1966), as well as declines in fertility rates and increases in life span (Certified General Accountants Association of Canada, 2005). The aging of a sizable proportion of the Canadian population represents one of the most significant public health policy issues facing society (Satariano, 2006). The central concern is that the increase in older adults will result in an increased prevalence of chronic health conditions that put a strain on an already overtaxed health care system and place a burden on aging individuals and those charged with their care (Satariano, 2006). In response to concerns about the economic and social demands posed by an aging society, the World Health Organization states that, “in all countries, and in developing countries   2in particular, measures to help older people remain healthy and active are a necessity, not a luxury” (World Health Organization, 2002a).    Mobility as a key determinant of healthy and active aging Active ageing is defined as, “the process of optimizing opportunities for health, participation and security in order to enhance quality of life as people age” (World Health Organization, 2002a). Mobility is broadly defined as the ability of individuals to move themselves (either independently or by using assistive devices or transportation) within community environments (Webber, Porter, & Menec, 2010). Mobility is a vital component of active ageing because it enables older adults to maintain their health and independence and participate in society. Mobility refers to both functional capability (e.g., underlying capacity to be mobile, like lower extremity function) as well as enacted function in the real world (e.g., actual behaviour, such as walking) (Glass, 1998). In order to distinguish between these two constructs, I refer to the former as capacity to be mobile and the latter as mobility. Mobility takes places in community environments that expand from the home (e.g., walking from room-to-room or climbing stairs) to the neighbourhood and beyond (Webber et al., 2010). Since the focus of this dissertation is outdoor mobility, I use the term mobility to refer to mobility that takes place outside of the home unless otherwise specified. Mobility can be non-motorized (e.g., walking, biking) and/or motorized (e.g., travel by car, scooter). Both forms of mobility may be necessary for accessing commodities, making use of neighbourhood amenities, and for participation in meaningful cultural, physical, and social activities (Rantanen, 2013). Collectively, the how (e.g., walking, biking) and why of people’s movement through space (e.g., mobility) is referred to as travel behaviour.   3Physical activity is defined as any bodily movement produced by skeletal muscles that results in energy expenditure (Caspersen, Powell, & Christenson, 1985). It is a form of mobility. Engaging in regular physical activity helps older adults to maintain their physical (Brach et al., 2004; Morey et al., 2008) and cognitive function (Bowen, 2012; Laurin, Verreault, Lindsay, MacPherson, & Rockwood, 2001) and reduces risk of morbidity (Hollmann, Strueder, Tagarakis, & King, 2007) and all-cause mortality (Wen et al., 2011).   The capacity to be mobile and physical activity are inextricably linked. Our research group previously made the helpful distinction between these two different but interconnected terms; “The capacity to be mobile is a ‘potential’ whereas physical activity is something that a person ‘does or does not do.’ By analogy, a person is either ‘able to walk’ or ‘unable to walk’ (i.e., mobility or mobility-disability). A person capable of walking may choose to walk or not (i.e., physically active/inactive). A person unable to walk may take steps to be able to walk (i.e., undertake physical activity despite mobility-disability). Intuitively, limited capacity to be mobile, particularly mobility-disability, is associated with lower levels of physical activity.”   Given the many benefits of mobility to health and everyday life, it comes as no surprise that according to the World Health Organization, mobility (including physical activity) is a fundamental component of healthy and active ageing (World Health Organization, 2002a, 2007).    The link between the built environment and older adult mobility The [outdoor] built environment is defined as human-made infrastructure that comprises urban design, land use, and transportation systems (S. L. Handy, Boarnet, Ewing, & Killingsworth,   42002). ‘Urban design’ typically refers to the arrangement and appearance of the city and physical elements within it, such as buildings, trees, and benches that line city streets, as well as the function and appeal of public spaces (S. L. Handy et al., 2002); or, more simply, the character of the spaces between buildings (Ewing & Cervero, 2001). ‘Land use’ typically refers to the density and diversity of different commercial, industrial and residential activities (Frank & Engelke, 2001; S. L. Handy et al., 2002). ‘Transportation system’ typically refers to physical infrastructure that provides connections between destinations (e.g., roads, sidewalks, bike paths), as well as the levels of service it provides (e.g., directness and quality of travel, bus frequencies) (Frank & Engelke, 2001; S. L. Handy et al., 2002). The built environment is central to mobility because it is the setting in which mobility occurs. Features of the built environment shape access to opportunities for mobility (e.g., presence of sidewalks provides the opportunity to walk, destinations such as grocery stores or restaurants provide opportunities to leave the home) (Brownson, Hoehner, Day, Forsyth, & Sallis, 2009). Figure 1.1 displays two types of built environments. The photo on the left is a car-oriented suburban neighbourhood. The absence of sidewalks leaves few alternatives to travel by car, as does the fact that there are no visible destinations that one can walk to. In contrast, the photo on the right shows a pedestrian-oriented (walkable) built environment characterized by the presence of sidewalks, traffic lights, bus stops and high density of destinations, residences, and offices. The contrasting features of these built environments illustrate how much the built environment shapes our mobility.   5 Figure 1.1. Comparison of a car-oriented suburban vs. pedestrian oriented urban built environment. Map data: Google.  As illustrated above, the built environment influences mobility across the lifespan by affecting how feasible it is to walk, drive, and/or take transit to reach destinations of interest. This is especially important in older age, as loss of a driver’s license increase older adults’ reliance on pedestrian-oriented neighbourhood built environments for independent living, meeting day-to-day needs and participation in society. This is especially true of older adults with little social or community-based support. With age-related declines in health and function, features of the built environment may more strongly hinder or facilitate mobility (Lawton, 1989; Lawton & Nahemow, 1973). For example, a poor quality built environment (e.g., cracked streets, potholes, broken curbs) may exacerbate age-related mobility limitations (Clarke, Ailshire, Bader, Morenoff, & House, 2008) and result in an older adult being unable to partake in physical activity, especially if the older adult already experiences difficulty walking (Rantakokko et al., 2010). On the other hand, a built environment that is free of barriers to moving about may slow down the age-related trajectory of mobility loss (Clarke, Ailshire, & Lantz, 2009) and thereby enable an older adult to remain active and live independently in his/her home longer. Remaining   6in the home as long as possible is one goal that an overwhelming number of older adults share (AARP, 2000). Given the importance of the built environment to active ageing and older adult mobility, the built environment represents a population-level exposure that is amenable to policy change and may help alleviate the public health burden associated with the aging of the baby boomers. The extent to which the built environment influences mobility depends on features of the built environment and on the capacity of the individual. However, much more work is needed to discern these interactions and to design built environments that best support older adult mobility. An important starting point in the study of the association between the built environment and older adult mobility is the adoption of a framework that guides the research.   Mobility is influenced by the person-environment interaction In the following sections, I provide an overview of the models and frameworks that I used to conceptualize: i) the built environment – person interaction as it influences older adult mobility and ii) individual (person) level variables that influence the capacity to be mobile.   1.1.4.1 The Press-Competence Model  According to the Ecological Theory of Aging, a person’s behaviour (e.g., mobility) results from the interaction between the individual’s capacity (referred to as individual competence) and the supports and pressures present in his/her environment (referred to as environmental press) (Lawton & Nahemow, 1973). In the context of older adult mobility and the built environment, examples of individual competence include person-level characteristics such as cognitive and physical function, self-efficacy for walking outdoors, social support, financial resources, etc. (Webber et al., 2010). Examples of environmental press include quality of sidewalks and roads,   7presence or absence of benches, distance to key destinations such as grocery stores or restaurants, etc. I apply the Press-Competence Model to guide my understanding of the environmental press – individual competence interaction (Figure 1.2).    Figure 1.2. The Press-Competence Model. Reprinted with permission of the American Psychological Association from Lawton and Nahemow (1973).  I now provide an overview of the Press-Competence Model (Lawton & Nahemow, 1973). In the model, individual competence lies on the y-axis and spans the continuum from low to high.   8Environmental press lies on the x-axis and spans the continuum from weak to strong. The point at which individual competence matches environmental press is known as the adaptive level. Although the person and environment are optimally matched at the adaptive level, departure from this point of balance does not necessarily have a negative impact on mobility. Specifically, the zones of maximum comfort and performance potential represent instances where positive behavioural outcomes occur despite a mismatch between individual competence and environmental press. In the zone of maximum comfort, the strength of environmental press is incrementally weaker than individual competence. Here, the individual is able to favourably maintain his/her behaviour and function with little stress. In the zone of maximum performance potential, the strength of the environmental press is incrementally stronger than individual competence. Since the mismatch between individual competence and environmental press is relatively small, it elicits stimulation and motivation for the individual to adapt in order to meet the demands of the environment and successfully carry out the behaviour (i.e., be mobile). Beyond the zones of maximum comfort and maximum performance potential lie zones where an individual no longer successfully functions, either from boredom and atrophy of skills or inability to meet the demands imposed by the environment. These are the zones where mobility-disability occurs (i.e., individual is unable to be mobile in his/her built environment). Applied to older adult mobility in the built environment, the Press-Competence Model posits that whether a given built environment facilitates or impedes mobility is not uniform across ‘all’ older adults. Mobility results from the interplay between features of the built environment and individual capacity. If the environment press is much greater than an older adult’s functional capacity, the older adult is likely to stop being mobile in the built environment (Noreau & Boschen, 2010; Shumway-Cook et al., 2003).    91.1.4.2 Framework of person-level variables that influence older adult mobility Webber et al. (2010) propose a framework of older adult mobility that identifies domains (cognitive, financial, physical, psychosocial) of person-level variables that contribute to individual competence and interact with the environment [press] to influence mobility. Examples of cognitive variables include memory and mental status; financial variables include income and economic resources; and physical variables include lower-extremity functioning and comorbidities. The framework conceptualizes the psychosocial domain as consisting of: i) variables that exist at the individual level and are likely to result from the process of socialization (e.g., thoughts and feelings such as self-efficacy or depression); and ii) variables that exist at a wider structural level (e.g., interpersonal relationships) (Singh-Manoux, 2003). However, in this dissertation, I differentiate between measures of thought and feelings as residing within a psychosocial domain and measures of interpersonal relationships (e.g., social networks, social support) and neighbourhood characteristics (e.g., social cohesion, neighbourhood physical and social disorder) as residing within a social environment domain (McNeill, Kreuter, & Subramanian, 2006). I do this to better differentiate between variables within (person-level) and outside (environment-level) the individual.   The specific types of variables that comprise each domain of Webber et. al’s (2010) framework are context specific. That is, they vary according to mobility type (travel mode, e.g., walking, transit, car) and where mobility is taking place (e.g., directly outside one’s home vs. in the neighbourhood vs. in the community). Since pressures (internal and external) that an individual faces increases the further he/she travels from home, the identity and complexity of variables within each domain increases the further an older adult is from his/her home. In addition to the   10identified domains, gender, culture, and biographical factors (e.g., personal life history) also influence mobility by shaping older adults’ opportunities, experiences, and behaviours. Taken together, Webber et al.’s (2010) framework of older adult mobility identifies domains (cognitive, financial, physical, psychosocial) of person-level variables that contribute to individual competence (the capacity to be mobile) and mobility. The framework also suggests that the specific types of person and environment –level variables that influence older adult mobility vary according to the type of mobility the person engages in.   Webber et al.’s (2010) framework of older adult mobility helps ensure that researchers conceptualize person-and environment level factors that influence older adults’ mobility comprehensively, across multiple person and environment-level domains. The framework would be strengthened, however, by expanding its consideration of mobility beyond type and location to include a biopsychosocial model of mobility-disability. A biopsychosocial model of disability, which is the basis for the International Classification of Functioning, Disability and Health, acknowledges that, “disability is a complex phenomena that is both a problem at the level of a person's body, and a complex and primarily social phenomena” (World Health Organization, 2002b). Taking the International Classification of Functioning, Disability and Health as an example, this means that mobility-disability may present as impairment in body functions or structures (e.g., decreased functional capacity), difficulty in carrying out activities (activity limitation, e.g., difficulty walking), and/or trouble participating in life situations (participation restriction, e.g., trouble walking to the store) (World Health Organization, 2002b). Adopting this model of mobility-disability within Webber et al.’s (2010) framework of older adult mobility would facilitate greater specificity in how mobility is conceptualized for a given research   11question, as well as consideration of how mobility (and mobility-disability) impacts on the older adults’ daily life.     SES and older adult mobility  Although individual competence is a product of multiple domains of variables, some are more common in the population and have a greater effect on health and mobility than others. Socioeconomic status (SES) is a broad and multidimensional construct that reflects an individual’s relative standing in society based on economic resources, power and prestige (Braveman et al., 2005). In the research setting, low SES is typically operationalized using measures of income, education and/or occupational status (Grundy & Holt, 2001). Importantly, older adults of low SES represent 12% of the Canadian older adult population, as estimated by Statistics Canada’s low income cut-off measure, and this figure is projected to rise (The Conference Board of Canada, 2013). The association between low SES and poor health outcomes that influence the capacity to be mobile is well established (Institute of Medicine (US) Committee on Health and Behavior, 2001; Marmot et al., 1991; Mustard, Derksen, Berthelot, Wolfson, & Roos, 1997; Reid et al., 1974). For example, older adults of low SES are at increased risk of morbidity, poor physical function and incident mobility impairment (Huisman, Kunst, & Mackenbach, 2003; Koster, Bosma, van Groenou, et al., 2006; Koster et al., 2005; Shumway-Cook, Ciol, Yorkston, Hoffman, & Chan, 2005). Any one of these adverse health outcomes decreases an older adult’s individual capacity and thereby potentially increases the press that he/she feels when attempting to be mobile (i.e., walk) in a given built environment.  Although older adults of low SES are at greater risk of adverse health outcomes that affect their capacity to walk, they are also more likely to experience difficulties with motorized-  12transportation. For example, older adults of low SES make less trips and travel less miles by car (Cao, Mokhtarian, & Handy, 2010; Frank, Kerr, Rosenberg, & King, 2010; Turcotte, 2012), as well as rely on others more in order to get to places out of walking distance (Turcotte, 2012). Older adults of low SES are also more likely to self-report that they miss out on activities due to lack of transportation (Kim, 2011). This decreased utilization of motorized transportation may be a result of financial restrictions that prohibit this subgroup from owning a car and/or as an effort to preserve financial resources. Regardless of the cause, living in pedestrian-oriented built environments may be especially important for older adults of low SES because it may enable them to access destinations by foot and engage in activities that they might otherwise miss out on.   Thus, older adults of low SES may be at increased risk of adverse health outcomes that make it harder for them to walk in unsupportive built environments. At the same time, they may be less likely to utilize other (motorized) forms of transportation. Therefore, built environments that support walking may be especially important to active ageing and mobility in this subgroup of older adults.   Summary In sum, mobility is integral to active ageing. The built environment is central to older adult mobility as it is the setting where mobility occurs. Features of the built environment (environment press) interact with the functional capacity (individual competence) of the individual to influence mobility. Older adults of low SES are at greater risk of reduced functional capacity and limitations in motorized transportation that potentially increase their reliance on   13neighbourhood built environments that support walking. Despite this, older adults of low SES are underrepresented in research (Schnirer & Stack-Cutler, 2011), including physical activity and aging research (S. L. Hughes et al., 2011). Thus, the association between the built environment and non-motorized mobility of older adults of low SES warrants further research.    Dissertation aims and outline The primary aim of my dissertation is to describe the mobility (capacity, travel behaviour, and physical activity) and to investigate the association between the built environment and non-motorized mobility (physical activity and walking for transportation) of older adults of low SES (as measured by income) residing in Metro Vancouver, Canada. My secondary aims are to contribute more broadly to the built environment and older adult mobility literature by: i) investigating a novel approach (virtual audits) to the measurement of built environment features that influence older adult’s walking; and ii) describing the destinations that older adults living on low income most commonly travel to. In order to achieve these aims I conducted a cross-sectional study of 161 older adults living on low income across Metro Vancouver and conducted audits of built environments across Metro Vancouver.   My dissertation is organized as follows. Chapter 2 provides a detailed description of methods I employed in the cross-sectional study. Chapter 3 details a study that I conducted to investigate a novel method (virtual audits) for measuring built environment features that influence older adults’ walking. Chapter 4 describes the mobility (capacity, travel behaviour, physical activity) of older adults who participated in the cross-sectional study. Chapter 5 describes participants’ travel behaviour (trip frequency, mode, purpose, and destinations most commonly visited) and   14the association between the built environment and participants’ walking for transportation. Chapter 6 investigates the association between the built environment and participants’ physical activity and walking for transportation. Chapter 7 provides an integrated discussion that links the four studies that comprise my dissertation, and presents unique aspects, implications, and the impact of my work, as well as overall strengths and limitations, and closes with overall conclusions.   I review the relevant literature in the next section of this chapter including measurement of: i) travel behaviour, ii) physical activity, and iii) the built environment. I then review existing literature on the association between the built environment and older adults’ walking for transportation and physical activity. I conclude Chapter 1 by stating my specific aims, rationale and objectives for each of the studies that comprise this dissertation.     151.2 Literature review This literature review is divided into four sections. The first two sections review the measurable dimensions of travel behaviour, physical activity and the built environment, as well as provide an overview of the approaches I used to measure these concepts in my dissertation. The third section reviews the literature on the association between the built environment and older adults’ walking for transportation and physical activity. I conclude by outlining the aims, rationale, and objectives of each study that makes up this dissertation    How mobility (travel behaviour and physical activity) is measured In this section, I review the measureable dimensions of travel behaviour and physical activity. I then go on to discuss the approaches that I used to measure travel behaviour and physical activity in this dissertation.   1.2.1.1 Measureable dimensions of travel behaviour A trip is defined as movement from one street address to another (S. L. Handy et al., 2002), or more simply, one-way travel between two destinations. Researchers typically measure travel behaviour with respect to trips (S. L. Handy et al., 2002; Meyer & Miller, 1984). Table 1.1 describes measureable components (dimensions) of travel behaviour (S. L. Handy et al., 2002; Meyer & Miller, 1984).      16Table 1.1. Measureable dimensions of travel behaviour Dimension What it measures Example Frequency Number of times trips were made in a defined time frame Number of trips/week Origin Start location  Typically classified into home, linked, return categories Destination End location Store, city center Time Time of day travel took place Noon, evening Length  Distance travelled Vehicle kilometers travelled, kilometers walked  Mode Means by which travel took place Car, walking Purpose Reason for travel (usually in reference to the activity undertaken at the destination) Shopping, recreation Cost Utility/disutility associated with travel Financial cost, travel time  As shown in the table above, mode is but one dimension of travel behaviour. Travel behaviour also encompasses other characteristics of a trip, such as start and end location and purpose. As such, the study of travel behaviour paints a picture of how and why people are mobile.   1.2.1.2 Approaches used to measure travel behaviour A travel diary is a self-report instrument that measures an individual’s movements over a defined period of time (Behrens & Masaoe, 2009). A global positioning system (GPS) is an electronic instrument that continuously measures an individual’s position in time and space (Shoval et al., 2010). Travel diaries and GPS are approaches used by researchers to measure travel behaviour. I now discuss the strengths and limitations of travel diaries, as they are the approach that I used to measure travel behaviour in this dissertation.    171.2.1.2.1 Travel diaries Federal and regional governments and researchers commonly measure travel behaviour with travel surveys (S. L. Handy et al., 2002; McCormack, 1999). Most household travel surveys use a diary format to collect data via telephone or face-to-face interviews or with a mail-out survey (Stopher, Kockelman, Greaves, & Clifford, 2008). For example, regional travel surveys conducted by transportation planners and/or transit authorities typically obtain travel diary data using computer assisted phone interviews where they ask participants to recall predetermined domains of their travel on a previous day. These data are used for regional planning and sometimes made available to researchers. Researchers have also used travel diaries to measure participants’ travel behaviour as it happens (Frank et al., 2010; Kemperman & Timmermans, 2009; Winters, Voss, et al., 2015). With this approach, researchers request that participants record predetermined dimensions of their trips across a specified number of days. I used this latter approach in my dissertation.   Measurement of travel behaviour with travel diaries has many strengths. Namely, travel diaries are a rich source of travel behaviour data (Behrens & Masaoe, 2009). For example, they can provide information on the purpose and destination of travel, by mode, throughout the day (Behrens & Masaoe, 2009). Further, participants are typically asked to record their travel behaviour on the day that it occurs. Thus, if faithfully maintained, travel diaries are less susceptible to recall bias than methods that rely on participants’ recall of past travel behaviour (Dishman, Washburn, & Schoeller, 2001). Finally, (unlike travel behaviour data collected with GPS), researchers do not need technical expertise to process and analyze data collected by travel diaries.   18Travel diaries also have limitations as a data source for travel behaviour. First, they rely on participant self-report and are therefore susceptible to information biases that include: i) response (social desirability) bias - participants may alter their regular travel behaviour while in the study to present themselves in a more positive light (Dishman et al., 2001) and ii) recall bias - missed trips (including under-reporting of short trips), item non-response (e.g., missing dimensions such as purpose), recording errors, and incomplete diary period recall (Behrens & Masaoe, 2009; Stopher & Greaves, 2007). Second, since travel diaries measure travel behaviour in trips, which are defined relative to travel between destinations, they assume that travel is a derived demand. That is, people travel in order to reach destinations or activities (S. L. Handy et al., 2002). As a result, travel diaries typically miss travel that is not tied to a specific destination, such as travel done for the purpose of exercise. Thus, data related to travelling by automobile is generally more complete than data related to walking trips (S. L. Handy et al., 2002). Finally, another limitation of travel diaries is respondent burden associated with the comprehensive nature of data collected by diaries (Behrens & Masaoe, 2009).  1.2.1.3 Measureable dimensions of physical activity Physical activity is a broad construct that encompasses all activities that result in energy expenditure above resting (Caspersen et al., 1985). Walking is the most common type of physical activity that older adults engage in outdoors (Spinney, 2013). Other examples of preferred physical activity include housework, gardening and playing golf. Commonly measured dimensions of physical activity are frequency, intensity, type and time (duration) – referred to as the FITT Principle. Table 1.2 describes measureable components (dimensions) of physical activity (Bauman, Phongsavan, Schoeppe, & Owen, 2006).   19Table 1.2. Measureable dimensions of physical activitya Dimension What it measures Example Frequency Number of times a physical activity is undertaken, usually expressed in a defined time frame Number of walking trips/week Intensity Rate of energy expenditure (exertion), based on self-reported perceptions, established energy expenditure values assigned to specific physical activities, or direct measurement of energy expenditure  Typically classified into light or moderate-to-vigorous intensity categories   Time/Duration Total time a physical activity is engaged in, expressed over a defined time frame Typically expressed as average time/day or total time/week Type Specific physical activity engaged in Walking, housework, gardening Domain Context in which physical activity occurs Typically classified into: i) leisure time physical activity (recreational physical activity, such as jogging or weight lifting); ii) occupational physical activity (physical activity obtained during work); iii) transportation-related physical activity (physical activity for the purpose of travel, such as walking to the bus stop or cycling to the store); and iv) household physical activity (domestic physical activity such as housework or gardening) (Sun, Norman, & While, 2013) aBauman et al. (2006)  Table 1.2 demonstrates that physical activity can be a broad or specific construct, dependent upon the dimension(s) that a researcher is interested in measuring. National physical activity guidelines recommend that older adults engage in > 150 minutes (time) of moderate-to-vigorous physical activity (MVPA, type) per week (frequency) in order to reduce risk of morbidity and mortality and to maintain health, functional independence and mobility (Chodzko-Zajko,   20Proctor, Fiatarone Singh, et al., 2009; Tremblay et al., 2011). Based on these guidelines, measurement of MVPA specifically is common in studies of older adult health and mobility (including the built environment). National physical activity guidelines were originally developed to enhance cardiorespiratory fitness and body composition in adults; they continue to be based in large part on the close association of MVPA with cardiovascular health and mortality in healthy adults and older adults (Physical Activity Guidelines Committee, 2008). However, this link between being moderately or vigorously active and cardiovascular health failed to recognize additional emotional, social and/or mental benefits older adults gain by engaging in light physical activities like walking or gardening. Emerging evidence shows that physical activity at levels below guidelines is beneficial to older adults’ self-reported general health and psychosocial well-being (e.g., stress, isolation, depression) (Buman et al., 2010), as well as cardiometabolic markers (e.g., triglycerides, systolic blood pressure, waist circumference) of adults (aged > 20 years) (Wolff-Hughes, Fitzhugh, Bassett, & Churilla, 2015). Engagement in all kinds of physical activity may also be a more realistic goal for older adults (I. Lee, 2015; Sparling, Howard, Dunstan, & Owen, 2015).   In the context of the built environment, researchers also commonly examine the association between the built environment and walking (Rosso, Auchincloss, & Michael, 2011; Van Cauwenberg et al., 2011). However, conceptually the neighbourhood built environment may not have a uniform influence across domains of physical activity (Sallis et al., 2006; Van Holle et al., 2012). For example, if travel is conceptualized as a derived demand, features of the built environment such as the presence of neighbourhood destinations may particularly influence transportation-related physical activity, and especially walking for transportation.    211.2.1.4 Approaches used to measure physical activity Researchers commonly measure physical activity with self-report questionnaires, accelerometers, or by direct observation. Choice of measurement approach is influenced by the research question being addressed [including dimension(s) of physical activity being measured], feasibility, and the strengths and limitations of each measurement tool. Below I discuss the strengths and limitations of self-report physical activity questionnaires and accelerometry, as they are the approaches that I used to measure physical activity in this dissertation.    1.2.1.4.1 Self-report questionnaires Researchers traditionally measured physical activity with self-report questionnaires.  Questionnaires can be self-administered or interviewer-administered. Self-report questionnaires: i) offer a cost-effective approach to collect data from a large number of participants (Dishman et al., 2001; Sallis & Saelens, 2000); ii) can be easily adapted to reflect the interests of researchers and/or any special needs or considerations of the population under study (e.g., wording of questions, inclusion of physical activities relevant to a population) (Sallis & Saelens, 2000); iii) allow researchers to measure the different dimensions of physical activity, including the types of physical activity that participants engage in (Sallis & Saelens, 2000); and iv) do not alter the behaviour under study (e.g., interfere with usual physical activity) since they rely on participants’ recall of past behaviour (Dishman et al., 2001; Sallis & Saelens, 2000). On the other hand, self-report questionnaires require respondents and researchers to share understanding of ambiguous terms, such as "physical activity" or "moderate-to-vigorous intensity." This source of error may be more pronounced among older adults, since limitations in lower extremity functioning influence perceptions of the intensity of physical activities (Rikli, 2000). Self-report   22questionnaires are also susceptible to response bias (e.g., over-reporting of physical activity due to social desirability) and recall bias (i.e., inaccurate recollection of past activity) (Dishman et al., 2001; Sallis & Saelens, 2000). Risk of recall bias increases with the length of recall period for activities that are light intensity (which are the activities that older adults most frequently engage in) and may be greater for participants that engage in physical activity sporadically (as opposed to participants who have regular physical activity routines or do not engage in physical activity) and/or have memory problems (Dishman et al., 2001; Sallis & Saelens, 2000; Washburn, 2000). Fluctuations in mood, anxiety, and level of depression may also affect participants’ responses (Rikli, 2000). Finally, self-report questionnaires do not capture patterns of physical activity throughout the day or week (Murphy, 2009) and measures designed for adult populations may not accurately measure the physical activity of older adult populations (Washburn, 2000).  1.2.1.4.2 Accelerometry An accelerometer is a small device that directly measures physical activity by integrating data on accelerations caused by body movement over a user-defined sampling period (epoch) to obtain a summed value of activity counts (Esliger, Copeland, Barnes, & Tremblay, 2005). These activity counts are used to provide estimates of movement intensity (Esliger et al., 2005). Researchers typically instruct participants to wear an accelerometer on the hip or wrist during waking hours over a specified period of days; this provides data on participants’ physical activity as it occurs in their lives (Murphy, 2009). Therefore, unlike self-report questionnaires, accelerometers are not susceptible to recall bias. Other strengths of accelerometry include the ability to: i) capture activity patterns such as when (e.g., time of day and/or day of the week) and how (e.g., sporadic,   23long bouts) physical activity occurs (Esliger et al., 2005; Murphy, 2009); ii) estimate exercise volume at different levels of intensity, which is in line with public health recommendations for physical activity (Esliger et al., 2005); and iii) measure inactivity. A major limitation of accelerometers is their cost, which potentially limits the feasibility of their use, especially in large studies (Esliger et al., 2005). Further, although accelerometry is a direct measure of physical activity, it is not necessarily an objective measure of physical activity as participants may change their behaviour while wearing these monitors due to social desirability and quality of data relies on participant compliance with measurement protocols (Esliger et al., 2005; Murphy, 2009). Data output from accelerometers is also influenced by decisions and assumptions that researchers make when gathering and analyzing the data; for example cut-points used to classify intensity of physical activity and number of days that accelerometers are to be worn, placement of the device (e.g., hip, wrist, side of the body) and model of accelerometer (Esliger et al., 2005; Gorman et al., 2014; Murphy, 2009). Finally, accelerometers inaccurately measure physical activities that are not step-based (e.g., swimming, cycling, weight-lifting) and do not capture the full energy cost of certain activities such as walking while carrying a heavy object or upper body movements (e.g., resistance training) or climbing a hill (elevation gain) (Colley et al., 2011; Esliger et al., 2005; Murphy, 2009).  In sum, self-report questionnaires and accelerometers are commonly used to measure physical activity. Importantly, accelerometers allow researchers to directly measure physical activity as it occurs in participants’ lives, whereas self-report questionnaires allow researchers to obtain more detailed information (context) on specific domains of physical activity. I am interested in   24participants’ general physical activity, as well as walking for transportation. Thus, in my research I used both approaches to measure physical activity.   How the built environment is measured In this section, I discuss the measureable dimensions of the built environment. I then discuss approaches that I used to measure the built environment in the studies I present in this dissertation.   1.2.2.1 Measureable dimensions of the built environment Table 1.3 describes the measurable components (dimensions) of the built environment (Ewing & Cervero, 2010; Ewing, Meakins, Bjarnson, & Hilton, 2011). These are known as the 5 Ds (dimensions) of urban form; they include density, diversity, design, destination accessibility, and distance to transit (Ewing & Cervero, 2010; Ewing et al., 2011). These dimensions impact mobility by influencing the degree to which the built environment supports walking vs. driving (Ewing & Cervero, 2010; Ewing et al., 2011). In this, the 5Ds influence travel behaviour, physical activity, and health.      25Table 1.3. Measureable dimensions of the built environmenta  Dimension What it measures Example Density Land use - compactness of an area  Number of residents per square kilometer (population density); floor area ratio (commercial floor area/lot area) Diversity Land use - extent to which different land use types are integrated spatially Ratio of different land uses per square kilometer (land use mix); distance from home to nearest destination  Design Transportation systems (street network characteristics) and urban design (character of space between buildings) (Ewing & Cervero, 2001) Typically classified into i) street connectivity (e.g., average block length, number of intersections per square kilometer) and ii) urban design categories (e.g., presence of pedestrian amenities such as crosswalks, landscaping variables such as trees) Destination accessibility Land use - the number of destinations reachable within a defined travel time Distance to downtown, distance from home to closest grocery store Distance to transit Transportation systems - extent of transit service in an area Distance from home to nearest bus stop a Ewing & Cervero (2010) and Ewing et al. (2011)  These built environment dimensions highlight that an individual’s mobility is influenced by both the neighbourhood-level spatial structure (macroscale) and the street-level character of the space between buildings (microscale) where an individual lives (Susan L. Handy, 1993). Although they are organized into separate categories, the 5 Ds are interrelated and overlap. Therefore, similar to domains of mobility, the effect of one dimension of the built environment on mobility is likely a by-product of the synergistic effect of other built environment dimensions on this outcome (Ewing & Cervero, 2010). Further, of the 5 Ds, pedestrians and cyclists (rather than motorists)   26are most aware of urban design features (Frank & Engelke, 2001). This is because pedestrians and cyclists travel at speeds that are slow enough to take note of many urban design features (Frank & Engelke, 2001) and urban design features may directly impede (e.g., poor sidewalk quality) or support (e.g., pedestrian crosswalks) walking or cycling. Urban design features that influence walking and cycling are further organized into four dimensions; I present these in Table 1.4.   Table 1.4. Measureable dimensions of urban design features that influence walking and cyclinga  Dimension What is measures Example Functionality Physical attributes of streets and sidewalks Quality of sidewalks and/or streets  Safety  Actual or perceived freedom from risk of harm Typically organized into: i) traffic (e.g., presence of traffic calming devices and/or pedestrian crossing devices) and ii) personal (e.g., adequate lighting, presence of stray dogs) categories Aesthetics  Presence of features that contribute to an interesting and pleasing built environment Design of buildings, landscaping features like shrubs and trees Destinations Presence of community and commercial facilities  Stores, parks a T. Pikora, Giles-Corti, Bull, Jamrozik, and Donovan (2003)  Urban design features exist at the scale of the pedestrian (and cyclist). Measureable dimensions presented in Table 1.4 include both objective and innately perceived features of the built environment. Functionality and destinations overlap with the 5 Ds. Once again, this highlights the interconnectedness between built environment dimensions.   271.2.2.2 Approaches used to measure the built environment  In the next section, I discuss approaches used by researchers to measure the built environment. These include: i) objective [geographic information systems (GIS) based], ii) self-report (perceived), and iii) observational (audits) approaches (Brownson et al., 2009). In my research, I used tools from within each category to measure the built environment.  1.2.2.2.1 Objective approach: GIS- based measures  Geographic information systems (GIS) are computer systems that integrate, analyze, and map geographically referenced information (Butler, Ambs, Reedy, & Bowles, 2011). Therefore, GIS-based measures of the built environment are “derived primarily from existing data sources… [and] have some spatial reference (e.g., address or census boundary identification)” (Brownson et al., 2009). These measures exist at multiple levels (e.g., municipal, regional, national) and are usually available through public (government) sources, such as the census or local transportation or planning agencies (Brownson et al., 2009; Institute of Medicine, 2012). Commercial databases also exist (Brownson et al., 2009; Institute of Medicine, 2012). Researchers typically use GIS-based measures to assess a broad range of built environment dimensions at the macroscale (e.g., density, diversity, street connectivity). These features may be assessed individually or as part of a composite index that represents the overall walkability of an area. As GIS-based measures use data from existing sources and are available across a range of geographic scales, this source of data is usually the most feasible option for studies with large numbers of participants and/or that span large geographic areas (Brownson et al., 2009; Institute of Medicine, 2012). That being said, GIS-based data are typically collected by non-health sectors. As a result, available data do not necessarily reflect the needs and/or interests of health researchers who, for example, are more   28interested in elements at the microstructure level (e.g., benches, crosswalks). These kinds of urban design features are rarely available in GIS databases (Brownson et al., 2009; Institute of Medicine, 2012). Specificity and quality of data available to researchers (e.g., accuracy and completeness of data; little standardization for data collection and entry) varies across GIS sources and between regions (Brownson et al., 2009; Institute of Medicine, 2012). Finally, as GIS-based measures typically rely on archival data, they may not reflect current built environment conditions (Brownson et al., 2009; Institute of Medicine, 2012).  1.2.2.2.2 Self-report approach: Perceived measures Perceived measures of the built environment assess participants’ perceptions across the range of built environment dimensions (Brownson et al., 2009). They are typically self-administered or administered by an interviewer in person or over the phone (Brownson et al., 2009). Perceived measures of the built environment: i) are relatively inexpensive; ii) can be used to gather data on built environment variables that are not readily available through other GIS-based measures (e.g., urban design features such as sidewalk quality or aesthetics); and iii) are feasible for use in studies with large numbers of participants and/or that span large geographic areas (Brownson et al., 2009; Institute of Medicine, 2012). An important limitation of this approach is that respondent’s mobility and/or individual capacity can introduce measurement bias (Brownson et al., 2009; Diez Roux, 2007). For example, it is unlikely that an individual who seldom goes out into his/her neighbourhood can accurately report on features of the neighbourhood’s built environment. Further, an individual’s functional capacity may affect his/her perceptions of the built environment (e.g., an older adult that has lower extremity impairment may perceive a built environment to be less pedestrian friendly than an older adult with no lower extremity   29impairments). That being said, it can be argued that when it comes to measures of the built environment, “perceptions are the reality,” especially when measuring urban design features such as aesthetics (Brownson et al., 2009). Regardless, it is important that researchers in this field are aware of the potential discordance between objective and perceived measures of the built environment (Arvidsson, Kawakami, Ohlsson, & Sundquist, 2012; Brownson et al., 2009; Gebel, Bauman, & Owen, 2009), as it affects the comparability of results across studies using different approaches to measure the built environment. Researchers are best to view these approaches as providing complementary data on features of the built environment that affect mobility (Arvidsson et al., 2012; Brownson et al., 2009).   1.2.2.2.3 Observational approach: Audits Audits measure the built environment through systematic observation (Brownson et al., 2009; Institute of Medicine, 2012). They are typically conducted by trained individuals that use standardized forms to collect data through in-person observation (Brownson et al., 2009; Institute of Medicine, 2012). A less common approach is to conduct audits through virtual mediums like Google Earth’s Street View feature (Google Inc., CA). A major strength of audits is that they allow researchers to directly measure urban design features that are not readily available through GIS and may be especially important to older adult mobility (e.g., sidewalk quality, curb cuts) (Institute of Medicine, 2012). Audits are a rich source of data across this and other dimensions of the built environment. That being said, there is a lack of scoring protocols to guide how best to summarize and use the data collected by audits (Institute of Medicine, 2012).There is also a paucity of data that links features of the built environment as measured by audits and older adults’ mobility outcomes. Further, audits are resource (time and money) intensive (Brownson et   30al., 2009; Institute of Medicine, 2012). Although they can be used to capture data on macroscale built environment features, they are generally less efficient at doing so than GIS-based measures (Brownson et al., 2009; Institute of Medicine, 2012). This limits the feasibility of audits in large studies conducted across large geographic areas (Brownson et al., 2009; Institute of Medicine, 2012).   In sum, researchers commonly assess the built environment using objective (GIS-based), (ii) self-report (perceived), and (iii) observational (audits) approaches (Brownson et al., 2009). Geographic information systems-based measures offer an efficient approach to the measurement of predominantly macroscale features of the built environment. Since they are objective measures, their findings are more closely comparable across studies and more easily translated into concrete policy recommendations. Perceived measures and audits can be used to assess both macro- and microscale features of the built environment. Since an individual’s perceptions are a reflection of the individual as well as his/her surroundings, perceived and objective measures may disagree (Arvidsson et al., 2012; Gebel et al., 2009). However, they provide complimentary data on features of the neighbourhood built environment (Arvidsson et al., 2012; Brownson et al., 2009). Finally, due to the resource intensive nature of audits, they may be most appropriate for measuring microscale features of the built environment not readily available in GIS. Although they are a rich source of built environment data, there is currently little evidence to guide how to best utilize the vast amounts of data gathered using this approach.     31 Existing literature on the association between the built environment and older adult physical activity and walking for transportation I now review existing literature on the association between the built environment and older adults’ physical activity and walking for transportation. I begin by discussing the most recent relevant reviews and then synthesize the findings of studies published since these reviews. Among the physical activity studies, I focus on the ones that measure physical activity objectively, as these are most relevant to my dissertation.   1.2.3.1 Recent systematic reviews The most recent systematic reviews on older adult physical activity and walking (including walking for transportation) were published in 2011 (Rosso et al., 2011; Van Cauwenberg et al., 2011). Both reviews based their findings on level three evidence (cohort and case-control studies) (Sackett, 1989). Van Cauwenberg et al. (2011) investigated the association between the built environment and older adult physical activity. They included 31 observational (29 cross-sectional and two prospective) studies of community-dwelling older adults aged > 65 years, published between 2000 and March 2010. Rosso et al. (2011) investigated the association between the objectively measured built environment and mobility (i.e., walking) or disability and physical functioning of community-dwelling older adults aged > 60 years; they included studies published between 1990 and December 2010. Of the 17 included studies (14 cross-sectional and three prospective), 14 investigated walking and three reported disability outcomes. Nine of the 14 studies were also included in the review by Van Cauwenberg et al. (2011). Of the studies included in both reviews, only two used accelerometers to measure physical activity; seven studies measured walking for transportation by self-report. Both Van Cauwenberg et al. (2011)   32nor Rosso et al. (2011) were unable to reach definitive conclusions regarding built environment features associated with older adults’ physical activity and/or walking for transportation. I provide further detail on the findings of both reviews below, as well as implications for future research.  Since built environment features that influence physical activity vary by physical activity domain, Van Cauwenberg et al. (2011) presented their findings separately for overall physical activity as well as by domain of physical activity (including walking for transportation), where possible. Only two of the studies they included used objective measures of physical activity (Berke et al., 2006; Morris, McAuley, & Motl, 2008). Morris et al. (2008) reported that  perceptions of street connectivity were associated with older adult women’s total physical activity (counts/day as measured by accelerometry). Berke et al. (2006) found that older adults that lived closer to a fitness facility were more likely to participate in any (vs. no) structured and unstructured physical activity programs (direct measure of physical activity) and were also more likely to participate in unstructured programs more frequently (direct measure of physical activity) than those living further away from fitness facilities. Van Cauwenberg et al. (2011) included six studies that assessed the association between the built environment and walking for transportation (all measured by self-report) (Borst et al., 2009; Frank et al., 2010; Mendes de Leon et al., 2009; Patterson & Chapman, 2004; Shigematsu et al., 2009; Su, Schmöcker, & Bell, 2009). They were unable to reach definitive conclusions regarding the nature of associations; this was also the case for all other self-reported physical activity outcomes, including total physical activity, recreational physical activity, total walking and cycling, and recreational walking. Van   33Cauwenberg et al. (2011) state that, “results were inconsistent but most of the studied environmental characteristics were reported not to be related to physical activity.”   The review by Rosso et al. (2011) took a different approach to synthesizing their findings; they present results by built environment domain instead of by mobility outcome. They included one study that measured physical activity objectively (steps/day as measured by accelerometry) (Hall & McAuley, 2010) and two that investigated walking for transportation (measured by self-report) (D. King, 2008; Patterson & Chapman, 2004). The remainder of included studies investigated disability and other walking outcomes (e.g., total walking, walking for exercise, walking to meet physical activity guidelines). Hall and McAuley (2010) reported that older adults that were less likely to walk 10,000 steps/day had significantly fewer walking paths within 1 km of their home (measured objectively with GIS) and perceived their neighbourhood built environments to have less street connectivity and to be more unsafe (pedestrian and traffic safety). D. King (2008) found that curb cuts, crosswalks, density of retail, and presence of physical incivilities (all measured with audits) predicted greater frequency of walking for transportation (times/week). Finally, Patterson and Chapman (2004) reported that walkability (as measured by the New Urbanism Index, calculated using data collected by audits) was associated with walking for transportation. When looking across all included studies, Van Cauwenberg et al. (2011) reported inconsistent findings across built environment domains. They noted that the most “promising evidence” lends support for an association between older adult mobility and street connectivity, street and traffic conditions, and proximity to green spaces and retail destinations.     34According to both Van Cauwenberg et al. (2011) and Rosso et al. (2011), reasons for inconclusive findings include between-study heterogeneity in instruments used to measure the built environment and mobility, as well as between-study heterogeneity in built environment and mobility domains being measured. In my opinion, key recommendations made by Van Cauwenberg et al. (2011) and Rosso et al. (2011) for future research include studies that: i) are guided by a strong theoretical framework; ii) are conducted outside of the USA; iii) use physical activity measures with established psychometric properties and that focus on specific physical activity domains and/or broader measures of mobility (e.g., lifespace); iv) use objective and perceived measures of specific (e.g., micrsoscale) built environment features within a standardized geographic scale; v) investigate the moderating effects of, e.g., age, gender, and/or SES in order to identify subpopulations who mobility is most closely associated with features of the built environment; vi) investigate associations among socially disadvantaged subpopulations, such as so women, minorities and individuals with low income; and vii) that employ a longitudinal design. Importantly, unless these future studies are of a higher level of evidence, this will still not provide definitive answers/conclusions.   Randomized controlled trials (RCTs) provide the highest quality of research evidence because they are designed to minimize bias and risk of systematic error and can establish causal conclusions (Burns, Rohrich, & Chung, 2011; Sackett, 1989). In a RCT, participants are randomly allocated to an intervention (experimental group) or no intervention (control group) and followed over time to determine whether they experience the study’s outcome of interest (Roberts & Dicenso, 1999). Randomised controlled trials are the preferred method to reduce confounding, as random allocation ensures that known and unknown factors (confounders) have   35comparable frequencies across levels of exposure (e.g., different built environments) (Katz, 2006a). Further, the longitudinal nature of this design helps ensure that exposure (intervention) precedes outcome (Roberts & Dicenso, 1999). These two key characteristics explain why RCTs are used to establish causal relations, whereas cross-sectional studies provide evidence of associations (Roberts & Dicenso, 1999). Unfortunately, RCTs are rarely feasible in built environment and mobility research, as they require individuals to be randomized to move to specific residential locations (built environments). Natural experiments are a type of study in which there is a natural or predetermined allocation of an intervention/ independent variable (Petticrew et al., 2005). As a result, they offer an important step forward in the absence of RCTs (Petticrew et al., 2005); they have the potential to move what we know about older adult mobility and the built environment away from associations and closer towards causal relations. For example, our research team partnered with the City of Vancouver to assess how changes to the built environment in Vancouver’s downtown core effect the health and mobility of residents across the lifespan (http://www.hiphealth.ca/media/ASAP(1).pdf).   1.2.3.2 Recent realist review A realist review is a theory-driven method of evidence synthesis that aims to identify patterns and mechanisms that underlie associations (Pawson, Greenhalgh, Harvey, & Walshe, 2005; Yen, Flood, Thompson, Anderson, & Wong, 2014). Yen et al. (2014) conducted a realist review to determine how built environment features supported and/or hindered older adult mobility (defined as walking and physical activity). The review included a younger population (aged > 50 years) than the standard age I use to classify older adults (> 65 years). Regardless, I discuss this work as it makes a unique theoretical contribution to the field of built environment and mobility   36research. Yen et al. (2014) drew conclusions regarding factors associated with older adults’ mobility decisions (whether older adults were mobile), but were unable to draw conclusions regarding factors that contributed to increased mobility (causation) due to a dearth of studies that investigated this outcome. Similar to Rosso et. al.’s (2011) work, land use (including presence of destinations) and connectivity (characteristics of street networks and pedestrian infrastructure) were built environment features that most consistently influenced older adults’ mobility decisions. Yen et al. (2014) also identified aesthetics, physical capacity, and safety as playing an important role.   The review by Yen et al. (2014) is particularly unique. It not only identifies built environment features associated with older adult mobility, it also proposes a mechanism [perceptions of safety (personal and traffic)] that helps explain why associations are not straightforward and why they do not exist across all studies. Understanding underlying mechanisms is important because it allows for an in-depth understanding of relations beyond statistical associations (Sheehan, Sobolev, Chudyk, Stephens, & Guy, 2016). Examples (of the mechanism) provided by Yen et al. (2014) now follow. Greater land use mix provides more destinations and as a result may promote mobility by offering older adults a reason to leave home. However, if greater land use mix (and more destinations) were associated with perceptions of an unsafe built environment (e.g., due to crowded streets and/or heavy traffic), these dimensions did not promote mobility. Similarly, greater street connectivity facilitated mobility through its association with short block lengths and land use density. However, greater street connectivity was a barrier to mobility if it was perceived as unsafe. For example, a street network with lots of intersections, but no traffic calming measures negatively influenced mobility of older adults. Finally, if an older adult   37perceived a built environment to be aesthetically displeasing, he/she was also likely to perceive the built environment as unsafe; built environments perceived as unsafe hindered mobility.   In summary, Yen et al. (2014) conducted a realist review that identified mechanisms through which features of the built environment influence older adults’ mobility decisions. Their work supports the theoretical frameworks I adopted for my research. That is -- older adult mobility represents the dynamic interplay between the person and their environment. Whether a built environment supports or hinders mobility is not straightforward. It must be understood within the context of many external and internal factors. The extent to which a given built environment supports older adult mobility, therefore, will vary between groups of individuals, all of whom are different across mental, financial, psychosocial and physical health parameters.    1.2.3.2.1 Overview of relevant studies published since the most recent systematic reviews I now provide an overview of the literature that was published after the reviews by Van Cauwenberg et al. (2011) and Rosso et al. (2011) and that investigates the association between the built environment and my mobility outcomes of interest (physical activity and walking for transportation). I focus on studies that measured physical activity with accelerometry, since this is the instrument I used to measure my dissertation’s primary physical activity outcomes. Further, since no studies targeted recruitment at older adults of low income, this literature review focuses on studies of the general older adult population (community-dwelling older adults aged > 65 years).     381.2.3.3 Studies of the association between the built environment and physical activity (as measured by accelerometry) Since the systematic reviews by Van Cauwenberg et al. (2011) and Rosso et al. (2011), there has been a proliferation of studies that investigated physical activity outcomes as measured by accelerometry. Table 1.5 describes these studies.     39Table 1.5. Studies that investigated the association between the built environment and older adult physical activity (as measured by accelerometry)a Authors and countryb Objective(s) Study design and sample Built environment measurement approach, tool, domainc Physical activity outcomed Relevant built environment-related resultsd Bracy et al. (2014).  USA. 1. Test the hypothesis that the association between the built environment and physical activity is moderated by perceptions of safety (crime, traffic, and/or pedestrian). Design: cross-sectional.  Sample: 718 older adults aged > 66 yrs. 1. GIS: Walkability Index; >1 park present; > 1 recreational facility present.  2. Perceived (NEWS): crime safety; pedestrian safety; traffic safety. 1. MVPA (mins/wk). 1. No interactions present.  2. Main effects: a. Walkability (β= 5.57-6.82) and parks (β = 31.64-33.28) consistently associated with MVPA (mins/wk) across models. b. Crime safety (β = 14.89, 95% CI = 0.12, 29.66) associated with MVPA (mins/wk) in a model that included the Walkability Index.  c. Pedestrian safety (β= 7.24, 95% CI = −10.58, 25.06) associated with MVPA (mins/wk) in a model that included recreation facilities nearby. Carlson et al. (2012).  USA. 1. To explore how the built environment and psychosocial variables interact in explaining physical activity. Design: cross-sectional.  Sample: 718 older adults aged > 66 yrs. 1. GIS: Walkability Index; >1 park present; > 1 recreational facility present.   2. Perceived (NEWS-A): aesthetics (presence of trees; interesting things to look at; atractive natural sights; attractive buildings/homes) and presence of walking/cycling facilities (presence of sidewalks on most streets; sidewalks separated from road by parked cars; presence of buffer between road and 1. MVPA (mins/wk). 1. Interactions:  a. Walkability interacted with social support in explaining MVPA (mins/wk, β= 13.7, 95% CI = 4.73, 22.70). b. Aesthetics interacted with psychosocial barriers to physical activity in explaining MVPA (mins/wk, β= −12.20, 95% CI =  -20.00, -4.41).  2. Main effects: a. Walkability associated with MVPA (mins/wk, β= 13.83, 95% CI = 5.02, 22.63).   40Authors and countryb Objective(s) Study design and sample Built environment measurement approach, tool, domainc Physical activity outcomed Relevant built environment-related resultsd sidewalk). Ding et al. (2014).  USA. 1. To examine whether driving status moderates the association between the built environment and physical activity. Design: cross-sectional.  Sample: 880 older adults aged > 66 yrs. 1. GIS: Walkability Index; number of nearby locations for recreational physical activity (total number of parks and private recreational facilities).   2. Perceived (NEWS): microscale built environment (summary measure created by averaging scores on walking–cycling infrastructure, aesthetics, traffic safety, pedestrian safety structures, and transit access subscales); subscales that comprise the microscale built environment summary measure; residential density; land use mix; street connectivity. 1. MVPA (mins/day).  2. Volume (cpmpd ). 1. No interactions present for MVPA (mins/d).  2. No interactions present for physical activity volume (cpmpd). A. C. King et al. (2011).  USA. 1. To evaluate the relations among objectively measured neighbourhood design and active transportation, moderate-to-vigorous physical activity (MVPA) and body weight.  2. To evaluate the Design: prospective (2 time points 6 months apart to control for seasonal variation in physical activity).   Sample: 647 older adults aged 1. GIS: Walkability Index 1. MVPA (mins/wk) 1. No associations between the built environment and MVPA (mins/wk).  2. No interaction present.    41Authors and countryb Objective(s) Study design and sample Built environment measurement approach, tool, domainc Physical activity outcomed Relevant built environment-related resultsd moderating effect of mobility impairment on these relations. > 66 yrs. McMurdo et al. (2012).  Scotland. 1. To determine which individual, social and environment factors explain person to person variation in daily physical activity. Design: cross-sectional.  Sample: 547 older adults aged > 65 yrs. 1. GIS: Road distances from residence to nearest grocery store/supermarket; % of greenspaces.  2. Perceived (OPAL): Perceptions of local area surroundings; streets; personal safety; overall satisfaction with local area. 1. Volume (counts/d) 1. No associations between the built environment and physical activity volume (counts/d).                                                                                                   42Authors and countryb Objective(s) Study design and sample Built environment measurement approach, tool, domainc Physical activity outcomed Relevant built environment-related resultsd Van Holle et al. (2014).  Belgium   1. To investigate the relation between neighbourhourhood walkability and older adults' physical activity.  2. To examine the moderating effect of neighbourhood income on these relations. Design: cross-sectional.  Sample: 438 older adults aged > 65 yrs. 1. GIS: Walkability Index 1. MVPA (mins/wk)  2. Low-light physical activity (mins/wk)  3. High-light physical activity (mins/wk) 1. Main effects:  a. Walkability associated with MVPA (mins/wk, β= 2.06, 95% CI = 0.68, 3.44). b. Walkability associated with low-light physical activity (mins/wk, β= -1.38, 95% CI = -2.59, -0.18). c. No associations between walkability and high-light physical activity (mins/wk).  2. Interactions: a. Walkability interacted with neighbourhood income (p<0.10) in explaining MVPA (mins/wk, β= −3.516, −0.134). b. No interactions present for low-light physical activity (mins/wk). c. No interactions present for high-light physical activity (mins/wk). aIncluded studies published after the reviews by Van Cauwenberg et al. (2011) and Rosso et al. (2011) that used multivariable analyses to investigate the association between the built environment and physical activity (as measured by accelerometry) of community-dwelling older adults aged > 65 years. bBracy et al. (2014), Carlson et al. (2012), Ding et al. (2014) and A. C. King et al. (2011) analyze data from the same study [the Senior Neighbourhood Quality of Life Study (SNQLs)]. cGIS = Geographic Information Systems; NEWS = New England Walkability Scale; NEWS-A = New England Walkability Scale Abbreviated; OPAL = Older People's Active Living questionnaire; dcpmpd = counts per minute per day; MVPA = moderate-to-vigorous physical activity    43All of the studies in Table 1.5 were set in the USA (Bracy et al., 2014; Carlson et al., 2012; Ding et al., 2014; Hall & McAuley, 2010; A. C. King et al., 2011) or Europe (Hall & McAuley, 2010; McMurdo et al., 2012) and all but one (A. C. King et al., 2011) had a cross-sectional design. A. C. King et al. (2011) collected physical activity data at two time points (six months apart) to control for seasonal variation; there was no seasonal variation in physical activity and subsequent studies of this dataset (Bracy et al., 2014; Carlson et al., 2012; Ding et al., 2014) analyzed data at a single time point. All but two studies in Table 1.5 (Hall & McAuley, 2010; McMurdo et al., 2012) measured the built environment using a GIS-based composite index of neighbourhood walkability; three of these studies (all using the same pool of data) complemented objective measures of walkability with GIS-based measures of single built environment features and perceived measures of the built environment (Bracy et al., 2014; Carlson et al., 2012; Ding et al., 2014). Two studies included GIS-based measures of single built environment features and perceived measures of the built environment (Hall & McAuley, 2010; McMurdo et al., 2012). Finally, the majority of studies (n = 5) in this literature review investigated the association between the built environment and MVPA (Bracy et al., 2014; Carlson et al., 2012; Ding et al., 2014; A. C. King et al., 2011; Van Holle et al., 2014); one investigated light physical activity (Van Holle et al., 2014) and three investigated total physical activity volume (Ding et al., 2014; Hall & McAuley, 2010; McMurdo et al., 2012). I now discuss the findings of these studies, organized by approach used to measure the built environment and physical activity dimension.  1.2.3.3.1 Associations between composite indices of walkability and MVPA  The studies that investigated the association between a composite index of walkability (Walkability Index) and MVPA reported mixed findings (Bracy et al., 2014; Carlson et al., 2012;   44Ding et al., 2014; A. C. King et al., 2011; Van Holle et al., 2014). Two studies reported a main effect of walkability on older adults’ MVPA (Bracy et al., 2014; Carlson et al., 2012). In addition, the study by Carlson et al. (2012) reported a positive interaction between walkability and social support, as well as a negative interaction between perceived neighbourhood aesthetics and psychosocial barriers (e.g., time constraints, discomfort). That is, living in more walkable neighbourhoods was associated with more MVPA only among participants that had higher levels of social support. Living in aesthetically pleasing built environments was associated with more MVPA only among those with few psychosocial barriers and was associated with less MVPA among those with more psychosocial barriers. Two other studies also found an association between walkability and older adults’ MVPA, but only after stratifying by a variable of interest (Ding et al., 2014; Van Holle et al., 2014). For example, Ding et al. (2014) examined whether driving status moderated the association between the built environment and several physical activity outcomes. Although they did not find a walkability by driving status interaction, the authors proceeded with separate analyses of physical activity outcomes by driving status. They noted that MVPA was only associated with walkability among drivers. Van Holle et al. (2014) did not find a main effect of walkability on MVPA. However, they did report a walkability by neighbourhood income interaction whereby residents of neighbourhoods with high walkability/low neighbourhood income engaged in more MVPA than residents of neighbourhoods with low walkability/low neighbourhood income. Finally, the study by A. C. King et al. (2011) did not find an association between walkability and MVPA. Furthermore, mobility impairment did not moderate the association between walkability and MVPA.      451.2.3.3.2 Associations between objective measures of the built environment and MVPA  Other objectively measured features of the built environment positively associated with MVPA included number of parks (Bracy et al., 2014), land use mix diversity (Ding et al., 2014), and land use mix access (among drivers only) (Ding et al., 2014). The only objective feature of the built environment negatively associated with MVPA was street connectivity (among non-drivers only) (Ding et al., 2014).   1.2.3.3.3 Associations between perceived measures of the built environment and MVPA  Of the four studies that included perceived measures of the built environment (Bracy et al., 2014; Carlson et al., 2012; Ding et al., 2014; McMurdo et al., 2012), only one found an association with MVPA. Bracy et al. (2014) reported that perceptions of safety (crime and pedestrian) were associated with older adults’ MVPA, but that the presence of this association depended on the objectively measured built environment variable present in the models (GIS-based composite index of walkability vs. number of recreational facilities). That is, the perception of safety from crime was significantly associated with MVPA when walkability (as measured by the Walkability Index) was also included in the model, whereas the perception of pedestrian safety (e.g., presence of crosswalks, sidewalks) was significantly associated with MVPA when number of recreational facilities was included in the model.   1.2.3.3.4 Associations between composite indices of walkability and light physical activity  It seems important to also consider benefits of light physical activity, especially given that light physical activity was associated with older adults’ physical health and well-being (Buman et al.,   462010) and older adults with mobility impairments may be unable to perform more intense activities. However, to the best of my knowledge, only Van Holle et al. (2014) investigated older adults’ light physical activity (as measured by accelerometry), which they subdivided into low-light and high-light intensity. They found no associations between a GIS-based composite index (Walkability Index) of walkability and high-light physical activity. They reported a negative association between low-light physical activity and walkability. The authors proposed that this unexpected negative association may be a result of older adults who live in less walkable neighbourhoods spending more time engaged in indoor activities, such as housework (and thereby spending more time in low-light physical activity).  1.2.3.3.5 Associations between objective and perceived measures of the built environment and total physical activity volume Emerging evidence demonstrates that engaging in any physical activity, regardless of intensity, is beneficial to the health of adults (including older adults). For example, total physical activity volume may have stronger associations with cardiometabolic biomarkers than MVPA accumulated in bouts (Wolff-Hughes et al., 2015). However, similar to light physical activity, there is a dearth of studies that investigated the association between the built environment and total physical activity volume (as measured by accelerometry). Ding et al. (2014) found that total physical activity volume (counts/day) was associated with a composite index of walkability (Walkability Index) and perceptions of land use mix diversity among drivers only. The authors did not propose any explanations for this counterintuitive finding. I suspect that drivers are more free to choose whether they walk, since they have access to alternative modes of transportation (i.e., car), and so features of the built environment have a greater influence on their walking (and   47incidental physical activity). In contrast, McMurdo et al. (2012) found no associations between objective or perceived built environment features and older adults’ physical activity volume (counts/day).   1.2.3.3.6 Summary of findings In sum, published studies of the association between the built environment and older adult physical activity (as measured by accelerometry): i) were conducted in the USA or Europe; ii) were observational; iii) predominantly measured the built environment using a GIS-based composite indexes (Walkability Index); and iv) predominantly investigated MVPA outcomes. No studies: i) were conducted in Canada; ii) used audits to measure urban design built environment features; and iii) focused on individuals of low income. Collectively, the mixed findings across studies suggest that the association between the built environment and older adults’ MVPA is complex and may be influenced by a number of moderating factors. Further, until associations between the built environment and objectively measured light physical activity and total physical activity volume are evaluated more thoroughly, it is not possible to draw convincing conclusions about the nature of these associations. I address this gap and the limitations described above (with the exception of observational study design) in studies that comprise this dissertation.    481.2.3.3.7 Studies of the association between the built environment and walking for transportation (as measured by self-report) Since the systematic reviews by Van Cauwenberg et al. (2011) and Rosso et al. (2011), nine published studies investigated the association between the built environment and walking for transportation of community-dwelling older adults aged > 65 years. I describe these in Table 1.6.   49Table 1.6. Studies that investigated the association between the built environment and older adult walking for transportation (as measured by self-report)a.  Authors and countryb Objective(s) Study design and sample Built environment measurement approach, tool, domainc Measure: Outcomed Relevant built environment-related results Bracy et al. (2014).  USA. 1. Test the hypothesis that the association between the built environment and physical activity is moderated by perceptions of safety (crime, traffic, and/or pedestrian). Design: cross-sectional.  Sample: 718 older adults aged > 66 yrs. 1. GIS: Walkability Index.   2. Perceived (NEWS): crime safety; pedestrian safety; traffic safety. 1. CHAMPS: Walking for transportation (mins/wk) 1. No interactions present.  2. Main effects: a. Walkability (β= 8.89-9.46) consistently associated with walking for transportation (mins/wk) across models.  Cao et al. (2010).  USA.  1. To explore if neighbourhood design preserves accessibility for older adults by enabling a shift from driving to transit and walking, while controlling for neighbourhood preferences and attitudes towards transportation.  2. To examine whether neighbourhood design impacts the    Design: cross-sectional.  Sample: 1682 adults and older adults; 251 older adults aged > 65 yrs. 1. GIS: network distances between residences and commercial establishments (institutional; maintenance; eating-out; leisure)  2. Perceived: 34 characteristics evaluated on a 4-point scale. 1. Mailed survey: Walking for transportation (n times walked to store in past month) 1. Distance to the closest grocery store (β= -0.60) associated with walking for transportation (n times walked to store in past month).  2. Perceptions of safety (crime and traffic,  β= -0.33) and of shopping areas within walking distance (β= 0.29) associated with walking for transportation (n times walked to store in past month).    50Authors and countryb Objective(s) Study design and sample Built environment measurement approach, tool, domainc Measure: Outcomed Relevant built environment-related results travel of older adults more than younger people and to compare neighbourhood preferences of older adults and younger people  Carlson et al. (2012).  USA. 1. To explore how the built environment and psychosocial variables interact in explaining physical activity. Design: cross-sectional.  Sample: 718 older adults aged > 66 yrs. 1. GIS: Walkability Index; >1 park present; > 1 recreational facility present.   2. Perceived (NEWS-A): aesthetics (presence of trees; interesting things to look at; attractive natural sights; attractive buildings/homes) and presence of walking/cycling facilities (presence of sidewalks on most streets; sidewalks separated from road by parked cars; presence of buffer between road and sidewalk). 1. CHAMPS: Walking for transportation (mins/wk). 1. Interactions:  a. Walkability interacted with social support in explaining walking for transportation (mins/wk, β= 7.90, 95% CI = 1.42, 14.37). b. Walkability interacted with self-efficacy in explaining walking for transportation (mins/wk, β= 7.66, 95% CI = 1.21, 14.11). c. Walkability interacted with psychosocial barriers to physical activity in explaining walking for transportation (mins/wk, β= -8.26, 95% CI = -14.80, -1.72).  2. Main effects: a. Walkability associated with walking for transportation (mins/wk, β= 21.52, 95% CI = 14.91, 28.14).   Cerin et al. (2013).  China. 1. To examine associations of objectively measured prevalence and diversity of Design: cross-sectional.  Sample:484 older adults aged > 65 yrs. Audits (Environment in Asia Scan Tool – Hong Kong version): prevalence and diversity of destinations (public transit points; recreational; places of 1. IPAQ-Long form Chinese version: walking for transportation (overall, 1. Main effects: a. Associations between prevalence and diversity of destinations and walking for transportation were stronger and more consistent for within the neighbourhood (than overall) walking for transportation   51Authors and countryb Objective(s) Study design and sample Built environment measurement approach, tool, domainc Measure: Outcomed Relevant built environment-related results destination categories with overall and within-neighbourhood walking for transportation.  2. To examine the moderating effects of neighbourhood safety and pedestrian infrastructure on the above associations. worship; health clinics; government/public; entertainment; non-food retail and services; food and grocery stores and; restaurants); safety (stray animals; street lights; signs of crime/disorder; pedestrian safety) and; infrastructure (street slope; public facilities; good path conditions; path obstructions). mins/wk).  2. Neighbourhood Walking Questionnaire - Chinese version for Seniors: walking for transportation (within the neighbourhood, mins/wk). (mins/wk).  b. Signs of crime/disorder (β= 1.22, 95% CI = 1.01, 1.47) and diversity of recreational destinations and public transit points (β= 1.13, 95% CI = 1.02, 1.25) were associated with overall walking for transportation (mins/wk). c. Overall walking for transportation (mins/wk) was associated with a destination index comprised of the sumof the z-scores of the destination categories associated with this outcome (β= 1.06, 95% CI = 1.02, 1.10).  d. Signs of crime/disorder (β= 1.28, 95% CI = 1.05, 1.57), prevalence of street lights (β= 1.02, 95% CI = 1.01, 1.03), prevalence of stray animals (β= 0.97, 95% CI = 0.94, 1.00), prevalence of non-food retail and services (β= 1.01, 95% CI =1.00, 1.01), prevalence of food and grocery stores (β= 1.02, 95% CI = 1.00, 1.03), prevalence of restaurants (β= 1.01, 95% CI = 1.01, 1.02), diversity of health clinics/services (β= 1.45, 95% CI = 1.06, 1.99), diversity of places of worship (β= 1.45, 95% CI = 1.09, 1.92) and diversity of recreation (β= 1.16, 95% CI = 1.04, 1.31) associated with within-neighbourhood walking for transportation (mins/wk).  2. Interactions: e. Prevalence of public transit points interacted with prevalence of stray animals in explaining overall walking for   52Authors and countryb Objective(s) Study design and sample Built environment measurement approach, tool, domainc Measure: Outcomed Relevant built environment-related results transportation (mins/wk).  f. Diversity of recreation destinations interacted with signs of crime/disorder, diversity of recreation destinations interacted with stray animals, diversity of entertainment destinations interacted with signs of crime/disorder, prevalence of non-food retail and services interacted with path obstructions, prevalence of non-food retail and services interacted with sloping streets, prevalence of food and grocery stores interacted with path obstructions, and prevalence of food and grocery stores interacted with sloping streets in explaining within the neighbourhood walking for transportation (mins/wk). g. A destination index comprised of the sumof the z-scores of the destination categories associated with within the neighbourhood walking for transportation (mins/wk) interacted with path obstructions in explaining within the neighbourhood walking for transportation (mins/wk).  Ding et al. (2014).  USA.   1. To examine whether driving status moderates the association between the built environment and physical activity. Design: cross-sectional.  Sample: 880 older adults aged > 66 yrs. 1. GIS: Walkability Index; number of nearby locations for recreational physical activity (total number of parks and private recreational facilities).   2. Perceived (NEWS): microscale built environment (summary 1. CHAMPS: Walking for transportation (any vs. none). 1. No interactions present for walking for transportation (any vs. none).   53Authors and countryb Objective(s) Study design and sample Built environment measurement approach, tool, domainc Measure: Outcomed Relevant built environment-related results measure created by averaging scores on walking–cycling infrastructure, aesthetics, traffic safety, pedestrian safety structures, and transit access subscales); subscales that comprise the microscale built environment summary measure; residential density; land use mix; street connectivity. Inoue et al. (2011).  Japan 1. To examine the association between perceptions of the built environment and walking for specific purposes. Design: cross-sectional.  Sample: 1921 older adults aged 65-74 yrs. 1. Perceived (IPAQ-E): residential density; access to shops; public transport; sidewalks; bicycle lanes; access to exercise facilities; crime safety; traffic safety; social environment; aesthetics Study’s questionnaire: Walking for transportation for daily activity excluding commuting to work (<60 mins/wk vs. > 60 mins/wk). 1. Bicycle lanes (β= 1.26, 95% CI = 1.03, 1.54), access to exercise facilities (β= 1.26, 95% CI = 1.03, 1.55), social environment (β= 1.31, 95% CI = 1.06, 1.61), and aesthetics (β= 1.31, 95% CI = 1.07, 1.61) associated with odds of engaging in > 60 mins/wk of active transportation in the overall sample of participants.  2. Bicycle lanes (β= 1.41, 95% CI = 1.06, 1.87), crime safety (β= 0.69, 0.51, 0.94), traffic safety (β= 0.71, 95% CI = 0.53, 0.94), and aesthetics (β= 1.33, 95% CI = 1.00, 1.76) associated with odds of engaging in > 60 mins/wk of active transportation in subgroup analysis of men.  3. Access to shops (β= 1.57, 95% CI = 1.16, 2.13), access to exercise facilities (β= 1.39, 95% CI = 1.03, 1.88) and social environment (β= 1.42, 95% CI = 1.06, 1.92) associated with odds of engaging in > 60 mins/wk of active transportation in subgroup analysis of women. A. C. King et al. 1. To evaluate the relations among Design: prospective (2 1. GIS: Walkability Index 1. CHAMPS: Active 1. Walkability associated with walking for transportation (mins/wk, F= 32.82).   54Authors and countryb Objective(s) Study design and sample Built environment measurement approach, tool, domainc Measure: Outcomed Relevant built environment-related results (2011).  USA. objectively measured neighbourhood design and active transportation, moderate-to-vigorous physical activity (MVPA) and body weight.  2. To evaluate the moderating effect of mobility impairment on these relations. time points 6 months apart to control for seasonal variation in physical activity).  Sample: 647 older adults aged > 66 yrs. transportation (walking or biking for transportation, mins/wk).   2. Walkability interacted with mobility impairment in explaining walking for transportation (mins/wk). .  Van Cauwenberg et al. (2013).  Belgium.  1. To investigate the cumulative influence of perceived favourable environmental factors on older adults’ walking for transportation.  2. To investigate the moderating effect of perceived distance to destinations on this relation. Design: cross-sectional.  Sample: 50,658 older adults aged > 65 yrs. 1. GIS: residential density (rural, semi-urban, urban).   2. Perceived (study’s tool): absence of high curbs; number of shops ; presence of: benches; crossings; bus stops; street lighting; safety from crime (measured using the Elderly Feelings of Safety scale); environmental index score created by combining the above variables; access to destinations (distance from home). 1. 5-pt scale: Walking for transportation (almost daily vs. less than almost daily). 1. Urban (vs. rural) area of residence (β= 0.306, 95% CI = 0.19, 0.43) associated with almost daily walking for transportation.   2. Presence of bus stops (β= 1.29, 95% CI = 1.22, 1.37), street lighting (β= 1.12, 95% CI = 1.04, 1.19), number of shops (β= 1.20, 95% CI = 1.15, 1.26) and safety from crime (β= 1.08, 95% CI = 1.03, 1.13) associated with almost daily walking for transportation.  3. Independent of the presence of other favourable environmental factors, “large” distance to destinations (vs. small) associated with almost daily walking for transportation (β= 0.63, 95% CI = 0.23, 1.02).  4. Distance to destinations interacted with the environmental index in explaining almost daily walking for transportation.   55Authors and countryb Objective(s) Study design and sample Built environment measurement approach, tool, domainc Measure: Outcomed Relevant built environment-related results a. Within “short” distances to destinations, odds of almost daily walking for transportation higher for 7 (β= 0.41, 95% CI = 0.39,0.43) vs. 3 (β= 0.36, 95% CI = 0.34, 0.38), 4 (β= 0.36, 95% CI = 0.34, 0.38), or 5 (β= 0.37, 95% CI = 0.35, 0.38) favourable environmental factors. b. Within “medium” distances to destinations, odds of almost daily walking for transportation increased from 0.22 (95% CI = 0.16, 0.28) for 0 to 0.31 (95% CI = 0.29, 0.33) for 4 and 0.30 (95% CI = 0.29, 0.32) for 5 favourable environmental factors.  c. Within “large” distances to destinations, there was no association between almost daily walking for transportation and number of favourable environmental factors.                                                                           56Authors and countryb Objective(s) Study design and sample Built environment measurement approach, tool, domainc Measure: Outcomed Relevant built environment-related results Van Holle et al. (2014).  Belgium   1. To investigate the relation between neighbourhourhood walkability and older adults' physical activity.  2. To examine the moderating effect of neighbourhood income on these relations. Design: cross-sectional.  Sample: 438 older adults aged > 65 yrs. 1. GIS: Walkability Index IPAQ: walking for transportation (mins/wk). 1. Walkability associated with walking for transportation (mins/wk, β= 4.63, 95% CI = 3.58, 5.68). aIncluded studies published after the reviews by Van Cauwenberg et al. (2011) and Rosso et al. (2011) that used multivariable analyses to investigate the association between the built environment and walking for transportation of community-dwelling older adults aged > 65 years. bBracy et al. (2014), Carlson et al. (2012), Ding et al. (2014) and A. C. King et al. (2011) analyze data from the same study [the Senior Neighbourhood Quality of Life Study (SNQLs)]. cGIS = Geographic Information Systems; IPAQ-E = International Physical Activity Questionnaire Environmental Module; NEWS = New England Walkability Scale; NEWS-A = New England Walkability Scale Abbreviated; OPAL = Older People's Active Living questionnaire. dCHAMPS = Community Healthy Activities Model Program for Seniors physical activity questionnaire; IPAQ = International Physical Activity Questionnaire. 57Five of the studies in Table 1.6 were set in the USA (Bracy et al., 2014; Cao et al., 2010; Carlson et al., 2012; Ding et al., 2014; A. C. King et al., 2011), two in Europe (Van Cauwenberg et al., 2013; Van Holle et al., 2014), and two in Asia (Cerin et al., 2013; Inoue et al., 2011). As with the studies in Table 1.5, all but one (A. C. King et al., 2011) had a cross-sectional design. There is heterogeneity among the studies in approach used to measure the built environment; notably, one (Cerin et al., 2013) used audits to measure features of the built environment. The majority (six) investigated the association between the built environment and duration of walking for transportation (Bracy et al., 2014; Carlson et al., 2012; Cerin et al., 2013; Inoue et al., 2011; A. C. King et al., 2011; Van Holle et al., 2014); three investigated the association between the built environment and frequency of walking for transportation (Cao et al., 2010; Ding et al., 2014; Van Cauwenberg et al., 2013). I now discuss key findings from these studies, organized by approach used to measure the built environment and dimension of walking for transportation.  1.2.3.3.8 Associations between composite indices of walkability and duration of walking for transportation Studies consistently reported a positive association between a GIS-based composite index of walkability (Walkability Index) and duration of self-reported walking for transportation (Bracy et al., 2014; Carlson et al., 2012; A. C. King et al., 2011; Van Holle et al., 2014). Moreover, the study by Carlson et al. (2012) also noted a positive interaction between walkability, self-efficacy and social support and a negative interaction between walkability and psychosocial barriers in explaining this outcome.    581.2.3.3.9 Associations between audit measures of the built environment and duration of walking for transportation One study provided preliminary evidence for types of destinations associated with older adults’ walking for transportation, as well as the urban design (pedestrian infrastructure such as sidewalk quality and sloping streets) and safety (traffic and crime) variables that may moderate these associations (Cerin et al., 2013). Unlike others, this study used a self-report measure that differentiated between within-neighbourhood and overall walking for transportation and used audits to measure built environment features. Walking for transportation within one’s neighbourhood was associated with the prevalence or diversity of six destination categories (non-food retail and services, food and grocery stores, restaurants, health clinics/services, places of worship, and recreation). There were seven interactions between sidewalk obstructions, sloping streets, and/or crime/disorder and the destination categories and an interaction between a destination-based composite index (calculated using the Environment in Asia Scan Tool – Hong Kong version) and path obstructions. In terms of overall walking for transportation, there were positive associations between diversity of recreational destinations, prevalence of public transportation points and a composite index of the two destinations variables; there was an interaction effect between prevalence of public transit points and prevalence of stray animals.   1.2.3.3.10 Associations between perceived measures of the built environment and duration of walking for transportation Of studies that used perceived measures of the built environment, only Inoue et al. (2011) reported associations with duration of walking for transportation. Perceptions of the presence of bicycle lanes, access to exercise facilities, pleasing aesthetics, and the social environment (e.g.,  59seeing people walking) were positively associated with engaging in > 60 minutes of walking for transportation/week (vs. < 60 minutes of transportation walking/week). For men, presence of bicycle lanes and pleasing aesthetics were positively and safety from crime and safety from traffic were negatively associated with engaging in > 60 minutes of walking for transportation/week (vs. < 60 minutes of walking for transportation/week). For women, access to shops, exercise facilities, and the social environment were positively associated with engaging in > 60 minutes walking for transportation/week /week (vs. < 60 minutes of walking for transportation/week).   1.2.3.3.11 Associations between the built environment and frequency of walking for transportation All three studies that investigated the association between the built environment and frequency of walking for transportation used perceived measures of the built environment (Cao et al., 2010; Ding et al., 2014; Van Cauwenberg et al., 2013). In addition, Ding et al. (2014) used GIS-based measures of single built environment features and a composite index of walkability (Walkability Index), while Cao et al. (2010) used GIS-based measures of single built environment features. Ding et al. (2014) found that a GIS-based composite index of walkability and perceptions of residential density, land use mix diversity and access, walking or cycling infrastructure, aesthetics, pedestrian safety structures, transit access and overall microscale attributes were associated with frequency of walking for transportation (any vs none) among drivers and non-drivers; in addition, street connectivity was associated with frequency of walking for transportation (any vs. none) among drivers only. Van Cauwenberg et al. (2013) found that the presence of bus stops, street lighting, shops and services, safety from crime, and living within  60short distance (vs. medium or large distances) were associated with frequency of walking for transportation. Further, there was a significant interaction between perceived distance to destinations and the cumulative influence of built environment features on older adults’ frequency of walking for transportation. Specifically, among participants who perceived that they were living a short distance from destinations, the probability of walking for transportation increased when > seven favourable built environment features were present. Among participants who perceived they were living a medium distance from destinations, there was an increase in the probability of walking for transportation in the presence of > four favourable built environment features. Among participants who perceived they were living long distance from destinations, there was no association between the cumulative presence of built environment features and walking for transportation. Finally, Cao et al. (2010) found a negative association between frequency of walking for transportation (to the store) and a GIS-based measure of distance to the closest grocery store and perception of safety. There was also a positive association between frequency of walking for transportation and perception of shopping areas within walking distance.  1.2.3.3.12 Summary of findings  In sum, studies that investigated the association between the built environment and older adults’ walking for transportation were set in USA, Europe, or Asia, all but one had a cross-sectional design, and none targeted individuals of low income. Studies most frequently used GIS-based measures of the built environment; only one study used environmental audits. Geographic Information Systems-based composite-indexes of neighbourhood walkability were consistently associated with duration of walking for transportation. Destinations and perceptions of aesthetics  61and safety may also be associated with duration and frequency of walking for transportation. There is some evidence to suggest that select person and environment-level variables may moderate the association between the built environment and older adults’ duration and frequency of walking for transportation.  1.2.3.4 Overall summary of the findings and direction for future research  I identified several gaps in the built environment-older adult mobility literature. Importantly, no Canadian studies investigated the association between the built environment and physical activity (as measured by accelerometry) and walking for transportation of older adults aged > 65 years. To the best of my knowledge, no studies targeted older adults of low income. Further, published studies provided only low levels of evidence as most had a cross-sectional design; none were controlled or RCTs. Thus, high quality studies are sorely needed to investigate causation and pathways by which the built environment can influence older adult (including older adults of low income) physical activity and walking for transportation. In terms of approaches used to measure the built environment, the most evidence exists for the association between GIS-based measures of walkability, destinations, street connectivity and perceptions of safety and aesthetics and physical activity and walking for transportation. Only one study used audits to measure urban design-based features of the built environment. This most likely reflects the resource intensive nature of audits. Finally, with respect to approaches used to measure older adults walking for transportation, most studies used self-report questionnaires that relied upon participants’ recall of past events. Moreover, almost all studies that used accelerometry to assess physical activity focused on MVPA. Given that many older adults may be challenged by mobility-disability, there is a clear need to investigate the association between the built environment and a wider spectrum of physical activity (including light physical activity and total  62physical activity volume). It also seems important to consider incidental physical activity such as walking for transportation using measures such as travel diaries so as to minimize recall bias and more comprehensively measure the outcome across domains of travel behaviour.   Aim, rationale, and objectives In this final section of Chapter 1, I outline the aims, rationale, and objectives of the four studies that make up this dissertation. These four studies, together, represent an integrated whole.  1.2.4.1 Study 1. Agreement between virtual and in-the-field environmental audits of Assisted Living sites Rationale. Features of the built environment that affect older adults’ walking can be measured at the macroscale (neighbourhood) level (e.g., land use mix, street connectivity) and the microscale (street) level (e.g., sidewalk and road characteristics). Almost all studies that investigated the association between objectively measured features of the built environment and older adults’ walking focused on macroscale features as they are more easily captured using GIS (Rosso et al., 2011). However, microscale features can support or pose a direct challenge to older adults’ ability to walk outdoors (e.g., presence or absence of benches and uneven or cracked sidewalks). These challenges are exacerbated if a person has trouble walking or lacks confidence in their walking ability. When the ‘environmental’ challenge is greater than an older adult’s perceived or real physical capacity, they may be less likely to engage with the environment (i.e., go outdoors) (Noreau & Boschen, 2010). This is of utmost concern in neighbourhoods with a high proportion of older adults with limited mobility; for example, in proximity to assisted living facilities.   63Environmental audits use in-person observation to directly and systematically measure microscale features of the built environment (Brownson et al., 2009). However, the resource intensive nature of audits (i.e., time and cost to travel to sites to conduct audits) limits their feasibility, especially across large geographic areas (Brownson et al., 2009). Google Earth® is a free, internet-based application that provides panoramic imagery of cities worldwide (Google Inc., CA). Street View feature within Google Earth provides high-resolution, street-level images that have potential to support virtual audits of micro- and macroscale features of neighbourhoods and communities. Importantly, they do not require many of the resources necessary to gather data in-the-field. Seven studies assessed the agreement between virtual and in-the-field environmental audits (Badland, Opit, Witten, Kearns, & Mavoa, 2010; Ben-Joseph, Lee, Cromley, Laden, & Troped, 2013; Clarke, Ailshire, Melendez, Bader, & Morenoff, 2010; C. M. Kelly, Wilson, Baker, Miller, & Schootman, 2012; Rundle, Bader, Richards, Neckerman, & Teitler, 2011; B. T. Taylor et al., 2011; Wilson et al., 2012). However, no previous study evaluated agreement between virtual and in-the-field administration of an environmental audit tool designed specifically to assess microscale features of the built environment that influence older adults’ walking.   Therefore, for this study I focused on geographic areas in proximity to assisted living facilities, as this is a setting where a high concentration of older adults with mobility impairments may reside (Giuliani et al., 2008). Further, as many of these older people may have mobility-impairments, the built environment could have great impact on their walking habits (A. C. King et al., 2011; Rantakokko et al., 2010; Rantakokko et al., 2009; Satariano et al., 2010; Shumway-Cook et al., 2003). Thus, developing an effective methodology to assess these geographic areas  64may have clinical relevance and policy implications. For example, cities could adapt the microscale built environment in these neighbourhoods as a priority.  Objectives. My primary objective is to compare the absolute agreement between environmental audits of assisted living sites conducted virtually (using Google Earth’s Street View feature) versus in-the-field audits, with a focus on items that are especially relevant for older adults.  Hypothesis. Absolute agreement will be > 80% for items evaluated by virtual and in-the-field audits. The exceptions are temporally unstable items and items that evaluate fine-grain details.  Contribution. To my knowledge, this is the first study to evaluate whether Google Earth’s Street View feature is a feasible and reliable option for gathering information about features of the built environment that support older adults’ walking. This study will provide insights into specific features of the built environment that virtual audits capture and identify potential barriers to using this approach. These findings may provide researchers in the area of the built environment and older adult mobility with a feasible and affordable option to assess the role of neighbourhoods on older adult walking. Further, results can inform the transportation sector of municipalities about areas where there are barriers to large populations of older people engaging with their communities.   651.2.4.2 Study 2. Walk the Talk: Characterizing mobility of older adults living on low income Rationale. Mobility is a fundamental component of active aging. It enables older people to engage in daily activities (e.g., housework, shopping), physical activity (e.g., walking for exercise), and generally participate in society (e.g., getting places by car or public transport) (Canadian Institutes of Health Research, n.d.; Satariano et al., 2012). The built environment plays an important role in older adult mobility as it provides the context where outdoor mobility takes place. However, older adult outdoor mobility is a result of the dynamic interplay between the individual’s capacity and the supports and pressures present in the environment – a two way interaction (Lawton & Nahemow, 1973). Financial status is one domain of person level factors that influence the capacity to be mobile (cognitive, physical, and psychosocial are others) (Webber et al., 2010). Specifically, older adults of low SES may be at increased risk of functional (Huisman et al., 2003; Koster, Bosma, van Groenou, et al., 2006; Koster et al., 2005; Shumway-Cook et al., 2005) and economic limitations that interfere with their capacity to be mobile. They may also be less likely to travel to places by car (Cao et al., 2010; Frank et al., 2010; Turcotte, 2012). Despite their unique mobility-related needs and characteristics, older adults of low SES represent an understudied segment of the older adult population. Thus, I address the need to better understand the potentially unique characteristics, including mobility and health profiles, of older adults of low SES (as defined by income).   Objectives. My objective is to comprehensively describe person and environment-level characteristics and mobility of older adults living on low income across a diverse range of built environments in Metro Vancouver.  66Contribution. To the best of my knowledge, this is the first study to describe mobility and its person and environment-level domains in Canadian older adults living on low income. My findings will describe this often-overlooked population, including the scope of their mobility. I envision that what I learn will inform future research studies and provide insight for those who interact directly with, and provide services for, this population [e.g., BC Housing, municipalities (e.g., parks and rec managers, social planners, transportation planners)].   1.2.4.3 Study 3. Destinations matter: The association between where older adults live and their travel behaviour Rationale. Retaining the relative freedom to walk or drive is integral to healthy aging and quality of life (Satariano et al., 2012). A small amount of regular walking can play a key role to maintain functional independence in old age (Simonsick, Guralnik, Volpato, Balfour, & Fried, 2005). As older adults typically leave their homes to travel to specific destinations, a high prevalence of destinations in a neighbourhood may encourage older adults to leave their homes, as well provide an opportunity to walk, instead of drive – the potential result is a consequent increase in incidental physical activity. There was a positive association between the presence and proximity of various destinations and older adults’ walking (Cao et al., 2010; Gauvin et al., 2012; Michael, Beard, Choi, Farquhar, & Carlson, 2006; Michael et al., 2010; Nagel, Carlson, Bosworth, & Michael, 2008; Nathan et al., 2012), walking for transportation (Cao et al., 2010; Cerin et al., 2013; D. King, 2008), and physical activity (Cao et al., 2010; Cerin et al., 2013; D. King, 2008). However, the specific types of destinations that encouraged walking were not consistent. This may be due to broad categories used to classify destinations, and inconsistency as to specific types of destinations included in destination categories. Thus, we know  67surprisingly little about the specific types of destinations older adults deem relevant. We know even less about destinations relevant to older adults living on low income. We also do not know whether the association between prevalence of neighbourhood destinations and mobility holds true in a group of older adults living on low income. As older adults living on low income are less likely to travel by car (Cao et al., 2010; Frank et al., 2010; Turcotte, 2012) and may also be at greater risk of functional limitations that impede walking (Koster, Bosma, van Groenou, et al., 2006; Koster et al., 2005), accessible neighbourhood destinations may be especially important to them.  Objectives. My primary objective is to describe specific destinations that older adults living on low income most commonly travel to in a week. My secondary objective is to determine the association between the prevalence of neighbourhood destinations and the number of walking for transportation trips (average/day) these older adults make.  Contribution. This is the first study to use seven-day travel diaries to assess specific travel destinations of older adults. Specifically, I provide preliminary evidence regarding destinations most relevant to a population of older adults living on low income. These outcomes fill a gap in the existing literature as no studies investigated the association between neighbourhood destinations and walking for transportation in this population. My results can be used to inform those who directly interact with older adults living on low income including their families, housing organizations (e.g., BC Housing), municipal social and transportation planners, and Ministries of Transportation and Health. Municipalities can utilize my findings to retrofit and develop neighbourhoods that support the mobility, health, and independence of older adults.  68Finally, my results are hypotheses generating and will inform design of intervention studies and guide researchers to develop destination-based built environment measures that are relevant to older adults living on low income.  1.2.4.4 Study 4. Can older adults’ neighbourhood walkability promote a walk to the shops Rationale. Current physical activity guidelines recommend that older adults engage in > 150 minutes of MVPA/week (Chodzko-Zajko, Proctor, Singh, et al., 2009; Tremblay et al., 2011). However, engaging in any physical activity (including light physical activity) is increasingly recognized as an important public health goal, especially for older adults (I. Lee, 2015; Sparling et al., 2015). Walking, and outdoor walking specifically, is the most common form of physical activity for older adults (Dai et al., 2015). Consequently, the built environment may represent either an important facilitator or barrier to older adults’ physical activity. Findings are mixed regarding specific features of the built environment most strongly associated with older adult physical activity and specifically outdoor walking (Hanson, Ashe, McKay, & Winters, 2012; Rosso et al., 2011; Van Cauwenberg et al., 2011; Yen et al., 2014). Older adults are diverse in terms of their physical, psychosocial and cognitive characteristics (Nelson & Dannefer, 1992) (among many other differences), thus mixed findings are not at all surprising. Further, there are no studies that targeted older adults living on low income. It seems important to do so as there are likely unique person-level characteristics that influence their capacity to be physically active, different types of physical activities they may engage in, and they may rely less heavily on travel by car to meet their daily needs. This part of my dissertation acknowledges both this gap and the  69need to better understand the association between features of the built environment and physical activity habits of older adults living on low income.  Objectives. My primary objective is to evaluate the association between the built environment and total physical activity of older adults living on low income. My secondary objective is to describe the association between the built environment, light physical activity, MVPA, and self-reported walking for transportation.   Contribution. Most studies that evaluated the association between physical activity and features of the built environment focused on MVPA. I extend this body of literature by assessing physical activity objectively (with accelerometry) across the spectrum of light to moderate-to-vigorous intensity activities. The importance of activities across this spectrum has become increasingly recognized as important for older adults; yet it is currently understudied. As per my other studies, my focus on older adults living on low income is unique and long overdue. My results can be used to inform those who directly interact with older adults living on low income, including their families and those who provide services or infrastructure to serve the needs of older British Columbians - municipal parks and recreation departments, social and transportation planners and the Ministry of Health. Municipalities can utilize my findings to support or adapt transportation plans that aim to encourage more active modes of transportation and infrastructure that supports this goal. Finally, my results are hypotheses generating and will inform design of intervention studies that aim to enhance features of the built environment that encourage physical activity and promote the health of older adults living on low income.   70Chapter  2: Methods In this chapter I describe the study design and methods common to Chapters 4-6.  2.1 Walk the Talk overview Walk the Talk: Transforming the Built Environment to Enhance Mobility in Seniors (Walk the Talk) is a cross-sectional study that used a mixed-methods approach to evaluate the association between the built environment and mobility and health of community-dwelling older adults living on low income across Metro Vancouver. Only the study’s quantitative methods relevant to my dissertation are discussed herein. Walk the Talk was approved by the University of British Columbia’s Clinical Research Ethics Board.   Design of Walk the Talk: Engaging community stakeholders  In April 2010, the Walk the Talk research group held a community forum in Vancouver, British Columbia to identify key research priorities related to the built environment and healthy aging. Forum participants represented a broad range of community stakeholders: community members, health care providers, researchers and representatives from municipal and provincial government, non-for profit organizations, and industry. Among issues raised at the forum was the importance of understanding the needs of vulnerable subgroups of older adults and the need to conduct research that addresses existing community needs. One community stakeholder, BC Housing, noted that we know very little about those in receipt of rental subsidies and they expressly communicated their interest in collaborating with our research group.   71 BC Housing as key community partner BC Housing is a provincial crown agency that provides affordable housing options across a continuum that extends from emergency shelters to public housing and rental assistance in the private market (BC Housing, n.d.-a). BC Housing services the most vulnerable members of society – the homeless, low income older adults and families, individuals with disabilities, women and children at risk of violence, and individuals who are Aboriginal (BC Housing, n.d.-a). BC Housing’s main intentions for partnership were; i) to obtain data on the sociodemographic, health, and neighbourhood profiles of its older adult clients, namely recipients of a Shelter Aid for Elderly Renters (SAFER) rental subsidy; ii) to better understand how neighbourhood (built environment) features may affect clients’ mobility; and iii) to apply results of i) and ii) to inform future investments in housing developments/re-developments. A close partnership with BC Housing was established, and provided an opportunity to: i) study aspects of the built environment and aging that reflected research priorities identified as important by community stakeholders; ii) gain access to a well-defined source population of older adults living on low income; and iii) focus upon a subgroup of the population that is under-represented in current research, yet potentially more reliant on features of the built environment that facilitate walking.    Integrated knowledge translation: approach to collaboration with BC Housing Integrated knowledge translation is a research approach that involves collaboration between researchers and knowledge users throughout the research process, with the aim of engaging in a mutually relevant and beneficial project (Canadian Institutes of Health Research, 2012; Kothari & Wathen, 2013). The Walk the Talk team’s partnership with BC Housing exemplified the  72integrated knowledge translation approach. We began our collaboration with BC Housing early on, during the development of the grant proposal that led to the subsequent funding of the Walk the Talk study. BC Housing helped us to develop our research question, including the identification of our source population, and was an active participant in the development of the study design, and especially sampling and recruitment. For example, we worked closely with BC Housing to develop and execute a sampling and recruitment protocol that satisfied their privacy concerns and carried out participant recruitment from within BC Housing’s office. We also selected data collection tools that reflected BC Housing’s research questions of interest (e.g., travel diaries for identification of destinations relevant to their clients). Finally, once we collected and began to analyze participants’ data, we presented our findings to BC Housing to disseminate the research results, as well as solicit input for their interpretation.         The Shelter Aid for Elderly Renters (SAFER) program BC Housing’s SAFER program aims to help make housing affordable through provision of a monthly cash rental subsidy to eligible older adult residents of BC. The program is based on a sliding scale that reimburses part of the difference between 30% of a recipient’s total income and the money the recipient spends on rent, with the most money given to recipients with the smallest income (BC Housing, n.d.-b). Eligibility criteria for SAFER are listed in Table 2.1 (BC Housing, n.d.-b).    73Table 2.1. Eligibility criteria for the Shelter Aid for Senior Renters (SAFER) programa Eligibility criteriab Ineligibility criteriac Aged > 60 years. Reside in subsidized housing or a residential care facility funded by the Ministry of Health. Lived in British Columbia for > 12 months immediately preceding SAFER application. Reside in co-operative housing and are a shareholder. Canadian citizen, resident, refugee status applicant, or individual for whom private sponsorship has broken downd. Receive (personally or through family) income assistance through the BC Employment and Assistance Act or the Employment and Assistance for Persons with Disabilities Act (excluding Medical Services only). Pay > 30% of gross monthly household income towards rent of residence. Resident of Metro Vancouver with a gross monthly income that exceeds $2550 (singles), $2750 (couples), $1776 (shared accommodatione).  Resident outside of Metro Vancouver with a gross monthly income that exceeds $2223 (singles), $2423 (couples), $1776 (shared accommodatione). aData obtained from (BC Housing, n.d.-b).  bAn individual must meet all these criteria to be eligible.  cPresence of any of these criteria deems an individual ineligible.  dApplies to applicant and his/her spouse with whom he/she is living.  eSingle older adult sharing a residence with one other adult.   The income of SAFER recipients relative to markers of low SES in Canada The average before-tax household income of SAFER recipients was approximately $18,000 in 2011 (Chudyk, personal communication, August 23, 2013). This is approximately three times lower than the 2011 Canadian average after-tax income for families where the major income earner is > 65 years old ($57,700) (Statistics Canada). This is also approximately one-and-a-half to two times as low as the average after-tax income for unattached Canadians aged > 65 years ($34,400 for unattached males and $29,700 for unattached females) (Statistics Canada, 2013). Further, the low income cut-off (LICO) is a before-tax income threshold used by Statistics Canada to identify individuals and families that are likely to devote a larger share of their income on the necessities of food, shelter and clothing than the average family (> 20% than the average  74family) (Statistics Canada, 2015a). Low income cut-offs vary according to the size of a family unit and size of the community of residence (Table 2.2). The mean before-tax income of SAFER recipients falls below the LICO for one and two persons families that reside in census agglomerations and census metropolitan areas (Statistics Canada, 2015a).   Table 2.2. 2011 Low income cut-offs by size of family unit and communitya Community size (no. inhabitants) Census Agglomeration (CA) Census Metropolitan Area (CMA) Size of family unit < 30,000 30,000 - 99,999 100,000 - 499,999 > 500,0001 person 18,246 19,941 20,065 23,2982 persons 22,714 24,824 24,978 29,004aData obtained from (Statistics Canada, 2015a).  2.2 Study design In this section I discuss all aspects of Walk the Talk’s setting, sampling frame, recruitment, and data collection, processing and analyses. I only discuss the aspects relevant to this dissertation. I was directly involved in all of the discussed aspects unless otherwise specified.    Setting Walk the Talk was set in Metro Vancouver, BC, Canada. Metro Vancouver comprises 21 urban and suburban municipalities (Statistics Canada, 2012), and was home to 2,313,330 people in 2011; approximately 13.5% are aged > 65 years (Statistics Canada, 2012). Study recruitment was restricted to municipalities within a one -hour travel time from the research center, for practical purposes. The study area (Figure 2.1) consisted of eight municipalities (Burnaby, New Westminster, North Vancouver, Richmond, Surrey, Vancouver, West Vancouver, White Rock).   75 Figure 2.1. Map of study and surrounding area. Base map source: iMapBC. Copyright Province of British Columbia. All rights reserved. Reproduced/adapted with permission of the Province of British Columbia.   Inclusion and exclusion criteria Walk the Talk included community-dwelling older adults who were aged > 65 years, current recipients of a SAFER rental subsidy, and residents of the study area (Burnaby, New Westminster, North Vancouver, Richmond, Surrey, Vancouver, West Vancouver, White Rock). We excluded individuals who were unable to understand and/or speak English, had been diagnosed with dementia, left their house to go into the community < once in a typical week,  76were unable to walk > 10 meters with or without a mobility aid (e.g., cane, walker), and/or were unable to participate in a mobility assessment where they were asked to walk four meters.    Pilot study Prior to sampling, we carried out a pilot study to assess the feasibility of our recruitment approach, ensure that measurement sessions were approximately two hours in duration, and review and refine study measures, as necessary. We recruited participants for the pilot study in September 2011; measurement sessions took place November 7th and 8th, 2011 and were conducted at two sites (Chuck Bailey Centre, 13458 107th Avenue, Surrey, BC and the West End Community Centre, 870 Denman Street, Vancouver, BC) to maximize convenience for participants. Six participants took part in the pilot study.    Sampling approach for main study Walk Score® is a publicly available index that calculates the walkability of a residence on a scale of 0-100, based on its distance to surrounding destination categories (www.walkscore.com). Walk the Talk aimed to sample participants across the range of walkability (as measured by the Walk Score) within the study area. Thus, the Walk the Talk team (Dr. M. Winters) generated a stratified sample of SAFER recipient postcodes using the following steps, in order to limit privacy concerns: 1. Identified postcodes of SAFER recipients aged > 65 years that were current residents of the sampling area. BC Housing provided these postcodes and assigned a unique identifier to each recipient so they were unidentifiable. 2. Generated a random sample of 2500 postcodes (six-digit) within the study area.   773. Sent the random sample of postcodes and SAFER post codes (with no identifiers) to Walk Score to obtain Walk Scores for each post code.  4. Used Walk Scores from a random sample of postcodes (Step 2) to determine cutpoints for Walk Score deciles in the study area. The upper cut-points (deciles) were 100(1), 93(2), 87(3), 78(4), 72(5), 67(6), 60(7), 52(8), 43(9), and 32(10).  5. Used the SAFER recipient list (Step 1) to place each SAFER recipient into deciles of Walk Score (from Step 4). 6. Examined the distribution of SAFER recipients within each decile of Walk Score. Since we aimed to evaluate approximately 20 participants from each decile of Walk Score, and estimated a 10% recruitment rate, we defined adequate distribution as a minimum of 150 SAFER recipients per decile. The estimated recruitment rate was based on results from our pilot study. 7. Randomly sampled 200 individuals identified in Step 5 from within each decile of Walk Score (ntotal = 2000 individuals).  8. Provided a list of randomly sampled individuals to BC Housing so they could link anonymized data back to personal mailing information, via the unique identifier (Step 1).    Recruitment  I now discuss Walk the Talk’s recruitment strategy, including recruitment target and process.   2.2.5.1 Recruitment target Walk the Talk’s recruitment target was 400 participants across two years (200/year). This corresponds to approximately 5% of SAFER recipients residing in Metro Vancouver. The  78sample size represents an estimate of the number of individuals that could be recruited during the study period with available study resources. This sample size was deemed adequate to detect modest correlations between older adults’ mobility (travel behaviour and physical activity) and the built environment. With 400 participants, there was > 80% power to detect a bivariate correlation of 0.15 or higher, assuming alpha = 0.05 (two-sided) and up to 10% missing or incomplete data.  On August 1, 2012, we decided to forego a second year of data collection, at which time 161 participants had consented to participate. Given the difficulties encountered in recruiting the first 161 participants, we thought it very unlikely we could reach our target of 400. In addition, although there was a lack of any associations between physical activity outcomes (as measured by accelerometry) and the built environment (assessed using Walk Score), we deemed that there were sufficient numbers and variability in the data collected to facilitate other analyses (e.g., travel behaviour outcomes).   2.2.5.2 Recruitment process overview We mailed packages (Appendix A) and placed follow-up recruitment phone calls to potential participants. BC Housing staff generated labels with names and mailing addresses of individuals identified in Step 8 of the sampling approach. To respect BC Housing’s privacy considerations, we traveled to the BC Housing main office to place labels on recruitment packages. BC Housing staff mailed out recruitment packages in batches of 500 (four mail-outs in total). This ‘stagger’ was designed to manage the timing of follow-up recruitment calls we made to invited individuals. Packages included study consent forms and letters from BC Housing and the Walk  79the Talk team to introduce the study. We placed follow-up recruitment phone calls a minimum of three working days following a mail-out.  2.2.5.3 Follow-up recruitment phone calls Prior to the first day of follow-up recruitment phone calls, BC Housing staff prepared a document with the name, address, phone number, and unique identifier of each individual sent a recruitment package; a separate list was created for each of the four mail-out batches. At the request of BC Housing, we managed lists using a paper-based system and kept them under lock-and-key at the BC Housing Head Office. We took individuals’ contact information off-site to our study centre (Centre for Hip Health and Mobility, CHHM) only after an individual gave verbal consent to participate in the study.   The purpose of recruitment calls was fourfold: i) to briefly explain the study; ii) answer questions; iii) gauge interest; and iv) confirm eligibility of potential study participants. We attempted to contact each sampled individual a maximum of two times by phone for recruitment into the study. We recorded all attempts to establish contact in a hand-written log; recorded information included date of the phone call, initials of the team member that attempted to make contact, and whether or not contact was made. We documented the result of each unsuccessful attempted phone contact [i.e., wrong number/number not in service; “non-responder” (unable to reach by phone but phone is in service); and no direct way to contact (e.g., lives in a hotel, lives in a residence with a switchboard operator, etc.)]. For the “non-responders” we also further documented whether: i) there was no answer and no answering machine; ii) we left a message on a machine; or iii) we left a message with a friend/family member. Of the individuals we reached  80by phone, we documented whether the individual: i) agreed to participate; ii) did not meet inclusion/exclusion criteria [e.g., did not speak English; stated that he/she was unable to walk > 10 meters with or without a mobility aid, etc.); and iii) declined participation. For those that declined participation, we recorded reason given (e.g., not interested, poor health, etc.). Finally, we also recorded when a family member/friend informed us that a potential study participant was deceased.  We conducted phone calls until we contacted every individual mailed a recruitment letter or made two unsuccessful attempts to contact them. At the completion of recruitment, BC Housing staff removed individuals’ personal identifiers (i.e., name, address, telephone) from the logs we used to document recruitment phone calls. We took de-identified logs back to CHHM to document flow of participants into the study.   Informed consent process We included an informed consent form (Appendix A) in the recruitment package mailed to all potential participants. We reviewed the informed consent form with each individual during the follow-up recruitment phone call and on the day of his/her measurement session. We obtained written informed consent from each participant on the day of his/her measurement session, prior to him/her taking part in any study-related assessments. As part of the International Committee on Harmonization guidelines, we gave each participant a copy of the informed consent form signed and dated by the participant and a research team member. We also kept a signed and dated copy of the informed consent form in locked storage cabinets as part of study records.   81 Incentives We provided participants with a $20 honorarium (gift card to a grocery store) for their involvement in the study.    Data collection Below I describe Walk the Talk’s data collection methods, including measures collected by the study.  2.2.8.1 Data collection overview  Each participant attended one measurement session during the study period, conducted March – April 2012. We also offered a single measurement session in May for those (n = 4) unable to take part during original data collection. We held measurement sessions at CHHM and, where possible, at community centres near the participants’ residences (Table 2.3). A research assistant provided participants with transportation in a study rental van to/from measurement sessions, as needed.   Table 2.3. Name and location of data collection sites Site name Address Bonsor Seniors Centre 6550 Bonsor Avenue, Burnaby, BC Confederation Seniors Centre 4585 Albert Street, Burnaby, BC Centre for Hip Health and Mobility 2635 Laurel Street, Vancouver, BC Elim Housing Society 9067 160th Street, Surrey, BC Harry Jerome Community Centre 123 East 23rd Street, North Vancouver, BC Killarney Community Centre 6260 Killarney Street, Vancouver, BC  Each measurement session lasted approximately two hours. Participants completed self-report questionnaires about their neighbourhood built environment, physical activity, health and social  82environment. We also directly assessed participants’ body composition, cognition, height, lower-extremity function, and weight. At the end of each measurement session, we gave participants travel diaries, GPS, accelerometers, and accelerometry logs and instructed participants on how to use/complete them. We used these measures to gather data on participants’ mobility (travel behaviour and physical activity) during the seven days following measurement sessions. A courier service retrieved travel diaries, GPS, accelerometers, and accelerometry logs from participants after the seven-day monitoring period. I discuss all measures relevant to my dissertation in more detail, below. I also include copies of all the measures, with the exception of accelerometry, in Appendices B and C.  2.2.8.2 Measures of mobility (travel behaviour and physical activity) I measured participants’ mobility (travel behaviour and physical activity) outcomes with travel diaries, accelerometry, and a self-report physical activity questionnaire, as detailed below.  2.2.8.2.1 Travel behaviour – travel diaries Travel behaviour and walking for transportation: I used travel diaries to measure participants’ travel behaviour in the seven-days (week) immediately following their measurement sessions. We gave participants travel diaries and provided instructions at the end of each measurement session Within their diaries, participants recorded their daily trips including: start location and time; end location (destination) and time; trip purpose; travel mode; and whether someone accompanied them on the trip. For my dissertation, I use the data collected with the travel diaries to describe participants’ travel behaviour, including: i) frequency (ntrips/day), ii) destinations (% trips by destination type), iii) purpose (% trips by purpose), and iv)  83travel mode (% trips by mode). I also use the data collected with travel diaries to report frequency of walking for transportation trips (ntrips/day).  2.2.8.2.2 Physical activity – accelerometry Physical activity: I used ActiGraph GT3X+ (LLC, Fort Walton Beach, FL) tri-axial accelerometers to directly measure participants’ physical activity. We gave participants their accelerometers and briefed them on their use at the end of each measurement session. We requested participants wear the accelerometer on their right hip, during waking hours, during the following week and that they remove the accelerometer during any water-based activities.  The accelerometer recorded data continuously (at 30 Hz) and I reintegrated the data to 60-second epochs. I considered more than 60 minutes of continuous zeroes as non-wear time. For analyses, I included data with three or more valid days (> 8 hours wear time/day) of wear time. I used cut-points proposed by C. E. Matthews et al. (2008) to classify time (minutes/day) spent in sedentary behaviour [<100 counts/minute (CPM)] and the cut-points proposed by Freedson, Melanson, and Sirard (1998) to classify time spent in light (100-1951 CPM) and MVPA ( > 1952 CPM). I also calculated time (minutes/day) spent in bouts of > 10 minutes of MVPA, allowing for a 1-2 minute interruption. I derived time (minutes/day) spent in physical activity of varying intensities and sedentary behaviour, as well as physical activity volume (counts/day), using batch processing with ActiLife software version 6.5.4 (LLC, Fort Walton Beach, FL). 842.2.8.2.3 Physical activity – self-report questionnaire Physical activity and walking for transportation: I measured participants’ self-reported physical activity with the Community Healthy Activities Model Program for Seniors (CHAMPS) physical activity questionnaire (Stewart et al., 2001). Community Healthy Activities Model Program for Seniors physical activity questionnaire evaluates type, frequency (times/week) and duration (hours/week) of physical activities that respondents engaged in over the preceding month. Frequency of physical activity is measured as a discrete variable, while duration of physical activity is categorical (< 1 hour, 1-2.5 hours, 3-4.5 hours, 5-6.5 hours, 7-8.5 hours, > 9 hours). Items include pre-defined physical activities that are of light (e.g., light gardening, stretching, light housework), moderate (e.g., water exercises, heavy housework) and vigorous (e.g., jogging, walking uphill, moderate-to-heavy strength training) intensity, and items specific to walking for errands and walking for leisure. Validity and reliability of CHAMPS is well established in community-dwelling older adults (Colbert, Matthews, Havighurst, Kim, & Schoeller, 2011; Harada, Chiu, King, & Stewart, 2001; Stewart et al., 2001). For my dissertation, I used CHAMPS to describe physical activities that participants most commonly engaged in, as well as to measure dimensions of walking for transportation (none vs. any; frequency; duration).   2.2.8.3 Measures of person and environment-level characteristics of participants I applied Webber’s (2010) framework of older adult mobility to comprehensively measure participants’ capacity to be mobile across multi-level domains (environmental, cognitive, physical, and psychosocial). Although Webber’s framework refers to these “domains” as “determinants,” I use the term “domains” to underscore the multi-directional nature of associations between variables and mobility. Further, although psychosocial variables include  85psychological attributes that exist at the individual level and are likely to result from the process of socialization (e.g., thoughts and feelings), as well as variables that exist at a wider structural level (e.g., interpersonal relationships) (Singh-Manoux, 2003), I distinguish the two as separate domains to differentiate between variables within (person-level) and outside of the individual (environment-level). Specifically, I group measures of interpersonal relationships (e.g., social networks, social support) and neighbourhood characteristics (e.g., social cohesion, neighbourhood physical and social disorder) into a social environment domain (McNeill et al., 2006) and measures of thought and feelings into a psychosocial domain. I describe measures used in this study in detail, below.   2.2.8.4 Environment-level variables My environment-level variables encompass measures of the built environment and social environment, described below.   2.2.8.4.1 Built environment – GIS-based measure Walkability: As described in Section 2.2.4.1, I used Walk Score during sampling to measure walkability of neighbourhoods where participants resided. Walk Score is a publicly available index that measures the walkability of the built environment surrounding an address on a scale of 0 (low walkability) to 100 (high walkability) based on distances to nearby destinations (www.walkscore.com). Street Smart Walk Score® (www.walkscore.com) was developed by Walk Score after recruitment of Walk the Talk participants in 2012. Street Smart Walk Score uses updated methodology that better reflects empirical research and is more closely associated with time spent in MVPA for adult and older adult populations (Frank, 2013). I compare  86methods used to derive Walk Score and Street Smart Walk Score in Table 2.4. I use categories of walkability (Street Smart Walk Score range), as provided by the manufacturer: “Car-dependent” (0-49), “Somewhat walkable” (50-69), “Very walkable” (70 to 89) and “Walker’s paradise” (90-100). Given Street Smart Walk Score’s improved methods to measure walkability, I use Street Smart Walk Score for my analyses. The association between participants’ Street Smart Walk Score and Walk Score at time of recruitment was r (159) = 0.92, p < 0.001.      87Table 2.4. Comparison of Walk Score and Street Smart Walk Score methodologya   Walk Score Street Smart Walk Score Data source includes Google.  Data sources include Google, Localeze, OpenStreetMap, Education.com and Walk Score users  Destination categories factored into the score include grocery, restaurant, coffee, bars, movies, schools, parks, libraries, book stores, fitness, drug stores, hardware, and retail.  Destination categories factored into the score include grocery; restaurant/bar; shopping; coffee; banks; parks; schools; books; and entertainment. Destination categories are weighted equally when score is calculated. Destinations categories are assigned different weights when score is calculated (weight - destination category): 3 – grocery; restaurant/bar 2 – shopping , coffee 1 – banks, parks, schools, books, entertainment.  Only the closest result in each destination category contributes to the score. Multiple results are counted in certain categories (maximum number of results counted – destination category): 10 – restaurant/bar 5 – shopping  2 – coffee 1 – grocery, banks, parks, schools, books, entertainment.  Distance from residence to destination categories measured using “crow flies" distance.  Distance from residence to destination categories measured using network walking distance.  Maximum points are awarded for destinations within 0.25 miles (400 meters) of a residence, with fewer points awarded at 0.5, 0.75 and 1 mile (800, 1200, 1600 meters, respectively). No points are awarded after 1 mile. Maximum points are awarded for destinations within 0.25 miles (400 meters) of a residence, after which score decreases with distance smoothly. At 1 mile (1600 meters), destinations receive only 12% of the points they would have received if they were within 0.25 miles (400 meters) of a residence. No points are awarded after 1.5 miles (2400 meters).       88Walk Score Street Smart Walk Score Road connectivity metrics are not factored into the score.  A residence is penalized up to 10% of its total score for poor road connectivity metrics (intersection density and block length). a I obtained this information through personal correspondence with Walk Score staff (Aleisha Jacobson). This methodology is specific to the Walk Scores and Street Smart Walk Scores that we purchased at the time of the study and does not necessarily reflect current Walk Score and Street Smart Walk Score methodology.    2.2.8.4.2 Built environment – perceived measure Perceptions of neighbourhood built environment features related to walking: I used a modified version of the Neighbourhood Environment Walkability Scale – abbreviated (NEWS-A) to measure participants’ perceptions of neighbourhood built environment features related to walking (Cerin, Saelens, Sallis, & Frank, 2006). Subscales of NEWS-A include residential density, land use mix - diversity, land use mix – access, street connectivity, infrastructure and safety for walking, aesthetics, traffic hazards, crime, lack of parking, lack of cul-de-sacs, hilliness, and physical barriers. However, as I did not collect all items that comprise the infrastructure and safety for walking subscale I do not present these outcomes. All subscales, with the exception of residential density and land use mix – diversity, use a four-point scale from 1 (strongly disagree) to 4 (strongly agree). The residential density subscale uses a five-point scale from 1 (none) to 5 (all), converted to a sub-score that ranges between 173 and 865. The land use mix – diversity subscale uses a five-point scale from 1 (1-5 minutes) to 5 (> 31 minutes/don’t know) to measure participants’ perceptions of time required to walk from home to select destinations. Reliability of individual items and validity of NEWS-A was established for adults who resided in the USA (Brownson et al., 2004; Cerin, Conway, Saelens, Frank, & Sallis, 2009; Cerin et al., 2006).     892.2.8.4.3 Social environment Interpersonal relationships: I assessed dimensions of participants’ interpersonal relationships (marital status, living arrangement, and perceived presence of people that offer physical and/or social support to go outside) using a self-report questionnaire, as well as a three-item measure of social interaction. The three-item measure of social interaction was drawn from Veroff et. al. (Veroff, Kulka, & Douvan, 1981). Two items asked participants to indicate how often (>1/week, 1/week, 2-3 times/month, about 1/month, <1/month, or never) they: i) get together with friends, neighbours or relatives to go out or visit in each other’s homes; and ii) attend meetings or programs of groups, clubs, or organizations that they belong to. A third item asked how often in a typical week (>1/day, 1/day, 2-3 times/week, about 1/week, <1/week, or never) they talk on the telephone or exchange emails with friends, neighbours or relatives. The scale has established validity (Clarke, Ailshire, Nieuwenhuijsen, & de Kleijn-de Vrankrijker, 2011; Musick, Herzog, & House, 1999) and reliability in samples of community dwelling older adults (House, 1994).   Neighbourhood social environment: I measured neighbourhood social environment characteristics using a five-item measure of social cohesion and trust (Sampson, Raudenbush, & Earls, 1997) and a five-item measure of physical and social disorder drawn from the Project on Human Development in Chicago Neighbourhoods (Sampson, 2012). The five-item measure of social cohesion and trust asks participants to indicate their agreement with statements that tap mutual trust and cohesion in the neighbourhood (each assessed on a five-point scale) (Sampson et al., 1997). The measure has demonstrated reliability in community-dwelling samples (Raudenbush & Sampson, 1999; Sampson et al., 1997) and validity with respect to both individual and community-level outcomes (Cradock, Kawachi, Colditz, Gortmaker, & Buka,  902009; Lochner, Kawachi, Brennan, & Buka, 2003; Sampson et al., 1997). The five-item measure of neighbourhood physical and social disorder [drawn from the Project on Human Development in Chicago Neighbourhoods (Sampson, 2012)] asks participants to indicate on a four-point scale how much (none, a little, some, a lot): i) broken glass or trash they see on neighbourhood sidewalks and streets; ii) graffiti they see on neighbourhood buildings and walls; iii) how many (none, a little, some, a lot) vacant/deserted houses or storefront they see in their neighbourhood; and how often (never, not very often, sometimes, very often) they see: iv) people drinking in public places in their neighbourhood; and v) unsupervised children hanging out on the street in their neighbourhood. This measure has established validity and within-neighbourhood reliability in community-dwelling samples (Mair, Roux, & Morenoff, 2010; Raudenbush & Sampson, 1999; Sampson & Raudenbush, 1999).   2.2.8.5 Person-level variables  My individual (person-level) variables encompass sociodemographic factors, as well as cognitive, physical, and psychosocial domains, as described below.   2.2.8.5.1 Sociodemographic information I measured participants’ age, gender, culture (ethnicity, self-identify as visible minority), highest education level attained, years lived in residence, whether they possessed a valid driver’s license, whether they had access to a vehicle in the seven days preceding study participation, and whether they owned a dog with a self-report questionnaire.    912.2.8.5.2 Cognitive domain Mild cognitive impairment: I used the Montreal Cognitive Assessment (MoCA), a brief clinical screening tool with high sensitivity and specificity to detect mild cognitive impairment (MCI), to screen for possible MCI (Nasreddine et al., 2005). Montreal Cognitive Assessment is scored out of 30 points; a total score < 26 used as a cut-off for suspected MCI (Nasreddine et al., 2005).  2.2.8.5.3 Physical domain  Body mass index: I used a TANITA Electronic Scale Model BWB-800 and Seca Stadiometer Model 242 to objectively measure participants’ weight (kg) and height (cm), respectively. I use these data to calculate participants’ body mass index (BMI, kg/m2).   Limitation in lower extremity functioning and gait speed: I objectively measured participants’ limitations in lower-extremity functioning using the Short Physical Performance Battery (SPPB) (Guralnik et al., 1994). I also used a self-report questionnaire to determine participants’ use of mobility aids (number, type) and falls history in the past six months. Short Physical Performance Battery is a reliable and valid measure of older adults’ mobility and balance that consists of standing balance tests, a four-meter walk at usual pace, and a sit-to-stand test (Freire, Guerra, Alvarado, Guralnik, & Zunzunegui, 2012; Guralnik, Ferrucci, Simonsick, Salive, & Wallace, 1995; Guralnik et al., 1994; Ostir, Volpato, Fried, Chaves, & Guralnik, 2002). I calculated participants’ gait speed (m/s) based on the time taken to walk 4-meters at usual pace.   92Health: I used self-report questionnaires to measure participants’ global health [European Quality of Life-5 Dimensions visual analogue scale (EQ-VAS) (Herdman et al., 2011)] and comorbidities [Functional Comorbidity Index (FCI) (Groll, To, Bombardier, & Wright, 2005)]. EQ-5D is a generic measure of health status, widely used across countries and clinical populations (http://www.euroqol.org). EQ–VAS asks participants to indicate their current health status on a scale from 0 ("the worst health you can imagine") to 100 ("the best health you can imagine"). Functional Comorbidity Index was designed to assess the presence of 18 comorbidities associated with physical function (Groll et al., 2005). Reliability and validity of the FCI was established in clinical adult populations (Fan et al., 2012; Fortin et al., 2005; Groll et al., 2005).  2.2.8.5.4 Psychosocial domain Self-efficacy for walking: I used self-report questionnaires to measure participants’ self-efficacy for walking in different environment situations [Ambulatory Self-Confidence Questionnaire (ASCQ) (Asano, Miller, & Eng, 2007)] and walking in the neighbourhood (five-point scale, where 1 = “not at all” and 5 = “very much”). Ambulatory Self-Confidence Questionnaire is a 22-item questionnaire that asks respondents to rate on a 10-point scale, where 1 = “not at all confident” and 10 = “extremely confident,” how confident they are in their ambulatory abilities across different environmental situations. Validity and reliability of the ASCQ was established for community-dwelling older adults (Asano et al., 2007).   Attitudes towards walking: I measured how much participants like to walk outside by self-report (five-point scale, where 1 = “not at all” and 5 = “very much”).  93Stress: I used the Perceived Stress Scale (PSS), a popular self-report questionnaire, to measure psychological stress (Cohen, Kamarck, & Mermelstein, 1983)]. Perceived Stress Scale asks participants about the frequency (never, almost never, sometimes, fairly often, and very often) of thoughts and feelings they had in the last month (Cohen et al., 1983). Validity of PSS was established in community-dwelling older adults (Ezzati et al., 2014) and reliability of PSS was established in a variety of adult populations (E. H. Lee, 2012).  Loneliness: I measured loneliness with 11 items drawn from the Revised UCLA Loneliness Scale, a validated and reliable self-report questionnaire that taps general feelings of social isolation, loneliness, and dissatisfaction with one’s social interactions (Hawkley, Browne, & Cacioppo, 2005; M. E. Hughes, Waite, Hawkley, & Cacioppo, 2004; Russell, 1996). The 11 item scale has established reliability in community dwelling older adults (J. Smith et al., 2013).    Data entry Express Data Limited is a professional data entry firm located in Vancouver, BC (www.expressdata.shawbiz.ca). They used specialized data entry software to enter data gathered during measurement sessions (with the exception of the MoCA). Express Data Limited ensured quality of data entry through double entry of data from each questionnaire. The Walk the Talk team manually entered MoCA, travel diary and accelerometry log data into Excel spreadsheets. I ensured data quality through a random 10% check of all entered data.    94  Statistical analysis In this section I provide an overview of the statistical methods I used in this dissertation, listed in alphabetical order. I further describe the specific analyses in the methods section of each relevant Chapter (3-6). I analyzed all data using Stata version 13.0. The exception is Chapter 3, where I used Stata version 10.0 (StataCorp, College Station, TX) because Stata version 13.0 was not yet available.   2.2.10.1 Absolute agreement Absolute agreement can be used to quantify the frequency with which two observers (A and B) assign the same rating to an observation measured on a binary scale (e.g., present vs. absent).  If the ratings assigned by these two observers are arranged in a traditional 2 X 2 contingency table (Table 2.5):   Table 2.5. Example contingency table of binary ratings by two observers   Observer A Present Absent Observer B   Present a b Absent c d  Then absolute agreement can be summarized as (Feinstein & Cicchetti, 1990):  P0 = (a+d) / (a+b+c+d)     (Equation 1)   95Thus, in my dissertation, I calculate absolute agreement as the number of times two observers (or two observational approaches) agree on binary ratings over the total number of ratings. Specifically, in Chapter 3, I use Stata to calculate absolute agreement between built environment audits conducted virtually (using Google Earth’s Street View feature) and in-the-field.  2.2.10.2 Linear regression Linear regression is a statistical method that models a linear relation between a continuous dependent variable y (outcome) and > 1 independent variable(s) xp [predictor(s)]. Simple linear regression refers to a model with one predictor, whereas multiple linear regression refers to a model with > 1 predictor. For a multiple regression model with p predictors, the regression line is expressed as (Vittinghoff, Glidden, Shiboski, & McCulloch, 2006b):  yi = B0 + B1x1 + B2x2 + … + Bpxp + ei   (Equation 2)  In Equation 2, the intercept B0 represents the value of the outcome when the value of all of the predictors = 0 (Vittinghoff et al., 2006b). The coefficient Bj, where j = 1 … p, represents the change in the outcome given a one-unit change in a given predictor and holding the other predictors constant (Vittinghoff et al., 2006b). Finally, ei represents the random error term; it accounts for the fact that for any given combination of predictors, a particular subject’s outcome will considerably vary due to e.g., measurement error, natural fluctuations, unmeasured determinants of the dependent variable, etc. (Vittinghoff, Glidden, Shiboski, & McCulloch, 2006a).    96The statistical assumptions underlying linear regression models concern the distribution of ei. Specifically, the model assumes that (Vittinghoff et al., 2006b):  ei ~ i.i.d. N (0, ∂2e)      (Equation 3)  Where “i.i.d.” stands for “independently and identically distributed” (Vittinghoff et al., 2006a). Collectively, Equation 3 means that when fitting a linear regression model we assume that the error term is normally distributed with a mean = 0, constant variance ∂2e at every value of the predictor, and that the predictors are statistically independent (Vittinghoff et al., 2006b).  In Chapter 6 of my dissertation, I used Stata to fit linear regression models for: i) physical activity volume (operationalized as steps and TAC), ii) light physical activity, iii) MVPA, and iv) duration of walking for transportation. I measured i) – iii) with accelerometry and iv) using self-report measures.    2.2.10.3 Logistic regression Logistic regression is a statistical approach that models a linear relation between the logarithm of the odds of a binary outcome y and > 1 predictor(s) xp. If we assume that y can take on the values of 0 or 1, P(x1, x2, …, xp) represents the probability that Y = 1 for a set of predictors with particular values of xp (Vittinghoff et al., 2006a). Then, for a logistic regression model with p predictors, the regression line is defined as logistic transformation of the probability P (Vittinghoff, Glidden, Shiboski, & McCulloch, 2006c):   97log[P(x1, x2, …, xp) / 1- P(x1, x2, …, xp)] = B0 + B1x1 + B2x2 + … + Bpxp  (Equation 4)  In Equation 4, the intercept B0 represents the value of the log odds of the outcome when the value of all of the predictors = 0 (Vittinghoff et al., 2006c). The coefficient Bj, where j = 1 … p, represents the change in the log odds of the outcome given a one-unit change in a given predictor and holding the other predictors constant (Vittinghoff et al., 2006c). For interpretation, we exponentiate the coefficients in Equation 4 and report the odds ratio. Specifically, Equation 4 can be re-written to express the logistic model in terms of the odds of the outcome for predictors xp,(Vittinghoff et al., 2006c):  P(x1, x2, …, xp) / 1- P(x1, x2, …, xp) = exp(B0 + B1x1 + B2x2 + … + Bpxp)  (Equation 5)  In Equation 5, the intercept B0 represents the odds of the outcome when the value of all of the predictors = 0 (Vittinghoff et al., 2006c). The coefficient Bj, where j = 1 … p, represents the change in the odds of the outcome given a one-unit change in a given predictor and holding the other predictors constant (Vittinghoff et al., 2006c).   The statistical assumptions underlying logistic regression models concern the outcome yi (Vittinghoff et al., 2006c). Logistic regression models assume that; i) yi follows a binomial distribution; ii) E[y|x] = the mean expected value of the outcome for a given value of x = P(x) is given by the logistic function (Equation 4); and iii) values of the outcome are statistically independent (Vittinghoff et al., 2006c).   98In Chapter 6 of my dissertation, I used Stata to fit logistic regression models for the outcome walking for transportation (any vs. none) as measured by a self-report questionnaire (CHAMPS).    2.2.10.4 Poisson regression Poisson regression is a statistical approach used to model outcomes that are counts of an event or occurrence over a defined period of time (Oxford Journals, n.d.). Poisson regression models a linear relation between the natural logarithm of an outcome and > 1 predictor(s) xp (Oxford Journals, n.d.):  loge(y) = B0 + B1x1 + B2x2 + … + Bpxp    (Equation 6)  In Equation 6, the intercept B0 represents the value of the natural logarithm of the outcome when the value of all of the predictors = 0. The coefficient Bj, where j = 1 … p, represents the change in the natural logarithm of the outcome given a one-unit change in a given predictor and holding the other predictors constant. In order to make the interpretation of the results of Poisson regression more intuitive, we exponentiate the coefficients in Equation 6 and interpret the outcome as an incidence rate ratio (IRR):  y = exp(B0 + B1x1 + B2x2 + … + Bpxp)    (Equation 7)  In Equation 7, the intercept B0 represents the incidence rate of the outcome when the value of all of the predictors = 0. The coefficient Bj, where j = 1 … p, represents the change in the incidence  99rate of the outcome given a one-unit change in a given predictor and holding the other predictors constant.   Since it is impossible for counts to take on negative values, Poisson regression models assume that the outcome has a Poisson distribution (rather than an e.g., Normal distribution) (Katz, 2006b). Other characteristics of the Poisson distribution include: i) its variance is equal to its mean and ii) it is skewed to the right (Katz, 2006b). Poisson regression models assume that the i) probability of the outcome is constant over time and ii) occurrences of the outcome are independent from each other (Katz, 2006b).  In Chapter 5 of my dissertation, I used Stata to fit Poisson regression models for the outcome frequency of walking for transportation as measured by a self-report questionnaire. Since I only measured this outcome among participants that made > 1 walking for transportation trip, I modelled the data using zero-truncated Poisson models, which use a Poisson distribution adjusted to account for the absence of zero-values in the outcome (Hilbe, 2011b).   2.2.10.5 Negative Binomial regression Negative Binomial regression is a statistical approach used to accommodate Poisson models in which the variance of the outcome of the regression model is greater than the mean; this is referred to as over-dispersion (Hilbe, 2011a). The Negative Binomial regression is defined using the same equations as for Poisson regression (Equations 6 and 7) (Hilbe, 2011c). However, the underlying probability distribution function is the Negative Binomial, rather than the Poisson (Hilbe, 2011a). There are actually 13 separate types of derivations for the negative binomial  100distribution; I refer to and use the traditional parameterization of the negative binomial model in this dissertation, commonly known as NB2 (Hilbe, 2011a).   In Chapter 6 of my dissertation, I used Stata to fit Negative Binomial regression models for the outcome frequency of walking for transportation as measured by travel diaries.  101Chapter  3: Agreement between virtual and in-the-field environmental audits  3.1 Introduction Engaging in regular physical activity has many health benefits for older adults, including reduced risk of all-cause mortality (Wen et al., 2011), morbidity (Hollmann et al., 2007), and maintenance of physical (Brach et al., 2004; Morey et al., 2008) and cognitive function (Bowen, 2012; Laurin et al., 2001). Despite the many known benefits of regular physical activity in later life, the older adult age group was least likely (Troiano et al., 2008) to meet public health recommendations to accumulate 30 minutes of at-least moderate physical activity on most, if not all, days of the week (Pate et al., 1995; U.S. Department of Health and Human Services, 1996). For the purpose of this dissertation, the built environment is defined as human-made infrastructure comprising urban design, land use, and the transportation system (S. L. Handy et al., 2002). Given that walking outdoors is older adults’ primary form of physical activity (Ashe, Miller, Eng, Noreau, & Physical Activity & Chronic Conditions Research Team, 2009; Centers for Disease Control and Prevention, 1999) the built environment has the potential to positively influence physical activity levels of older adults. A poorly designed built environment was associated with unmet physical activity needs, particularly among older adults who experienced difficulties walking (Rantakokko et al., 2010) and may exacerbate age-related mobility limitations (Clarke et al., 2008). Features of the built environment that affect older adults’ walking can be measured at both the macroscale (neighbourhood-level, e.g., land use mix, street connectivity) and microscale (street-level, e.g., sidewalk and road characteristics). Most studies to date that investigated the association between objectively measured features of the built  102environment and older adults’ walking focused on macroscale features as they are more easily captured using Geographic Information Systems (GIS) (Rosso et al., 2011).   Microscale features of the built environment may pose a direct challenge to older adults’ ability to walk outdoors. If the challenge is greater than an older adults’ perceived or real physical capacity, the individual may be less likely to engage with the built environment (i.e., go outdoors) (Noreau & Boschen, 2010). For example, Rantakokko et al. (2010) noted that when older adults perceived they were unable to partake in physical activity (because of unfavourable environment features) they were less likely to do so. Further, older adults fear of moving outdoors was associated with a variety of factors such as low socioeconomic status (SES), musculoskeletal diseases, slow walking speed, the presence of poor street conditions, hills in the nearby environment, and noisy traffic (Rantakokko et al., 2009). Similarly, Shumway-Cook et al. (2003) found that older adults with a mobility-disability (defined as “requiring assistance to walk half a mile or climb stairs”) were more likely than older adults without a mobility-disability to avoid built environments that posed a challenge to moving about effectively, such as those with curbs, uneven surfaces, and stairs.   Environmental audits provide a direct and systematic approach to measure microscale features of the built environment through in-person observation (Brownson et al., 2009). However, the resource intensive nature of audits (i.e., time and cost to travel to sites and conduct audits) limits their feasibility, especially in studies set across large geographic areas (Brownson et al., 2009). To my knowledge, only two studies (D. King, 2008; Michael et al., 2006) used environmental audits to investigate the association between the microscale built environment and walking  103behaviour of older adults aged > 65 years. D. King (2008) found specific microscale features (presence of curb cuts, cross walks, or retail density) were positively associated with frequency of walking for errands. However, Michael et al. (2006) did not find a significant association between sidewalks and sidewalk obstructions and neighbourhood walking. Both studies found that if graffiti was present it was negatively associated with walking.   Google Earth® is a free, internet-based application that provides panoramic imagery of cities worldwide (Google Inc., CA). The Street View feature within Google Earth provides high-resolution, street-level images that offer the potential to conduct audits of micro and macroscale features of neighbourhoods and communities virtually, without many of the resources necessary to gather the data in-the-field. A small number of studies assessed agreement between virtual and in-the-field environmental audits and found moderate-to-high levels of agreement between modes (Badland et al., 2010; Ben-Joseph et al., 2013; Clarke et al., 2010; C. M. Kelly et al., 2012; Rundle et al., 2011; Wilson et al., 2012). However, items addressing aesthetics (i.e., litter; graffiti) and items requiring a fine level of detail (i.e., sidewalk characteristics, street condition) generally resulted in lower agreement (Ben-Joseph et al., 2013; Clarke et al., 2010; Rundle et al., 2011; B. T. Taylor et al., 2011; Wilson et al., 2012). Four studies (Badland et al., 2010; Ben-Joseph et al., 2013; C. M. Kelly et al., 2012; Wilson et al., 2012) used instruments designed to audit environmental correlates of physical activity in the general population. Physical activity was defined as walking and cycling (Badland et al., 2010); walking, cycling, or moderate and vigorous leisure-time activity (C. M. Kelly et al., 2012; Wilson et al., 2012); and generally as “activity” (Ben-Joseph et al., 2013). To the best of my knowledge, no previous studies evaluated agreement between virtual and in-the-field administration of an environmental audit tool  104designed specifically to assess macro and microscale built environment features that influence older adults’ walking.  Given our aging demographic, devising a method that reliably identifies microscale features of the built environment that support or hinder older adults’ walking is an important research priority. Further, studies (A. C. King et al., 2011; Rantakokko et al., 2010; Rantakokko et al., 2009; Satariano et al., 2010; Shumway-Cook et al., 2003) also support that the built environment may have the greatest impact on the physical activity, and specifically, walking habits, of older adults with mobility-impairment. Thus, the clinical relevance of developing an effective methodology is likely greater in geographic areas where a high concentration of older adults with mobility impairments reside, such as assisted living (AL) sites.   I define AL as ‘a semi-independent housing option that typically provides transitional housing for older adults who require greater assistance with health or activities of daily living than can be provided by social networks and/or homecare service providers’ (McGrail et al., 2013). Importantly, these older persons do not yet require more intense services typically provided in long-term care or nursing home settings (McGrail et al., 2013). Residents of AL are generally older women who have one or more major chronic conditions (McGrail et al., 2013). Compared with community-dwelling older adults with disabilities, those in AL are similarly independent but have a higher prevalence of mobility impairments (Giuliani et al., 2008). Despite the poorer health and the higher prevalence of mobility impairments in AL residents, Lu (2010) found that 16% reported walking outdoors as a dominant mode of exercise. Wang and Lee (2010) found that 54% of AL residents walked outdoors > once/day. Thus, this crucial time marks a crossroads  105for these individuals -- between living independently longer or transition into higher levels of residential care. Being as physically active and mobile as possible (despite some mobility impairment) is key to their independence. Thus, AL sites represent a setting where a high concentration of vulnerable older adults reside and their physical activity, and specifically walking habits, may be particularly influenced by the design of their surrounding built environment.  Therefore, my objective was to compare the absolute agreement between environmental audits of AL sites conducted virtually (using Google Earth’s Street View feature) vs. environmental audits conducted in-the-field, with a focus on items especially relevant for older adults. My secondary objective was to use virtual audits to describe the built environment surrounding AL sites.  3.2 Methods This study compares agreement between environmental audits conducted virtually vs. in-the-field at four AL sites located in Metro Vancouver, Canada. The four AL sites contained publicly funded units that represented selected neighbourhoods at the extremes of walkability as measured by the Walk Score® (www.walkscore.com). Walk Score is a publicly available index that uses distances between a given address and select destinations to determine the walkability of the address. I did not use the Street Smart Walk Score to measure walkability as it was not yet available at the time of this study. Two of the audited sites were in urban neighbourhoods with high walkability (High 1, Walk Score = 92 and High 2, Walk Score = 93) and two were in suburban neighbourhoods with low walkability (Low 3, Walk Score = 48 and Low 4, Walk  106Score = 52). Site Low 4 had approximately 100 residential units; others sites contained approximately 50 residential units each.   Sampling of segments Using ArcGIS 10.1 (ESRI, CA) and the Digital Road Network shapefile, the Walk the Talk team (Dr. M. Winters) created a 400-meter aerial buffer surrounding each AL site. This distance was chosen as it is commonly used as a marker for mobility-disability in the older adult literature (Pahor et al., 2006). A segment is defined as a, “section of road between two intersections” (T. J. Pikora et al., 2002). Dr. M. Winters randomly selected segments around each site until she identified 12 segments per site visible on Street View (Street View does not have imagery for segments like restricted roads, strata zones, etc.). Coverage (availability of virtual images) was limited in some urban and suburban settings. Thus, four randomly selected segments were unavailable in Low 3 and seven were unavailable in High 2. All randomly selected segments were visible using Street View in High 1 and Low 4.    Audits Two trained raters conducted audits of each segment using the Seniors’ Walking Environmental Assessment Tool – Revised (SWEAT-R), an audit tool developed to examine microscale features that could potentially influence the walking behaviours of older adults (Michael et al., 2009). This instrument consists of 165 questions categorized within four domains: functionality (structural aspects, i.e., building use, sidewalks); safety (personal and traffic conditions, e.g., presence of street lights, cross walks); aesthetics (visual appeal and quality of microscale urban  107design); and destinations (presence of resources and services, e.g., public transportation, gathering places).  Rater training included one classroom-based session to review the protocol (Michael & McGregor, 2007), a practice audit in-the-field, and another using Street View. For in-the-field audits, raters travelled to each site together but conducted audits independently except for items that involved measurement, as per the published protocol (Chaudhury et al., 2011). Raters conducted virtual audits independently on different computers. Raters conducted in-the-field audits in September and November 2011 and virtual audits in January 2012.  3.3 Statistical analysis Two research assistants, who were not otherwise involved in the study, entered data into a master database. I established quality of data entry by checking the accuracy of a random sample of 10% of the segments. I examined inter-mode agreement using absolute agreement (%) for categorical items (n=155) and discrete items (number of: benches; street lights; mature trees). Unlike the Kappa statistic, absolute agreement does not account for agreement that occurs due to chance. However, I did not calculate Kappa statistics as most sample distributions were limited and skewed (i.e., majority of responses recorded as “no”). This would lead to the paradox of high absolute agreement but low kappa (Feinstein & Cicchetti, 1990). I did not include seven items that required reviewers to directly measure built environment features (i.e., sidewalk width, buffer zone width, curb height, duration of traffic lights) since SWEAT-R protocol dictated that raters work together to obtain these measurements (that is, raters’ observations were not independent for these items); further, it was impossible to assess those items using virtual audits.   108I described the built environment around each AL site using select features that reportedly have the greatest potential to impact older adults’ walking (Chaudhury et al., 2011; Hanson et al., 2012; D. King, 2008; Moniruzzaman & Paez, 2012); I dichotomized (presence vs. absence of the feature) and combined individual SWEAT-R items to reflect features identified in these studies. I considered a segment to have mixed-land use if it contained > two land use categories. When creating the mixed-land use item, I combined individual items that measured land use into six different land use categories (housing; office/institutional; restaurant/café/commercial; industrial; vacant/undeveloped; and recreation) based on Pedestrian Environmental Data Scan methodology (Clifton, Smith, & Rodriguez, 2007). I considered a segment to have > one traffic control device if it contained any of: traffic signal; stop sign; traffic circle; speed bumps/humps; or sidewalk extensions (Clifton et al., 2007). Finally, I considered a segment to have > one crossing aid present if it contained any of: marked crossing area; crosswalk; pedestrian crossing sign; sign for pedestrian/children/etc.; signs for traffic activity; pedestrian signal; pedestrian overpass/underpass/bridge; or median (Clifton et al., 2007).   I categorized absolute agreement (> 95%; 80-84%; 70-79%; <70%) for each individual SWEAT-R item and present it separately by rater, based on SWEAT-R domain (functionality, destinations, aesthetics, safety and comfort). I present interrater agreement by mode in Table D.1 (Appendix D). I conducted all analyses using Stata version 10 (Stata Corp, TX).  3.4 Results  Audits covered 48 segments (12 segments/site) both virtually and in-the-field. Time required to audit segments was similar between virtual (9 minutes/segment, on average) and in-the-field (11  109minutes/segment, on average) approaches. However, in-the-field audits required additional travel time, as sites were 5-30 km from the research centre (20– 80 minutes round trip); travel was also required between segments once at the site (approximately 4 minutes/segment).  Table 3.1 describes features of the built environment that surround audited AL sites. Data in this table are based on Rater 1; I provide similar data for Rater 2 in Table D.2 (Appendix D). The prevalence of features associated with older adults’ walking varied based on AL site and SWEAT-R domain. All sites had a high prevalence of street segments without litter/graffiti/broken facilities, streets with more than one lane, and sidewalks on one or both sides of the street. There were public restrooms/washrooms in only one segment (in High 1) and > one bench in 25% of segments in High 1, High 2, and Low 4. There were no benches in any segment in Low 3. I noted mixed land use in 33% of segments in High 1, High 2, and Low 3. However, there were no specific destinations that supported older adults’ walking in the majority of segments we audited. Safety features were present in <50% of sites; the exceptions were intended safety crossings at both ends of the segment and > 1 traffic control devices in High 1 and the presence of > 1 crossing aids in High 2. Segments within each site had a mix of sidewalk features that support walking, although arcade/awnings coverings were present in only one segment (in High 1).  Table 3.1 also specifies the prevalence of features identified by each audit method. Aesthetic features (absence of litter/graffiti/broken facilities), four destination items (restaurants, grocery/convenience stores, pharmacy/drug stores, and health clinics/medical facilities), and three sidewalk features (presence, curb cuts, arcades and/or awnings) were similarly captured by both audit methods. Amenities and certain destinations (mixed and vertical mixed use, retail  110stores, bank/financial services, service facilities, transit stop), safety (> 1 traffic control device); sidewalks (continuity, condition, slope), and streets (>1 lane) related features were recorded with higher prevalence using in-the-field audits. Presence of senior housing, senior’s activities and two of three safety features (intended crossing at both segment ends and presence of > one crossing aid) were more often captured using Street View. Overall, discrepancies were rare and typically occurred on one feature/segment.  Agreement between virtual and in-the-field audits was > 80% (Rundle et al., 2011) for 92% (23/25) of the features included in Table 3.1. Agreement between virtual and in-the-field audits for individual SWEAT-R items is displayed in Table 3.2. Ninety percent (121/135) and 92% (125/136) of items displayed > 80% agreement for Rater 1 and 2, respectively. Twenty-three items (Rater 1) and 22 items (Rater 2) did not display heterogeneity (variability in the characteristics of an item across segments) and were therefore not included for one or both raters in Table 3.2. These items fell within the following SWEAT-R domains; buildings (15 items: mobile homes; other residential buildings; other recreational buildings; museum, auditorium, concert hall, or theater; post office; other type of public building; hospital; other institutional building; hotel/hospitality; car dealership; offices; other office/service; industrial/manufacturing; harbor/marina/boat launch; agricultural land, ranch, farming), destinations (five items: “corner” store; art galleries, museums, theatres; farmers market; shopping mall; outdoor mall), street features (five items: proportion of the street under repair; four items assessing presence of different materials used to construct the street), aesthetics (two items: amount of litter, graffiti, broken glass on the segment; amount of abandoned buildings on the segment), sidewalk  111characteristics (one item: minimum sidewalk width), safety (one item: presence of a designated bike lane), and street life (one item: public garden).  112Table 3.1. Prevalence of select built environment features related to older adults’ walking at four Assisted Living sites   Assisted Living site’s % agreement between Street View and in-the-field No. segments with item in Street Viewb/no. segments with  item in-the-fieldb  prevalence in Street View (%) Built environment feature  HW1 (%)a  HW2 (%)a   LW3 (%)a   LW (%)4a  AESTHETICS       No litter, graffiti, or broken facilitiesc 100 100 100 100 100 48/48 AMENITIES       >1 benchc,d  25 25 0 25 94 9/10 Public restrooms/washroom 8 0 0 0 98 1/2 DESTINATIONS       Mixed land usee 33 33 8 33 79 13/17 Vertical-mixed use buildingsc  8 0 0 0 98 1/2 Restaurants 8 25 0 0 100 4/4 Grocery / Convenience store 8 25 0 0 100 4/4 Retail stores 8 8 0 0 98 2/3 Bank/financial service 8 25 0 0 98 4/5 Pharmacy/Drug Store 8 0 0 0 96 1/1 Health clinics, Medical facilities/offices  8 8 0 0 100 2/2 Service facilities  8 0 0 0 98 1/2 Senior housingc 0 8 0 17 94 3/0 Senior activities (e.g., senior centers)c 17 17 0 8 90 5/0 Transit stop 8 25 0 8 98 5/6 SAFETY       Intended crossing at both segment endse 83 33 0 17 90 16/15 >1 traffic control devicee 50 25 8 42 81 15/22 >1 crossing aid presente   25 58 17 17 90 14/13  113 Assisted Living site’s % agreement between Street View and in-the-field No. segments with item in Street Viewb/no. segments with  item in-the-fieldb  prevalence in Street View (%) Built environment feature  HW1 (%)a  HW2 (%)a   LW3 (%)a   LW (%)4a  SIDEWALKS       Sidewalks present on 1 or both sidese 100 83 100 83 100 44/44 Sidewalks continuous on 1 or both sidese 50 67 92 67 85 33/38 Good sidewalk conditionc 75 58 83 67 73 34/35 Flat slopec 50 100 100 58 85 37/40 2 ramps/curb cuts at each ende 100 50 0 17 92 20/20 Some/most of segment covered with arcades, awnings, or bothc 8 0 0 0 100 1/1 STREETS        >1 lanec 100 92 100 100 98 47/48 Note. Data based on Rater 1. HW = high walkable; LW = low walkable; SWEAT-R = Seniors’ Walking Environmental Assessment Tool-Revised. aWalk Score used to define the walkability (pedestrian-friendliness) of AL sites where: HW1 - Walk Score = 92; HW2 - Walk Score = 93; LW3 - Walk Score = 48; LW4 - Walk Score = 52.  bntotal = 48 segments (12 segments/site).  cResponse categories for these features were dichotomized; response categories for remaining items are dichotomous in SWEAT-R.  dOriginally assessed as a discrete item (number of benches) using the SWEAT-R.  eFeatures derived by combining individual SWEAT-R items.  114Table 3.2. Comparison of virtual vs. in-the-field audits by rater  aRefer to Chaudhury et al. (2011) for a list of Seniors’ Walking Environmental Assessment Tool-Revised (SWEAT-R) items falling in each domain. bTwenty-three items (Rater 1) and 22 items (Rater 2) not included in the table because they did not display any variability across segments; seven items not included because they could not be assessed by Street View.  cItems evaluating building characteristics (low-rise family housing, predominant building height), sidewalk characteristics (grooves, color contrast, broad-apron curb cuts, continuity, condition), presence of traffic signal, stop sign, mature tree counts.  dItems evaluating vertical mixed use, street condition, presence of houses with front porches, street light counts.  eItems evaluating sidewalk characteristics (condition, slope), mature trees on one side of segment, mature tree counts, bench counts, street light counts.  fItems evaluating sidewalk characteristics (buffer zone, ramps/curb cuts), street condition, presence of houses with front porches, segment ends in a cul-de-sac or dead end.  Domaina   Functionality Destinations and facilities Aesthetics Safety Total itemsb  Absolute agreement Buildings Sidewalks Street features Street life Rater 1 >95% 20 4 1 4 12 2 22 65 80-94% 5 7 0 10 6 2 26 56 70-79%c 2 2 0 1 0 0 5 10 < 70%d 1 0 1 2 0 0 0 4 total items 28 13 2 17 18 4 53 135 Rater 2 >95% 25 2 2 4 8 3 20 64 80-94% 5 9 0 7 4 3 33 61 70-79%e 0 2 0 4 0 0 0 6 < 70%f 0 0 1 3 0 0 1 5 total items 30 13 3 18 12 6 54 136    1153.5 Discussion Agreement between virtual and in-the-field audits was > 80% for most items assessed using the SWEAT-R. I also noted four key benefits to using virtual versus in-the-field audits. First, the overall time required to conduct virtual audits was, on average, two minutes less/segment. Second, raters did not need to travel to the audit site. Third, unlike in-the-field audits, weather, seasonality, or the presence of daylight did not influence the ability to conduct virtual audits. Fourth, virtual audits cost less (compensation for rater’s time and travel cost) and therefore may be more feasible than in-the-field audits.  Consistent with other studies (Clarke et al., 2010; Rundle et al., 2011; B. T. Taylor et al., 2011; Wilson et al., 2012), items that required fine levels of detail (i.e., sidewalks – grooves or bumps, color contrast, curb cuts) and/or for raters to make subjective judgments (presence of mature trees, proportion of houses with front porches, sidewalk and street condition) displayed lower agreement. There was also lower agreement for some items related to building characteristics (i.e., height, mixed-use, bars on windows) and pedestrian safety (traffic signal, stop sign, buffer zone) and for items that required counts of features (counts of mature trees, benches, street lights). However, when I transformed items that required counts into dichotomous variables (feature present vs. absent), absolute agreement was > 80%.   The presence and condition of sidewalks, benches, and destinations, as well as curb height, adequate lighting, and lack of graffiti are microscale features of the built environment that may be associated with older adults’ walking (Hanson et al., 2012; D. King, 2008). My findings indicate the use of Street View may accurately identify the presence of sidewalks, amenities    116(benches, public washrooms), and destinations. However, the use of Street View may not be as accurate as in-the-field audits to identify details associated with certain items, such as counts of street lights; curb cuts - presence, features, and height; sidewalk - continuity, condition, slope. Although others have reported that virtual audits may not accurately capture items for litter and graffiti (Clarke et al., 2010; C. M. Kelly et al., 2012; Rundle et al., 2011; B. T. Taylor et al., 2011; Wilson et al., 2012), I did not find this. Perhaps this was due to the low prevalence of street segments with undesirable aesthetics (only present on one segment, audited “in-the-field” by Rater 2).  I chose to audit street segments surrounding AL sites. These areas have high populations of older adults, including a higher proportion of those with mobility impairments whose activity may be more strongly influenced by an unsupportive built environment as a result of their mobility impairments (A. C. King et al., 2011; Rantakokko et al., 2010; Rantakokko et al., 2009; Satariano et al., 2010; Shumway-Cook et al., 2003). Based on the Press-Competence model (Lawton, 1989; Lawton & Nahemow, 1973), whether or not a given built environment facilitates an older adult’s walking is a product of the match between person-level (e.g., physical functioning, self-efficacy for walking and/or positive attitudes towards physical activity) and environment-level factors. Further, objective measures of the built environment do not necessarily match perceptions of the built environment, and both measures may differentially affect walking behaviour (Arvidsson et al., 2012). Therefore, virtual audits may identify the presence or absence of built environment features that support walking. However, to ascertain whether or not a given built environment is walkable for a given older adult, features of the    117environment, the individuals’ perceptions of the built environment and person-level variables that influence the capacity to walk need to be measured and considered.   The prevalence of built environment features associated with older adults’ walking varied across AL sites. All sites had a high prevalence of segments without litter/graffiti/broken facilities, streets with more than one lane, and sidewalks on one or both sides of the street, and a mixed prevalence of many other features that support older adults’ walking. However, the prevalence of destinations that may be relevant for older adults, such as grocery/convenience stores, restaurants, pharmacy/drug stores, bank/financial services, and transit stops was < 25% across AL sites. Therefore, residents, and especially those with mobility impairments or that are unable to drive, may need to rely on others, such as family members or staff from the AL sites, to access key services. This may be especially relevant in terms of food, since AL sites may only provide lunch and dinner, and residents are responsible for making their own breakfast (Vancouver Coastal Health, 2012). This creates a barrier to older adult independence and their ability to engage with everyday activities within their communities.  I acknowledge three barriers to the use of virtual audits. First, limited coverage was an issue when imagery was unavailable for private strata zones and some smaller side-roads. Second, virtual audits could not measure certain features, such as sidewalk width, buffer zone width, curb height, and timing of pedestrian signals, all of which may be important to older adults’ walking. To evaluate segments not visible on Street View, to assess highly detailed sidewalk features (e.g., condition, curb cut texture) and to obtain data on features that require measurement (e.g., sidewalk width, timing of pedestrian signals), future studies would be strengthened by    118supplementing virtual audits using Street View with in-the-field audits. Finally, Street View does not provide information on when images were last obtained and updated. Many features of the microscale built environment may be stable over time, but this limitation would affect the utility of Street View to assess items that do have temporal variability, or for research that assesses new residential developments or redevelopments. Future studies can partially address this limitation by contacting local authorities to determine when environmental modifications last occurred in the area they wish to audit.  3.6 Conclusions My study extends the emerging body of literature surrounding the use of virtual audits by investigating the appropriateness of Google Earth’s Street View feature for gathering information about built environment features that support older adults’ walking. Street View may be an efficient and appropriate tool to identify the presence of microscale features that potentially support older adults’ walking, such as the presence of sidewalks, amenities (i.e., benches, public washrooms) and destinations. However, Street View may not be as reliable as in-the-field audits to identify details associated with more fine-grained items such as counts of mature trees and street lights, the presence, features, and height of curb cuts, or the continuity, condition, and slope of sidewalks. Thus, Street View may present a convenient, reliable and more affordable option. However, whether or not it is an appropriate choice to virtually audit microscale built environment features associated with older adults’ walking largely depends on the purpose of the audits. Specifically, it is a suitable and attractive alternative to in-the -field audits unless the aim is to identify fine-grain characteristics of the microscale built environment.       119Chapter  4: Walk the Talk: Characterizing mobility in older adults living on low income  4.1 Introduction Mobility is broadly defined as the ability of individuals to move themselves within community environments (Webber et al., 2010). Mobility is a fundamental component of healthy and active aging because it enables engagement in daily activities (e.g., housework, shopping), including physical activity (e.g., walking for exercise) and participation in society (e.g., getting places by car or public transport) (Canadian Institutes of Health Research, n.d.; Satariano et al., 2012). Through hindrance or support of these activities, mobility may also have an influence on social and economic independence (e.g., inability to engage in social activities and/or maintain employment due to lack of transportation) as well as physical and mental health (Canadian Institutes of Health Research, n.d.; Centers for Disease Control and Prevention, n.d.; Satariano et al., 2012). Mobility-disability is defined as difficulty walking up and down a flight of stairs, standing in one spot for 20 minutes or moving from one room to another (Human Resources and Skills Development Canada, 2011). Despite the importance of mobility to everyday life, mobility-disability is the most common type of disability experienced by Canadian older adults (Human Resources and Skills Development Canada, 2011). Approximately 1/3 of Canadians aged > 65 years have a mobility-related disability, with higher prevalence among women and with older age (Statistics Canada, 2008). Given the significance of mobility to healthy and active aging, and the prevalence of mobility-disability among older adults, a thorough understanding of factors associated with older adults’ mobility is a public health priority.    120The built environment is defined as human-made infrastructure that comprises urban design, land use, and transportation systems (S. L. Handy et al., 2002). The built environment plays an important role in older adults’ mobility because it is the context in which outdoor mobility occurs. Further, age-related declines in health and mobility can make it more difficult for older adults to be mobile in the built environment (Noreau & Boschen, 2010; Shumway-Cook et al., 2003). Lawton and Nahemow’s (1973) ecological model of adaptation and aging states that the extent to which an individual successfully functions in his/her environment is a result of the interplay between the individual’s capacity (referred to as individual competence) and the supports and pressures present in his/her environment (referred to as environmental press). Applied to older adult mobility, the model posits that if the pressure imposed by a built environment is greater than an older adult’s functional capacity, the older adult is likely to stop engaging with the outdoor environment (Noreau & Boschen, 2010; Shumway-Cook et al., 2003). Therefore, it is necessary to consider person (individual)-level variables, built environment variables, and their interaction when studying older adults’ outdoor mobility.   Webber et al. (2010) developed a framework of older adult mobility that organizes person (individual)-level variables that influence the capacity to be mobile into four categories of determinants (cognitive, financial, physical, and psychosocial). These categories interact with a fifth (environmental) category to influence older adult mobility. The framework also acknowledges that gender, culture and personal life history indirectly influence mobility through their association with categories of mobility determinants, as well as by shaping individuals’ experiences, opportunities, and behaviours. While Webber et al.’s (2010) framework helps to ensure a holistic approach to measurement of person-level variables that influence the capacity to    121be mobile, another important consideration is the multi-directional association between mobility and its determinants. Mobility limitations can have a negative influence on determinants of mobility, which may in turn directly or indirectly (e.g., through mediating variables such as health) bring about further declines in mobility and health (Satariano et al., 2012). For example, walking difficulties may precipitate increased loneliness in older adults due to decreased autonomy to participate outdoors (e.g., autonomy to make trips and travel, meet other people) (Rantakokko et al., 2014). Loneliness, in turn, may be associated with motor and functional decline in old age (Buchman et al., 2010; Perissinotto, Stijacic Cenzer, & Covinsky, 2012), which may bring about further declines in mobility. On the other hand, positive mobility outcomes/behaviours may directly or indirectly (e.g., through mediating variables such as good health) help promote future mobility. For example, engagement in regular physical activity may reduce perceptions of stress in older adults (Rueggeberg, Wrosch, & Miller, 2012). Lower levels of perceived stress can protect against developing chronic health problems in older adults (Rueggeberg et al., 2012) and thereby have a protective effect against subsequent mobility loss. A comprehensive study of older adult mobility therefore requires that one considers multi-directional and interconnected influences of factors across the individual, environment, and mobility.  Older adults of low SES represent an understudied segment of the older adult population with potentially unique mobility-related needs and characteristics. This segment comprises approximately 12% of Canadian older adults, as estimated by Statistics Canada’s low income cut-off measure, and is on the rise (The Conference Board of Canada, 2013). Older adults of low SES may have an increased reliance on walking to get to places to meet their day-to-day (i.e.,    122basic, social, medical) needs in order to preserve financial resources or as a result of financial restrictions that prohibit them from owning a car or utilizing other travel options (e.g., taxi, bus). Indeed, the association between low SES and decreased likelihood of travel by car is well established in older adults (Cao et al., 2010; Frank et al., 2010; Turcotte, 2012). At the same time, the association between poor health outcomes and low SES is well established in epidemiologic studies (Institute of Medicine (US) Committee on Health and Behavior, 2001; Marmot et al., 1991; Mustard et al., 1997; Reid et al., 1974). Further, studies of older adults report that low SES is an independent risk factor for morbidity, difficulty walking, and incident mobility-disability (Huisman et al., 2003; Koster et al., 2005; Nilsson, Avlund, & Lund, 2010; Shumway-Cook et al., 2005). In addition, there is some evidence to suggest that biomedical factors (e.g., high BMI, high serum levels of inflammatory markers), behavioural factors (e.g., lower physical activity levels), and psychosocial factors (e.g., smaller social networks and lower feelings of self-efficacy) may mediate the association between low SES and mobility-disability, poor function and/or health (Koster, Bosma, van Groenou, et al., 2006; Koster et al., 2005; Ovrum, Gustavsen, & Rickertsen, 2014). Clearly, this presents a complex landscape. However, it speaks to the likelihood that older adults of low SES may rely more on walking to meet their day-to-day needs. At the same time they may be at increased risk of functional limitations that interfere with walking, especially if there is a mismatch between their individual capacity and the pressures exerted by the environment. We need to elucidate characteristics and mobility profiles of older adults living of low SES to better understand and support older adult mobility. Therefore, the primary objective of this study is to comprehensively describe person and environment-level characteristics and mobility of older adults of low SES (as measured by income) across a diverse range of built environments in Metro Vancouver.    1234.2 Methods I provide detailed methods in Chapter 2 of this dissertation and highlight measures relevant to this study below. Briefly, I applied Webber et al.’s (2010) framework to comprehensively measure participants’ capacity to be mobile across person (cognitive, financial, physical, and psychosocial) and environment-level (built and social) domains. I also conceptualized participants’ mobility (enacted function) as physical activity and travel behaviour. I present measures relevant to this study in Table 4.1 below. In addition to the measures in Table 4.1, I measured participants’ sociodemographic characteristics [age, gender, culture (ethnicity, self-identify as visible minority), highest education level attained, years lived in residence, whether they possessed a valid driver’s license, whether they had access to a vehicle in the seven days preceding study participation, and whether they owned a dog] with a self-report questionnaire. I also used a self-report questionnaire to obtain data on participants’ use of mobility aids (number, type) and falls history in the past six months.      124Table 4.1. Select person-level, environment-level and mobility measures used in the study Domain Tool What the tool measures PERSON-LEVELCognitive Montreal Cognitive Assessment [MoCA, (Nasreddine et al., 2005)]  Presence of mild cognitive impairment (MCI). Score of <26 indicates presence of  Physical TANITA Electronic Scale BWB-800 Body mass (kg). Physical Seca Stadiometer Model 242 Height (cm). Physical Short Physical Performance Battery [SPPB, (Guralnik et al., 1994)]  Limitations in lower-extremity functioning. Includes static balance, gait speed, and chair-stand subscales. Individual subscale scores range between 0-4 points and are combined into a summary score that ranges from 0-12. Physical European Quality of Life-5 Dimensions Visual Analogue Scale [EQ-VAS, (Herdman et al., 2011)] Global health. Measured with a visual analogue scale that ranges from 0-100. Physical Functional Comorbidity Index [FCI, (Groll, To, Bombardier, & Wright, 2005)] Presence of eighteen comorbid diseases associated with physical function. Psychosocial Ambulatory Self-Confidence Questionnaire [ASCQ, (Asano, Miller, & Eng, 2007)] Perceived self-efficacy to walk in twenty-two different environment situations. Items are measured on a 10-point scale. Psychosocial Loneliness Questionnaire [adapted from Russel (1996)] General feelings of social isolation and dissatisfaction with social interactions. Items are measured on a three-point scale.  Psychosocial Perceived Stress Scale [PSS, (Cohen, Kamarck, & Mermelstein, 1983)] Degree to which situations that occurred in the last month were perceived as stressful. Items are measured on a five-point scale. ENVIRONMENT-LEVEL Social environment Three-item measure of social interaction [drawn from work by Veroff, Kulka & Douvan (1981)]  Frequency of interaction with friends, neighbours, relatives and/or groups. Items are measured on a five-point scale. Social environment Five-item measure of social cohesion and trust (Sampson, Raudenbush, & Earls, 1997) Perceptions of neighbourhood social cohesion and trust of neighbours. Items are measured on a five-point scale.          125Domain Tool What the tool measures Social environment Five-item measure of physical and social disorder [drawn from the Project on Human Development in Chicago Neighbourhoods (Sampson, 2012)] Perceptions of neighbourhood social disorder and violence. Items are measured on a four-point scale. Built environment Street Smart Walk Score Neighbourhood walkability. Calculated as an index score that ranges from 0-100. Built environment Neighbourhood Environment Walkability Scale – Abbreviated [NEWS-A, (Cerin, Saelens, Sallis, & Frank, 2006)] Perceptions of neighbourhood features related to walking. Items are measured on four-point scale for all but two subscales, which measure items on a five-point scale. MOBILITY  Accelerometry Time (minutes/day) spent in sedentary behaviour, light physical activity, and moderate-to-vigorous physical activity.  Community Healthy Activities Model Program for Seniors (CHAMPS) physical activity questionnaire (Stewart et al., 2001) Type, frequency (times/week) and duration (hours/week) of physical activities that respondents engaged in over the preceding month.  Travel diaries (seven-day) Frequency and characteristics of daily trips (e.g., purpose, travel mode, destination) made over a seven-day period.   Measures of person-level characteristics I used the Montreal Cognitive Assessment (MoCA) as a screening tool for mild cognitive impairment (MCI); a total score < 26 indicates suspected MCI (Nasreddine et al., 2005). I directly measured participants’: i) limitations in lower-extremity functioning and gait speed with the Short Physical Performance Battery (SPPB) (Guralnik et al., 1994), ii) height (cm), and iii) weight (kg). I used ii) and iii) to calculate participants’ body mass index (BMI, kg/m2).      126I used self-report questionnaires to measure participants’: i) global health [European Quality of Life-5 Dimensions visual analogue scale (EQ-VAS) (Herdman et al., 2011)]; ii) comorbidities [Functional Comorbidity Index (FCI) (Groll et al., 2005)]; iii) self-efficacy for walking in different environment situations [Ambulatory Self-Confidence Questionnaire (ASCQ) (Asano et al., 2007) and walking in the neighbourhood (five-point scale, where 1 = “not at all” and 5 = “very much”)]; iv) preference for walking outside (five-point scale, where 1 = “not at all” and 5 = “very much”); v) stress [Perceived Stress Scale (PSS) (Cohen et al., 1983)]; and vi) loneliness [Loneliness Questionnaire adapted from work by Russell (1996)]. In addition to calculating mean scores for each measure, I also list items (environment situations) that participants reported the least confidence with as measured by the ASCQ (Asano et al., 2007) and describe participants’ outcomes for the three-item short-form embedded in the loneliness questionnaire (M. E. Hughes et al., 2004).    Measures of environment-level characteristics  My environment-level variables encompass measures of the social and built environments, described below.  4.2.2.1 Social environment domain I assessed participants’ interpersonal relationships (marital status, living arrangement, and perceived presence of people that offer physical and/or social support to go outside) with a self-report questionnaire and a three-item measure of social interaction [drawn from Veroff et al. (1981)]. I measured neighbourhood social environment characteristics with a five-item measure of social cohesion and trust (Sampson et al., 1997) and a five-item measure of physical and    127social disorder drawn from the Project on Human Development in Chicago Neighbourhoods (Sampson, 2012). I calculated a mean score for each measure.  4.2.2.2 Built environment domain  I measured the walkability of participants’ neighbourhood built environment objectively (Walk Score® and Street Smart Walk Score®) and with a self-report questionnaire [a modified version of the Neighbourhood Environment Walkability Scale – abbreviated (NEWS-A) (Cerin et al., 2006)]. Walk Score and Street Smart Walk Score (www.walkscore.com) are publicly available indices that measure the walkability of the built environment surrounding an address on a scale of 0 (low walkability) to 100 (high walkability) based on distances to nearby destinations. Street Smart Walk Score uses updated methodology (from Walk Score) that better reflects empirical research and is more closely associated with time spent in MVPA for adult and older adult populations (Frank, 2013). I use Walk Score to examine the distribution of participants’ neighbourhood built environments across the range of walkability within the study area.  I use Street Smart Walk Score to examine the distribution of participants’ neighbourhood built environments across categories of walkability, as provided by the manufacturer. These categories (Street Smart Walk Score range) are: “Car-dependent” (0-49), “Somewhat walkable” (50-69), “Very walkable” (70 to 89) and “Walker’s paradise” (90-100). I used a modified version of the NEWS-A to measure participants’ perceptions of neighbourhood built environment features related to walking (Cerin et al., 2006). Subscales of the modified NEWS-A include residential density, land-use mix – diversity, land-use mix – access, street connectivity, aesthetics, traffic hazards, crime, lack of parking, lack of cul-de-sacs, hilliness, and physical barriers.      128 Measures of participants’ mobility  I measured participants’ mobility (physical activity and travel behaviour) by accelerometry, self-report questionnaire, and travel diary, as described below.  4.2.3.1 Physical activity  Accelerometry: At the end of the in-person measurement session, I provided participants with ActiGraph GT3X+ (LLC, Fort Walton Beach, FL) tri-axial accelerometers to objectively measure participants’ physical activity patterns for the seven days following their measurements sessions. I requested participants wear the accelerometer on their right hip, during waking hours, during the following week; the accelerometer was removed during any water-based activities. Participants received a log to record the dates and times they wore the accelerometer.  The accelerometer recorded data continuously (at 30 Hz) and I reintegrated the data to 60-second epochs. I considered more than 60 minutes of continuous zeroes as non-wear time. For analyses, I included data with three or more valid days (> 8 hours wear time/day) of wear time. I used cut-points proposed by C. E. Matthews et al. (2008) to classify time (minutes) spent in sedentary behaviour [<100 counts/minute (CPM)] and the cut-points proposed by Freedson et al. (1998) to classify time (minutes) spent in light (100-1951 CPM) and MVPA ( > 1952 CPM). I also estimated time (minutes) spent in bouts of > 10 minutes of MVPA, allowing for a 1-2 minute interruption. I derived time (minutes) spent in sedentary behaviour, light physical activity, and MVPA using batch processing with ActiLife software version 6.5.4 (LLC, Fort Walton Beach, FL).     129Community Healthy Activities Model Program for Seniors physical activity questionnaire:  I identified the physical activities that participants most frequently reported participating in with the Community Healthy Activities Model Program for Seniors (CHAMPS) physical activity questionnaire (Stewart et al., 2001). CHAMPS captures the type, frequency (times/week) and duration (hours/week) of physical activities that respondents engaged in over the preceding month. Items include pre-defined physical activities that are of light (e.g., light gardening, stretching, light housework), moderate (e.g., water exercises, heavy housework) and vigorous (e.g., jogging, walking uphill, moderate-to-heavy strength training) intensity, and items specific to walking for errands and walking for leisure.   4.2.3.2 Travel behaviour  Travel diaries: I measured participants’ self-reported travel behaviour with travel diaries. Participants recorded their daily trips including start location and time, end location and time, reason, travel mode and social accompaniment. I defined a trip as one-way travel between two locations. Participants filled out their travel diaries in the week immediately following measurement sessions and concurrently with accelerometry data. Detailed information on travel diary data cleaning are found in Chapter 5.  4.3 Statistical analysis I describe participants’ person and environment-level characteristics using means (SD) and counts (percent). The exception is participants’ years lived at current residence, which I present as a median value (25th and 75th percentiles) as these data were skewed (Altman & Bland, 1994). I present Pearson’s correlations coefficients among measures within each person and    130environment-level domain in Appendix E. I describe participants’ mobility (physical activity and travel behaviour) outcomes using medians (25th and 75th percentiles) or counts (percent), as trip frequency and MVPA (minutes/day) data were also skewed. I calculated average daily time (minutes/day) in sedentary behaviour, light physical activity, and MVPA as total time (minutes) divided by valid accelerometry days. I calculated weekly time (minutes/week) in MVPA and whether participants met physical activity guidelines [engaged in > 150 minutes of MVPA/week, accumulated in bouts of > 10 minutes (Tremblay et al., 2011)] by multiplying daily estimates of MVPA (minutes/day) by seven. I standardized estimates of daily time spent in different physical activity intensities (e.g., light, MVPA) and sedentary behaviour to a 13-hour wear day (Herrmann, Barreira, Kang, & Ainsworth, 2013). Trajectories of aging vary between men and women, as does physical activity and travel behaviour (McPherson & Wister, 2008; Spirduso, Francis, & MacRae, 2005; Sun et al., 2013; Turcotte, 2012). Therefore, while not a longitudinal study, I present data separately for men and women. I used Stata version 13.0 for analysis (Stata Corp, TX).   4.4 Results Figure 4.1 represents the flow of participants into the study. One hundred and sixty one individuals participated in the study and overall study recruitment rate was eight percent (161 participants/ 1995 mailed invitations).      131 Figure 4.1. Flow of participants into the study aHouseholds in my study area (Burnaby, New Westminster, North Vancouver, Richmond, Surrey, Vancouver, West Vancouver, White Rock) that receive a Shelter Aid for Elderly Renters (SAFER) rental subsidy from BC Housing, have a head of household aged > 65 years, and a telephone number on file with BC Housing. bCould not be reached again after expression of interest in study participation Total consented n=161 participants Reasons for not participating:  Not interested  n = 706  Could not reach by     telephone   n = 581  Ineligible  n = 334  Declined because of    health problems  n = 118  Away during study n = 40  Lost after telephone           contactb     n = 40  Deceased  n = 7  Other n = 8 Contacted for participation n = 1995 households Random sample  n = 2000 households 5 households excluded because they were contacted previously for participation in the pilot study component of this main study Source populationa n = 5806 households    132 Person and environment-level characteristics of participants I present descriptive statistics for select person and environment-level characteristics of participants, by domain of mobility, below.  4.4.1.1 Sociodemographic information I provide descriptive statistics for select sociodemographic variables in Table 4.2.  Table 4.2. Descriptive statistics for select sociodemographic characteristics Characteristic  nmen/nwomen Men Women Both Age (yrs), mean (SD) 59/102 74.2 (6.3) 74.4 (6.2) 74.3 (6.2) White, % 59/102 73 80 78 Self-identify as visible minority, % 59/101 19 15 16 Educational attainment, % 59/102    Less than secondary school  10 14 12 Secondary school  17 22 20 Some trade/technical school or college   10  16  14  Trade/technical school or college   25 16 19 Some university   19 15  16 University degree or higher (graduate work)  19 19 19 Years lived at current residence, median (p25, p75) 59/102 6.0 (3.0,12.0) 6.6 (3.0, 13.0) 6.2 (3.0, 12.0) Possesses valid driver’s license, % 59/102 78 68 71 Had vehicle at disposal in last 7 days, % 59/100 59 50 53 Owns a dog, % 59/102 8 12 11    133Participants were aged 74.3 (SD=6.3) years, on average. They were predominantly White [men:women (%), 73:80]. Most participants completed > a secondary school education [men:women (%), 90:86]; 19% of participants (men and women) obtained a university degree or higher (e.g., graduate work). Participants lived at their current residence for 6.2 (3.0,12.0; median P25, P75) years. Approximately three-quarters of participants [men:women (%), 78:68] possessed a valid drivers’ license; however, only about half [men:women (%), 59:50] had a vehicle at their disposal in the seven days preceding study participation.   I present descriptive statistics for select person and environment-level characteristics of participants in Table 4.3.       134Table 4.3. Descriptive statistics for select measures by domain of mobility Domain Tool/subscale nmen/nwomen Men mean (SD) Women mean (SD)  Both mean (SD)  PERSON-LEVELCognition MoCAa  59/97 22.4(4.4) 23.3(3.4) 22.9(3.8) Physical BMIb (kg/m2) 59/102 26.9(4.6) 27.0(5.7) 27.0(5.3) Physical FCIc 57/101 2.8(2.0) 3.0(2.2) 2.9(2.1) Physical SPPB total scored 59/102 9.7(1.8) 9.7(2.0) 9.7(1.9)  Gait speed (m/s)e 59/102 1.0(0.2) 1.0(0.3) 1.0(0.3) Physical EQ VASf 58/101 79.6(14.5) 80.3(16.2) 80.1(15.6) Psychosocial ASCQg 59/102 8.6(1.4) 8.2(1.8) 8.4(1.7) Psychosocial PSSh 58/100 12.1(7.6) 12.4(6.9) 12.3(7.1) Psychosocial Loneliness questionnairei 58/102 1.7(0.5) 1.5(0.4) 1.6(0.4) ENVIRONMENT-LEVEL Social environment SC-3PTj 59/102 4.0(1.3) 4.7(1.0) 4.4(1.2) Social environment SC-5PTk 56/101 3.3(0.7) 3.5(0.7) 3.4(0.7) Social environment Neighbourhood disorderl 57/102 1.6(0.6) 1.5(0.4) 1.5(0.5) Built environment NEWS-Am      Residential density 58/100 324.6 (168.2) 333.6 (153.1) 330.3 (158.3)  Land-use mix (diversity) 59/100 2.8(0.9) 2.8(0.9) 2.8(0.9)  Land-use mix (access) 57/101 3.3(0.8) 3.4(0.7) 3.4(0.8)  Street connectivity 57/99 2.9(0.8) 3.2(0.7) 3.1(0.8)    135Domain Tool/subscale nmen/nwomen Men mean (SD) Women mean (SD)  Both mean (SD)   Aesthetics 58/102 2.9(2.6) 3.3(0.6) 3.2(0.7)  Traffic hazardsn 57/98 2.6(0.6) 2.6(0.6) 2.6(0.6)  Crimen 56/96 1.7(0.7) 1.7(0.7) 1.7(0.7)  Lack of parking 45/89 2.2(1.1) 2.2(1.1) 2.2(1.1)  Lack of cul-de-sacs 58/101 3.0(0.9) 3.0(1.1) 3.0(1.0)  Hillinessn 57/102 2.0(1.0) 2.0(1.1) 2.0(1.0)  Physical barriersn 57/102 1.4(0.7) 1.4(0.9) 1.4(0.8) aMoCA= Montreal Cognitive Assessment; scale range 0-30; total score < 26 indicates suspected mild cognitive impairment (Nasreddine et al., 2005).  bBMI= Body Mass Index. cFCI = Functional Comorbidity Index; scale range 0-18. dSPPB = Short Physical Performance Battery; scale range 0-12; total score of 10-12 indicates minimal limitations and a score of 7-9 indicates mild limitations in lower extremity function (Guralnik et al., 1995).  eGait speed calculated from the SPPB; a gait speed of > 0.8 m/s is required for community ambulation (Fritz & Lusardi, 2009), while a gait speed of > 1.2 m/s is needed to cross the street (Asher, Aresu, Falaschetti, & Mindell, 2012; Montufar, Arango, Porter, & Nakagawa, 2007). fEQ VAS = European Quality of Life-5 Dimensions Visual Analogue Scale; scale range 0-100.  gASCQ = Ambulatory Self-Confidence Questionnaire; scale range 1-10. hPSS = Perceived Stress Scale; scale range 0-40. iLoneliness Questionnaire = 11 item questionnaire drawn from the Revised UCLA Loneliness Scale; scale range 1-3. jSC-3PT = 3-item measure of social interaction; scale range 1-6. kSC-5PT = 5-item measure of social cohesion and trust; scale range 1-5. lNeighbourhood Disorder = 5-item measure of neighbourhood physical and social disorder; scale range1-4. mNEWS-A= Neighbourhood Environment Walkability Survey-Abbreviated; residential density sub-scale range 173-865; land-use mix diversity sub-scale range 1-5; range of other subscales 1-4.  nfor these NEWS-A subscales, higher scores indicates worse walkability.      1364.4.1.2 Cognitive domain One hundred and fifteen of the 156 participants (75%) that completed the MoCA scored below the cut-off for suspected MCI [total score < 26 (Nasreddine et al., 2005)].  4.4.1.3 Physical domain  Participants were diagnosed with three chronic conditions, on average – most commonly arthritis (48%), visual impairment (43%), and/or obesity (23%). Most participants had few physical limitations, as indicated by the SPPB, gait speed, use of mobility aids, and falls history. Specifically, sixty percent of participants [men:women (%), 59:61] had minimal (SPPB score 10-12), 33% of participants [men:women (%), 36:31] had mild (SPPB score 7-9) and seven percent of participants [men:women (%), 5:8] had moderate (SPPB score 4-6) limitations in lower extremity function. Further, 79% of participants [men:women (%), 76:80] had a gait speed that supported community ambulation (> 0.8 m/s) (Fritz & Lusardi, 2009). Eighty-four percent of participants [men:women (%), 88:81] reported that they did not use a mobility aid when walking; 81% of participants [men:women (%), 86:77] did not report a fall during the six months preceding study participation.  4.4.1.4 Psychosocial domain Most participants liked to walk outside (88% scored > four on a five-point scale) and were confident walking in their neighbourhood (93% scored > four on a five-point scale). Participants reported being least confident walking on a moving bus, walking in the dark or at night, and walking on uneven or slippery ground (as measured by the ASCQ). Participants’ average score on the Loneliness Questionnaire indicates that they were most typically lonely “hardly ever or    137never” or “some of the time.” The three-item short form of the Loneliness Questionnaire indicated participants (%) “hardly ever or never,” “some of the time,” and “often,” respectively, felt they lacked companionship (48, 41, 11), left out (68, 26, 6), or isolated from others (64, 25, 11). Participants’ mean (SD) perceived stress scores were similar to reference population norms for American older adults [12.3 (7.1) vs. 12.0 (6.3) (Cohen, 1988)].   4.4.1.5 Social environment domain Most participants were unmarried [never married, widowed, separated, or divorced; men:women (%), 81:97] and lived alone [men:women (%), 68:88]. Forty-nine percent of men and 70% of women felt people in their lives offered physical and/or social support when they went outside. Most participants reported frequent social interactions, particularly with friends or relatives. Approximately 2/3 of participants [men:women (%), 56:75] visited with friends or relatives > once a week and approximately 2/3 of participants [men:women (%), 64:72] talked on the telephone or exchanged emails > once a day. Approximately 60% of participants [men:women (%), 51:65] attended group programs, clubs, or organizations (they belonged to) > once a month. The vast majority (~90%) of participants tended to have neutral or positive views about the social cohesiveness of their neighbourhoods, except when asked about shared values among neighbourhood residents; for this item, answers were evenly distributed among strongly disagreed/disagreed, neutral and agreed/strongly disagreed categories. Finally, participants generally reported no (2/3 of participants) or a little (1/4 of participants) neighbourhood physical and social disorder, except when asked about how much broken glass or trash they see on neighbourhood sidewalks and streets; responses for this item were more evenly distributed    138among the first two answer categories, where approximately half of participants reported “no” and 1/3 of participants reported “a little” disorder.  4.4.1.6 Built environment domain Participants’ neighbourhoods spanned the walkability continuum, as measured by Walk Score and Street Smart Walk Score. Table 4.4 displays the distribution of participants across strata (deciles) of Walk Score. As percentages, nineteen (both men and women) resided in Car Dependent, 21 in Somewhat Walkable (men:women, 22:20), 25 in Very Walkable (men:women, 20:28), and 35 in Walker’s Paradise (men:women, 39:33) neighbourhoods. Further, participants generally reported positive perceptions of walkability (mean scores were in the positive half of the scale range e.g., > 2.5 on a four-point scale; higher scores reflect a more pedestrian friendly environment) across NEWS-A subscales (range = 59%-87% of participants). Exceptions were residential density, traffic hazards, and lack of parking subscales, where participants (%) with positive perceptions were 15 (men:women, 17:14), 42 (men:women, 47:39), and 42 (men:women, 40:43), respectively.     139Table 4.4. Distribution of participants (n) across strata of Walk Score Decile Walk Score rangea  n 1 94-100 14 2 88-93 21 3 79-87 20 4 73-78 21 5 68-72 12 6 61-67 22 7 53-60 17 8 44-52 10 9 33-43 12 10 0-32 12 abased on the distribution of Walk Scores of a random sample of post codes (6-digit, n = 2500) within the study area.      Mobility  I provide descriptive data for objectively measured physical activity and travel behaviour in Table 4.5.      140Table 4.5. Descriptive statistics for select mobility outcomes  Predictor  nmen/nwomen Men median (p25, p75) Women median (p25, p75) Both median (p25, p75) Physical activitya     Sedentary behaviour (minutes/day) 49/92 570.5  (531.9, 612.9)  533.7  (486.0, 578.0)  546.9  (494.2, 596.1)  Light physical activity (minutes/day) 49/92 191.5 (142.5, 217.9) 232.1  (183.0, 277.2)  207.9  (170.3, 261.7)  MVPAb (minutes/day) 49/92 18.5  (3.7, 36.2) 9.7  (2.7, 28.8) 11.6  (2.9, 31.2) Travel behaviour     Trip frequency (trips/day) 48/99 3 (2,5) 3 (2,5) 3 (2,5) Trip mode (%) 48/99    walk  41 36 38 car  37 43 41 transit  17 17 17 other  5 4 4 Trip purpose (%) 48/99    Shopping/errands  51 49 50 Social/entertainment/food  23 24 24 Exercise  16 14 15 Other  10 13 11 aAs measured by accelerometry (ActiGraph GT3X+, 60 second epochs), based on > 3 days with > 480 minutes/day valid wear time. Estimates have been standardized to a 13-hour wear day.  bMVPA = moderate-to-vigorous physical activity  4.4.2.1 Physical activity Participants spent most of their day engaged in sedentary behaviour [men:women (%), 73%:68%]. Men spent 24% of their day engaged in light physical activity and 3% in MVPA. Women spent 30% of their day engaged in light physical activity and 2% in MVPA. Importantly, MVPA incorporates both moderate and vigorous physical activity. Men and women spent a negligible percentage of their day in vigorous physical activity (0.05 and 0.03, respectively). Fifty-four participants (39%) engaged in > 150 minutes of MVPA/week [men:women (%), 49:33]; twenty-five (18%) participants engaged in > 150 minutes of MVPA/week, accumulated    141in bouts of 10 minutes or more [men:women (%), 22:15]. Percentage of time spent by participants in physical activity and sedentary behaviour was similar to data for Canadian older adults (aged 60-79 years) assessed as part of in the Canadian Health Measures Survey (accelerometry; 2007-2009) (Colley et al., 2011).  The most common types of physical activity participants engaged in (as measured by CHAMPS) were light work around the house (93%); walking for errands (77%); walking leisurely for exercise or pleasure (66%); stretching/flexibility exercises (51%); walking/hiking uphill (48%); light gardening (45%); and walking fast or briskly for exercise (38%). Less than 25% engaged in other activities.  4.4.2.2 Travel behaviour I provide an in-depth analysis of participants’ common travel destinations in Chapter 5. Participants made three (2, 5; median P25, P75) trips/day. Approximately half of all trips were made to run errands/go shopping, 24% were for social/entertainment/eating out, 15% for exercise, and 11% for other purposes (e.g., medical appointments, volunteering/work, to attend a place of worship). Forty-one percent of trips were by car [men:women (%), 37:43], 38% by foot [men:women (%), 41:36], 17% by transit, and 4% by other travel modes (e.g., bicycle, taxi). To compare, regional travel survey data showed that older adults (65-79 years old) in Metro Vancouver made a far smaller proportion of trips by foot (8%) and a higher proportion of trips by car (82%) (TransLink, 2010).    1424.5 Discussion I extend and enhance the scant literature on older adults living on low income by providing an in-depth description of their person and environment-level characteristics as they relate to mobility. I found that participants tended to make a high proportion of trips by foot, as compared to regional travel survey data. However, I did not see, on average, physical activity levels above population-level norms. Further, although older adults of low SES may be at increased risk of morbidity, poor physical function and incident mobility impairment (Huisman et al., 2003; Koster, Bosma, van Groenou, et al., 2006; Koster et al., 2005; Nilsson et al., 2010; Ovrum et al., 2014; Shumway-Cook et al., 2005), this was not the case in this cohort of participants. For example, participants’ self-reported health, as measured by the EQ-VAS, is similar to population norms available for older adult residents of Alberta, Canada (80% vs. 76% for older adults aged 65-74 years and 68% for older adults aged > 75 years) (Johnson & Pickard, 2000) and their average self-reported number of comorbidities (three) is similar to population norms available for Canadian older adults (Wister, Levasseur, Griffith, & Fyffe, 2015). Furthermore, over 90% of participants have minimal-to-mild limitations in lower extremity function as assessed by the SPPB (Guralnik et al., 1995) and almost 80% have a gait speed (> 0.8 m/s) that supports community (Fritz & Lusardi, 2009). I consider three factors that could result in differences between my cohort and other samples of older adults of low SES. First, I do not know how long participants were in a low socioeconomic strata, as my data do not provide a person’s socioeconomic history. Thus, low income could be either a recent circumstance or a lifelong situation. Second, I consider the potential influence of synergy among mobility domains and overall competence of participants. Finally, I am unable to rule out selection bias. I discuss each below.     143First, socioeconomic indicators are diverse and typically include measures of income, education level, and/or occupational status (Grundy & Holt, 2001). Education and occupational status may represent social class at an earlier stage of the life course, whereas income may represent social class at a later life stage (Koster, Bosma, Kempen, et al., 2006). Both measures of SES are associated with health outcomes in old age (Grundy & Holt, 2001; Huisman et al., 2003). Although study participants were considered of low SES as measured by income, they were highly educated as compared with normative data for Canadian older adults. Eighty-eight percent of participants completed secondary school and 19% completed post-secondary education (obtained a bachelor’s degree or higher). In comparison, approximately 57% of Canadian older adults aged > 65 years completed secondary school and 10% completed post-secondary education (obtained a bachelor’s degree or higher) (Statistics Canada, 2009). This potential discrepancy between participants’ present SES and that of earlier life stages needs to be addressed. In future, studies of older adults living on low income might assess SES across the lifespan. This would shed some light on what may be different roles for more recent versus lifelong low SES on older adult mobility.  Second, participants generally displayed positive profiles across person-level and environment-level domains of mobility. While I cannot ascertain causality, synergy between and within person and environment-level factors most likely contributed to the positive profiles across most domains of mobility. For example, features of the social environment such as social networks, social interactions and social support are associated with older adults’ physical functioning and health, as well as mobility patterns and time spent in physical activity (Annear et al., 2014; Hanson et al., 2012). Psychosocial resources such as control beliefs, coping styles,    144positive/negative emotion states, and social support may partially mediate the association between SES and health (K. A. Matthews, Gallo, & Taylor, 2010; S. E. Taylor & Seeman, 1999). Participants’ high level of competence across domains may also have allowed them to better utilize and/or adapt their personal and environmental resources to meet their needs and maintain their mobility and independence (Lawton, 1989). For example, although most participants were single and lived alone, they regularly socialized with friends, family, and/or neighbours and generally did not report being lonely or stressed.  While my current study focused on mobility as it relates to person and environment-level resources and barriers, I pause here to consider how behaviour driven agency might shape the person-environment interaction (Wahl, Iwarsson, & Oswald, 2012). Older adults’ mobility is heavily influenced by the interplay between person and environment-level characteristics. However, older adults are not passive recipients of the environment’s demand characteristics but proactively shape their environment, tasks or self to meet their needs and to maintain their independence (Lawton, 1989). To illustrate, older adults may decide to walk with a mobility aid to compensate for loss of physical function. This active choice promotes their continued mobility in an environment that would otherwise not be walkable. They might also otherwise self-restrict their activities to within environments and familiar places they perceive to suit their capacities and needs. I did not measure older adults’ proactive behaviours directly. However, in a companion study, Walk the Talk researchers used qualitative methods and a strengths based approach to assess a subgroup of Walk the Talk participants to better understand factors that facilitated physical activity among highly active participants (Franke et al., 2013). Resourcefulness (e.g., engagement in self-help strategies such as self-efficacy, self-control and    145adaptability) was a key facilitator to physical activity, despite personal challenges. The nuanced and multi-faceted nature of the intersection between person and environment-level factors and older adult mobility is worthy of further investigation.   I am drawn to one exception to participants’ generally positive profiles across the domains of mobility. A relatively high proportion of participants (75%) scored below the cut-off point for MCI (MoCA score < 26) (Nasreddine et al., 2005). Interestingly, recent community-based studies reported similar rates and suggested low scores might be due to either poor specificity of the measurement tool or a true higher proportion of older adults with cognitive dysfunction across a range of populations (Freitas, Simoes, Alves, & Santana, 2011; Fujiwara et al., 2013; Narazaki et al., 2013; Rossetti, Lacritz, Cullum, & Weiner, 2011). Dementia is a decline in cognitive function severe enough to affect social or occupational functioning (American Psychiatric Association, 1994). However, MCI typically does not interfere with independence in everyday life (Lin, O’Connor, Rossom, Perdue, & Eckstrom, 2013). Nonetheless those with MCI are at an increased risk of developing dementia (Dong et al., 2012; Petersen, 2004; Petersen et al., 1999). Importantly, physical activity was associated with reduced risk of cognitive decline and dementia (Blondell, Hammersley-Mather, & Veerman, 2014).This speaks to an even greater need for environments that support physical activity of older adults with suspected MCI.   A third factor that warrants consideration when discussing my study findings is selection bias. Individuals of low SES are an understudied subgroup. There are many reasons for this – for example, they are more difficult to access and are more concerned than others their age about participating in studies (Schnirer & Stack-Cutler, 2011). Walk the Talk employed several    146strategies to address these considerations so as to enhance participation. We established a key partnership with BC Housing who provided a sampling frame of community-dwelling older adults living on low income. BC Housing had established relationships with many older adults who met eligibility criteria. They served as knowledge brokers and addressed participant questions or concerns as a means to enhance participant buy-in. We also sought to diminish barriers to study participation and reduce non-response bias (e.g., lack of vehicle access, concerns with transportation costs, trouble walking). In the first instance, we conducted measurement sessions at local community centres that were in close proximity to older adult participants who lived outside Vancouver. We also offered free rides to/from measurement sessions. We enlisted a courier service to retrieve take home measures from participants during the week following measurement so as not to inconvenience them with ‘mail ins’ or ‘drop offs.’   Despite these best efforts, recruitment rate into my study was 8%. Selection bias may be present if healthy participants were more likely to participate. Although other studies that evaluated the association between the mobility of community-dwelling older adults and the built environment reported recruitment rates as high as 20-25%, they did not target low income or other vulnerable populations [e.g., (Davis et al., 2011; A. C. King et al., 2011; Rosso, Grubesic, Auchincloss, Tabb, & Michael, 2013; Winters, Voss, et al., 2015)]. Others document that older adults of low SES are less likely to participate (Martinson et al., 2010). It is possible that my relatively low recruitment introduced selection bias that influenced health and mobility outcomes. When I compared participants to those in the sampling frame, the gender distribution was similar although participants in my study were slightly younger (74 vs. 77 years).      147I note that my study had a number of strengths. These include: i) a theoretical framework that guided selection of variables across person and environmental-level domains of mobility (Webber et al., 2010); ii) a community partner who guided recruitment of a well-defined source population of older adults living on low income; iii) an objective measure of the physical determinants of mobility (SPPB) to complement self-report questionnaires; iv) objective measurement of physical activity and sedentary behaviour (accelerometry) to complement a self-report questionnaire on specific physical activities participants engaged in; and v) a seven day travel diary to measure travel behaviour.   I also acknowledge that my study had several limitations. As the primary aim of Walk the Talk was to assess the association between the built environment and older adult mobility, participants were required to have a minimum level of mobility to be eligible for study participation. These inclusion criteria prevent me from generalizing my findings to a larger population of older adults living on low income, those that have severe mobility impairments and those unable to leave their homes. I did not assess a higher income comparison group, but used population-level comparators when available. Finally, the cross-sectional design prevents me from drawing causal inferences from my findings.  4.6 Conclusions Unlike others, my findings did not generally support that older adults living on low income were at increased risk of mobility limitations (Koster, Bosma, van Groenou, et al., 2006; Koster et al., 2005; Nilsson et al., 2010; Ovrum et al., 2014; Shumway-Cook et al., 2005). Despite an economic disadvantage, participants presented with positive profiles across person and    148environment-level domains of mobility. Participants also made a high proportion of trips by foot, although these did not together serve to meet physical activity guidelines for most. I challenge future researchers to focus on innovative strategies to recruit this difficult to access population into studies, to consider the influence of SES across the lifespan, as well as to consider the role of behaviour driven agency when investigating the association between the person, environment, and older adult mobility.      149Chapter  5: Destinations matter: The association between where older adults live and their travel behaviour  5.1 Introduction The world’s population is rapidly aging, with individuals aged > 65 years projected to account for 16% of the world’s population by 2050 (United Nations Department of Economic and Social Affairs, n.d.-b). As individuals age, health declines and the prevalence and severity of most types of disabilities increases (Brault, 2012; Statistics Canada, 2006). Mobility, broadly defined as the ability of individuals to move themselves (e.g., either independently or by using assistive devices or transportation) within community environments (Webber et al., 2010), is especially affected by age. For example, approximately 40% of American older adults aged > 65 years experience difficulty walking, climbing stairs, or using a wheelchair, cane, crutches, or walker (Brault, 2012). Similarly, mobility limitation affects one in three Canadian older adults (Statistics Canada, 2006). Relative freedom in walking or driving is integral to healthy aging and quality of life (Satariano et al., 2012), and even a small amount of regular walking can play a key role in the maintenance of functional independence in old age (Simonsick et al., 2005). In contrast, mobility limitations are associated with decreased social participation (James, Boyle, Buchman, & Bennett, 2011), increased annual health care costs (Hardy, Kang, Studenski, & Degenholtz, 2011), risk of depression (Ragland, Satariano, & MacLeod, 2005) and mortality (Hardy et al., 2011). Given the prevalence and consequences of mobility limitations with increased age, a better understanding of factors that support older adults’ mobility, and especially walking, is crucial to support good health in an aging population.    150Person (i.e., financial, psychosocial, physical, and cognitive) and environment-level factors influence older adult mobility (Webber et al., 2010). The built environment, defined as urban design, land use, and transportation systems (S. L. Handy et al., 2002), is a central environment-level determinant of mobility as it is the setting where mobility occurs. Importantly, according to travel demand theory, travel is a ‘derived’ demand, which means that individuals are typically mobile in order to reach destinations (Crane, 1996). Therefore, an understanding of destinations that are relevant to older adults is critical to the design of built environments that support older adult mobility. Living in neighbourhoods that are close to destinations may also provide the opportunity for older adults to walk, instead of drive, and thereby obtain incidental physical activity. Some studies found a positive association between the presence and proximity of various types of destinations and older adults’ walking (Cao et al., 2010; Gauvin et al., 2012; Michael et al., 2006; Michael et al., 2010; Nagel et al., 2008; Nathan et al., 2012), walking for transportation (Cao et al., 2010; Cerin et al., 2013; D. King, 2008), and for physical activity (Cao et al., 2010; Cerin et al., 2013; D. King, 2008). However, findings regarding the specific types of destinations associated with walking are not consistent. This may be due to broad categories used to classify destinations, as well as inconsistency between studies in the specific types of destinations included in destination categories. This makes it difficult to identify which specific destinations encourage older adults’ mobility. Additionally, broad categories limit the ability to translate findings into targeted changes and to specifically influence urban planning practice. Therefore, we know relatively little about the specific types of destinations that are relevant to older adults.  We know even less about destinations that support the mobility of older adults of low socioeconomic status (SES) or where they travel to meet their day-to-day needs. This is    151important as limited financial resources may affect car ownership and/or the affordability of taxis or even transit. Studies show older adults of low SES have a decreased reliance on travel by car (Cao et al., 2010; Frank et al., 2010; Turcotte, 2012). However, individuals of low SES are also at increased risk of functional impairment (Koster, Bosma, van Groenou, et al., 2006; Koster et al., 2005). Thus, this population may have a greater reliance on walking, and at the same time, may be faced with greater challenges walking. Living in close proximity to relevant destinations may be especially important to this population’s ability to live independently.   Therefore, this paper uses seven-day travel diary data to measure the mobility (travel behaviour) of older adults living low income. My specific objectives are to: i) describe the types of destinations that older adults living on low income most commonly travel to in a week and ii) determine the association between the prevalence of neighbourhood destinations and the number of walking for transportation trips (average/day) these older adults make.   5.2 Methods I provide detailed methods in Chapter 2 of this dissertation. I highlight measures, as well expand on methods specific to this study, below.     Study measures Participants took part in one, two-hour measurement session conducted March - May 2012. I applied Webber et al.’s (2010) framework to comprehensively measure participants’ capacity to be mobile across person (cognitive, financial, physical, and psychosocial) and environment-level (built and social) domains. I provide details of measures relevant to this study below.     1525.2.1.1 Independent variables I used the Street Smart Walk Score® as an objective measure of the built environment. This was my main effect of interest and served as a proxy measure for prevalence of neighbourhood destinations. Street Smart Walk Score measures walkability of an address on a scale of 0 (low) to 100 (high) based on network distances from the address to nine different amenity (destination) categories (e.g., grocery stores, restaurants, shopping); it also takes into account the presence of street network characteristics (intersection density and block length) that do not support pedestrian friendliness. I used a self-report questionnaire to measure participants’ sociodemographic (e.g., age, gender, marital status, living arrangement, dog ownership) and select mobility (e.g., preferences for walking, vehicle ownership, falls history) characteristics. I obtained data on the presence of comorbidities with the Functional Comorbidity Index (FCI) (Groll et al., 2005). I calculated participants’ gait speed as part of the Short Physical Performance Battery (SPPB); the SPPB measures gait speed based on the time taken to walk 4-meters at usual pace (Guralnik et al., 1994). Finally, I measured participants’ ambulatory confidence with the Ambulatory Self-Confidence Questionnaire (ASCQ) (Asano et al., 2007).   5.2.1.2 Travel behaviour I used travel diaries to prospectively gather data on trips participants made in the week following measurement sessions. I asked participants to record all trips, where I defined a trip as one-way travel between two destinations. Data included start location (address or intersection) and time, end location and time, destination name, trip purpose (walk, volunteer, exercise, education shopping/errands, social/entertainment, health appointment, other), travel mode (walking, bicycle, wheelchair, scooter, transit, taxi, car, “other”), and accompaniment (alone, spouse,    153sibling, child, friend, neighbour, volunteer, other). A research assistant entered travel diary data into Excel. I established quality of data entry by checking a random sample of 10% of the entered trips.  5.2.1.3 Destinations I used the four-digit 2012 North American Industry Classification System (NAICS) to systematically code the names of destinations visited by participants into 72 destination categories (referred to as destinations herein). The NAICS provides a framework for classification of businesses according to type of economic activity. I modified the NAICS by collapsing small and large grocery stores into a single destination and by the addition of five destinations reflective of participants’ travel: i) private residence (other than the participants’ residences); ii) nursing/care home; iii) park, beach, trail; iv) pleasure drive (trips made by car that began and ended at a participant’s residence); and v) neighbourhood stroll (walks for exercise or pleasure that began and ended at a participant’s residence). Some NAICS destinations have non-intuitive titles, and I refer to these throughout the manuscript according to the most common type of business represented by the destination. These include [NAICS (business)]: lessor of real estate (mall); depository credit intermediation (bank); religious organization (place of worship); other information services (library); individual and family services (seniors' center); other amusement and recreation industry (recreation centers). Given my focus on daily travel, I excluded trips with missing travel modes and trips that were for tourism-related activities [i.e., sight-seeing, travel outside of the study area (e.g., Vancouver Island)] from my analyses.     1545.3 Statistical analysis I summarized participants’ sociodemographic and mobility (capacity) characteristics, overall travel behaviour (frequency, purpose, travel mode, number destinations visited/week), and travel behaviour (frequency, purpose) stratified by destination as medians (25th and 75th percentiles) or counts (percent). The denominator for frequency of travel to a given destination is participants who made > 1 trip to a given destination. Although I measured the variable “likes to walk outside” with a five-point scale, I dichotomized this variable for my analyses given there were very few responses in categories aside from “very much like to walk outside.” I present descriptive data for destinations that > 20% of participants traveled to during the observation week. I refer to these collectively as most common destinations.  I used a negative binomial model to determine the association between Street Smart Walk Score (used as a proxy measure for the prevalence of neighbourhood destinations) and number (counts) of walking for transportation trips; I included number of days of travel diary data as an exposure variable. I excluded walking trips that were clearly for leisure, where I defined leisure as those trips that were going for a walk, or exercise on the trip in and of itself (e.g., neighbourhood strolls, where the start and end location was the participant’s house), which accounted for <10% of walking trips. I fitted three multivariable models: one that estimated the main effect of Street Smart Walk Score on number (counts) of walking for transportation trips (average/day); a second model identical to the first but that also controlled for the potential effect of age and gender; and a final model where I controlled for sociodemographic and mobility characteristics that had bivariate associations with the outcome at p < 0.20 (Vittinghoff, Glidden, Shiboski, & McCulloch, 2006d). Bivariate analyses included t-tests for dichotomous data and Pearson’s    155correlation coefficients for continuous data. I chose sociodemographic and mobility characteristic variables based on their known association with travel behaviour and mobility (Brown & Flood, 2013; Moniruzzaman, Páez, Nurul Habib, & Morency, 2013; Turcotte, 2012). In order to test the effect of the exclusion of walking for leisure trips on my estimates of effect, I ran sensitivity analyses with ‘all walking trips’ as the outcome and obtained similar results. I present select sociodemographic and mobility characteristics, bivariate associations between each variable and number of walking trips (average/day), and estimates from negative binomial regression analyses for number of walking trips (average/day) in Tables F.1 and F.2 of Appendix F. Although I did not use a cluster sampling design, I acknowledge that there is a potential for clustering within city (i.e., dependence in observations between participants that live in the same city); if present, clustering undermines the statistical assumption of independency of observations (Cerin, 2011). To investigate the potential effect of clustering within city I ran two more models: first, a model that included fixed effects for city, and second, a model with a cluster robust variance (Tables F.3 and F.4 of Appendix F) (Cerin, 2011). The point estimate for a 10-point change in Street Smart Walk Score for the model with city fixed effects was similar to the simplest model without adjusting for clustering. There are limitations using these two models when cluster size varies or is small (as in my study). As such I present the simplest model (without adjusting for clustering) in my dissertation. I considered p < 0.05 to be statistically significant in multivariable analyses. I carried out all analyses with Stata version 13.0 (Stata Corp, TX).     1565.4 Results A detailed description of the flow of participants into the study is found in Chapter 4. Briefly, of 5806 households in my source population, I randomly sampled 2000 individuals (from 2000 households) and contacted 1995 individuals (from 1995 households) for study participation. Of these, 161 individuals (102 women and 59 men) signed consent forms and participated in measurement sessions. The recruitment rate (contacted for participation/signed consent form) was 8%. Of the 1834 individuals who did not participate in the study, I could not reach approximately 32% in-person. Reasons for non-contact included wrong/inactive telephone numbers (11%), inability to be contacted directly at the place of residence (<1%, e.g., resided at a hotel), and failure to return telephone messages (20%). Of the remainder (68%) of individuals that did not participate, 38% declined because they were not interested in study participation, 18% did not meet inclusion/exclusion criteria, 6.4 % declined because of health problems, and 5.6% did not participate for other reasons (e.g., away during study measurement sessions, deceased).   Ninety three percent (150/161) of participants contributed > one day of travel diary data for analysis. Of the 11 travel diaries not included in my analysis, eight were returned blank, one was illegible, and one was excluded because a participant was not based at their home. One participant refused to fill out the travel diaries due to language barriers. Participants that completed travel diaries logged trips for a median of seven days. Table 5.1 shows participants’ sociodemographic and mobility characteristics. Participants were 74 years old (70, 79; median P25, P75), approximately two-thirds were female, and 80% lived    157alone. Participants had few physical mobility limitations, as assessed by self-report and objective measures. Only half of participants stated that they had a vehicle at their disposal. Street Smart Walk Score and ten sociodemographic and mobility characteristics were associated with number of walking for transportation trips.      158Table 5.1. Descriptive statistics for select sociodemographic and mobility characteristics and bivariate association between each variable and number of walking for transportation trips (average/day)   Characteristic n % P a Age 150 74 (70, 79)b 0.112 Gender   0.449 Men 51 34  Women 99 66  Married   0.698 Yes 13 9  No 137 91  Living arrangement   0.113 Lives alone 121 81  Lives with others 29 19  Likes to walk outside   <0.001 Very much (5 on a 5-point scale) 104 69  Less than very much (1-4 on a 5-point scale) 46 31  Use mobility aid   0.043 Yes 24 16  No 126 84  Had vehicle at disposal in last 7 days (n=148)   <0.001 Yes 79 53  No 69 47  Owns a dog   0.296 Yes 15 10  No 135 90  Fell in previous 6 months   0.028 Yes 31 21  No 119 79  Comorbiditiesc (n = 148) 148 3 (1,4)b <0.001 Gait speed (m/s) 150 1.0 (0.8, 1.2)b 0.076 Community ambulator (gait speed > 0.8 m/s)    No 32 21 0.198 Yes 118 79  Ambulatory Self-Confidence Questionnaire 150 8.9 (7.5, 9.7) b 0.001        159Characteristic n % P a Street Smart Walk Score 150 80 (54, 92)b <0.001 abivariate analyses included t-tests for dichotomous data and Pearson’s correlation coefficients for continuous data bmedian (p25, p75) ctotal number; measured with the Functional Comorbidity Index  Participants recorded a total of 3334 trips. Participants made three (2,5; median P25, P75) trips/day. Thirty-four percent of participants reported that they stayed at home on > one day of the observation week. Most trips were made by car (41%) or by walking (38%). Public transit accounted for 17% of all trips. Only 4% of trips were made by other travel modes such as bicycle or taxi. The most common trip purpose was shopping/errands (50% of all trips), social/entertainment/eating out (24% of all trips), and exercise (15% of all trips). The remainder of trips were made to attend health appointments, for volunteer/work, or other (e.g., religious, education).  Participants made trips to six different destinations/week (5, 9; median P25, P75) and made 1-2 trips/week (median) to each destination that they visited. Figure 5.1 displays the number of participants that made > one trip to the most common destinations. More than 75% of participants made a trip to the grocery store and over 50% of participants made a trip to a restaurant/café and/or a mall during the week data were collected. Between 20-42% of participants made > one trip to nine other most common destinations.      160  Figure 5.1. Number of participants that made > 1 trip/week to most common destinations (n=150)b aOther (n): library (38); neighbourhood [stroll] (38); seniors' center (37); natural environment (34); recreation centers (31). bI present only the destinations that > 20% of participants made > 1 trip to.  I define most frequented destinations as destinations visited by the greatest proportion of participants. The most frequented destinations for the purpose of shopping/running errands were the grocery store, mall, bank, health and personal care store, and library. The most frequented destinations for the purpose of socialization/entertainment/eating out were restaurants/cafés and private residences. The most frequented destinations for the purpose of exercise were neighbourhoods, natural environments, and recreation centers. A similar proportion of participants made a trip to exercise in a natural environment such as a park or the beach, and/or 0 10 20 30 40 50 60 70 80 90 100 110 120 130 140 150OtherPlace of worshipHealth and personal care storeBankPrivate residenceRestaurant/caféMallGrocery storeParticipants (n)a   161around their neighbourhood as made a trip to exercise in a more formal exercise setting (i.e., recreation centers). Participants most frequently visited seniors’ centers for the purpose of exercise as well as for the purpose of socialization/entertainment/eating out.  Table 5.2 displays negative binomial models fitted to determine the association between Street Smart Walk Score and number of walking for transportation trips. There was a significant positive association between Street Smart Walk Score and number of walking for transportation trips in all three models. The final model demonstrates that for each 10-point increase in Street Smart Walk Score, the number of walking for transportation trips increased by 20% (IRR = 1.20, 95% CI = 1.12, 1.29). There was a significant positive association between number of walking for transportation trips and liking to walk very much (IRR = 2.10, 95% CI = 1.40, 3.13) and a significant negative association between number of walking for transportation trips and vehicle availability (IRR = 0.54, 95% CI = 0.39, 0.75) and number of comorbidities (IRR = 0.88, 95% CI = 0.80, 0.96). Liking to walk very much [vs. 1-4 (not at liking to walk to somewhat liking to walk) on a five-point scale] increased the number of walking for transportation trips by 110%, while having a vehicle available in the week before assessment decreased the number of walking for transportation trips by 46%. Each additional comorbidity decreased the number of walking for transportation trips by 12%.      162Table 5.2. Estimates from negative binomial regression analyses for number of walking for transportation trips (average/day)   Unadjusted Incidence  Rate Ratio (95% CI) Adjusted Incidence Rate Ratio  (95% CI)    P  P  (n = 145) Model 1 (n=145)  Model 2a (n=141)  Street Smart Walk Score (10-point change) 1.29 (1.19, 1.39) 1.30 (1.20, 1.40) <0.001 1.20 (1.12, 1.29) <0.001 Age (10-year change) 0.74 (0.54, 1.00) 0.66 (0.54, 0.90) 0.020 0.78 (0.58, 1.03) 0.081 Female 0.86 (0.56, 1.30) 1.00 (0.69, 1.46) 1.000 0.80 (0.54, 1.16) 0.240 Lives alone 1.54 (0.93, 2.54) -  1.62 (1.05, 2.50) 0.030 Very much likes to walkb 3.20 (2.11, 4.86) -  2.10 (1.4, 3.09) <0.001 Vehicle available 0.57 (0.39, 0.85) -  0.56 (0.40, 0.77) <0.001 Comorbiditiesc 0.84 (0.77, 0.93) -  0.86 (0.79, 0.94) 0.001 aOnly present variables that are significant at p < 0.05 b5 (very much likes to walk) vs. 1-4 (not at all liking to walk to somewhat liking to walk) on a 5-point scale. cTotal number; measured with the Functional Comorbidity Index      1635.5 Discussion I extend the literature on older adults’ travel behaviour by describing the types of destinations that older adults most commonly travel to in a week and by investigating the association between the prevalence of neighbourhood destinations (as measured by Street Smart Walk Score) and the frequency of older adults’ walking for transportation. My approach has a novel component in that I captured specific destinations older adults travel to and collected these data in a population of older adults living on low income. Destinations most relevant (common) to older adults living on low income were: grocery stores, malls, and restaurants/cafés. Importantly, prevalence of neighbourhood destinations was positively associated with the number of walking for transportation trips taken by this population.  Although low SES may be a risk factor for functional limitations in older adults (Koster, Bosma, van Groenou, et al., 2006; Koster et al., 2005), I found that participants reported few functional limitations, as measured by self-report and objective measures of mobility and health. Most participants (~80%) did not require a mobility aid when walking outside and participants’ gait speed was 1.0 m/s (median); this is greater than the 0.8 m/s gait speed considered necessary for community ambulation (Fritz & Lusardi, 2009) but is less than the 1.2 m/s gait speed required to cross a street (Asher et al., 2012; Montufar et al., 2007). Living on low income may have affected participants’ travel behaviour. Half did not have access to a car and perhaps related, participants reported a relatively high proportion of walking trips (38%). Regionally, only ~8% of trips are by foot for older adults (TransLink, 2010) – almost five-fold less than for participants in my study. Only 41% of participants’ trips in my study were taken by car, a stark contrast with studies of older adults from the United States where almost 90% of trips were by car    164(Boschmann & Brady, 2013; Lynott, McAuley, & McCutcheon, 2009). My results concur with other studies that reported how low SES might increase the reliance of older adults on walking as a travel mode (Cao et al., 2010; Kemperman & Timmermans, 2009; Turcotte, 2012).   I believe this is the first study to measure specific destinations to which older adults travel to prospectively, using seven-day travel diaries. Two studies reported older adults’ frequency of travel to destinations retrospectively, over longer time periods (W. C. King et al., 2003; G. C. Smith & Sylvestre, 2001). G. C. Smith and Sylvestre (2001) asked older adult participants to recall retrospectively the frequency with which they made trips to a pre-determined list of eight destinations in the past year (banks; grocery stores; friends’/relatives’ homes; pharmacies; recreation centres; place of worship; volunteering/work; senior centres). More than 90% of participants reported visiting a grocery store > once/week; frequency of travel to other destinations varied based on participants’ person-level characteristics (gender, comorbidities, living arrangements, income). W. C. King et al. (2003) asked about walking trips only, asking participants to recall (again retrospective) the frequency of walking trips made in the last month to 11 destinations. More than 20% of participants reported walking to a convenience/deli/grocery store and to parks. The third most commonly visited destination (walked to by 18% of participants) was a restaurant/pub/bar. W. C. King et al. (2003) only provide information on destinations visited by foot, but it is probable that if trips made by modes other than walking were recorded by participants, the frequency of visits to each destination would change.   Living in neighbourhoods with a greater prevalence of destinations was associated with making more walking for transportation trips. This suggests that given the opportunity to travel to    165destinations of interest nearby, older adults living on low income may be willing to walk instead of drive to reach them. Two recent studies found that the presence of destinations, measured in terms of destination diversity (Cerin et al., 2013; Rosso et al., 2013) and prevalence (Cerin et al., 2013), was associated with older adults’ within-neighbourhood mobility. Neither these analyses, nor my study, used destination-based metrics of walkability tailored to older adults. Destinations identified using travel diaries provide preliminary evidence of types of destinations that should be included in future studies that assess older adults’ walking for transportation.  The Press-Competence model posits that features of the environment (e.g., presence of destinations) and person-level factors (e.g., physical function, health, self-efficacy) interact to influence behaviour (Lawton & Nahemow, 1973). Similarly, the Theory of Planned Behaviour states that individuals’ attitudes, norms, and perceived behavioural control influence whether or not they intend to walk and thereby influence whether or not they do walk (Ajzen, 1991). In my study, person-level factors associated with making more walking for transportation trips were: lack of vehicle access, a lower number of comorbidities, and enjoyment of walking. This highlights the importance of relevant neighbourhood destinations for older adults without access to a car (such as those with low income). Municipal planners might also consider the specific needs of older adults with health disparities and older adults with less favourable attitudes towards walking in their pedestrian planning model.  My study has several strengths. First, I measured participants’ travel behaviour prospectively with travel diaries. Most studies that investigated the association between destinations and older adults’ mobility relied on participants’ retrospective self-report of past travel behaviour. These    166methodologies may be susceptible to recall bias. To the best of my knowledge, no studies have asked older adult participants to record their habitual travel behaviour over the span of a week. Second, adherence was high; 93% (150/161) of participants filled out travel diaries for a median of seven days. I used a classification system (NAICS) common to transportation research in order to systematically classify destinations. Third, I recruited participants who lived across a range of built environment settings so as to capture the differences between them.   I also acknowledge that my study had several limitations. First, we experienced a low recruitment rate. Eight percent of the older adults living on low income invited to participate in my study agreed to do so. Other surveys of older adults and the built environment reported higher recruitment rates, in the order of ~20-25% [e.g., (Davis et al., 2011; A. C. King et al., 2011; Rosso et al., 2013; Winters, Voss, et al., 2015)]. My focus was a lower SES population that is traditionally more difficult to recruit into research studies (Martinson et al., 2010; Schnirer & Stack-Cutler, 2011). Reasons suggested for this greater challenge are: access barriers (e.g., lack of awareness, associated “out of pocket costs”), participation concerns (e.g., privacy, trust), and demographic barriers (life stresses, illiteracy) (Schnirer & Stack-Cutler, 2011). The gender distribution in my sample is comparable to individuals in the sampling frame, although the age of participants was somewhat younger (median age 74 versus 77 years old). Second, my study includes restrictions presented by NAICS coding. For example, I was unable to determine the specific destinations that participants travelled to within the mall. This may have resulted in misclassification if trips were made to visit a specific destination within the mall. I also cannot state whether trips to health and personal care stores were for medical-related purpose, as a wide variety of home and food items are now typically available at these destinations. Finally, in    167reality, a trip may be made in order to both reach a destination, as well as for exercise. In my dataset, these trips are categorized as walking for transportation.   5.6 Conclusions Given the shift towards an ageing population, it is somewhat surprising how little we know about the influence of neighbourhoods on older adults’ mobility and independence. Municipal planning, transportation and parks and recreation sectors represent key partners in developing and implementing thoughtful evidence-based neighbourhood design. My findings provide preliminary evidence that identifies destinations that may be most relevant to older adults and suggest that the presence of neighbourhood destinations may encourage walking. As we approach a new era of healthy city benchmarks, my findings might guide policy makers and developers to retrofit and develop communities that support the mobility, health, and independence of older adults. Specific suggestions include the avoidance of food deserts and zoning to include more destinations that are relevant to older adults (especially grocery stores) in neighbourhoods with a high proportion of older adult residents (e.g., assisted living sites and retirement villages). Finally, I envision that my findings might also encourage researchers to conduct longer term prospective and intervention studies to evaluate the effect of changes to the built environment on older adults’ mobility.     168Chapter  6: Can older adults’ neighbourhood walkability promote a walk to the shops  6.1 Introduction Despite the many benefits of a lifestyle that includes regular activity, many older adults fall far below the recommended guidelines of > 150 minutes of moderate-to-vigorous physical activity (MVPA) per week (Chodzko-Zajko, Proctor, Singh, et al., 2009; Tremblay et al., 2011). Of note, adults aged > 60 years represent the least active age group; only 2.4% of older adults in the United States and 4.5% of older adults in Canada attain sufficient physical activity to meet public health recommendations (Colley et al., 2011; Troiano et al., 2008). Barriers to engaging in physical activity in older adults include poor health, unsupportive physical environments (e.g., no sidewalks, parks or recreation centres), lack of knowledge about the relationship between physical activity and health and negative experiences with exercise earlier in life (Schutzer & Graves, 2004). Meeting recommended physical activity guidelines is important for older adults’ health and wellbeing. However, physical activity at levels below guidelines is often overlooked as important and may also be key to the general health and community engagement of older people. For example, total physical activity volume may have stronger associations with cardiometabolic biomarkers than MVPA accumulated in bouts (Wolff-Hughes et al., 2015). Further, light-intensity physical activity was associated with older adults’ physical health and well-being, independent of MVPA (Buman et al., 2010). It is also plausible that individuals who are more physically active outdoors have more opportunities for community engagement. Thus, although current physical activity guidelines are important for health, encouraging any physical    169activity, including light physical activity, is being increasingly recognized as important (I. Lee, 2015; Sparling et al., 2015). Moreover, given the broad range of mobility limitations for some older adults, this may be more a more realistic public health goal for older adults.  Walking, and in particular, outdoor walking, is the most common form of physical activity for older adults (Dai et al., 2015). Walking requires minimal equipment, and intensity is dictated by the individual. Further, within a supportive outdoor environment, walking can be incorporated relatively easily into daily life routines as either structured or incidental activity. The built environment, defined as urban design, land use, and transportation systems (S. L. Handy et al., 2002), can be an important facilitator or a barrier to outdoor walking. For example, a built environment rich with destinations relevant to older adults provides an opportunity to walk for daily travel (Cao et al., 2010; Cerin et al., 2013; Chudyk et al., 2015; D. King, 2008). Yet findings are mixed regarding specific built environment features associated with older adult walking, and physical activity in general (Hanson et al., 2012; Rosso et al., 2011; Van Cauwenberg et al., 2011; Yen et al., 2014). Built environment features most consistently associated with older adult walking and physical activity include indices of neighbourhood walkability, street connectivity, access to destinations (e.g., shops, restaurants) and features related to perceived safety [e.g., good lighting, absence of crime, presence of crosswalks (Rosso et al., 2011; Yen et al., 2014)]. The extent to which an individual successfully navigates his/her environment is a result of the match between the pressures exerted by the environment (e.g., features of the built environment) and the competence (e.g., capacity) of the individual (Lawton, 1989). Person-level factors that contribute to older adults’ capacity to be active in their neighbourhood include cognitive, physical, psychosocial, and financial domains; the social    170environment also plays an important role (McNeill et al., 2006; Webber et al., 2010). Importantly, older adults span diverse physical, psychosocial and cognitive abilities and person-level factors play a key role in the person-environment interaction. Thus, it is relevant to focus on distinct subgroups within the older adult population ([or example those with mobility limitations or of low socioeconomic status (SES)] to better understand the association between the built environment and older adult physical activity.   Older adults of low SES are understudied in physical activity and aging research (S. L. Hughes et al., 2011). The built environment may more strongly influence the physical activity habits of this population specifically, as they have less disposable income and as a result may rely more upon unstructured (and free) physical activities, such as outdoor walking. Further, older adults of low SES are more likely to walk or take public transit, instead of drive, as their main form of transportation (Cao et al., 2010; Turcotte, 2012). In doing so they may accrue incidental physical activity, as well as engage with the built environment and potentially others. Conversely, individu