Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Natural history of balance confidence : its significance and relationship with social participation in… Yiu, Jeanne Yeung Chun 2010

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

Item Metadata

Download

Media
24-ubc_2010_spring_yiu_jeanne.pdf [ 2.7MB ]
Metadata
JSON: 24-1.0069229.json
JSON-LD: 24-1.0069229-ld.json
RDF/XML (Pretty): 24-1.0069229-rdf.xml
RDF/JSON: 24-1.0069229-rdf.json
Turtle: 24-1.0069229-turtle.txt
N-Triples: 24-1.0069229-rdf-ntriples.txt
Original Record: 24-1.0069229-source.json
Full Text
24-1.0069229-fulltext.txt
Citation
24-1.0069229.ris

Full Text

NATURAL HISTORY OF BALANCE CONFIDENCE: ITS SIGNIFICANCE AND RELATIONSHIP WITH SOCIAL PARTICIPATION IN INDIVIDUALS WITH STROKE  by  Jeanne Yeung Chun Yiu B.Sc., The University of Alberta, 1986  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE  in  THE FACULTY OF GRADUATE STUDIES  (Rehabilitation Sciences)  THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver)  February 2010  © Jeanne Yeung Chun Yiu, 2010  Abstract Background Balance confidence may be an important factor affecting the recovery and rehabilitation of individuals with stroke. Little is known about how balance confidence changes over time and whether a relationship exists between balance confidence and critical outcomes such as social participation in individuals with stroke. No study has investigated the influence of balance confidence on social participation while controlling for important covariates. If balance confidence is an important predictor of social participation, treating reduced balance confidence may enhance an individual’s social participation. Purpose 1) To compare how balance confidence changed over 1 year in individuals with stroke and controls 2) To determine if stroke status was an important predictor of balance confidence and explore stroke specific factors affecting balance confidence 3) To compare how social participation changed over 1 year in individuals with stroke and controls 4) To determine if balance confidence was an important predictor of social participation Methods In this longitudinal study, 98 individuals with stroke and 98 age- and sex-matched controls were evaluated at baseline (discharge from in-patient rehabilitation for stroke subjects) 3, 6 and 12 months post baseline. Subjects were recruited from 5 communities in British Columbia. Multilevel modeling and multiple regression analyses were used to answer our research questions.  ii  Results Balance confidence scores improved slightly over 1 year in individuals with stroke however while the change was statistically important it was not considered clinically meaningful. Balance confidence remained significantly lower in these individuals compared to controls (p<0.001). Stroke status was the most important predictor of balance confidence even after controlling for covariates (p<0.001) and stoke status interacted with symptoms of depression and walking capacity when predicting balance confidence. The level of social participation did not change over 1 year in either group but it was statistically lower in individuals with stroke (p<0.001). Balance confidence interacted with balance performance (p=0.023) in predicting social participation. Conclusions Reduced balance confidence is a persistent and serious problem in individuals with stroke. Having better balance confidence may enhance an individual’s social participation. Clinicians working in stroke rehabilitation should incorporate assessment and treatment for reduced balance confidence into their rehabilitation regime.  iii  Table of Contents Abstract ..................................................................................................................................... ii Table of Contents ..................................................................................................................... iv List of Tables ......................................................................................................................... viii List of Figures .......................................................................................................................... ix List of Abbreviations ................................................................................................................ x Acknowledgements .................................................................................................................. xi Co-Authorship Statement........................................................................................................ xii Chapter One: Introduction and Literature Review.................................................................... 1 1.1 Epidemiology of Stroke .................................................................................................. 1 1.2 Impact of Stroke on Social Participation......................................................................... 1 1.3 Balance Confidence and Stroke ...................................................................................... 2 1.4 Covariates of Balance Confidence .................................................................................. 3 1.5 Balance Confidence and Social Participation ................................................................. 4 1.6 Covariates of Social Participation ................................................................................... 5 1.7 Reduced Balance Confidence is Remediable .................................................................. 5 1.8 Theoretical Background .................................................................................................. 6 1.8.1 Perceived Self-efficacy ............................................................................................. 6 1.8.2 Perceived Self-efficacy and Fear and Avoidance Behavior ..................................... 8 1.8.3 Postulated Relationship between Balance Confidence and Social Participation Based on Self-efficacy Concepts ....................................................................................... 8 1.9 Rationale of the Study ..................................................................................................... 9 1.10 Purpose of the Study ..................................................................................................... 9 1.11 Research Questions and Hypotheses ........................................................................... 10 1.11.1 Research Questions and Hypotheses for Chapter 2 .............................................. 10 1.11.2 Research Questions and Hypotheses for Chapter 3 .............................................. 10 Reference List ......................................................................................................................... 13 Chapter Two: Longitudinal Analysis of Balance Confidence in Stroke Survivors Using a Multilevel Model for Change.................................................................................................. 20 2.1 Introduction ................................................................................................................. 20 2.2 Methods ....................................................................................................................... 23 iv  2.2.1 Participants ............................................................................................................. 24 2.2.2 Outcome Measures ................................................................................................. 25 2.2.3 Statistical Analysis ................................................................................................. 28 2.3 Results ........................................................................................................................... 34 2.3.1 Variable Descriptions ............................................................................................ 34 2.3.2 Sample Sizes and Missing Data............................................................................. 35 2.3.3 Results of Outlier Analyses ................................................................................... 37 2.3.4 Multicollinearity .................................................................................................... 38 2.3.5 Data Exploration .................................................................................................... 38 2.3.6 Preparation of Variables ........................................................................................ 39 2.3.7 Results of Multilevel Analyses .............................................................................. 39 2.4 Discussion ..................................................................................................................... 42 2.4.1 Limitations of the Study ......................................................................................... 47 2.5 Conclusions ................................................................................................................... 48 Reference List ......................................................................................................................... 63 Chapter Three: Relationship between Balance Confidence and Social Participation in Individuals with First Stroke: A One Year Follow Up Study................................................. 72 3.1 Introduction ................................................................................................................... 72 3.2 Methods ......................................................................................................................... 77 3.2.1 Participants ............................................................................................................. 77 3.2.2 Procedure ................................................................................................................ 77 3.2.3 Outcome Measures ................................................................................................. 78 3.2.4 Statistical Analysis ................................................................................................. 82 3.3 Results ........................................................................................................................... 86 3.3.1 Description of Sample Size, Variables and Group Characteristics ........................ 86 3.3.2 Results of Multilevel and Multiple Regression Analyses ...................................... 87 3.4 Discussion ..................................................................................................................... 90 Reference List ....................................................................................................................... 104 Chapter Four: Discussion, Conclusions and Future Direction.............................................. 113 4.1 Overview ..................................................................................................................... 113  v  4.2 Reduced Balance Confidence Is a Persistent and Serious Problem in Individuals with Stroke ................................................................................................................................ 113 4.3 Having Better Balance Confidence May Enhance an Individual’s Social Participation ........................................................................................................................................... 115 4.4 Strengths and Limitations ........................................................................................... 116 4.5 Significance of the Study and Recommendations for Practice ................................... 117 Reference List ....................................................................................................................... 119 APPENDICES ...................................................................................................................... 121 Appendix I: Ethics Certificate ........................................................................................... 121 Appendix II: Psychometric Information of Outcome Measures Used in the Study.......... 122 Appendix III: The Berg Balance Scale ............................................................................. 124 Appendix IV: The Timed “Up and Go” Test .................................................................... 128 Appendix V: The Six Minute Walk Test .......................................................................... 129 Appendix VI: The Activities-Specific Balance Confidence Scale (ABC Scale) .............. 130 Appendix VII: The Center for Epidemiologic Studies Depression Scale ......................... 131 Appendix VIII: The State Anxiety Inventory for Adults .................................................. 132 Appendix IX: Cognitive Capacity Screening Examination .............................................. 133 Appendix X: The 6-item Interpersonal Support Evaluation List (ISEL) .......................... 134 Appendix XI: Chronic Condition Questionnaire .............................................................. 135 Appendix XII: A Person-Period Data Set ......................................................................... 137 Appendix XIII: Centering Variables ................................................................................. 138 Appendix XIV: Creating an Unconditional Growth Model .............................................. 139 Appendix XV: Pseudo R Squares Calculations ................................................................ 140 Appendix XVI: Histograms Showing Frequency Distributions of Variables ................... 141 Appendix XVII: Reasons for Missing Data ..................................................................... 143 Appendix XVIII: Differences between Dropouts and Participants ................................... 144 Appendix XIX: Selected SPSS Output for Regression Analysis with All Predictors on Collinearity Diagnostics (Tolerance, VIF and Condition Index) ...................................... 146 Appendix XX: Scatter Plots of Individual Changes in Balance Confidence over Time .. 148 Appendix XXI: Written Models, SPSS Syntax and Selected Output for (a): Unconditional Mean Model, (b) Unconditional Growth Model and (c) Model A.................................... 152 Appendix XXII: Written Model, Syntax and Selected Output for Model B..................... 158 vi  Appendix XXIII: Written Model, Syntax and Selected Output for Model C ................... 161 Appendix XXIV: The Frenchay Activities Index (FAI) ................................................... 165 Appendix XXV: Selected SPSS Output for (a) the Unconditional Means Model (UMM), (b) the Unconditional Growth Model (UGM) and (c) the Final Model with Total FAI as the Dependent Variable ........................................................................................................... 166 Appendix XXVI: Selected SPSS Output for (a) the Unconditional Means Model (UMM), (b) Unconditional Growth Model (UGM) and (c) the Final Model with Domestic Subdomain of FAI as the Dependent Variable ........................................................................ 169 Appendix XXVII: Selected SPSS Output for (a) the Unconditional Means Model (UMM), (b) Unconditional Growth Model (UGM) and (c) the Final Model with Leisure/Work Subdomain of FAI as the Dependent Variable ........................................................................ 172 Appendix XXVIII: Selected SPSS Output for (a) the Unconditional Means Model (UMM), (b) Unconditional Growth Model (UGM) and (c) the Final Model with Outdoors Subdomain of FAI as the Dependent Variable ........................................................................ 175 Appendix XXIX: Test of Assumptions of Normality, Linearity and Homoscedasticity of Residuals ........................................................................................................................... 178 Appendix XXX: Outliers Analyses (Regression Analysis of FAI on ABC and Covariates) ........................................................................................................................................... 179 Appendix XXXI: Collinearity Diagnostics: Regression Analysis of FAI on ABC and Covariates .......................................................................................................................... 185  vii  List of Tables Table 2-1 Table 2-2 Table 2-3 Table 2-4 Table 2-5 Table 2-6 Table 3-1 Table 3-2  Table 3-3  Table 3-4  Table 3-5  Table 3-6 Table 3-7 Table 4 Table 5 Table 6 Table 7 Table 8 Table 9 Table 10 Table 11  Demographic and Stroke Characteristics of Participants................................49 Participants’ Profile at Baseline in the Stroke and Control Groups with Normative Data...............................................................................................50 Results of Fitting the Unconditional Means Model (UMN), the Unconditional Growth Model (UGM) and Model A to Data..................................................51 Results of Fitting Model B to Data.................................................................52 Results of Fitting Model C to Data.................................................................54 Percentage of Variance Explained by Each Variable Computed from the Unconditional Growth Model, Model A and Model B...................................56 Mean Scores of Total FAI and Sub-domains of FAI from 3 to 12 Months after Baseline in the Stroke and Control Groups.....................................................96 Results of Fitting the Unconditional Means Model (UMM), the Unconditional Growth Model (UGM) and the Final Model to Data with Total FAI Score as the Dependent Variable...................................................................................97 Results of Fitting the Unconditional Means Model (UMM), the Unconditional Growth Model (UGM) and the Final Model to Data with Domestic Subdomain FAI Score as the Dependent Variable................................................98 Results of Fitting the Unconditional Means Model (UMM), the Unconditional Growth Model (UGM) and the Final Model to Data with Leisure/Work Subdomain FAI Score as the Dependent Variable................................................99 Results of Fitting the Unconditional Means Model (UMM), the Unconditional Growth Model (UGM) and the Final Model to Data with Outdoors Subdomain FAI Score as the Dependent Variable..............................................100 Correlation Matrix of Continuous Predictor Variables and Outcome Variable.........................................................................................................101 Multiple Regression of Social Participation (FAI) at 12 Months on Balance Confidence.....................................................................................................102 Method of Centering Variables in the Study.................................................138 Mean Difference of Variables at Baseline between Dropouts and Participants in the Stroke Group.......................................................................................144 Mean Difference of Variables at Baseline Between Dropouts and Participants in the Control Group......................................................................................145 Extreme Values of Baseline 6MWT, BBS, TUG, CES-D, STAI, ABC, ISEL and FAI..........................................................................................................182 Coefficients, Tolerance and VIF Values of the Unadjusted Model..............185 Condition Index and Variance Proportions of the Unadjusted Model..........185 Coefficients, Tolerance and VIF Values of the Adjusted Model..................186 Condition Index and Variance Proportions of the Adjusted model..............187  viii  List of Figures Figure 1-1 Figure 2-1 Figure 2-2 Figure 2-3 Figure 2-4 Figure 2-5 Figure 2-6 Figure 3-1 Figure 4 Figure 5 Figure 6 Figure 7 Figure 8 Figure 9 Figure 10 Figure 11 Figure 12 Figure 13 Figure 14  Postulated Relationships between Balance Confidence, Component Skills and Contextual Factors Based on Social Cognitive Theory..........................12 Diagram Illustrating Final Level-2 Sample Size...........................................57 Diagram Illustrating Final Level-1 Sample Size...........................................58 Prototypical Trajectories from the Unconditional Growth Model and Model A Displaying the Effect of Stroke on Balance Confidence (ABC)...............59 Prototypical Trajectories from Model B Displaying the Controlled Effect of Stroke on Balance Confidence.......................................................................60 Prototypical Trajectories from Model C of 4 Individuals Showing the Interaction between Stroke Status and Depression Symptoms......................61 Prototypical Trajectories Derived from Model C of 4 Individuals Showing the Interaction between Stroke Status and 6MW Distance...........................62 Scatter Plot Illustrating the Interaction Effect of Balance Performance on the Association between Balance Confidence and Social Participation............103 Schematic Diagram Showing the Arrangement/Set up for the Timed “Up and Go” Test.......................................................................................................128 Part of a “Person-Period” Data Set for Records of 4 Variables (Age, Sex, ABC and BBS) of 4 Selected Participants..................................................137 Sample Output Table for Creating an Unconditional Growth Model.........139 Frequency Distributions of Time Varying Predictors- TUG, Transformed TUG, CES-D, 6MW, STAI and Transformed STAI...................................141 Frequency Distributions of Time-invariant Predictors- BBS, Transformed BBS, ABC, Transformed ABC and ISEL...................................................142 Scatter Plots of Individual Changes in ABC over Time in the First 49 Stroke Subjects........................................................................................................148 Scatter Plots of Individual Changes in ABC over Time in the Second 49 Stroke Subjects............................................................................................149 Scatter Plots of Individual Changes in ABC over Time in the First 49 Controls........................................................................................................150 Scatter Plots of Individual Changes in ABC over Time in the Second 49 Controls........................................................................................................151 Plot of Predicted Values of the FAI Against Residuals...............................178 Histograms Showing Distribution of Predictors (a) 6MW, (b) BBS, (c) TUG & Transformed TUG, (d) CES-D, (e) STAI, (f) ABC, (g) ISEL and (h) FAI...............................................................................................................180  ix  List of Abbreviations 6MWT  Six-minute Walk Test  ABC  Activities Specific Balance Confidence Scale  ADL  Activities of Daily Living  BBS  Berg Balance Scale  CCSE  Cognitive Capacity Screening Examination  CES-D  Center for Epidemiological Studies Depression Scale  FAI  Frenchay Activity Index  FES  Falls Efficacy Scale  FIM  Functional Independence Measure  ICC  Intraclass Correlation Coefficient  ICF  International Classification of Functioning, Disability and Health  ISEL  6-item Interpersonal Support Evaluation List  LHS  London Handicap Scale  Mahal  Mahalanobis Distance  MSE  Mean Square Error  REML  Restricted Maximum Likelihood  SD  Standard Deviation  SRM  Standardized Response Mean  STAI  State-Trait Anxiety Inventory for Adults  TUG  Timed Up and Go  UGM  Unconditional Growth Model  UMM  Unconditional Means Model  VIF  Variance Inflation Factor  WHO  World Health Organization  x  Acknowledgements I would like to thank my supervisor, Dr. Bill Miller, for allowing me to take part in this project and for his continued support, guidance and encouragement through the course of my Master’s study. He has been a great teacher and mentor and has enabled me to learn a great deal. I would also like to thank my thesis committee, Dr. Janice Eng and Dr. Tal Jarus who have been supportive. I really appreciate the feedback and the effort that they put in assisting me to complete this thesis. In addition, I would like to thank Dr. Bruno Zumbo and Yan Liu for sharing their knowledge and assistance with some of the data analyses. I would also like to thank the research coordinators of the G. F. Strong Lab for their assistance with various activities that were important to my project progress; in particular, I would like to thank Elmira Chan for her prompt responses to my requests for help. I would like to thank all the study participants who were willing to offer their time and be part of this project. In addition, I would like to thank all my colleagues at Vancouver General Hospital and G. F. Strong Rehabilitation Centre for their kind words of encouragement and for sharing and listening; in particular, I would like to express gratitude to Deirdre Lee, Alison McLean and Jade Lew. Most of all, I would like to thank my family and would like to dedicate this thesis to them. The tremendous support and encouragement I receive from my husband Douglas have made my study possible. I would also like to thank my children Sam and Gabriel for their understanding and patience. I have not heard any complaints but rather reminders like “Go do your thesis!”, “How much have you done today?” I would also like to thank my mother and Douglas’ mother for their continued supports throughout the years. Finally I am grateful to the Canadian Occupational Therapy Foundation, the VGH and UBCH Foundation and the Disability and Health Research Network for their financial support.  xi  Co-Authorship Statement Specific chapters of this thesis are in preparation for publication in peer-review journals. All of these manuscripts have multiple authors. The details of authorship contributions are listed below. Chapter 2 and 3: Co-authors Jeanne Yiu, Dr. William C. Miller, Dr. Janice Eng and Dr. Tal Jarus: Dr. Miller was responsible for conceptualizing the study, obtaining grant funding, executing the study and reviewing and editing the manuscript that will arise from this chapter. Dr. Eng contributed to the study design, consulted in the conduction of the study and reviewed and edited the manuscript that will arise from this chapter. Dr. Jarus provided feedback on the manuscript. Jeanne Yiu collected some of the follow up data, conducted all the analyses, and is the primary author for the manuscript.  xii  Chapter One: Introduction and Literature Review 1.1 Epidemiology of Stroke Stroke is one of the leading causes of death and disability in Canada (Canadian Institute for Health Information, 2007). Each year more than 50,000 individuals suffer a stroke (Heart and Stroke Foundation, 2010). In 2004/05, there were 46570 individuals with stroke admitted to acute care hospitals (Canadian Institute for Health Information, 2007). Today over 300,000 Canadians are living with the effects of stroke (Heart and Stroke Foundation, 2010). In 2006/07, 25% of individuals with stroke who attended an inpatient rehabilitation program were not able to return to their previous living environment upon discharge, and instead required a more supportive environment such as assisted living or long term care facility (Canadian Institute for Health Information, 2008); over 40% of individuals with stroke who were discharged from inpatient rehabilitation required paid health services in the home after discharge (Canadian Institute for Health Information, 2008). 1.2 Impact of Stroke on Social Participation According to the World Health Organization (WHO) (2003), participation is involvement in a life situation. Participation restrictions are difficulties an individual may experience in involvement in life situations. Stroke results not only in impairment of body functions and limitation in daily activities, it also restricts participation in social roles and activities meaningful to the individual even years after stroke (Gresham et al., 1979; Patel et al., 2006; Pound, Gompertz, & Ebrahim, 1998). Studies have shown that individuals with stroke who are more dependent in mobility and physical activities experience a greater restriction in participation (Clarke, Black, Badley, Lawrence, & Williams, 1999; D'Alisa, Baudo, Mauro, & Miscio, 2005). Physical impairments related to mobility, such as decreased gait speed and poor balance, are highly prevalent at 3 months post stroke (Mayo et al., 1999). 1  While physical impairments may contribute to activity limitations in stroke , other factors such as reduced walking capacity has been found to have a more long-term effect on participation and quality of life in community dwelling individuals with stroke (Mayo et al., 1999; Muren, Hutler, & Hooper, 2008). Moreover, individuals with stroke are at risk for experiencing reduced tolerance for activity (Mol & Baker, 1991). The limitation in activity tolerance and walking capacity may in the long term promote a sedentary lifestyle. In addition, physical impairments may undermine an individual’s confidence to perform activity thus contributing to the cycle of inactivity, sedentary lifestyle, decreased quality of life, and decreased life satisfaction. 1.3 Balance Confidence and Stroke Balance confidence could prove to be a serious problem in individuals with stroke. It is defined as the belief in one’s ability to maintain balance while performing selected activities (Powell & Myers, 1995). The beliefs in one’s capabilities to organize and execute the skills required to produce given attainments is referred to as perceived self-efficacy (Bandura, 1997). Therefore balance confidence can be viewed as self-efficacious beliefs in one’s capabilities to execute balance skills required to participate in activities that demand them. For the purpose of this thesis balance confidence and balance self-efficacy were used interchangeably. Individuals who have had a stroke may be susceptible to developing reduced balance confidence (Belgen, Beninato, Sullivan, & Narielwalla, 2006; Salbach et al., 2006; Schmid & Rittman, 2007). Balance confidence was found to be low in a sample of stroke subjects discharged from rehabilitation, as measured by the Activities-specific Balance Confidence (ABC) scale (summary score ranges from 0 to 100; higher score indicates higher balance confidence) (mean=59, 95%CI, 55-64) (Salbach et al., 2006). When compared to healthy 2  older adults of similar age range, other studies found that the mean ABC scores were 69.8±10.2 (Brouwer, Musselman, & Culham, 2004) and 78.87±19.08 (Hatch, Gill-Body, & Portney, 2003). ABC scores lower than 50 are indicative of low functioning older adults; ABC scores between 50 and 80 are indicative of a moderate level of functioning whereas ABC scores above 80 are indicative of high level of functioning, usually physically active older adults (Myers, Fletcher, Myers, & Sherk, 1998). To our knowledge, there has not been a study on balance confidence in stroke that has incorporated a non-disabled control group for comparison. Reduced balance confidence among individuals with stroke could prove to be a major psychological barrier that may lead to decreased participation in daily and social activities which, in turn, may result in decreased quality of life and life satisfaction. 1.4 Covariates of Balance Confidence Risk factors associated with reduced balance confidence have been identified among older adults include impaired balance (Hatch et al., 2003), decreased ability to weight shift (Binda, Culham, & Brouwer, 2003), lower gait speed (Brouwer et al., 2004; Kressig et al., 2001), history of falls (Friedman, Munoz, West, Rubin, & Fried, 2002; Scheffer, Schuurmans, van Dijk, van der Hooft, & de Rooij, 2008), being female and being older (Scheffer et al., 2008). Some of these factors have also been identified among stroke survivors such as decreased balance ability (Hellstrom, Nilsson, & Fugl-Meyer, 2001; Hellstrom, Lindmark, Wahlberg, & Fugl-Meyer, 2003; Rosen, Sunnerhagen, & Kreuter, 2005), slower gait velocity (Rosen et al., 2005), and history of falls (Belgen et al., 2006; Pang & Eng, 2008; Schmid & Rittman, 2007; Watanabe, 2005). Lower extremity strength was found to be a determinant of balance confidence in chronic stroke (Belgen et al., 2006). Reduced balance confidence was also significantly associated with increased depressive symptoms, decreased walking capacity, increased social support and increased assistive 3  device use in individuals discharged from stroke rehabilitation (Salbach et al., 2006). Given that reduced balance confidence tends to be associated with many undesirable factors and having the confidence to main balance while performing daily activities is crucial to maintaining independent functioning, it would be important to understand how reduced balance confidence affects individuals with stroke and what factors are important in predicting balance confidence. 1.5 Balance Confidence and Social Participation Participation is defined as involvement of an individual in a life situation (World Health Organization, 2001). Social participation is the engagement in productive, social and leisure activities. There has been no study with individuals with stroke that directly investigated the relationship between balance confidence and social participation; therefore, little is known if a relationship exists between balance confidence and social participation in these individuals. In individuals with traumatic brain injury, perceived general and social self-efficacy [as measured by the Self-efficacy Scale (Sherer & Maddux, 1982)] has been found to predict social participation (Dumont, Gervais, Fougeyrollas, & Bertrand, 2005). Dumont et al (2005) found that perceived self-efficacy explained 40% of the variance in social participation. Other studies have found that balance confidence predicted physical function, self-perceived health status (Salbach et al., 2006), and satisfaction with community reintegration in community-dwelling individuals with stroke (Pang, Eng, & Miller, 2007). Although perceived health status and satisfaction with community reintegration are not equivalent to social participation, it is possible that the level of physical function, perceived health status and satisfaction with community reintegration and social participation are related. If balance confidence is indeed an important predictor of social participation, it would be important to address the problem of reduced balance confidence in the 4  rehabilitation of individuals with stroke. Balance confidence could prove to be a crucial factor in the recovery after stroke. 1.6 Covariates of Social Participation Many factors have been found to be associated with social participation. In individuals with stroke, depression (Desrosiers et al., 2008; Lo et al., 2008; Teoh, Sims, & Milgrom, 2009; Wade, Legh-Smith, & Langton Hewer, 1985), co-morbidity (Desrosiers, Noreau, Rochette, Bravo, & Boutin, 2002; Desrosiers et al., 2006), and age (Desrosiers et al., 2002; Desrosiers et al., 2006; Lo et al., 2008; Schepers, Visser-Meily, Ketelaar, & Lindeman, 2005) are negatively associated with social participation. Arm function (Desrosiers et al., 2006; Sveen, Bautz-Holter, Sodring, Wyller, & Laake, 1999), lower extremity function (Desrosiers et al., 2002), balance (Desrosiers et al., 2002), walking ability (Desrosiers et al., 2008), and ability to perform activities of daily living (ADL) (Beckley, 2006; Roth & Lovell, 2007; Schepers et al., 2005; Wade et al., 1985) are positively associated with social participation. In addition to the above, being male (Lo et al., 2008; Schepers et al., 2005; Wade et al., 1985), being married (Schepers et al., 2005), and having better financial status (Lo et al., 2008) are associated with a higher level of social participation. Moreover, Beckley (2006) found that subjective social support interacts with ADL limitation and predicts social participation. Among all these factors affecting social participation, no study has investigated whether balance confidence is also an important factor. 1.7 Reduced Balance Confidence is Remediable Interventions for reduced balance confidence have been developed for communityliving older people (Zijlstra et al., 2007); however, few interventions have targeted individuals with stroke. Community-based tai chi (Li et al., 2005), home-based exercise and walking program (Campbell et al., 1997), and home-based fall-related multi-factorial 5  interventions (Yates & Dunnagan, 2001) have been shown to improve balance confidence in community-living older people. One study (Salbach et al., 2005) has examined the effect of a task-oriented walking intervention explicitly aimed to improve balance confidence targeted individuals with stroke. The authors found that individuals with low balance confidence benefited the most from the task-oriented walking intervention, and that change in balance confidence was associated with change in functional walking capacity. Another study (Marigold et al., 2005) compared two types of exercise programs (agility vs stretching/weight-shifting program) aimed to improve balance, postural reflexes, and mobility in chronic stroke, and found that both interventions improved balance confidence. Moreover, home-based training aimed to improve gait was found to improve balance confidence in a small number of subjects with chronic stroke (Rodriquez et al., 1996). Since reduced balance confidence is remediable, it would be important to investigate the impact of this problem on individuals with stroke so that appropriate interventions can be provided to those who have reduced balance confidence. 1.8 Theoretical Background 1.8.1 Perceived Self-efficacy Balance confidence is conceptualized based on Bandura’s Social Cognitive theory. According to Bandura (1983), perceived self-efficacy refers to people’s beliefs or judgments of how well they can organize and execute skills they possess in dealing with prospective situations (Bandura, 1983). Efficacy expectation is “the conviction that one can successfully execute the behavior required to produce the outcomes” (Bandura, 1977, p. 193). Perceived self-efficacy influences people’s choice of activities and environmental settings, how much effort they expend, and how long they will persist in the face of obstacles and stressful situations (Bandura & Adams, 1977; Bandura, 1982). Perceived self-efficacy alone will not 6  produce the desired performance if the component capabilities are lacking or if there are no incentives to perform; however, the stronger the perceived self-efficacy, the more active, vigorous, and persistent the coping efforts (Bandura, 1977). Given appropriate skills and adequate incentives, perceived self-efficacy is a major determinant of people’s choice behavior (Bandura, 1977). Efficacy information comes from four major sources: performance accomplishments, vicarious experience, verbal persuasion, and physiological states (Bandura, 1977). Performance accomplishments provide the most influential efficacy information because it is based on personal experience of mastery (Bandura & Adams, 1977; Bandura, 1977). Vicarious experience provides less dependable efficacy information but many efficacy expectations are derived from seeing others perform challenging activities or deal with stressful situations successfully (Bandura, 1977). Verbal persuasion is leading people to believe that they can cope successfully through suggestions or by simply telling them that they can do it (Bandura, 1977). Lastly, people also judge their level of anxiety or ability to cope with stress by the state of their physiological arousal (Bandura & Adams, 1977). People process, weigh and integrate different sources of information regarding their capability, and they regulate their choice behavior and effort expenditure accordingly (Bandura, 1977). The impact of efficacy information on perceived level of self-efficacy depends on how it is cognitively appraised (Bandura, 1977). Changes in perceived self-efficacy result from how efficacy information is interpreted and weighted (Bandura, 1977). For example, performance successes generally raise the level of perceived self-efficacy; however, whether this will actually happen depends also on a person’s preconception of his or her capabilities, his or her perception of task difficulty, how much effort he or she expends, the amount of  7  help he or she receives and many other factors (Bandura, 1997). Contextual factors such as social, situational, and temporal circumstances in which events occur will influence a person’s appraisal of his or her capability (Bandura, 1977). A strong sense of self-efficacy is reinforced when one accomplishes a task that one considers difficult, under varied circumstances, and when one ascribes one’s accomplishment to his or her ability instead of to effort or fortune (Bandura, 1977). 1.8.2 Perceived Self-efficacy and Fear and Avoidance Behavior People avoid activities that they believe exceed their coping ability, but they undertake and perform activities that they judge themselves capable of managing (Bandura, 1977). Self-efficacy theory postulates that it is mainly perceived inefficacy in coping with potentially stressful situations that makes people fearful (Bandura, 1983). Evidence suggests that when perceived self-efficacy becomes stronger, fear declines (Bandura & Adams, 1977; Bandura, 1977), and fear arousal stems from perceived coping inefficacy (Bandura, 1982). There is evidence that perceived self-efficacy predicts fear arousal and behaviors whereas fear arousal does not predict behavior (Leland, 1983). Based on these findings and selfefficacy postulations, it would be reasonable to infer that self-efficacy has a role to play among fear (anxiety) and coping behavior. 1.8.3 Postulated Relationship between Balance Confidence and Social Participation Based on Self-efficacy Concepts If balance confidence is associated with component skills that are required for performing a certain activity, such as balance skills, mobility, and functional capacity, and if component skills are important contributors to social participation, balance confidence may have a relationship with social participation. Since balance confidence can be influenced by contextual factors such as personal and environmental factors which may also influence 8  social participation, successful participation in activities that require balance skills will depend on the combination of skills, balance confidence and contextual factors. This postulated relationship is presented in Figure 1.1 1.9 Rationale of the Study Cross sectional evidence suggests that balance confidence is an important factor affecting individuals with first stroke. It seems plausible that balance confidence may be related to social participation but there is no evidence of this found in the literature. Little is known how balance confidence in individuals with stroke changes over time. No study has included a control group for comparison while investigating covariates of balance confidence in individuals with stroke thus stroke-specific factors influencing balance confidence are unknown. Moreover, no study has investigated the influence of balance confidence on social participation while controlling for important covariates. Understanding the natural history of balance confidence in the first year post stroke rehabilitation and identifying the strokespecific factors for reduced balance confidence will help determine the need and timing of interventions for this problem. If balance confidence is found to be an important factor of social participation, treating reduced balance confidence may contribute to enhancing social participation in individuals with stroke. 1.10 Purpose of the Study The overall purpose of this study was to investigate the longitudinal relationships 1) between stroke status and balance confidence and 2) between balance confidence and social participation while controlling for physical performance, psychological and contextual variables in individuals with stroke in the first year after discharge from inpatient rehabilitation and to compare how these relationships are different in individuals who have not had a stroke. 9  1.11 Research Questions and Hypotheses 1.11.1 Research Questions and Hypotheses for Chapter 2 1) How does balance confidence change over time for individuals who have had a stroke compared to matched controls without a stroke? It was hypothesized that balance confidence would remain significantly lower in individuals with stroke than in individuals without a stroke over the 12 month study period. 2) Does stroke status remain an important predictor of balance confidence after controlling for age, sex, perceived social support, balance performance, basic mobility, walking capacity, anxiety state and depression symptoms? It was hypothesized that stroke status would remain a significant risk factor for reduced balance confidence after controlling for age, sex, perceived social support, balance performance, basic mobility, walking capacity, anxiety state and depression symptoms. 3) Does stroke status interact with other factors when predicting balance confidence?” It was hypothesized that stroke status would interact with at least one other predictor variable when predicting balance confidence. 1.11.2 Research Questions and Hypotheses for Chapter 3 1) How does social participation change over 1 year after discharge from stroke rehabilitation and is there a difference in the level of social participation overall and the sub-domains of social participation between individuals with stroke and individuals without stroke? It was hypothesized that individuals with stroke would have a consistently lower level of social participation overall and in all the sub-domains of social participation, as measured with the FAI, than individuals without a stroke over 1 year. 2) Is balance confidence at discharge from rehabilitation (baseline) an important predictor of social participation at 1 year after discharge after controlling for important covariates: age, 10  sex, partnering status (presence/absence of a spousal partner), number of chronic conditions, perceived social support, stroke status, balance performance, basic mobility, walking capacity, anxiety trait and depression symptoms? It was hypothesized that balance confidence would remain an independent predictor of social participation after controlling for age, sex, partnering status, number of chronic conditions, perceived social support, stroke status, balance performance, basic mobility, walking capacity, anxiety trait and depression symptoms.  11  Figure 1.1 Postulated Relationships between Balance Confidence, Component Capabilities and Contextual Factors Based on Social Cognitive Theory.  Component skills: Balance ability, physical endurance, walking ability, activities of daily living  Efficacy information  Performance accomplishment: (e.g. successful social participation)  Social Participation Vicarious experience: (e.g. peer modeling)  Verbal persuasion: (e.g. therapist’s feedback)  Balance Confidence  Contextual factors: personal, environmental, social, situational, and temporal  Physiological states: (e.g. fear of falling)  12  Reference List Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Reveiw, 84, 191-215. Bandura, A. (1982). Self-efficacy mechanism in human agency. American Psychologist, 37, 122-147. Bandura, A. (1983). Self-efficacy determinants of anticipated fears and calamities. Journal of Personality and Social Psychology, 45, 464-469. Bandura, A. (1997). Self-efficacy: The exercise of control. New York: W.H. Freeman and Company. Bandura, A., & Adams, N. E. (1977). Analysis of self-efficacy theory of behavioral change. Cognitive Therapy and Research, 1, 287-310. Beckley, M. N. (2006). Community participation following cerebrovascular accident: Impact of the buffering model of social support. American Journal of Occupational Therapy, 60, 129-135. Belgen, B., Beninato, M., Sullivan, P. E., & Narielwalla, K. (2006). The association of balance capacity and falls self-efficacy with history of falling in community-dwelling people with chronic stroke. Archives of Physical Medicine & Rehabilitation, 87, 554561. Binda, S. M., Culham, E. G., & Brouwer, B. (2003). Balance, muscle strength, and fear of falling in older adults. Experimental Aging Research, 29, 205-219. Brouwer, B., Musselman, K., & Culham, E. (2004). Physical function and health status among seniors with and without a fear of falling. Gerontology, 50, 135-141.  13  Campbell, A. J., Robertson, M. C., Gardner, M. M., Norton, R. N., Tilyard, M. W., & Buchner, D. M. (1997). Randomised controlled trial of a general practice programme of home based exercise to prevent falls in elderly women. British Medical Journal, 315, 1065-1069. Canadian Institute for Health Information. (2008). Inpatient rehabilitation in Canada 20062007 Canadian Institute for Health Information. (2007). The burden of neurological diseases, disorders and injuries in Canada. Ottawa: CIHI. Clarke, P. J., Black, S. E., Badley, E. M., Lawrence, J. M., & Williams, J. I. (1999). Handicap in stroke survivors. Disability and Rehabilitation, 21, 116-123. D'Alisa, S., Baudo, S., Mauro, A., & Miscio, G. (2005). How does stroke restrict participation in long-term post-stroke survivors? Acta Neurologica Scandinavica, 112, 157-162. Desrosiers, J., Demers, L., Robichaud, L., Vincent, C., Belleville, S., Ska, B., et al. (2008). Short-term changes in and predictors of participation of older adults after stroke following acute care or rehabilitation. Neurorehabilitation & Neural Repair, 22, 288297. Desrosiers, J., Noreau, L., Rochette, A., Bravo, G., & Boutin, C. (2002). Predictors of handicap situations following post-stroke rehabilitation. Disability and Rehabilitation, 24, 774-785. Desrosiers, J., Noreau, L., Rochette, A., Bourbonnais, D., Bravo, G., & Bourget, A. (2006). Predictors of long-term participation after stroke. Disability & Rehabilitation, 28, 221230.  14  Dumont, C., Gervais, M., Fougeyrollas, P., & Bertrand, R. (2005). Perceived self-efficacy as a factor associated with the social participation of adults who have sustained a traumatic brain injury. Canadian Journal of Occupational Therapy, 72, 222-233. Friedman, S. M., Munoz, B., West, S. K., Rubin, G. S., & Fried, L. P. (2002). Falls and fear of falling: Which comes first? A longitudinal prediction model suggests strategies for primary and secondary prevention. Journal of the American Geriatrics Society, 50, 1329-1335. Gresham, G. E., Phillips, T. F., Wolf, P. A., McNamara, P. M., Kannel, W. B., & Dawber, T. R. (1979). Epidemiologic profile of long-term stroke disability: The Framingham study. Archives of Physical Medicine & Rehabilitation, 60, 487-491. Hatch, J., Gill-Body, K. M., & Portney, L. G. (2003). Determinants of balance confidence in community-dwelling elderly people. Physical Therapy, 83, 1072-1079. Heart and Stroke Foundation. (2010). Stroke statistics. Retrieved 01/06, 2010, from http://www.heartandstroke.com/site/c.ikIQLcMWJtE/b.3483991/k.34A8/Statistics.htm Hellstrom, K., Nilsson, L., & Fugl-Meyer, A. R. (2001). Relationship of confidence in task performance with balance and motor function after stroke. Physiotherapy Theory and Practice, 17, 55-65. Hellstrom, K., Lindmark, B., Wahlberg, B., & Fugl-Meyer, A. R. (2003). Self-efficacy in relation to impairments and activities of daily living disability in elderly patients with stroke: A prospective investigation. Journal of Rehabilitation Medicine, 35, 202-207. Kressig, R. W., Wolf, S. L., Sattin, R. W., O'Grady, M., Greenspan, A., Curns, A., et al. (2001). Associations of demographic, functional, and behavioral characteristics with  15  activity-related fear of falling among older adults transitioning to frailty. Journal of the American Geriatrics Society, 49, 1456-1462. Leland, E. I. (1983). Relationship of self-efficacy and other factors to precompetitive anxiety in basketball players. Unpublished Doctoral, Stanford University, Li, F., Harmer, P., Fisher, K. J., McAuley, E., Chaumeton, N., Eckstrom, E., et al. (2005). Tai chi and fall reductions in older adults: A randomized controlled trial. Journal of Gerontology: Medical Sciences, 60A, 187-194. Lo, R. S., Cheng, J. O., Wong, E. M., Tang, W. K., Wong, L. K., Woo, J., et al. (2008). Handicap and its determinants of change in stroke survivors: One-year follow-up study. Stroke, 39, 148-153. Marigold, D. S., Eng, J. J., Dawson, A. S., Inglis, J. T., Harris, J. E., & Gylfadottir, S. (2005). Exercise leads to faster postural reflexes, improved balance and mobility, and fewer falls in older persons with chronic stroke. Journal of the American Geriatrics Society, 53, 416-423. Mayo, N. E., Wood-Dauphinee, S., Ahmed, S., Gordon, C., Higgins, J., McEwen, S., et al. (1999). Disablement following stroke. Disability & Rehabilitation, 21, 258-268. Mol, V., & Baker, C. A. (1991). Activity intolerance in the geriatric stroke patient. Rehabilitation Nursing, 16, 337-343. Muren, M. A., Hutler, M., & Hooper, J. (2008). Functional capacity and health-relatd quality of life in individuals post stroke. Topics in Stroke Rehabilitation, 15, 51-58. Myers, A. M., Fletcher, P. C., Myers, A. H., & Sherk, W. (1998). Discriminative and evaluative properties of the activities-specific balance confidence (ABC) scale. Journals of Gerontology Series A-Biological Sciences & Medical Sciences, 53, M287-94.  16  Pang, M. Y., & Eng, J. J. (2008). Fall-related self-efficacy, not balance and mobility performance, is related to accidental falls in chronic stroke survivors with low bone mineral density. Osteoporosis International, 19, 919-927. Pang, M. Y., Eng, J. J., & Miller, W. C. (2007). Determinants of satisfaction with community reintegration in older adults with chronic stroke: Role of balance self-efficacy. Physical Therapy, 87, 282-291. Patel, M. D., Tilling, K., Lawrence, E., Rudd, A. G., Wolfe, C. D. A., & McKevitt, C. (2006). Relationships between long-term stroke disability, handicap and health-related quality of life. Age and Ageing, 35, 273-279. Pound, P., Gompertz, P., & Ebrahim, S. (1998). A patient-centred study of the consequences of stroke. Clinical Rehabilitation, 12, 338-347. Powell, L. E., & Myers, A. M. (1995). The activities-specific balance confidence (ABC) scale. Journals of Gerontology Series A-Biological Sciences & Medical Sciences, 50A, M28-34. Rodriquez, A. A., Black, P. O., Kile, K. A., Sherman, J., Stellberg, B., McCormick, J., et al. (1996). Gait training efficacy using a home-based practice model in chronic hemiplegia. Archives of Physical Medicine & Rehabilitation, 77, 801-805. Rosen, E., Sunnerhagen, K. S., & Kreuter, M. (2005). Fear of falling, balance, and gait velocity in patients with stroke. Physiotherapy Theory & Practice, 21, 113-120. Roth, E. J., & Lovell, L. (2007). Community skill performance and its association with the ability to perform everyday tasks by stroke survivors one year following rehabilitation discharge. Topics in Stroke Rehabilitation, 14, 48-56.  17  Salbach, N. M., Mayo, N. E., Robichaud-Ekstrand, S., Hanley, J. A., Richards, C. L., & Wood-Dauphinee, S. (2005). The effect of a task-oriented walking intervention on improving balance self-efficacy poststroke: A randomized, controlled trial. Journal of the American Geriatrics Society, 53, 576-582. Salbach, N. M., Mayo, N. E., Robichaud-Ekstrand, S., Hanley, J. A., Richards, C. L., & Wood-Dauphinee, S. (2006). Balance self-efficacy and its relevance to physical function and perceived health status after stroke. Archives of Physical Medicine & Rehabilitation, 87, 364-370. Scheffer, A. C., Schuurmans, M. J., van Dijk, N., van der Hooft, T., & de Rooij, S. E. (2008). Fear of falling: Measurement strategy, prevalence, risk factors and consequences among older persons. Age & Ageing, 37, 19-24. Schepers, V. P., Visser-Meily, A. M., Ketelaar, M., & Lindeman, E. (2005). Prediction of social activity 1 year poststroke. Archives of Physical Medicine & Rehabilitation, 86, 1472-1476. Schmid, A. A., & Rittman, M. (2007). Fear of falling: An emerging issue after stroke. Topics in Stroke Rehabilitation, 14, 46-55. Sherer, M., & Maddux, J.E. (1982). The Self-Efficacy Scale: Construction and validation. Psychological Reports, 51, 663-671. Sveen, U., Bautz-Holter, E., Sodring, K. M., Wyller, T. B., & Laake, K. (1999). Association between impairments, self-care ability and social activities 1 year after stroke. Disability & Rehabilitation, 21, 372-377.  18  Teoh, V., Sims, J., & Milgrom, J. (2009). Psychosocial predictors of quality of life in a sample of community-dwelling stroke survivors: A longitudinal study. Topics in Stroke Rehabilitation, 16, 157-166. Wade, D. T., Legh-Smith, J., & Langton Hewer, R. (1985). Social activities after stroke: Measurement and natural history using the frenchay activities index. International Rehabilitation Medicine, 7, 176-181. Watanabe, Y. (2005). Fear of falling among stroke survivors after discharge from inpatient rehabilitation. International Journal of Rehabilitation Research, 28, 149-152. World Health Organization. (2001). International Classification of Functioning, Disability and Health:ICF. Geneva: World Health Organization, 2001. Yates, S. M., & Dunnagan, T. A. (2001). Evaluating the effectiveness of a home-based fall risk reduction program for rural community-dwelling older adults. Journal of Gerontology: Medical Sciences, 56A, M226-M230. Zijlstra, G. A., van Haastregt, J. C., van Rossum, E., van Eijk, J. T., Yardley, L., & Kempen, G. I. (2007). Interventions to reduce fear of falling in community-living older people: A systematic review. Journal of the American Geriatrics Society, 55, 603-615.  19  Chapter Two: Longitudinal Analysis of Balance Confidence in Stroke Survivors Using a Multilevel Model for Change 1  2.1 Introduction Stroke is the leading cause of neurologic impairment, disability and participation restriction in Canada (Heart and Stroke Foundation, 2003) and one of the leading causes of disease burden worldwide (World Health Organization, 2008). For individuals with stroke, there can be sudden changes in body function requiring continuous coping and adaptations at each stage of the recovery process. The ultimate goals of rehabilitation are to facilitate participation and enhance quality of life. In the early stages post-stroke, rehabilitation focuses on maximizing independence in basic mobility and self-care. For those individuals who can return to living in the community, recovery is still not stable in the first year post discharge from rehabilitation (Horgan, O'Regan, Cunningham, & Finn, 2009). To assist in the ongoing adjustment and maximizing quality of life in individuals with stroke, knowledge about factors contributing to an individual’s ongoing recovery is useful to inform rehabilitation. Recently there has been a growing interest in exploring self-efficacy as a factor influencing functional recovery and quality of life in individuals with stroke (Asakawa, Usuda, Mizukami, & Imura, 2009; Brock et al., 2009; Jones, Mandy, & Partridge, 2009). It has been suggested that achievement of goals may be associated with improved self-efficacy (Brock et al., 2009) and that the degree of self-efficacy influences the extent of activities or participation in individuals with stroke (Asakawa et al., 2009). Bandura suggested that self-efficacy has a strong influence on people’s coping efforts and choice of activities and that it is a stronger predictor of behavior than skills or ability 1  A version of this chapter will be submitted for publication. Yiu, J., Miller, W.C., Eng, J.J., and Jarus, T. (2010) Longitudinal Analysis of Balance Confidence in Stroke Survivors Using a Multilevel Model for Change.  20  (Bandura, 1977). Previous studies of older adults have found that those with lower fallrelated self-efficacy (defined as a person’s belief in his or her ability to perform a series of everyday tasks without falling) tend to decline more in their function and health (Cumming, Salkeld, Thomas, & Szonyi, 2000; Tinetti, Mendes de Leon, Doucette, & Baker, 1994). In individuals with stroke, it has been suggested that physical activity self-efficacy predicts capabilities in performing activities of daily living and quality of life (LeBrasseur, Sayers, Ouellette, & Fielding, 2006). Robinson-Smith, Johnston and Allen (2000) found that selfcare self-efficacy (the belief a person has in his or her ability to perform self-care activities) is strongly related to quality of life and to depression in individuals with stroke. Salbach et al. (2006) found that individuals living in the community 6 months to 1 year after stroke experienced reduced balance self-efficacy 2 [defined as a person’s belief in executing skills they possess to maintain balance while performing selected activities (Powell & Myers, 1995)] and suggested that it was relevant to include balance self-efficacy as an outcome of stroke rehabilitation. However, Salbach et al. (2006) did not include a control group of individuals without stroke. Moreover, in her study, outcome variables were measured after an intervention; therefore, the natural history of balance confidence and outcome variables could not be measured. Little is known about how balance confidence changes over time in individuals with stroke. So far, one longitudinal study has specifically examined this. Hellstrom, Lindmark, Wahlberg, and Fugl-Myer (2003) found that perceived confidence in task performance without falling, as measured with the Falls Efficacy Scale (FES), improved significantly from rehabilitation admission (mean days post stroke onset=22; SD=6.6) to discharge (from a mean FES score of 63.5±35.5 to 109.7±25.5 out of a possible 130) but not from discharge to 2  In this paper, we will use the terms balance confidence and balance self-efficacy interchangeably.  21  10 month follow up. In the same study, they also found that those with lower FES scores at discharge showed less pronounced improvement in balance, walking ability and activities of daily living (ADL) than those with high FES scores. Despite a weaker association between FES scores and balance at 10 month post stroke, self-efficacy was found to be a more powerful predictor of ADL than balance (Hellstrom et al., 2003). These findings suggest that balance confidence changes over time in the post rehabilitation phase of stroke recovery as well as balance, walking ability and ADL. However, Hellstrom et al. (2003)’s study did not include a control group, did not control for covariates of balance confidence and had a small sample (N=37); it would be difficult to conclude from Hellstrom et al. (2003)’s study if individuals with stroke had reduced balance confidence and if the changes in balance confidence are natural variation or related to stroke. Understanding the natural history of balance confidence and factors affecting changes in balance confidence during stroke recovery is useful to inform rehabilitation. Many factors have been found to be associated with balance confidence. Balance performance, motor function and walking ability have been found to be important covariates of balance confidence in both community dwelling elderly (Hatch, Gill-Body, & Portney, 2003) and individuals with stroke (Hellstrom, Nilsson, and Fugl-Myer, 2001; Hellstrom et al., 2003). Hellstrom et al., (2003) in their longitudinal study of individuals from 3 weeks to 10 months post stroke found that perceived confidence in task performance without falling, as measured with the Falls Efficacy Scale (FES), was associated with balance and motor function at rehabilitation admission and discharge; however, the associations were weaker at 10 months post stroke. Previous studies have also identified depression, social support and sex as factors associated with balance confidence in individuals with lower limb amputees (Miller &  22  Deathe, 2004), older adults (Gillespie & Friedman, 2007; Kressig {{1802 Kressig et al., 2001) and individuals with stroke (Salbach et al., 2006). It would be important to control these covariates of balance confidence while examining the relationship between stroke status and balance confidence. Little is known if individuals with stroke have reduced balance confidence and how much variation in balance confidence in these individuals is related to natural variation and how much is related to stroke. No study has examined the natural history of balance confidence in individuals with stroke using a comparison group of individuals who do not have a stroke. The overall objective of this study was to investigate the natural variation in balance confidence and the relationship between stroke status and balance confidence while controlling for physical performance, psychological and contextual variables in individuals with stroke in the first year after discharge from rehabilitation. Multilevel models for change were constructed to study how balance confidence varied in individuals with stroke over time compared to those who have not had a stroke and to identify important variables explaining these variations. We had 3 research questions: 1) Does balance confidence change over time for individuals who have had a stroke compared to matched controls without a stroke?; 2) Does stroke status remain an important predictor of balance confidence after controlling for age, sex, perceived social support, balance performance, basic mobility, walking capacity, anxiety and depressive symptoms?; 3) Does stroke status interact with other factors when predicting balance confidence?” 2.2 Methods We used a longitudinal design measuring variables at four time points in a group of individuals who had recently had a stroke and a control group of individuals who had not experienced a stroke. Multilevel models using mixed-effects that take into account each 23  individual’s growth trajectory (within-person change) and between-person differences in change were constructed to study how balance confidence varied in individuals with stroke over time compared to those who have not had a stroke and to identify important variables explaining these variations. 2.2.1 Participants Individuals with stroke and age- and sex-matched controls with no disability were recruited to participate in this cohort study. Stroke subjects were recruited through inpatient rehabilitation units in 5 hospitals within British Columbia and controls were recruited through advertisements in local newspapers and community centers. Stroke subjects were identified within 1 month of discharge from rehabilitation and followed up at 3, 6, and 12 months post discharge. Subjects were measured at all 4 time points on all variables except for demographic information and stroke characteristics which were only collected at baseline. Subjects were assessed by research assistants who were fully trained in the administration of all performance measures and questionnaires. To be included stroke subjects had to have a history of a single stroke, residual unilateral lower extremity weakness caused by their stroke, and were able to independently walk a minimum of 10 meters. All subjects were over 50 years of age and able to communicate in English. Those who had significant musculoskeletal or neurological conditions other than stroke, who lived greater than 50 km from the data collection centres, or those with significant cognitive or communication impairments were excluded. This study was approved by the University of British Columbia Clinical Research Ethics Board and the local health authorities (see Appendix I for ethics certificate). All eligible subjects gave written informed consent before participating in the study.  24  2.2.2 Outcome Measures (Appendix II presents the details of the psychometric properties of all the measures) Performance measures. The Berg Balance Scale (BBS) (Berg, Wood-Dauphinee, & Gayton, 1989; Berg, Wood-Dauphinee, Williams & Maki, 1992) was used to measure balance performance. The BBS is a 14-item scale designed to provide an indication of balance while sitting, standing or stepping. Each item is rated on a 5 point likert scale from 0 to 4. A summary score is obtained by summing all individual item scores. A higher score indicates better balance performance. Appendix III shows a sample of the BBS scoring form. A recent systematic review of the usefulness of the BBS in stroke rehabilitation reveals that the BBS has excellent internal consistency, interrater reliability, intra-rater reliability, and test-retest reliability (Blum & Korner-Bitensky, 2008). Evidence of validity has been demonstrated by having strong correlations with other balance, mobility, and functional measures, as well as gait speed (Blum & Korner-Bitensky, 2008). The Timed Up and Go (TUG) Test was used to measure basic mobility function including transfers and walking. The time taken to the nearest tenth of a second that is required to stand from a sitting position, walk a three-meter distance, turn, walk back to the chair and sit down is recorded. Appendix IV provides a visual of the TUG. The TUG has excellent test-retest and inter-rater reliability and evidence supporting validity has been reported for older adults (Mathias, Nayak, & Isaacs, 1986; Podsiadlo & Richardson, 1991). In terms of validity, the TUG has been found to correlate with walking speed (r=0.66), the Older Adults Resources and Services ADL Scale (-0.45) (Lin et al., 2004) and the Barthel  25  Index (r=-0.51) (Podsiadlo & Richardson, 1991). Responsiveness of the TUG has been investigated for the stroke population (SRM 3 range from 0.41 to 0.88) (Salbach et al., 2001). The 6MWT is a commonly used measure of exercise capacity in individuals with compromised ability (Guyatt et al., 1984; Guyatt et al., 1985). It was used in this study to measure walking capacity. The farthest distance walked to the nearest meter in 6 minutes is recorded. The individual may or may not use a gait aid or take a break during walking. The scoring form of the 6MWT is shown in Appendix V. Inter-rater reliability for the 6MWT is good and has been assessed with individuals with stroke, as well as intra-rater reliability and responsiveness to change (Kosak & Smith, 2005). Test-retest reliability in stroke has been established (ICC=0.99) (Eng, Dawson, & Chu, 2004). Psychological measures. The Activities-Specific Balance Confidence Scale (ABC) (Powell & Myers, 1995) was used to measure balance confidence. The ABC is a self-report questionnaire that was developed using a self-efficacy framework (Bandura, 1977). It assesses the confidence that one can engage in 16 activities of daily living without losing balance or becoming unsteady. These activities are meant to represent a wide spectrum of difficulty. The ABC uses a 100 point confidence scale ranging from 0 (no confidence) to 100 (completely confident). A summary score is derived by averaging the total item scores (see Appendix VI for a sample of the ABC scoring form). Psychometric properties of the ABC have been studied with individuals with stroke (Botner, Miller, & Eng, 2005). The ABC is internally consistent for individuals with stroke (α=0.94) (Botner et al., 2005). Test-retest reliability of the ABC is strong (ICC=0.85) (Botner et al., 2005; Powell & Myers, 1995). Validity of the ABC is established by demonstrating a correlation with performance tests  3  Standardized response mean is the ratio of the mean change to the standard deviation of the change scores (Streiner & Norman, 2003).  26  (Botner et al., 2005) and other measures of falls related confidence (Hotchkiss et al., 2004; Powell & Myers, 1995). The Center for Epidemiological Studies Depression (CES-D) Scale (Radloff, 1977) was used to assess the level of depression symptomatology. The CES-D Scale is a self-report questionnaire consisting of 20 questions asking how frequently the respondents experienced symptoms in the past week. Responses are rated on a 4-point Likert scale ranging from 0 (rarely or none of the time) to 3 (most or all of the time). Higher scores indicate a higher level of depressive symptomatology. Appendix VII shows the scoring form of the CES-D. The CES-D is designed to measure current state and to be responsive to changes over time (Radloff & Terri, 1986). The scale has been studied on different types of people of a wide range of ages, geographical area, and racial backgrounds (Radloff & Terri, 1986). Inter-rater reliability and validity for the CES-D are established in stroke patients (Shinar et al., 1986). The State-Trait Anxiety Inventory (STAI) for Adults (Spielberger, Gorsuch, & Lushene, 1970; Spielberger, 1983) was used to measure the level of anxiety at the moment (for State) and in general (for Trait). In this study, only the STAI State scale scores were used for the analysis. Since a person’s physiological state affects his or her perceived selfefficacy (Bandura, 1977), it is plausible that balance confidence could be influenced by a person’s anxiety state; therefore, it was important to control for it. Responses are rated on a 4-point Likert scale ranging from 1 to 4. Higher scores indicate higher levels of anxiety. Appendix VIII shows a copy of the STAI. The STAI has well-established psychometric properties (Barnes, Harp, & Jung, 2002; Hedberg, 1972; Metzger, 1976). For the purpose of screening, baseline cognitive status of the participants was measured using the Cognitive Capacity Screening Examination (CCSE) (Jacobs, Bernhard,  27  Delgado, & Strain, 1977) (Appendix IX). Validity of the CCSE has been examined with patients with a neurological diagnosis (Kaufman, Weinberger, Strain, & Jacobs, 1979). It has been found to correlate with full neurologic evaluation (Kaufman, Weinberger, Strain, & Jacobs, 1979). Measures of contextual variables. Data concerning stroke characteristics were extracted from facility medical records. Socio-demographic information (sex, age, partnering status, residence) was collected using study questionnaires. The 6-item Interpersonal Support Evaluation List (ISEL) (Cohen & Hoberman, 1983; Cohen, Mermelstein, Kamarck, & Hoberman, 1985) was used to measure perceived support resources (Appendix X). The ISEL demonstrates good internal consistency (Cronbach’s α= 0.45 to 0.75) and test/retest reliability (ICC= 0.63 to 0.85). A number of variables reflecting the subject’s health condition are collected. A questionnaire consisting of 31 questions, adapted from the Canadian Community Health Survey (Statistics Canada, 2000-2001), was used to identify any chronic condition that the participant may have (Appendix XI). The number of prescribed medications taken by each subject was also recorded. 2.2.3 Statistical Analysis Multilevel model for change. Multilevel models for change (Bryk & Raudenbush, 1987; Singer & Willett, 2003) were created to address our 3 research questions. These models describe patterns of individual changes in balance confidence over time and to identify predictors contributing to the differences in these changes. Multilevel models for change are mixed-effects models that take into account each individual’s growth trajectory (within-person change) and between-person differences in change (Singer & Willett, 2003). The fixed effects tell us the systematic between-person differences in change trajectory based on the values of the predictors; the random effects allowed the parameters of the growth 28  trajectories to scatter about the population average (Singer & Willett, 2003). A level-1 model describes how each individual changes over time (within-person change), and the level-1 residuals represent not just measurement error but also the variance in balance confidence not explained by time (Singer & Willett, 2003). A level-2 model describes how balance confidence changes systematically differently across individuals and the effects of timeinvariant characteristics of the individual, and the level-2 residuals represent variance in balance confidence remained unexplained by the level-2 predictors (Singer & Willett, 2003). Models were fitted and compared using a taxonomy as suggested by Singer and Willet (2003). Decisions to enter, retain and remove predictors were based on plausibility of the relationship, prior research, results of data exploration, hypothesis testing and comparison of model fit (Singer & Willett, 2003). All models were estimated using the method of Restricted Maximum Likelihood (REML). All the analyses were performed using SPSS Statistics (version 17) MIXED MODELS LINEAR with a significance level of 0.05 and 95% confidence intervals. Variable description. Descriptive statistics were used to present demographic data, participant characteristics, and baseline variables. Continuous variables were described in terms of means and standard deviations and categorical variables in terms of frequency and percentage. Baseline characteristics in the stroke group and control group were compared to normative data and using independent-samples t test for continuous variables and chi-square test for categorical variables. Distributions. Histograms were used to check distribution forms of the 2 groups combined. Variables with extremely skewed distributions (skewness > │1│) were log transformed using SPSS “transform” “compute variable” and “Ln”. Transformed variables  29  with improved distribution form were used in the multilevel analyses. The results were compared to results of models fitted with untransformed variables. If there are no significant differences in the results, the untransformed variables were used. Sample sizes and missing data. Sample sizes were calculated for each level of the multilevel model. The level-1 sample size is the number of repeated measures within each participant, and the level-2 sample size is the number of participants who completed testing (Bickel, 2007). Only those participants providing data from at least 2 testing occasions were included in the analyses (Tabachnick & Fidell, 2007). Baseline differences between dropouts and participants were compared using t test for continuous variables and chi-square test for categorical variables. To deal with issues regarding missing data, we used the following strategies. Hard copies of the data were checked to make sure that missing data were really missing and reasons for missing were noted. Real missing values were replaced with the average variable value for that participant (Tabachnick & Fidell, 2007). Multilevel analyses were conducted with missing values replaced or not replaced, and results of these analyses were compared. If there were no significant differences in the results, missing values would not be replaced. Outliers. Original hard copy data were first checked to make sure that there were no discrepancies between hard copies and electronic copies of data and that the values were plausible. Outliers were defined as those who: 1) had a value of the Mahalanobis Distance (Mahal) that is significantly higher than the others (as determined by the critical χ2 value for the respective df at α=0.001); (Mahal values were obtained using multiple regression analysis with predictors from each level) (Tabachnick & Fidell, 2007); 2) scored beyond 3 SDs as identified by box plots; and 3) scored in the top 5 highest or lowest for any variables. Each  30  outlier was checked to see if their background information is significantly different from the rest of the group or if they belong to a sub-population. Multilevel analyses were conducted with each of these outliers removed, one at a time. Decisions regarding whether all or only the extreme outliers or none would be removed was made based on the amount of variance remained unexplained by the model and how influential the outliers are on the results of the analyses. Multicollinearity. Multiple regression analyses were conducted with predictors at each level to determine if problems of multicollinearity existed (Tabachnick & Fidell, 2007). Corrective actions, such as centering variables or removing variables involved in multicollinearity, were considered if needed. Data exploration. Scatter plots of ABC as a function of time were used to explore individual changes over time and to confirm the appropriateness of a linear change trajectory. Descriptive statistics was used to present individual changes in balance confidence in terms of frequency and percentages of individuals whose change score exceed the minimally detectable change of the ABC. Preparation of dataset. The dataset was rearranged in the “Person-Period” format to prepare for the multilevel analysis. Appendix XII (Figure 5) shows an example of the “Person-Period” format of the dataset used for the multilevel analyses. In this format, each person has multiple records, one for each testing occasion. Preparation of variables. Predictors were centered to facilitate interpretation (Singer, 2003) (Table 4 in Appendix XIII describes how variables were centered). When a predictor is centered on its sample mean, the fitted intercepts represent the average fitted values of initial status or rate of change (Singer & Willett, 2003).  31  In order to identify time-varying and time-invariant independent variables, an unconditional growth model using time as a predictor and each independent variable as the dependent variable was created (see Appendix XIV for details of how to create an unconditional growth model). Only those variables that had a statistically significant change over the study period were treated as time varying; otherwise, time-invariant. Baseline values of time-invariant variables were used for the multilevel analysis. Multilevel analyses. Hypothesis 1 for our first research question was that balance confidence would remain significantly lower in individuals with stroke than in individuals without a stroke over the 12 month study period. To address our first research question, an unconditional growth model (UGM) was first fitted using time (months post baseline) as a level-1 predictor. This model would allow us to establish how much within-person or between-person variations in balance confidence were attributed to time alone (Singer & Willett, 2003). A second model was fitted by adding stroke status as a level-2 predictor to the UGM. We called this model “Model A” which would provide answers to our first research question. Time would be treated as random effect. We hypothesized that stroke status would remain a significant risk factor for reduced balance confidence after controlling for age, sex, perceived social support, balance performance, basic mobility, walking capacity, anxiety state and depression symptoms. To address this hypothesis, we first added all time varying predictors as level-1 predictors then all time-invariant variables as level-2 predictors to Model A. We called this model “Model B” which would provide answers to our second research question. The decision regarding treating which time varying variable as random effect would be made based on model fit, ability to converge and theoretical considerations.  32  Hypothesis 3 for our third research question was that stroke status would interact with at least one other predictor variable when predicting balance confidence. To address our third research question, we first decided which interaction terms to build. Since we were interested in whether stroke status interacts with other factors, we built interaction terms with stroke status. Both within level and cross level interactions were added one at a time to the model (Model B) after the main effects included in it because changing the order of entry can affect parameter estimates for fixed effects (Tabachnick & Fidell, 2007). Statistically significant interactions were re-entered into the model to create the final model (Model C). Prototypical change trajectories. Prototypical change trajectories were used to display the results of fitted multilevel models for change. A prototypical plot is a graph of the trajectory of a dependent variable for selected values of the predictors (Singer & Willett, 2003). It presents fitted trajectories for prototypical individuals derived from the models. If a predictor was categorical, separate lines are plotted for each group. If a predictor is continuous, lines are drawn for representative values. Sample mean values were selected as prototypical values of continuous predictors in model B while for Model C, the 25th and the 75th percentiles were selected to represent values of predictors that had significant interactions with stroke status. Effect size. Pseudo-R2 were calculated to compare contributions of each predictor in the models. Appendix XV describes how pseudo-R2 are computed. They quantify the proportional reduction in residual variance on the addition of predictors (Singer & Willett, 2003). Separate Pseudo-R2 was calculated for level 1 (within-person) and level 2 (betweenperson) variations. This is because time-invariant predictors cannot explain much within-  33  person variation whereas time-varying predictors can affect both within- and between-person variations (Singer & Willett, 2003). 2.3 Results 2.3.1 Variable Descriptions Demographic and characteristics of the participants in terms of means and standard deviations and frequency and percentages are shown in Table 2-1. Significant differences were observed between the 2 groups for place of residence, number of chronic conditions, number of medications and results of the Cognitive Capacity Screening Examination (all p values<0.001) but not for partnering status. Most stroke participants lived in a house whereas most control participants lived in an apartment. Stroke participants had a higher number of chronic conditions, took more medications and had lower cognitive screening scores. Table 2-2 shows the means and standard deviations of all the variables at baseline and available normative data of all the measures. When compared to controls, participants with stroke had significantly lower balance confidence (mean difference=32.25 points), poorer balance performance (mean difference=9.17 points), poorer basic mobility function (mean difference=13.2 seconds), lower walking capacity (mean difference=263.56m), had more symptoms of depression (mean difference=8.96 points) and higher state anxiety scores (mean difference=9.95 points). All differences between groups were significant at p<0.001. There was, however, no significant difference in terms of perceived social support (p=0.58). When compared to norms, the stroke group ABC scores showed reduced balance confidence indicative of moderate level of functioning older adults whereas the control group ABC scores showed that they belonged to a high functioning group (see Table 2-2). Moreover, the mean TUG time for the stroke group was outside of the norms of the oldest group (age 8099). 34  Distributions. Three variables, TUG, CES-D and STAI, have extreme positive skewness while the BBS and the ABC had extreme negative skewness (skewness ranged from 1.16 to 5.27 and from -2.21 to -1.29). TUG was much improved with logarithmic transformations (skewness decreased from 5.27 to 1.01) and STAI, to a lesser extent (skewness decreased from 1.16 to 0.56). CES-D could not be transformed due to presence of values equal or less than zero, and transformation of BBS and ABC increased the negative skewness. Therefore, the decision was made to leave these variables untransformed. Moreover, modeling with transformed variables that had improved distribution forms did not improve model fit and failed to converge when TUG or STAI were treated as random effects. In the interest of interpretability, the decision was made to model untransformed variables. The 6MWT, age and perceived social support are acceptably distributed. Figure 7 and 8 in Appendix XVI shows histograms for frequency distributions of all the variables, transformed and untransformed. 2.3.2 Sample Sizes and Missing Data Samples sizes. A total of 98 individuals with stroke and 98 age- and sex-matched controls without stroke were successfully recruited and participated in the study at baseline. Of these, 73 participants with stroke and 80 controls (N=153) were successfully followed up at 12 months. Three participants skipped one measurement occasion but participated again later. The level-2 sample size is the number of participants who completed testing. This varies at each measurement occasion. At baseline, 196 participants completed testing. At 3 months, 178 participants completed testing. At 6 months, 172 participants completed testing. At 12 months, 153 participants completed testing. Eleven participants with stroke and 4 controls were excluded from the analyses because they provided data from only 1 testing  35  occasion. After excluding these 15 participants, data collected from 181 participants were used for the multilevel analyses. Figure 2-1 illustrates the final level-2 sample size. The level-1 sample size is the number of repeated measures within each participant (Bickel, 2007) which is supposed to be 4 if no observations were missing. There were 683 observations collected from the 181 participants for the TUG, BBS, STAI and ABC; however, only 679 observations for the 6MWT, 680 for the CES-D and 682 for the ISEL. Due to attrition and other reasons for missing data, the actual number of repeated measures taken by each participant for the different variables varies from 1 to 4 with an average ranging from 3.75 to 3.77. Appendix XVII lists the reasons for missing data. Figure 2-2 illustrates the final level-1 sample size. In terms of baseline differences between dropouts and participants, for the stroke group, there were no significant differences between those who dropped out of the study and those who did not in terms of age, sex, walking capacity, basic mobility, perceived social support and all psychological measurements; however, differences were observed for balance confidence and balance performance. Dropouts in the stroke group had lower balance confidence (mean difference= 14.03 points; p=0.008) and poorer balance performance (mean difference=4.4 points; p=0.026) than the participants. For the controls, there were no significant differences between dropouts and participants for all variables except for anxiety state. Dropouts had higher anxiety state (mean difference=4.69 points; p =0.004) than participants in the control group. Table 5 and 6 in Appendix XVIII presents the mean difference of variables at baseline between dropouts and participants. Missing data. Checking hard copies of data did not reveal any errors in data entry in regards to missing data. When all missing data were replaced with the average values for that  36  participant and multilevel analyses were conducted, the results were similar for most of the estimates of the fixed effects and all significant levels, for the total between-person variance and total within-person variance explained by all the predictors; however, the effect of time was affected resulting in a decrease in the estimated rate of change in balance confidence and the differential difference in the estimated rate of change in the stroke participants. Since missing data were more likely to happen in the later testing occasions, it is not surprising to see a bias towards a “flatter” slope when missing data was replaced with the average for that participant. In the interest of minimizing this bias, the decision was made to conduct the multilevel analyses with missing data not replaced. 2.3.3 Results of Outlier Analyses We identified 11 participants who have Mahal values that were significantly higher than the rest of the sample. Of these, we corrected values of 4 observations which were entered incorrectly into the electronic files. One participant (ff-020) was found to have significantly longer time since stroke onset (623 days compared to group mean of 96.93days) than the rest of the sample, to be the only one living in a long term care facility at follow up and to have the highest scores which were beyond 3 SDs for the TUG (worst mobility). Three observations of the TUG from this outlier were removed for the multilevel analyses. Another 5 participants were identified to score beyond 3SDs in 7 observations (TUG-1 observation, BBS-1observation, ISEL-2 observations, CES-D-1 observation, STAI-1 observation and ABC-1 observation) and to have the highest or lowest score of the sample. Two of these 5 participants had extreme scores in 2 different measures. These 7 observations were removed one by one for the multilevel analyses. Of these, 4 observations were found to have very minimal influence on the fixed effects and variance component of the results of the analyses and therefore were not removed for the final analyses. Three other observations (ff37  102 TUG; c-251 ISEL; ff-417 ISEL) were considered as influential outliers and were removed for the final analyses. 2.3.4 Multicollinearity Collinearity diagnostics using multiple regression analysis that included the 9 predictors (using centered variables) from all levels were presented in Appendix XIX. The variance inflation factors (VIF), condition indices (CI) and tolerance were all within acceptable values (i.e. VIF<10, CI<30 and tolerance >0.1) indicating that collinearity was not a concern. 2.3.5 Data Exploration Inspection of scatter plots of individual changes over time in ABC (balance confidence) scores revealed that there was little change in balance confidence in most of the individuals without a stroke whereas balance confidence increased slightly or remained unchanged over time in many individuals from the stroke sample. Figures 9 to 12 in Appendix XX shows the plots of individual changes of the whole sample. In individuals with stroke, there appeared to be variations in the intercepts and, to a lesser extent, slopes of the fitted regression lines; whereas, in individuals without a stroke, there was little variation. The pattern of change appears to be linear for most of the sample. A linear change trajectory appears to be reasonable for many individuals in our sample. Further data explorations revealed that some individuals with stroke did change over time and they varied in their change patterns. At an individual level, 32 individuals with stroke had a change score that was greater than 13 points, which has been identified as being indicative of a minimally detectable change of the ABC (Steffen & Seney, 2008), over the 1 year study period; 25 of them had an increase and 7 had a decrease. This made up 45% of the stroke sample who had complete data for computing the change score. 38  2.3.6 Preparation of Variables Decisions regarding time varying and time-invariant variables. Unconditional growth modeling revealed that changes in balance performance and perceived social support in the whole sample were non-significant (p = 0.374 and 0.608 respectively). These variables were therefore treated as time-invariant and the values of baseline measures were used for the multilevel analysis. There was; however, a significant change in main effect for time in depression symptoms (p =0.025), walking capacity (p<0.001), anxiety state (p<0.001) and basic mobility function (p=0.001). These variables were then treated as time-varying in the multilevel analysis. 2.3.7 Results of Multilevel Analyses Hypothesis 1. The results of multilevel linear modeling that answer our first research question are presented in Table 2-3. We fitted 3 models: 1) the unconditional means model with no predictors, 2) the unconditional growth model (UGM) in which time (months post baseline) was the only predictor in the level-1 sub-model and 3) Model A in which stroke status was added as a level-2 predictor to the UGM (see Appendix XXI for details of how these models were created). When we compared the variance components of the 3 models, we found that time explained most of the within person variation in balance confidence (the within-person residuals decreased from 68.03 to 59.15 when time was added as a predictor to the level-1 sub-model. We also found that stroke status explained some of the betweenperson variation in initial status (residuals decreased from 453.51 to 245.35 when stroke status was added to the UGM) and almost all of the variation in rate of change in balance confidence (residuals became non-significant in model A compared to the UGM). Model A revealed that the estimated initial ABC score for the average individual without a stroke was 94.03; the estimated differential in initial ABC score between 39  individuals with stroke and without stroke was -28.90 which was significant (P<0.001). The estimated rate of change in ABC for an average individual without stroke was -0.12 which represented a non-significant decline (p =0.262). The estimated differential in rate of change between individuals with stroke and without stroke was 0.74 which was significant (P<0.001) and represented an increase in ABC score. Figure 2.3 shows prototypical trajectories derived from the UGM and Model A. Hypothesis 2. The results of multilevel linear modeling adjusting for all confounding variables are presented in Table 2-4. Model B answers our second research question (see Appendix XXII for details of how Model B was created). This model revealed that stroke status still remained a significant predictor of the intercept and slope of the estimated average change trajectory in balance confidence. Initial ABC score was still significantly lower in the average individual who had a stroke (8.30 points difference on the ABC) (p<0.001), and difference in average monthly rate of change in ABC between this individual and an average control subject was still statistically significant (p=0.005). Better balance performance, better basic mobility and better walking capacity were associated with higher initial balance confidence (p=0.001, p=0.003 and p<0.001 respectively). Presence of more depression symptoms (p<0.001), being female (p=0.045) and higher anxiety state (p=0.034) were associated with lower balance confidence. Age and perceived social support were not found to be associated with balance confidence (p=0.99 and 0.13 respectively). Monthly rate of decline in balance confidence remained non-significant in the average individual without a stroke (p=0.10). Figure 2.4 shows prototypical trajectories derived from Model B. Hypothesis 3. The results of multilevel linear modeling (Model C) that included all significant interaction terms are presented in Table 2-5 (see Appendix XXIII for details of  40  how Model C was created). This model revealed that there were significant interactions between stroke status and depression symptoms (p=0.002) and between stroke status and walking capacity (p<0.001). It appears that individuals with stroke who have a CES-D score greater than the grand mean (i.e. more depressed individuals with a positive score after centering) would have lower balance confidence since the estimated effect of stroke status X depression symptoms was -0.41, a negative estimate. Similarly those individuals with stroke who have a higher walking capacity (i.e. positive 6MWT score after centering) had better balance confidence since the estimated effect of stroke status X walking capacity was 0.53, a positive estimate. Figure 2-5 shows prototypical trajectories derived from Model C of 4 individuals demonstrating the interaction between stroke status and depression symptoms. It shows that an average individual with stroke who was depressed had a consistently lower level of balance confidence over time than an average individual with stroke who was not depressed; however, the effect of depression does not appear in the control individuals. Figure 2-6 shows prototypical trajectories derived from Model C of 4 individuals demonstrating the interaction between stroke status and walking capacity. It appears that an average individual with stroke who had a high walking capacity consistently have higher balance confidence over time than the average individual with stroke who had a low walking capacity; similarly, this effect impacts only the individuals with stroke. After adding the interactions, stroke status alone still remained as an independent predictor of balance confidence (p<0.001); initial balance confidence was still significantly lower in individuals who had a stroke than individuals without a stroke by 9.84 points. Balance performance, basic mobility and anxiety remained as significant predictors  41  (p=0.012, p=0.03 and p=0.039) of balance confidence. However, sex became non-significant (p=0.073) and the estimated differential in rate of change in balance confidence between individuals who had a stroke and individuals without a stroke also became non-significant (a differential of 0.25 point per month on the ABC) (p=0.078). As shown in both figure 2-4 and 2-5, the slopes of the linear trajectories between individuals with stroke and the controls appear more similar to each other although the stroke individuals still show an increase whereas the control individuals had a decline over time. These plots help visualize the nonsignificant estimated differential in rate of change in balance confidence in Model C. Effect size. Variables associated with balance confidence by Model B are shown in Table 2-6. Table 2-6 shows the percentage change in these pseudo- R2 which represents the amount of variance in ABC explained by each variable. In terms of within-person variance, months post baseline was the most important factor explaining 13.06% of the within-person variance. The next most important factor was basic mobility explaining 11.25% of the within-person variance; walking capacity explained 9.16%, and depression symptoms explained 6.04% of the within-person variance. These time-varying predictors also explained some percentages of the between-person variance ranging from 1.17% to 3.95%. Stroke status was the most important factor explaining the between-person variation in balance confidence (45.90%). The next most important factor was balance performance (19.28%). Sex explained 4.16%; perceived social support explained 3.91%, whereas age explained minimally the between-person variance. 2.4 Discussion The results of our study suggest that we cannot reject all of the a priori null hypotheses. Balance confidence remained significantly lower in individuals with stroke than in individuals without a stroke over a 1 year study period; stroke status remained a significant 42  risk factor for reduced balance confidence after controlling for the covariates; and stroke status interacted with depression symptoms and walking capacity when predicting balance confidence. The present study showed that patterns of change in balance confidence over a 1 year study period are different for those who have had a stroke and those without a stroke. Model A suggested that initially balance confidence in individuals who had a stroke was significantly lower than individuals without a stroke. Individuals without a stroke showed little to no change in balance confidence over the 1 year whereas individuals with stroke showed a statistically significant increase over 1 year after discharge from rehabilitation. Multilevel modeling for change in our study revealed that the there was a differential of 0.75 points in the monthly rate of change in ABC score between those who have had s stroke and those without a stroke. Although statistically significant, this value is very small considering the minimally detectable change of the ABC has been found to be 13 points (Steffen & Seney, 2008). Therefore the change in balance confidence was not necessarily clinically meaningful. Although some individuals with stroke had a change score of greater than 13 points, one third of them had a decrease in balance confidence. Our findings suggested that reduced balance confidence is a serious problem in individuals with stroke. As a group, the mean ABC score increased from 61.87 at baseline to 73.14 at 12 months. The fact that ABC scores changed over time in our sample suggests that there could a natural recovery in balance confidence. But the slow natural group improvement in individuals with stroke over time could also represent measurement error and was not clinically meaningful. It is therefore important to identify factors associated with this change for the purpose of designing interventions. Our findings also suggest there may be a  43  likelihood of greater improvement in balance confidence with interventions in individuals with stroke who have reduced balance confidence. Balance confidence did not change significantly in individuals who have not had a stroke. One reason for this would be that there was a ceiling effect of the ABC particularly in the control group; however, this also showed the stability of our control subjects. The within-person residuals remained significant in Model A and even Model B suggesting the presence of unexplained variance. Although this included measurement error, there could be other time-varying predictors for which we did not control such as falling during the study period. Due to sample size constraints we had to select substantial predictors to put in the models. Our study showed that individuals with stroke had significantly lower initial balance confidence even after controlling for age, sex, perceived social support, balance performance, basic mobility, walking capacity, anxiety state and depression symptoms. This is not surprising as previous studies with stroke have found reduced balance confidence in individuals with stroke (Salbach et al., 2006; Hellstrom et al., 2003). In our study, balance confidence remained significantly lower in individuals who have had a stroke over 1 year despite a steady increase in balance confidence in these individuals. This suggests that despite a natural improvement, reduced balance confidence appears to be a persistent problem in individuals with stroke suggesting the importance of providing interventions for this problem. The results of multilevel modeling for change by Model B were consistent with previous research findings. Aside from stroke status, we found that balance performance, basic mobility and walking capacity were among the important factors associated with  44  balance confidence with balance performance explaining the most (19.28%) of the variance in balance confidence in our sample. In a study by Hatch et al. (2003), they found that balance performance explained 57% of variance in balance confidence in a sample of community dwelling elderly people. Myers et al. (1996) also found significant relationships between balance confidence and balance performance, mobility and walking speed in a group of community dwelling elders. Our findings suggest that these significant relationships also exist in individuals with stroke and further support the theoretical claim that self-efficacy is associated with skills in an activity-specific context. In our study we controlled for anxiety state and found that it was significantly associated with balance confidence in those who have had a stroke and those without a stroke. This is a novel finding and may have important implications. Bandura (1977) suggested that efficacy information comes from 4 major sources one of which is physiological states. A high level of anxiety may produce somatic information that carries efficacy implications (Bandura, 1997). Moreover, van Haastregt, Zijlstra, van Rossum, van Eijk, and Kempen (2008) found that severe activity avoidance was associated with anxiety. Future studies should further investigate the relationship between anxiety state and balance confidence and the implications for activity avoidance. Our study found that being female and the presence of symptoms of depression were associated with lower balance confidence. This is also consistent with previous research findings. Kressig et al., 2001 found a significant association between the depression symptoms (CES-D) and balance confidence (ABC) in a small sample (N=20) of independent living seniors. Miller and Deathe (2004) found that being female and symptoms of depression predicted balance confidence (ABC) in individuals with lower limb amputation,  45  and Salbach et al. (2006) found that balance confidence (ABC) decreased with increasing depression symptoms in 91 individuals with chronic stroke (mean days post stroke=227±79). Our study added to previous findings in that being female and the presence of depression symptoms were also important predictors of balance confidence in individuals in the post acute phase of stroke rehabilitation. Our study discovered that there was a significant interaction between stroke status and depression symptoms in the prediction of balance confidence. Increasing depression symptoms appeared to further lower balance confidence in individuals who had a stroke versus those without a stroke. While previous studies reported significant associations between depression and balance confidence in stroke and non-stroke subjects, no studies have used a control group for comparisons. Our findings have important clinical implications. Depression has been known to be an important predictor of participation, satisfaction with community reintegration and quality of life post stroke (Kim, Warren, Madill, & Hadley, 1999; Lo et al., 2008; Pang, Eng, & Miller, 2007), and considerable effort has been placed on finding the best treatment for depression post stroke (Hackett, Anderson, House, & Xia, 2008). The finding suggests that treating balance confidence could potentially be an adjunct to treatment of depression in individuals with stroke as this may enhance the treatment effect of the latter. This of course requires further study. Another important finding in our study was that stroke status interacted with walking capacity when predicting balance confidence. Previous studies (Salbach et al., 2005; Salbach et al., 2006) suggested a relationship between balance confidence and walking capacity in individuals with stroke. Salbach et al. (2005) found that change in balance confidence in a sample of stroke was associated with change in walking capacity after a task specific walking  46  intervention in improving balance confidence. It may be that stroke-specific impairments are major limitations to the distance walked (Eng, Chu, Dawson, Kim, & Hepburn, 2002); therefore, those with low walking capacity are at a higher risk for developing reduced balance confidence. Our finding may have important implications for designing interventions to improve balance confidence. Interventions may involve providing experience of personal mastery in activities that demand walking capacity (i.e. distance walked and gait speed). As suggested by Bandura (1977) performance accomplishments provide the most influential efficacy information because it is based on personal experience of mastery. 2.4.1 Limitations of the Study The generalizability of the results of this study presents limitations. This study was carried out with individuals who had a stroke and who were able to go through a short course of inpatient rehabilitation. Only those individuals who were ambulatory were included. Those individuals who had had a mild stroke and were discharged home directly from acute care were not included. On the other hand, individuals with more severe deficits who were cognitively impaired or had severe communication difficulties were not included. Therefore our findings cannot be generalized to these individuals. Approximately seventy-three percent of our sample was men and this could have affected how psychological measures such as depressions, anxiety state and even balance confidence were reported. Although the proportion of men who have stroke is similar to that of women (Roquer, Campello, & Gomis, 2003), women were found to suffer aphasic disorders more often and those who survived a stroke remained more disabled than men (Roquer et al., 2003). Since individuals with aphasia and more severe deficits were not included in our study, this explains why a large proportion of our sample was men.  47  The dropouts had lower ABC and this means survival bias may be present; therefore, our results are conservative and likely underestimate the degree of the problem in stroke. There was a ceiling effect in ABC particularly in the control group. Since we had volunteers and not a random sample, these volunteer may actually have higher ABC scores. Although we found statistically significant relationships between the predictors and balance confidence, we could not claim a causal relationship. Although a longitudinal design is a strength but the temporal relationship between predictors and balance confidence could not be clearly established given the interpretive difficulties that time-varying predictors can present (Singer & Willett, 2003). All of our time-varying predictors (TUG, 6MWT, STAI, CES-D) describe an individual’s potential for change. Given that our variable of interest was stroke status, a time-invariant variable, it would have been ideal to measure balance confidence before stroke happened but this would be impossible. A very large population based sample would be required at considerable cost. Nevertheless, we had an age- and sexmatched control group which enabled us to compare despite not having an experimental design. 2.5 Conclusions Previous studies have suggested a negative impact of reduced balance confidence on functioning and quality of life of individuals with stroke. Results from the present study confirmed that reduced balance confidence is a persistent and serious problem in these individuals. Balance performance, basic mobility, walking capacity, depression symptoms and anxiety state were important factors associated with balance confidence over time. Depression and walking capacity appeared to be a stroke specific factor affecting level of balance confidence. Interventions designed to treat reduced balance confidence may include providing experience of balance-related performance accomplishment. 48  Table 2-1 Demographic and Stroke Characteristics of Participants Stroke Participants (N=98)  Controls (N=98)  Mean(SD) or n(%)  Mean(SD) or n(%)  67.41(10.13)  67.64(9.91)  Male  71(72.4%)  71(72.4%)  Female  27(27.6%)  27(27.6%)  Presence of a spousal partner  66(67.4%)  56(57.1%)  Absence of a spousal partner  32(32.6%)  42(42.9%)  Apartment  27(27.6%)  44(44.9%)  House  70(71.4%)  54(55.1%)  1(1.0%)  0  Number of chronic conditions**  4.13(1.88)  2.16(1.86)  Number of medications**  6.05(3.55)  3.45(4.14)  Cognitive Capacity Screening  25.51(4.23)  28.04(1.91)  Ischemic  68(69.4%)  -  Hemorrhagic  19(19.4%)  -  8(8.1%)  -  Left  47(48.0%)  -  Right  43(43.9%)  -  Variable Age Sex  Partnering status  Residence*  Care home  Examination** Type of stroke  Others & unknown Side of stroke  Others & unknown Time since stroke onset  8(8.2%) 96.9 days(69.0)  -  **p <0.001 by Independent- Samples t Test; *p <0.05 by Chi-Square Test  49  Table 2-2 Participants’ Profile at Baseline in the Stroke and Control Groups with Normative Data Variable  Stroke  Controls  Participants  (N=98)  (N=98)  Mean(SD)  Available Normative Data  Mean(SD) ABC*  61.87(23.89)  94.12(8.20)  <50= low functioning; >50 and <80 = moderate functioning; >80 = high functioning (Myers, Fletcher, Myers, & Sherk, 1998)  BBS*  45.29(8.91)  54.46(3.08)  <45 = falls risk (Thorbahn & Newton, 1996)  TUG*  21.38(17.89)  8.18(1.92)  age 60-69: 8.1(7.1-9.0) age 70 to 79: 9.2(8.2-10.2)  (sec)  age 80-99: 11.3(10.0-12.7) (Bohannon, 2006) 6MWT*  267.65(139.53)  531.21(90.75)  Norms (2SD) age 60-69: 354-756; age 70-79: 321-697;  (m)  age 80-89: 222-563 (Steffen, 2000) CES-D*  15.29(9.11)  6.33(5.69)  ≥16 are indicative of depression (Shinar et al. 1986)  STAI*  >39-40 = clinically significant symptoms of  State  36.06(11.18)  26.11(6.40)  anxiety state  Trait  37.48(10.12)  29.22(7.30)  >55/54 geriatric patients (Kvaal, Ulstein, Nordhus, & Engedal, 2005)  ISEL  13.27(2.37)  13.44(2.04)  not available  *P<0.001 by Independent- Samples t Test  50  Table 2-3 Results of Fitting the Unconditional Means Model (UMM), the Unconditional Growth Model (UGM) and Model A to Data (Appendix XXI (a) to (c) present the written models, SPSS syntax and selected output for these models) Fixed Effects Composite model Level 2 predictor Level 1 predictor  UMM  UGM  Model A  Intercept, 𝜋𝜋0𝑖𝑖 (initial status)  81.26***  80.13***  94.03***  Stroke status  -  -  -28.90***  Rate of change, 𝜋𝜋1𝑖𝑖 (months  -  0.24**  -0.12  Stroke status X months post  -  -  0.74***  68.03***  59.15***  59.12***  (p=0.262)  post baseline)  baseline Variance Components Level 1 Level 2  Within-person, 𝜀𝜀𝑖𝑖𝑖𝑖  In initial status, 𝜁𝜁0𝑖𝑖  391.65*** 453.51*** 245.35***  In rate of change, 𝜁𝜁1𝑖𝑖  -  Covariance, 𝜁𝜁0𝑖𝑖 and 𝜁𝜁1𝑖𝑖  -  0.33*  0.20 (p=0.12)  -7.46***  -2.11 (p=0.15)  ***p< 0.001; **p=0.004; *p=0.02  51  Table 2-4 Results of Fitting Model B to the Data (Appendix XXII presents the written model, SPSS syntax and selected output for Model B) Fixed Effects Composite model Level 2 predictors:  Parameter  Model B  Intercept (initial status), 𝜋𝜋0𝑖𝑖  γ00  86.45***  Stroke status  γ01  -8.30***  Sex  γ02  -3.89*  (time-invariant)  (p=0.045) Age  γ03  0.00 (p=0.99)  Social support  γ04  0.64 (p=0.13)  Balance performance  γ05  0.64*** (p=0.001)  Level 1 predictors: (time-varying)  Rate of change, 𝜋𝜋1𝑖𝑖  γ10  Stroke status X months post  γ11  (p=0.10)  (months post baseline)  baseline Basic mobility  -0.16  0.42** (p=0.005)  γ30  -0.40** (p=0.003)  Walking capacity  γ40  0.27***  Depression symptoms  γ50  -0.27***  Anxiety  γ60  -0.13*  52  Fixed Effects  Parameter  Model B (p=0.034)  Variance Components Level 1 Level 2  Parameter  Model B  Within-person, 𝜀𝜀𝑖𝑖𝑖𝑖  σ2ε  39.90***  In rate of change, 𝜁𝜁1𝑖𝑖  σ12  0.17  In initial status, 𝜁𝜁0𝑖𝑖  σ20  Basic mobility, 𝜁𝜁3𝑖𝑖  σ22  Walking capacity, 𝜁𝜁4𝑖𝑖  σ23  Anxiety, 𝜁𝜁6𝑖𝑖  σ24  Depression symptoms, 𝜁𝜁5𝑖𝑖  72.93***  (p=0.065) 0.15 (p=0.12) 0.12* (p=0.024) 0.08 (p=0.072) 0.05 (p=0.347)  ***p≤ 0.001; **p< 0.01; *p<0.05  53  Table 2-5 Results of Fitting Model C to Data (multilevel modeling that included all significant interaction terms; Appendix XXIII presents the written model, SPSS syntax and selected output for Model C) Fixed Effects Composite model Level 2 predictors:  Parameter  Model C  Intercept (initial status), 𝜋𝜋0𝑖𝑖  γ00  92.09 ***  Stroke status  γ01  -9.84***  Sex  γ02  -3.04  (time-invariant)  (p=0.073) Age  γ03  -0.09 (p=0.285)  Social support  γ04  0.51 (p=0.18)  Balance performance  γ05  0.49* (p=0.012)  Level 1 predictors: (time-varying)  Rate of change, 𝜋𝜋1𝑖𝑖  γ10  Stroke status X months post  γ11  (p=0.19)  (months post baseline)  baseline Basic mobility  -0.12  0.25 (p=0.078)  γ30  -0.30 (p=0.03)  Walking capacity  γ40  0.02 (p=0.80)  Depression symptoms  γ50  -0.04 54  Fixed Effects  Parameter  Model C (p=0.68)  Anxiety  γ60  -0.11* (p=0.039)  Interaction terms:  Stroke status X depression  γ70  symptoms Stroke status X walking capacity Variance Components Level 1 Level 2  -0.41** (p=0.002)  γ80  0.53***  Parameter Within-person, 𝜀𝜀𝑖𝑖𝑖𝑖  σ2ε  34.55***  In rate of change, 𝜁𝜁1𝑖𝑖  σ12  0.19  In initial status, 𝜁𝜁0𝑖𝑖  σ20  Basic mobility, 𝜁𝜁3𝑖𝑖  σ22  Anxiety, 𝜁𝜁6𝑖𝑖  σ24  Stroke Status X Depression symptoms, 𝜁𝜁8𝑖𝑖  Stroke Status X Walking capacity, 𝜁𝜁7𝑖𝑖  50.44***  (p=0.023) 0.16 (p=0.08) 0.04 (p=0.28) 0.20 (p=0.12)  σ23  0.51***  ***p≤ 0.001; **p< 0.01; *p<0.05  55  Table 2-6 Percentage of Variance Explained by Each Variable Computed from the Unconditional Growth Model, Model A and Model B 𝑹𝑹𝟐𝟐𝜺𝜺 ∆x100% (within-person) 13.06%  𝑹𝑹𝟐𝟐𝟎𝟎 ∆x100% (between-person)  Stroke status  ~  45.90%  Sex  ~  4.16%  Age  ~  0.5%  ISEL  ~  3.91%  BBS  ~  19.28%  TUG  11.25%  1.17%  6MW  9.16%  3.95%  CES-D  6.04%  3.40%  STAI  1.83%  1.70%  Model B  41.34%  83.92%  Variable Months post  -  baseline  𝑅𝑅𝜀𝜀2 ∆ indicates change in proportion of within-person variation “explained” by the variable; 𝑅𝑅02 ∆ indicates change in proportion of between-person variation “explained”.  56  Figure 2-1 Diagram Illustrating Final Level-2 Sample Size Stroke N=98  # assessed at baseline: N=98  # assessed at 3 months: N=86, 12 missing  At baseline, N=196  # assessed at 6 months: N=82, 16 missing  # assessed at more than 1 testing occasion: N = 87  # assessed at 12 months: N=73, 25 missing  # assessed at baseline: N=98  Total assessed at more than 1 testing occasion: N2 = 181 Total excluded= 15  Controls N=98  # assessed at 3 months: N=92, 6 missing  # assessed at 6 months: N=90, 8 missing  # assessed at 12 months: N=80, 18 missing  # assessed at more than 1 testing occasion: N=94  N2= 181 Legend: N = # of participants N2 = Level-2 sample size = 87+94 = 181  57  Figure 2-2 Diagram Illustrating Final Level-1 Sample Size  N2 = 181 181x4=724 observations (from 4 testing occasions for each variable if none missing)  150 participants had 4 repeated measures = 600 observations 21 participants had 3 repeated measures = 63 observations  Total observations: 600+63+20 = 683 Observations missing: 724-683 = 41  10 participants had 2 repeated measures = 20 observations  TUG: 683 observations N1= 3.77  BBS: 683 observations N1=3.77  STAI: 683 observations N1=3.77  ABC: 683 observations N1=3.77  More observations missing in these variables 6MW: 679 observations Missing=45 N1=3.75  CES-D: 680 observations Missing=44 N1=3.76  ISEL: 682 observations Missing=42 N1=3.77  N2 = Level-2 sample size N1 = Level-1 sample size = # of repeated measures (baseline, 3, 6 and 12 months post baseline) 58  Figure 2-3 Prototypical Trajectories from the Unconditional Growth Model (UGM) and Model A Displaying the Effect of Stroke on Balance Confidence (ABC)  Unconditional Growth Model 100 80  ABC  60 40  ABC scores  20 0 Baseline  3  6  9  12  Months Post Baseline  Model A 100 80  ABC  60 Stroke  40  Control 20 0 Baseline  3  6  9  12  Months Post Baseline  59  Figure 2-4 Prototypical Trajectories from Model B Displaying the Controlled Effect of Stroke on Balance Confidence  Model B 100 90  ABC  80 Stroke/female  70  Stroke/male 60  Control/female  50  Control/male  40 Baseline  3  6  9  12  Months Post Baseline  60  Figure 2-5 Prototypical Trajectories from Model C of 4 Individuals Showing the Interaction between Stroke Status and Depression Symptoms (All individuals shown have the mean score of all variables except for the CES-D. The individual who is labeled “depressed” has the 75th percentile CSE-D score (4.22) and the one who is labeled “not depressed” has the 25th percentile CES-D score (-6.78). All values were centered according to methods presented in appendix XIII.)  Model C 100 90  ABC  80 Stroke/depressed  70  Stroke/not depressed 60  control/depressed  50  control/not depressed  40 Baseline  3  6  9  12  Months Post Baseline  61  Figure 2-6 Prototypical Trajectories Derived from Model C of 4 Individuals Showing the Interaction between Stroke Status and 6MW Distance (All individuals shown have the mean score of all variables except for the 6MW. The individual who is labeled “low walking capacity” has the 25th percentile 6MW score (-13.08m) and the one who is labeled “high walking capacity” has the 75th percentile 6MW score (12.92m). All values were centered according to methods presented in appendix XIII.)  Model C 100  ABC  90 80  Stroke/low walking capacity  70  Stroke/high walking capacity  60  control/low walking capacity  50  control/high walking capacity  40 Baseline  3  6  9  12  Months Post Baseline  62  Reference List Asakawa, Y., Usuda, S., Mizukami, M., & Imura, S. (2009). Moderator and mediator effects of personal factors in patients with stroke. Journal of Physical Therapy Science, 21, 5563. Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Review, 84, 191-215. Bandura, A. (1997). Self-efficacy: The exercise of control. New York: W.H. Freeman and Company. Barnes, L. L. B., Harp, D., & Jung, W. S. (2002). Reliability generalization of scores on the spielberger state-trait anxiety inventory. Educational and Psychological Measurement, 62, 603-618. Berg, K., Wood-Dauphinee, S., Williams, J. I., & Gayton, D. (1989). Measuring balance in the elderly: Preliminary development of an instrument. Physiotherapy Canada, 41, 304311. Berg, K. O., Wood-Dauphinee, S. L., Williams, J. I., & Maki, B. (1992). Measuring balance in the elderly: Validation of an instrument. Canadian Journal of Public Health.Revue Canadienne De Sante Publique, 83(Suppl 2:S7-11), JuAug. Bickel, R. (2007). Multilevel analysis for applied research. New York, NY: The Guildford Press. Blum, L., & Korner-Bitensky, N. (2008). Usefulness of the berg balance scale in stroke rehabilitation: A systematic review. Physical Therapy, 88, 559-566. Bohannon, R. W. (2006). Reference values for the timed up and go test: A descriptive metaanalysis. Journal of Geriatric Physical Therapy, 29, 64-68. 63  Botner, E. M., Miller, W. C., & Eng, J. J. (2005). Measurement properties of the activitiesspecific balance confidence scale among individuals with stroke. Disability & Rehabilitation, 27, 156-163. Brock, K., Black, S., Cotton, S., Kennedy, G., Wilson, S., & Sutton, E. (2009). Goal achievement in the six months after inpatient rehabilitation for stroke. Disability & Rehabilitation, 31, 880-886. Bryk, A. S., & Raudenbush, S. W. (1987). Application of hierarchical linear models to assessing change. Psychological Bulletin, 101, 147-158. Cohen, S., & Hoberman, H. M. (1983). Positive events and social supports as buffers of life change stress. Journal of Applied Social Psychology, 12, 99-125. Cohen, S., Mermelstein, R., Kamarck, T., & Hoberman, H. M. (1985). Measuring the functional components of social supports. In I. G. Sarason, & B. R. Sarason (Eds.), Social support: Theory, research, and application (pp. 73-94). The Netherlands: Martinus Nijhoff. Cumming, R. G., Salkeld, G., Thomas, M., & Szonyi, G. (2000). Prospective study of the impact of fear of falling on activities of daily living, SF-36 scores, and nursing home admission. Journals of Gerontology Series A-Biological Sciences & Medical Sciences, 55, M299-305. Delistamati, E., Samakouri, M.A., Davis, E.A., Vorvolakos, T., Xenitidis, K., Livaditis, M. (2006). Interpersonal Support Evaluation List (ISEL)-college version: Validation and application in a Greek sample. International Journal of Social Psychiatry, 52, 552-560.  64  Eng, J. J., Chu, K., Dawson, A. S., Kim, C. M., & Hepburn, K. (2002). Functional walk tests in individuals with stroke: Relationship to perceived exertion and myocardial exertion. Stroke, 33, 756-761. Eng, J. J., Dawson, A. S., & Chu, K. (2004). Submaximal exercise in persons with stroke: Test-retest reliability and concurrent validity with maximal oxygen consumption. Archives of Physical Medicine & Rehabilitation, 85, 113-118. Gillespie, S. M., & Friedman, S. M. (2007). Fear of falling in new long-term care enrollees. Journal of the American Medical Directors Association, 8, 307-313. Guyatt, G. H., Pugsley, S. O., Sullivan, M. J., Thompson, P. J., Berman, L., Jones, N. L., et al. (1984). Effect of encouragement on walking test performance. Thorax, 39, 818-822. Guyatt, G. H., Sullivan, M. J., Thompson, P. J., Fallen, E. L., Pugsley, S. O., Taylor, D. W., et al. (1985). The 6-minute walk : A new measure of exercise capacity in patients with chronic heart failure. Canadian Medical Association Journal, 132, 919-923. Hackett, M.L., Anderson, C.S., House, A., & Xia, J. (2008). Interventions for treating depression after stroke. Cochrane Database of Systematic Reviews,4, 003437. Hatch, J., Gill-Body, K. M., & Portney, L. G. (2003). Determinants of balance confidence in community-dwelling elderly people. Physical Therapy, 83, 1072-1079. Heart and Stroke Foundation. (2003). The growing burden of heart disease and stroke in Canada 2003 Hedberg, A. G. (1972). Review of 'state-trait anxiety inventory'. Professional Psychology, 3, 389-390.  65  Hellstrom, K., Nilsson, L., & Fugl-Meyer, A. R. (2001). Relationship of confidence in task performance with balance and motor function after stroke. Physiotherapy Theory and Practice, 17, 55-65. Hellstrom, K., Lindmark, B., Wahlberg, B., & Fugl-Meyer, A. R. (2003). Self-efficacy in relation to impairments and activities of daily living disability in elderly patients with stroke: A prospective investigation. Journal of Rehabilitation Medicine, 35, 202-207. Horgan, N. F., O'Regan, M., Cunningham, C. J., & Finn, A. M. (2009). Recovery after stroke: A 1-year profile. Disability & Rehabilitation, 31, 831-839. Hotchkiss, A., Fisher, A., Robertson, R., Ruttencutter, A., Schuffert, J., & Barker, D. B. (2004). Convergent and predictive validity of three scales related to falls in the elderly. American Journal of Occupational Therapy, 58, 100-103. Jacobs, J. W., Bernhard, M. R., Delgado, A., & Strain, J. J. (1977). Screening for organic mental syndromes in the medically ill. Annals of Internal Medicine, 86, 40-46. Jones, F., Mandy, A., & Partridge, C. (2009). Changing self-efficacy in individuals following a first time stroke: Preliminary study of a novel self-management intervention. Clinical Rehabilitation, 23, 522-533. Kaufman, D. M., Weinberger, M., Strain, J. J., & Jacobs, J. W. (1979). Detection of cognitive deficits by a brief mental status examination. the cognitive capacity screening examination, a reappraisal and a review. General Hospital Psychiatry, 1, 247-255. Kim, P., Warren, S., Madill, H., & Hadley, M. (1999). Quality of life of stroke survivors. Quality of Life Research, 8, 293-301. Kosak, M., & Smith, T. (2005). Comparison of the 2-, 6-, and 12-minute walk tests in patients with stroke. Journal of Rehabilitation Research and Development, 42, 103-107.  66  Kressig, R. W., Wolf, S. L., Sattin, R. W., O'Grady, M., Greenspan, A., Curns, A., et al. (2001). Associations of demographic, functional, and behavioral characteristics with activity-related fear of falling among older adults transitioning to frailty. Journal of the American Geriatrics Society, 49, 1456-1462. Kvaal, K., Ulstein, I., Nordhus, I.H., & Engedal, K. (2005). The Spielberger State-Trait Anxiety Inventory (STAI): the state scale in detecting mental disorders in geriatric patients. International Journal of Geriatric Psychiatry, 20, 629-634. LeBrasseur, N. K., Sayers, S. P., Ouellette, M. M., & Fielding, R. A. (2006). Muscle impairments and behavioral factors mediate functional limitations and disability following stroke. Physical Therapy, 86, 1342-1350. Lin, M. R., Hwang, H. F., Hu, M. H., Wu, H. D. I., Wang, Y. W., & Huang, F. C. (2004). Psychometric comparisons of the timed up and go, one-leg stand, functional reach, and tinetti balance measures in community-dwelling older people. Journal of American Geriatrics Society, 52, 1343-1348. Lo, R. S., Cheng, J. O., Wong, E. M., Tang, W. K., Wong, L. K., Woo, J., et al. (2008). Handicap and its determinants of change in stroke survivors: One-year follow-up study. Stroke, 39, 148-153. Mathias, S., Nayak, U. S., & Isaacs, B. (1986). Balance in elderly patients: The "get-up and go" test. Archives of Physical Medicine and Rehabilitation, 67, 387-389. Metzger, R. L. (1976). A reliability and validity study of the state-trait anxiety inventory. Journal of Clinical Psychology, 32, 276-278.  67  Miller, W. C., & Deathe, A. B. (2004). A prospective study examining balance confidence among individuals with lower limb amputation. Disability & Rehabilitation, 26, 875881. Myers, A. M., Fletcher, P. C., Myers, A. H., & Sherk, W. (1998). Discriminative and evaluative properties of the activities-specific balance confidence (ABC) scale. Journals of Gerontology Series A-Biological Sciences & Medical Sciences, 53, M287-94. Myers, A. M., Powell, L. E., Maki, B. E., Holliday, P. J., Brawley, L. R., & Sherk, W. (1996). Psychological indicators of balance confidence: Relationship to actual and perceived abilities. Journals of Gerontology Series A-Biological Sciences & Medical Sciences, 51, M37-43. Pang, M. Y., Eng, J. J., & Miller, W. C. (2007). Determinants of satisfaction with community reintegration in older adults with chronic stroke: Role of balance self-efficacy. Physical Therapy, 87, 282-291. Podsiadlo, D., & Richardson, S. (1991). The timed "up and go": A test of basic functional mobility for frail elderly persons. Journal of the American Geriatrics Society, 39, 142148. Powell, L. E., & Myers, A. M. (1995). The activities-specific balance confidence (ABC) scale. Journals of Gerontology Series A-Biological Sciences & Medical Sciences, 50A, M28-34. Radloff, L. S. (1977). The CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1, 385-401. Radloff, L. S., & Teri, L. (1986). Use of the center for epidemiological studies-depression scale with older adults. Clinical Gerontologist, 5, 119-135.  68  Robinson-Smith, G., Johnston, M.V., & Allen, J. (2000). Self-care self-efficacy, quality of life, and depression after stroke. Archives of Physical Medicine and Rehabilitation, 81, 460-464. Roquer, J., Campello, A.R., & Gomis, M. (2003). Sex differences in first-ever stroke. Stroke, 34, 1581-1585. Salbach, N. M., Mayo, N. E., Higgins, J., Ahmed, S., Finch, L. E., & Richards, C. L. (2001). Responsiveness and predictability of gait speed and other disability measures in acute stroke. Archives of Physical Medicine and Rehabilitation, 82, 1204-1212. Salbach, N. M., Mayo, N. E., Robichaud-Ekstrand, S., Hanley, J. A., Richards, C. L., & Wood-Dauphinee, S. (2005). The effect of a task-oriented walking intervention on improving balance self-efficacy poststroke: A randomized, controlled trial. Journal of the American Geriatrics Society, 53, 576-582. Salbach, N. M., Mayo, N. E., Robichaud-Ekstrand, S., Hanley, J. A., Richards, C. L., & Wood-Dauphinee, S. (2006). Balance self-efficacy and its relevance to physical function and perceived health status after stroke. Archives of Physical Medicine & Rehabilitation, 87, 364-370. Shinar, D., Gross, C. R., Price, T. R., Banko, M., Bolduc, P. L., & Robinson, R. G. (1986). Screening for depression in stroke patients: The reliability and validity of the center for epidemiologic studies depression scale. Stroke, 17, 241-245. Singer, J. D., & Willett, J. B. (2003). Applied longitudinal data analysis. modeling change and event occurrence. New York: Oxford University Press, Inc. Spielberger, C. D. (1983). Manual for the state-trait anxiety inventory (STAI). PaloAlto, CA: Consulting Psychologists Press.  69  Spielberger, C. D., Gorsuch, R. L., & Lushene, R. E. (1970). Manual for the state-trait anxiety inventory. Palo Alto, Calif: Consulting Psychologists Press. Statistics Canada. (2000-2001). Canadian community health survey cycle 1.1. Retrieved May/18, 2008, from http://www.statcan.ca/english/sdds/instrument/3226_Q1_V1_E.pdf Steffen, T.M. (2000) Functional assessment: A literature review of four tools. Focus: Geriatric Physical Therapy: An Independent Home Study Course for Individual Continuing Education. Steffen, T., & Seney, M. (2008). Test-retest reliability and minimal detectable change on balance and ambulation tests, the 36-item short-form health survey, and the unified parkinson disease rating scale in people with parkinsonism. Physical Therapy, 88, 733746. Streiner, D.L., & Norman, G.R. (2003). Health Measurement Scales: A Practical Guide to their Development and Use (3rd ed.). New York: Oxford University Press. Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics (5th ed.). Boston, MA: Pearson Education, Inc. Thorbahn, L. D. B., & Newton, R. A. (1996). Use of the berg balance test to predict falls in elderly persons. Physical Therapy, 76(6), 576-583. Tinetti, M. E., Mendes de Leon, C. F., Doucette, J. T., & Baker, D. I. (1994). Fear of falling and fall-related efficacy in relationship to functioning among community-living elders. Journal of Gerontology, 49, M140-7. van Haastregt, J. C., Zijlstra, G. A., van Rossum, E., van Eijk, J. T., & Kempen, G. I. (2008). Feelings of anxiety and symptoms of depression in community-living older persons who avoid activity for fear of falling. American Journal of Geriatric Psychiatry, 16, 186-193.  70  World Health Organization. (2008). The global burden of disease: 2004 update. Geneva: World Health Organization. Zijlstra, G. A., van Haastregt, J. C., van Rossum, E., van Eijk, J. T., Yardley, L., & Kempen, G. I. (2007). Interventions to reduce fear of falling in community-living older people: A systematic review. Journal of the American Geriatrics Society, 55, 603-615.  71  Chapter Three: Relationship between Balance Confidence and Social Participation in Individuals with First Stroke: A One Year Follow Up Study4 3.1 Introduction Stroke has devastated many people’s lives leading to restrictions in participation. According to the International Classification of Functioning Disability and Health (ICF), (World Health Organization, 2001), participation is involvement of an individual in a life situation, and participation restrictions are difficulties an individual may experience in life situations. There are many forms of social participation (Bukov, Maas, & Lampert, 2002); in this paper, we concentrate on social participation as defined by engagement in productive (activities that generate an outcome), social and leisure activities. Several studies have used participation instruments that measure beyond just social participation; therefore, we have used the term social participation and participation interchangeably. Optimizing social participation is an important goal of stroke rehabilitation (Salter, Foley, Jutai, & Teasell, 2007). In fact, social participation is the highest level of outcome as described by the ICF (World Health Organization, 2001). Previous longitudinal studies (Isaacs, Neville, & Rushford, 1976; Lawrence & Christie, 1979; Niemi, Laaksonen, Kotila, & Waltimo, 1988; Robinson, Bolduc, Kubos, Starr, & Price, 1985) have identified that individuals with stroke experience a decline in the level of social participation up to 4 years after stroke. Problems related to social participation include inability to engage in household activities, disruption in roles within the family, inability to resume paid/unpaid occupation, social and leisure activities (Lawrence & Christie, 1979; Niemi et al., 1988). A lower level of 4  A version of this chapter will be submitted for publication. Yiu, J., Miller, W.C., Eng, J.J., and Jarus, T. (2010) Relationship between Balance Confidence and Social Participation in Individuals with First Stroke: A One Year Follow Up Study  72  social participation (measured with the Activity Card Sort and Reintegration to Normal Living-RNL) has been found to correlate with lower life satisfaction 1 year post stroke (Hartman-Maeir, Soroker, Ring, Avni, & Katz, 2007) and poorer quality of life [measured with the SF-36 and QOL-VAS(visual analog scale)] 2 years post stroke (Mayo, WoodDauphinee, Cote, Durcan, & Carlton, 2002). Recent studies reveal complex patterns of change in participation among individuals with stroke over time and different measures of participation were used in these studies. Desrosiers et al. (2008) found a significant increase in participation (measured with the Assessment of Life Habits, LIFE-H) for stroke survivors in the first 3 to 6 months after being discharged home from either acute or post-acute rehabilitation hospital. Lo et al. (2008) found no significant change in overall participation (measured with the London Handicap Scale, LHS) from 3 to 12 months post-stroke in their sample, although they found a significant improvement in the mobility5 and social interaction sub-categories. In contrast, findings of another study (Desrosiers et al., 2006) on long term changes in participation have revealed a decline in 4 categories [nutrition which includes grocery planning, meal preparation and eating meals, fitness (physical activities), personal care and housing (household maintenance)] of daily activities at 6 months and between 2 to 4 years poststroke; interestingly, in Desrosiers’ study, participation in the social roles did not change and they also found an improvement in the responsibilities (banking, making purchases, budgeting, care-giving) category. The different findings of these studies indicate that participation post-stroke appears to vary depending on the type of activities involved, the length of time since stroke onset and how participation was measured.  5  Mobility was measured as “participation” in the London Handicap Scale because it measures how mobility restricts the individual getting around or where he/she wants to go.  73  Many factors have been found to impact social participation in the general population of older adults and individuals who have suffered a stroke. For older adults, higher general activity level, being married, better functional skills, younger age are associated with a higher level of social participation (Bukov et al., 2002; Dahan-Oliel, Gelinas, & Mazer, 2008; Sorensen, Axelsen, & Avlund, 2002); whereas, depression (Sorensen et al., 2002), limited mobility, health problems and physical and mental disability (Bukov et al., 2002; Orbon, Satink, de Jong, & van der Gulden, 2009) are associated with a lower level of social participation. In individuals who suffered a stroke, depression (Desrosiers et al., 2008; Lo et al., 2008; Teoh, Sims, & Milgrom, 2009; Wade, Legh-Smith, & Langton Hewer, 1985), comorbidity (Desrosiers, Noreau, Rochette, Bravo, & Boutin, 2002; Desrosiers et al., 2006), and age (Desrosiers et al., 2002; Desrosiers et al., 2006; Lo et al., 2008; Schepers, VisserMeily, Ketelaar, & Lindeman, 2005) are negatively associated with social participation. Arm function (Desrosiers et al., 2006; Sveen, Bautz-Holter, Sodring, Wyller, & Laake, 1999), lower extremity function (Desrosiers et al., 2002), balance (Desrosiers et al., 2002), walking ability (Desrosiers et al., 2008), and ability to perform activities of daily living (ADL) (Beckley, 2006; Roth & Lovell, 2007; Schepers et al., 2005; Wade et al., 1985) are positively associated with social participation. In addition to the above, being male (Lo et al., 2008; Schepers et al., 2005; Wade et al., 1985), being married (Schepers et al., 2005), and having better financial status (Lo et al., 2008) are associated with a higher level of social participation. Moreover, Beckley (Beckley, 2006) found that subjective social support interacts with ADL limitation and predicts social participation in individuals with stroke. Although Desrosiers et al. (2008) and Lo et al. (2008) found that depression symptoms have the strongest association with social participation in individuals with stroke, most studies  74  have focused on studying the impact of physical and socio-demographic factors of social participation. The impact of psychological factors has received less research attention. Balance confidence is a psychological factor that has received growing interest in the literature. It has been defined as the belief that one can execute skills they possess to maintain balance while performing selected activities (Powell & Myers, 1995). Balance confidence could prove to be an important psychological factor of participation in individuals who have sustained a stroke. The concept of balance confidence is derived from Social Cognitive theory suggesting that self-efficacy has a strong influence on people’s choice of activities and that it is a stronger predictor of behavior than skills or ability (Bandura, 1977). Studies of older adults have found that those with lower balance confidence tend to decline more in their function and health (Cumming, Salkeld, Thomas, & Szonyi, 2000; Tinetti, Mendes de Leon, Doucette, & Baker, 1994). It has been suggested that stroke survivors experience reduced balance confidence (chapter 2 of this thesis), and that balance confidence predicts physical function and physical health (Salbach et al., 2006), but little is known about its contribution to restrictions in participation after stroke. In individuals with traumatic brain injury, perceived self-efficacy has been found to predict social participation (Dumont, Gervais, Fougeyrollas, & Bertrand, 2005). Dumont et al (2005) found that perceived selfefficacy explained 40% of the variance in social participation. If balance confidence is indeed an important factor of social participation, clinicians could incorporate appropriate interventions for reduced balance confidence in the rehabilitation of individuals with stroke which in theory could lead to better social participation. Change in social participation post-stroke appears to be complex and research findings have been controversial. Balance confidence may prove to be an important factor of  75  social participation in individuals with stroke, and no previous study has investigated the time varying relationship between balance confidence and social participation in individuals with stroke. The overall objective of this study was to examine the variation in social participation in individuals with stroke in the first year following discharge from stroke rehabilitation compared to individuals without a stroke and to determine the relationship between balance confidence and social participation over this 1 year period. Our specific research questions were: 1) How does social participation change over 1 year after discharge from stroke rehabilitation and is there a difference in the level of social participation overall and the sub-domains of social participation between individuals with stroke and individuals without a stroke?; 2) Is balance confidence at discharge from rehabilitation (baseline) an important predictor of social participation at 1 year after discharge after controlling for important covariates: age, sex, partnering status, number of chronic conditions, perceived social support, stroke status, balance performance, basic mobility, walking capacity, anxiety state and depression symptoms? We had 2 hypotheses. First it was hypothesized that individuals with stroke would have a consistently lower level of social participation overall and in all the sub-domains of social participation as measured with the FAI than individuals without a stroke over 1 year. Second, balance confidence would remain an independent predictor of social participation after controlling for age, sex, partnering status, number of chronic conditions, perceived social support, stroke status, balance performance, basic mobility, walking capacity, anxiety trait and depression symptoms.  76  3.2 Methods 3.2.1 Participants Individuals with first stroke and an age- (+ 2 years) and sex-matched group of controls without a stroke were recruited to participate in this cohort study. Details about the method of recruitment were reported in chapter 2. In brief, stroke subjects were recruited through inpatient rehabilitation units in 5 hospitals in British Columbia: Lion’s Gate Hospital, Kelowna General Hospital, Victoria General Hospital, Holy Family Hospital and GF Strong Rehabilitation Centre. Individuals without a stroke were recruited through advertisements in local newspapers and community centers. Inclusion/exclusion criteria for participation in the study were also reported in details in chapter 2. This study was approved by the University of British Columbia Clinical Research Ethics Board and the local health authorities, and all eligible subjects gave written informed consent before participating in the study. 3.2.2 Procedure Baseline testing was completed within 1 month of discharge from rehabilitation (baseline) for stroke participants and at recruitment for control participants. All participants were also followed up at 3, 6, and 12 months after baseline. Social participation was measured only at 3, 6 and 12 months. The reason we did not measure social participation at baseline was that the Frenchay Activity Index (FAI), a commonly used tool for stroke, has a time component to its responses. The FAI assesses the level of social participation in the last 3 to 6 months. Baseline measurements of participation were not taken because most participants with stroke would have had spent the preceding months in hospital due to stroke. Other measures were taken at all time points except for demographic information and stroke characteristics which were only collected at baseline. Subjects were assessed by research 77  assistants who were fully trained in the administration of all performance measures and questionnaires. 3.2.3 Outcome Measures We selected variables to control for based on previous studies looking at participation and on factors believed to be plausible confounders or covariates with balance confidence in predicting participation. Measure of social participation. The Frenchay Activity Index (FAI) (Holbrook & Skilbeck, 1983) was used to measure social participation. The FAI is a 15-item scale developed for use with stroke patients and intended to measure three domains of daily living: domestic chores, work/leisure and outdoor activities (Holbrook & Skilbeck, 1983; Schuling, de Haan, Limburg, & Groenier, 1993). Responses capture the frequency of participation in the past 3 or 6 months and range from 0 (never or none) to 3 (daily or weekly). Summary scores are derived by adding the items, with scores ranging from 0 (no activity) to 45 (very high participation). Psychometric properties of the FAI have been studied with stroke. The scale demonstrates sufficient internal consistency (α range from 0.78 to 0.87) (Schuling et al., 1993), good inter-rater reliability (ICC=0.90, CI: 0.82-0.94) (Post & de Witte, 2003), and responsiveness to change (effect size between 6 to 12 months post stroke=0.59) (Schepers, Ketelaar, Visser-Meily, Dekker, & Lindeman, 2006). In Schepers et al. (2006)’s study, they quantified responsiveness by calculating effect sizes for each measure by dividing the mean absolute change score by the standard deviation of the baseline score. Test-retest reliability has also been studied (Bland and Altman reliability coefficient=7.1 out of 45) (Green, Forster, & Young, 2001). The 2-3 week retest reliability in the general population has been reported to be good (r=0.96) (Turnbull et al., 2000). Construct validity for the stroke population has been demonstrated by correlating the FAI with the Barthel Index (r range 78  from 0.61-0.66), and parts of the Sickness Impact Profile (high correlation with household activities and physical functioning r range from -0.39 to -0.73 but low correlation with emotional and alertness behavior r=-0.15 and 0.14) (Schuling et al., 1993; Wade et al., 1985). Additionally, evidence from a general population study demonstrated the FAI was able to discriminate between groups based on rating of self-reported health and overall activity levels (p<0.001) (Turnbull et al., 2000). Appendix XXIV shows a copy of the FAI. Measure of balance confidence. The Activities-Specific Balance Confidence Scale (ABC scale) (Powell & Myers, 1995) (Appendix VI) was used to measure balance confidence. The ABC is a self-report questionnaire that was developed using a social learning theory framework (Bandura, 1977). It measures the confidence that one can engage in 16 activities of daily living without losing balance or becoming unsteady. These activities are meant to represent a wide spectrum of difficulty. The ABC uses a 100 point numerical scale ranging from 0 (no confidence) to 100 (completely confident). A summary score is obtained by averaging the sum of all item scores. Psychometric properties of the ABC have been demonstrated with both community seniors (Powell & Myers, 1995; Talley, Wyman, & Gross, 2008) and individuals with stroke (Botner, Miller, & Eng, 2005). The ABC shows good internal consistency for individuals with stroke (α=0.94) (Botner et al., 2005), and acceptable test-retest reliability (ICC=0.85) has been reported (Botner et al., 2005; Powell & Myers, 1995). Evidence of validity has been established by demonstrating a correlation with performance tests: such as the Berg Balance Scale (ρ = 0.36) and gait speed (ρ = 0.48) (Botner et al., 2005; Talley et al., 2008) and other measures of falls related confidence such as the Falls Efficacy Scale (r=0.86), the Survey and Fear of Falling in the Elderly in the  79  general older adult population (r=0.66 to 0.67) (Hotchkiss et al., 2004; Powell & Myers, 1995). Covariates. The Berg Balance Scale (BBS) (Berg, Wood-Dauphinee, Williams, & Gayton, 1989; Berg, Wood-Dauphinee, Williams, & Maki, 1992) was used to measure balance performance. The BBS is a 14-item scale designed to provide an indication of balance while sitting, standing or stepping. Each item is scored on a scale from 0 (poor balance) to 4 (good balance). A summary score is obtained by adding all the item scores. Total scores range from 0 to 56 with higher scores reflecting better balance. Evidence of internal consistency (α=0.92-0.98), inter-rater reliability ICCs=0.95-0.98), intra-rater reliability (ICC-0.97) and test-retest reliability (ICC=0.98) have been reported in a recent systematic review of the usefulness of the BBS in stroke rehabilitation (Blum & KornerBitensky, 2008). The BBS has demonstrated construct validity by having good correlations with the balance subscale of the Fugl-Meyer Assessment (range from r=0.62 to 0.94), basic ADL function Barthel Index (r=0.88) and the Functional Independence Measure (r=0.57 to 0.76) (Blum & Korner-Bitensky, 2008). Appendix III describes the items used in the BBS. The Timed Up and Go (TUG) Test (Appendix IV) was used to measure basic mobility function including transfers, walking and turning while walking. It measures the time taken to the nearest tenth of a second that is required to stand from a sitting position, walk a three-metre distance, turn, walk back to the chair and sit down. Excellent intra-rater and inter-rater reliability (range from r=0.93 to 0.99) and evidence supporting validity have been reported for older adults (Lin et al., 2004; Mathias, Nayak, & Isaacs, 1986; Podsiadlo & Richardson, 1991a). In terms of validity, the TUG has been found to be able to identify potential fallers in patients in a stroke unit (Andersson, Kamwendo, Seiger, & Appelros,  80  2006), correlate with walking speed (r=0.66), the Older Adults Resources and Services ADL Scale (-0.45) (Lin et al., 2004) and the Barthel Index (r=-0.51) (Podsiadlo & Richardson, 1991b). Responsiveness of the TUG has been investigated for the stroke population (SRM range from 0.41 to 0.88) (Salbach et al., 2001). The 6MWT (Appendix V) is a commonly used measure of exercise capacity or endurance in individuals with compromised ability (Guyatt et al., 1984; Guyatt et al., 1985). It was used in this study to measure walking capacity. The test measures the farthest distance walked to the nearest meter in 6 minutes. The individual was allowed to use a gait aid or take a break during walking if needed. The 6MWT has demonstrated evidence of reliability with ICC’s ranging from 0.91 to 0.92 following practice tests in individuals with chronic diseases (Guyatt, Thompson et al., 1985). Test-retest reliability in community-dwelling individuals with stroke has been established (ICC=0.99) (Eng, Dawson, & Chu, 2004). For inpatient stroke, inter-rater reliability was lower (ICC=0.78), as well as intra-rater reliability (ICC=0.74), and responsiveness to change was also studied with a SRM of 1.52 (Kosak & Smith, 2005). The Center for Epidemiological Studies Depression (CES-D) Scale (Radloff, 1977) (Appendix VII) was used to assess the level of depression symptomatology. The CES-D Scale is a self-report questionnaire consisting of 20 questions asking how frequently the respondents experienced symptoms in the past week. Responses are rated on a 4-point Likert scale ranging from 0 (rarely or none of the time) to 3 (most or all of the time). Higher scores indicate a higher level of depressive symptomatology. The CES-D is designed to measure current state and is expected to have lower test retest reliability (Radloff & Teri, 1986). The scale has been studied on various populations (Radloff & Teri, 1986). Inter-rater reliability  81  (r=0.76) and validity for the CES-D are established in stroke patients (Shinar et al., 1986). The CES-D correlates other measures of depression: Zung depression scale (Spearman’s rho=0.65) (Shinar et al., 1986). The State-Trait Anxiety Inventory (STAI) for Adults (Spielberger, 1983) (Appendix VIII) was used to measure the level of anxiety at the moment (for State) and in general (for Trait). The STAI Trait was used at baseline and STAI State was used at both baseline and follow ups. The STAI has 20 items with responses rated on a 4-point Likert scale ranging from 1 to 4 with lower scores indicating higher level of anxiety. A summary score is obtained by adding all items scores. The STAI has well-established psychometric properties (Barnes, Harp, & Jung, 2002; Spielberger, 1984). Data concerning stroke characteristics were extracted from facility medical records. Socio-demographic information (sex, age, partnering status, living situation and education level) as well as information on medication use was collected using study questionnaires. The Canadian Community Health Survey was used to identify the number of chronic conditions (Appendix XI). The 6-item Interpersonal Support Evaluation List (ISEL) (Cohen & Hoberman, 1983; Cohen, Mermelstein, Kamarck, & Hoberman, 1985) (Appendix X) was used to measure the perceived availability of support resources. 3.2.4 Statistical Analysis Description of sample size, variables and group characteristics. Sample size and reasons for missing data were reported. Descriptive statistics were used to present demographic data, participant characteristics, and baseline variables. Continuous variables were described in terms of means and standard deviations and categorical variables in terms of frequency and percentage. Baseline characteristics in the stroke group and control group  82  were compared by means of independent-samples t test for continuous variables and chisquare test for categorical variables. Hypothesis 1. Multilevel models for change were used to describe changes in social participation. Multilevel models for change are mixed-effects models that take into account each individual’s growth trajectory (within-person change) and between-person differences in change (Singer & Willett, 2003). An unconditional means model was first created to assess the amount of variation in social participation across individuals without consideration of time. Then an unconditional growth model was created to assess the amount of variation across both individuals and time. These models allowed us to quantify the amount of withinor between-person variation explained by time and decide whether there was systematic variation in social participation that was worth exploring (Singer & Willett, 2003). To assess the differences in change in social participation between individuals with stroke and controls, stroke status was added as a predictor to the unconditional growth model, and a final model was created. The same process was repeated using the 3 sub-domains of the FAI (domestic chores, work/leisure and outdoor activities) as dependent variables to assess changes in these sub-domains and differences in change between the stroke subjects and controls. The intraclass correlation coefficient was calculated from the variance component of the intercept only model to determine the proportion of within- and between-individuals variations and the need for multilevel analyses for hypothesis 2 (Tabachnick & Fidell, 2007). Hypothesis 2. To assess the relationship between balance confidence and social participation, contingent on the finding of testing our first hypothesis, we would continue to build multilevel models by adding balance confidence and other covariates to the models. If there were no statistically meaningful changes (p<0.05) in social participation over time and  83  multilevel modeling was not indicated, multiple regression analyses would be conducted to determine if balance confidence would remain as an independent predictor of social participation after controlling for age, sex, partnering status, number of chronic conditions, perceived social support, stroke status, balance performance, basic mobility, walking capacity, anxiety trait and depression symptoms. Variable selection was conducted by first reviewing the bivariate analyses to identify statistically significant predictors of social participation. Pearson’s product-moment correlation coefficients would be used for continuous variables (age, balance confidence, balance performance, walking capacity, basic mobility, depression symptom, and anxiety trait). For categorical variables (sex, partnering status, stroke status), a t test was used. Partnering status was collapsed into a discrete variable; individuals were considered having a spousal partner or not. Any variable with a correlation between the absolute values of 0.2 and 0.7 and categorical variables with a significant mean difference by t test would be entered into the model. This is because a correlation smaller than 0.2 is too weak (Portney & Watkins, 2000) and a correlation greater than 0.7 indicates possible collinearity (Kleinbaum, Kupper, Muller, & Nizam, 2008). To estimate the relationship between balance confidence and social participation, we first performed a regression of social participation on balance confidence. Then, we assessed the presence of effect modification between the association of balance confidence and social participation. Assessment of effect modification or interaction should always take precedence over assessment of confounding (634 Kleinbaum, D.G. 1998). To select interaction terms involving balance confidence and covariates to enter into the model, we followed the steps recommended by Kleinbaum et al. (2008). First we specified the maximum model which  84  included balance confidence, all control variables and interaction terms involving balance confidence and control variables. Next, our criterion for model selection was chosen. Specifically we used the largest R2 and the smallest mean square error (MSE) values as selection criteria. A backward elimination procedure was used to determine which interaction terms involving balance confidence, our variable of interest, would be included in the model (p<0.05). We first created an unadjusted model which included balance confidence, the interaction terms as well as the first order effects included in the interaction terms. Next we created an adjusted model by entering the covariates into the model. To facilitate the interpretation of the interaction effects, scatterplots would be used to illustrate the relationship between social participation (Y-axis) and balance confidence (X-axis); the variable involved in the interaction terms would be collapsed into a binary variable by selecting a meaningful cutoff score; two regression fit lines representing individuals who scored above and below the cutoff score would be graphed. To avoid problems related to collinearity, all continuous independent and dependent variables were centered. The scores were subtracted by subtracting the sample means. Finally we assessed potential confounding by the covariates using a forward approach. All analyses were performed using SPSS, version 17.0, a significance level of 0.05 indicating statistical difference and confidence intervals of 95%. Regression diagnostics. Assumptions of multiple regression would be evaluated in terms of distribution, outliers, normality, linearity and homoscedasticity of residuals. Residual plots of the predicted values of the FAI would be examined to determine if assumptions of normality, linearity and homoscedasticity of residuals were violated. Histograms of frequency distributions of the independent and dependent variables would be  85  examined for skewness. Skewed distributions were log or square root transformed to see if transformation would improve distributions. Outliers would be determined using SPSS EXPLORE and the extreme values (top 5 highest and lowest values) were examined to see if any of them were disconnected from the next extreme values. Multivariate outliers would be sought using Mahalanobis distance obtained from SPSS REGRESSION. Outliers would be defined as having the highest Mahalanobis distance greater than the critical χ2 value at α=0.001 for 12 df (=number of independent variables). Multicollinearity would be assessed by means of tolerances, the variance inflation factor (VIF) statistic and condition index. Any tolerance smaller than 0.1, VIF greater than 10.0 and condition index greater than 30 would be a concern indicating problems with collinearity (Kleinbaum, Kupper, Muller, & Nizam, 2008). 3.3 Results 3.3.1 Description of Sample Size, Variables and Group Characteristics Sample characteristics. A total of 98 individuals with stroke and 98 age- and sexmatched controls without a stroke volunteered. Of these, 73 participants (74.5%) with stroke and 80 controls (81.6%) were successfully followed for12 months. FAI scores for 3 participants at 12 months were missing (reason unknown). Two CES-D and 1 ISEL scores were missing at baseline. Therefore the total number of subjects used for the multiple regression analyses was 147, with 70 stroke and 77 control subjects. The different reasons for missing data are shown in Appendix XVII. Variables and group characteristics. Table 2-1 of chapter 2 presents demographic information and stroke characteristics of the participants. In brief the majority of subjects were male (72.4%). Mean age was 67.4 years (SD=10.13). Mean time since stroke onset was 96.93 days (SD=69.0). Stroke participants predominantly had ischemic stroke (69.4%), 86  hemorrhagic (19.4%) and other (11.2%). Fourty-eight percent had left hemisphere stroke, 43% had right hemisphere stroke and 9% had no information on area of stroke. Balance confidence and covariates differences at baseline between the stroke and control groups are shown in Table 2-2 in chapter 2. The mean ABC Scale score at baseline was statistically lower (61.87±23.89) for the stroke subjects versus the controls (94.12±8.20). The stroke participants had statistically more chronic conditions, took more medications, had lower cognitive status, had poorer balance performance, poorer basic mobility function, lower walking capacity, were more depressed and had higher anxiety trait (mean difference=8.26±1.26). All differences were significant at p<0.001. There was, however, no significant difference in terms of perceived social support (p=0.575). Mean scores in overall and 3 sub-domains of social participation from 3 to 12 months post baseline are presented in Table 3-1. At 3 months, the total FAI mean score was 20.46 ± 8.88 for the stroke subjects and 32.89 ± 6.29 for the controls. At 12 months the total FAI mean score was 22.26 ± 9.91 in the stroke subjects and 32.54 ± 6.70 in the controls. 3.3.2 Results of Multilevel and Multiple Regression Analyses Hypothesis 1. Multilevel models for change revealed that there was a significant difference in overall FAI score and all the sub-domain scores between individuals with stroke and controls. The results are summarized in Table 3-2 to 3-5 and presented in more details in Appendices XXV to XXVIII. Table 3-2 shows that individuals with stroke on average scored 12.68 points lower on the FAI total score than controls (p<0.001). In terms of sub-domain scores, Table 3-3 to 3-5 shows that individuals with stroke also scored lower (by 4.84, 3.62 and 4.35 points respectively) in the 3 sub-domains (domestic, leisure/work and outdoors) of FAI (p<0.001).  87  The unconditional means and unconditional growth models revealed that there was not a significant change in overall social participation (FAI total score) and all the subdomains (domestic, leisure/work and outdoors) of the FAI from 3 to 12 months. Table 3-2 shows that the monthly rate of change in FAI total score was 0.07 points (p=0.168) and Table 3-3 to 3-5 show that the monthly rate of change in the sub-domains of FAI ranged from 0.02 to 0.03 points (p ranged from 0.120 to 0.484). When stroke status was added to the unconditional growth models as a predictor, the results revealed a significant difference only in the monthly rate of change in the domestic FAI score in individuals with stroke (Table 33). The individual with stroke increased their domestic FAI score by a differential of 0.14 points (significant at p=0.002) per month; compared to controls, who showed a nonsignificant decline of 0.03 points in their domestic FAI score (p=0.277). Much of the variability in social participation was associated with between-individual differences; the intraclass correlation coefficient was 0.86 meaning that 86% of variability happened among individuals and 14% of variability within-individuals. Since there was little variability that was associated with within-individual differences, multilevel modeling was not used (Tabachnick & Fidell, 2007). Hypothesis 2. As there was no meaningful change in social participation from 3 months to 12 months after baseline in both the stroke subjects and controls, multiple regression analyses were used to assess the relationship between baseline balance confidence and social participation at 12 months. Residuals plots revealed that our data satisfy all multiple regression assumptions (see Figure 13 in Appendix XXIX). Appendix XXX presents the details of outlier analyses. No outliers among the cases were found as determined by the use of Mahalanobis distance (the critical χ2 value for 12 df is 32.910;  88  p<0.001). Multicollinearity was not a concern as determined by tolerance, VIF and condition index (see Appendix XXXI for details). For continuous variables, bivariate analyses revealed that the correlations between balance confidence and FAI and between other covariates and FAI were significant (p<0.01 except for age). Table 3-6 shows the correlation matrix of these variables. Except for age, all correlations were between the absolute values of 0.2 and 0.7; age and FAI had a correlation of -0.169 which was outside of our criteria for being included in the regression analyses despite that it was significant at p<0.05. Therefore age was not included in the multiple regression analyses. For categorical variables, t test analyses revealed that there was a significant difference in FAI based on partnering status (p<0.01) and stroke status (p<0.001) but not sex (p=0.072); thus sex was not included in the multiple regression analyses. The predictors of social participation at 12 months are presented in Table 3-7. Balance confidence at baseline remained as a strong independent predictor of social participation after controlling for covariates (age, sex, partnering status, number of chronic conditions, perceived social support, stroke status, balance performance, basic mobility, walking capacity, anxiety trait and depression symptoms). The adjusted model reveals that a higher level of social participation is explained by having higher balance confidence, better balance performance, partnering status and a higher perceived level of social support. We found that there was a significant interaction effect between balance confidence and balance performance (f change=4.21, p=0.042). Figure 3-1 illustrates this effect. Since balance performance was involved in the interaction terms, BBS scores were collapsed into a binary variable by using a cutoff of 45 (out of 56) (Thorbahn & Newton, 1996) when used in a scatterplot of the association between balance confidence and social participation. This cutoff  89  score was selected to distinguish high and low levels of balance performance based on data from a previous study (Thorbahn & Newton, 1996). Individuals with low balance confidence demonstrated a smaller difference in their levels of social participation even if they had different levels of balance performance; whereas, in individuals with high balance confidence, those who had higher level of balance performance showed a much higher level of social participation as compared to individuals with lower balance performance. As shown in Table 3-7, covariates included in the final model included absence of a spousal partner and perceived social support. The adjusted R2 value of 0.483 indicated that 48.3% of the variability in social participation in our sample was explained by the combination of balance confidence, balance performance, the interaction term and covariates. 3.4 Discussion In this study, we assessed 12 month changes and differences in social participation in individuals with stroke and age- and sex-matched individuals without stroke and examined the relationship between balance confidence and social participation. We hypothesized that individuals with stroke would have a consistently lower level of social participation overall and in all the sub-domains of social participation as measured with the FAI than individuals without a stroke over 1 year. The results of our study failed to reject the null hypothesis. The level of social participation in our stroke sample remained lower by an average of 12.68 points on the FAI total score in individuals with stroke than those without stroke. Previously Schuling et al. (1993) reported a difference of 10.67 points on the FAI score between stroke and control subjects at 6 months post stroke onset. Our stroke sample’s mean FAI score at 1year post discharge from rehabilitation was 22.26±9.91 which is higher than various mean 1-year FAI scores reported in previous studies (range 12.6±9.8 to 18.4±9.4) (Horgan, O'Regan, Cunningham, & Finn, 2009; Schepers et al., 2005; 90  Sveen et al., 1999; Wade et al., 1985). The reasons for this could be related to the differences in timing of baseline measurement and source of recruitment between these studies and ours. Most of the subjects in these previous studies were recruited through acute care hospitals or stroke registry through general practitioners. Baseline measurements were taken at 2 to 3 weeks post stroke onset and therefore 1 year measurement would have been taken close to 1 year post stroke onset. Many of these individuals may not have had a chance to go through rehabilitation due to candidacy issues and may not have been able to return to community living. In Wade et al. (1985) and Sveeb et al. (1999)’s studies, 12% and 15.4% of their sample lived in a hospital or care facility. In our study we recruited individuals who had gone through rehabilitation and were being discharged into the community; thus they may in fact be higher functioning than individuals who participated in previous studies. Our study revealed little change in the level of social participation over 1 year in either individuals with stroke or the controls except change in participation was significant in the domestic sub-domain of the FAI in individuals with stroke. Previous studies found that change in participation appeared to vary according to the type of activity involved (Desrosiers et al., 2006; Lo et al., 2008). Lo et al. (2008) found an improvement in the mobility6 category of their measure of participation from 3 to 12 months. Desrosiers et al. (2006) found that there was a decline in personal care and fitness but social roles were stable from 6 months to between 2 to 4 years. These differences could be explained by the fact that different measures were used for measuring participation. Our measure FAI does not have a category specifically related to mobility but most of the activities measured by the FAI require the use of some form of mobility. The FAI also does not include a category of  6  Mobility was measured as “participation” in the London Handicap Scale because it measures how mobility restricts the individual getting around or where he/she wants to go.  91  personal care; however, the work/leisure and outdoor sub-domains of the FAI could be comparable to social roles. Change in participation post stroke appears to be complex; consistent and valid measures of participation are needed for future studies (Noonan, 2009). We also hypothesized that balance confidence would remain an independent predictor of social participation after controlling for age, sex, partnering status, number of chronic conditions, perceived social support, stroke status, balance performance, basic mobility, walking capacity, anxiety trait and depression symptoms. The results of our multiple regression analyses supported our hypothesis. Few studies (Dumont et al., 2005; Miller, Deathe, Speechley, & Koval, 2001) that we know of have used balance confidence to predict social participation in non-stroke populations; therefore, this is an important finding for the stroke population. We found that balance confidence interacted with balance performance, in predicting social participation. For individuals with low balance confidence the level of balance performance does not have an effect on their level of social participation; whereas, in individuals with high balance confidence, those who had high balance performance showed a much higher level of social participation as compared to those with lower balance performance. These findings further support the theoretical claim that beliefs of personal efficacy (assessed as balance confidence in this study) are active contributors to behaviors (Bandura, 1977; Bandura, 1997). According to Bandura (1997), effective functioning requires both skills and efficacy beliefs to use the skills well. These beliefs are not just beliefs of having the sub-skills required to perform an activity but are beliefs of one’s capability to organize the skills and execute them under varied circumstances and in different contexts. According to the ICF, participation is the actual performance of a task in an individual’s current environment which brings in the context and could include all aspects of  92  the physical, social and attitudinal environment (World Health Organization, 2001). If efficacy beliefs are crucial to executing skills well, then having better balance confidence could enhance an individual’s performance of activities in real life situations or so called participation. Individuals who have a strong sense of self-efficacy are more motivated, committed to engage in the activity; therefore, together with skills, they are more likely to succeed. It is therefore equally important to treat reduced balance confidence as well as poor balance performance to prevent decline in social participation. The generalizability of the results of this study have several limitations. This study was carried out with individuals who have had a stroke and who were able to go through a short course of inpatient rehabilitation. Those individuals who had had a mild stroke and were discharged home directly from acute care were not included. On the other hand, individuals with more severe deficits who were cognitively impaired or had severe communication difficulties were not included. Also only those individuals who were ambulatory were included. Over 70% of our sample was men and this could have affected how psychological measures such as depression symptoms, anxiety state and even balance confidence were reported. There might be volunteer bias as our subjects were recruited through advertisement and therapist working with individuals with stroke. Among the participants who dropped out of the study, approximately 14% of them had specific reasons such as death or health reasons. Reasons for the majority others who dropped out of the study included: cancelled/missed appointments or decline to participate. It is possible that the dropouts had difficulty participating in the study either because they had difficulty getting to the appointments which could include reasons directly related to their  93  level of social participation. If this was the case, our missing data were not missing at random, and our results may be biased. Participation is a multidimensional concept (Noonan, 2009). How the concept of participation should be operationalized is still evolving (Noonan, 2009). Our measure of social participation includes mainly a measure of the frequency of engagement in activities. It does not measure difficulties, limitation, autonomy, satisfaction or importance. Noonan (2009) suggested that measuring these concepts will enhance the understanding of how participation is conceptualized and its relationship with quality of life or subjective well being. Social participation was not measured at discharge from rehabilitation or pre-stroke FAI score was not obtained (Wade et al., 1985). Pre-stroke FAI score has been shown to have a strong correlation with FAI score at 1 year (Wade et al., 1985). This would have also provided valuable information about how social participation changed since the onset of stroke compared to the control group. However one drawback would be that individuals would have no chance to experience what they would do after discharge before measuring, since the FAI asked questions related to activities in the past 3 and 6 months. Although we found statistically significant relationships between predictors and social participation, we cannot claim a causal relationship. In our study we could only describe the extent, direction and strength of the relationship between significant predictors and social participation because the temporal relationship between predictors and outcome variables was uncertain. However, with this being a longitudinal study that included a control group, the credibility of our findings is enhanced.  94  In conclusion, participation restriction is a serious consequence of stroke and it is not stable at the time of discharge from rehabilitation. There is a trend for participation to remain at a low level or even decline over time in the long term. This study has identified balance confidence as an important predictor of social participation. Together with balance performance, balance confidence has a strong impact on the level of social participation. The findings of this study suggest that it is equally important to provide interventions for reduced balance confidence as well as treat poor balance performance early on post stroke to facilitate a higher level of participation and prevent functional decline in individuals with stroke.  95  Table 3-1 Mean Scores of Total FAI and Sub-domains of FAI from 3 to 12 Months after Baseline in the Stroke and Control Groups FAI  3 Months  6 Months  12 Months  Mean(SD)  Mean(SD)  Mean(SD)  Stroke (n=85)  Stroke (n=70)  Stroke (n=70)  Control (n=92)  Control (n=80)  Control (n=80)  Stroke  7.42(4.44)  7.51(4.55)  8.61(4.65)  Control  11.72(3.37)  11.49(3.53)  11.39(3.63)  Stroke  5.32(3.13)  5.75(3.13)  5.77(3.26)  Control  8.93(2.78)  8.83(3.06)  8.88(3.21)  Stroke  7.69(3.12)  8.23(3.28)  8.14(3.53)  Control  12.24(2.15)  11.98(1.99)  12.31(2.06)  Stroke  20.46(8.88)  21.24(8.98)  22.26(9.91)  Control  32.89(6.29)  32.08(7.08)  32.54(6.70)  Domestic  Leisure/Work  Outdoors  Total score  96  Table 3-2 Results of Fitting the Unconditional Means Model (UMM), the Unconditional Growth Model (UGM) and the Final Model to Data with Total FAI Score as the Dependent Variable (Appendix XXV presents selected SPSS outputs for these models) Fixed Effects Composite model Level 2 predictor Level 1 predictor  UMM  UGM  Final Model  27.03***  26.58***  32.62***  Stroke status  -  -  -12.68***  Rate of change, 𝜋𝜋1𝑖𝑖 (months post baseline)  -  0.07 (p=0.168)  -0.02 (p=0.761)  -  -  0.19 (p=0.062)  13.82***  11.91***  11.99***  82.75***  86.39***  46.16***  -  0.09 (p=0.105)  0.08 (p=0.150)  -  -0.532 (p=0.381)  0.08 (p=0.867)  Intercept, 𝜋𝜋0𝑖𝑖 (initial status)  Stroke status X months post baseline Variance Components Level 1 Level 2  Within-person, 𝜀𝜀𝑖𝑖𝑖𝑖  In initial status, 𝜁𝜁0𝑖𝑖  In rate of change, 𝜁𝜁1𝑖𝑖 ***p< 0.001  Covariance, 𝜁𝜁0𝑖𝑖 and 𝜁𝜁1𝑖𝑖  97  Table 3-3 Results of Fitting the Unconditional Means Model (UMM), the Unconditional Growth Model (UGM) and the Final Model to Data with Domestic Sub-domain FAI Score as the Dependent Variable (Appendix XXVI presents selected SPSS outputs for these models) Fixed Effects Composite model Level 2 predictor Level 1 predictor  UMM  UGM  Final Model  9.70***  9.47***  11.78***  Stroke status  -  -  -4.84***  Rate of change, 𝜋𝜋1𝑖𝑖 (months post baseline)  -  0.03 (p=0.120)  -0.03 (p=0.277)  -  -  0.14 (p=0.002)  3.07***  3.01***  2.89***  16.85***  17.64***  12.85***  -  0.02 (p=0.852)  0.00 (p=0.619)  -  -0.07 (p=0.581)  -  Intercept, 𝜋𝜋0𝑖𝑖 (initial status)  Stroke status X months post baseline Variance Components Level 1 Level 2  Within-person, 𝜀𝜀𝑖𝑖𝑖𝑖  In initial status, 𝜁𝜁0𝑖𝑖  In rate of change, 𝜁𝜁1𝑖𝑖 ***p< 0.001  Covariance, 𝜁𝜁0𝑖𝑖 and 𝜁𝜁1𝑖𝑖  98  Table 3-4 Results of Fitting the Unconditional Means Model (UMM), the Unconditional Growth Model (UGM) and the Final Model to Data with Leisure/Work Sub-domain FAI Score as the Dependent Variable (Appendix XXVII presents selected SPSS outputs for these models) Fixed Effects Composite model Level 2 predictor Level 1 predictor  UMM  UGM  Final Model  7.29***  7.17***  8.90***  Stroke status  -  -  -3.62***  Rate of change, 𝜋𝜋1𝑖𝑖 (months post baseline)  -  0.02 (p=0.484)  -0.01 (p=0.876)  -  -  0.04 (p=0.369)  2.99***  2.37***  2.37***  9.22***  10.12***  6.86***  -  0.03 (p=0.011)  0.03 (p=0.012)  -  -0.15 (p=0.163)  -0.11 (p=0.262)  Intercept, 𝜋𝜋0𝑖𝑖 (initial status)  Stroke status X months post baseline Variance Components Level 1 Level 2  Within-person, 𝜀𝜀𝑖𝑖𝑖𝑖  In initial status, 𝜁𝜁0𝑖𝑖  In rate of change, 𝜁𝜁1𝑖𝑖 ***p< 0.001  Covariance, 𝜁𝜁0𝑖𝑖 and 𝜁𝜁1𝑖𝑖  99  Table 3-5 Results of Fitting the Unconditional Means Model (UMM), the Unconditional Growth Model (UGM) and the Final Model to Data with Outdoors Sub-domain FAI Score as the Dependent Variable (Appendix XXVIII presents selected SPSS outputs of these models) Fixed Effects Composite model Level 2 predictor Level 1 predictor  UMM  UGM  Final Model  10.17***  10.02***  12.07***  Stroke status  -  -  -4.35***  Rate of change, 𝜋𝜋1𝑖𝑖 (months post baseline)  -  0.02 (p=0.256)  -0.01 (p=0.645)  -  -  0.02 (p=0.600)  2.37***  2.04***  2.04***  9.50***  9.73***  4.97***  -  0.02 (p=0.084)  0.02 (p=0.086)  -  -0.06 (p=0.491)  -0.04 (p=0.606)  Intercept, 𝜋𝜋0𝑖𝑖 (initial status)  Stroke status X months post baseline Variance Components Level 1 Level 2  Within-person, 𝜀𝜀𝑖𝑖𝑖𝑖  In initial status, 𝜁𝜁0𝑖𝑖  In rate of change, 𝜁𝜁1𝑖𝑖 ***p< 0.001  Covariance, 𝜁𝜁0𝑖𝑖 and 𝜁𝜁1𝑖𝑖  100  Table 3-6 Correlation Matrix of Continuous Predictor Variables and Outcome Variable  FAI Age # of Chronic Conditions  6MWT BBS ABC TUG CES-D STAI ISEL  FAI  Age  # of Chronic Conditions  6MW  BBS  ABC  TUG  CES-D  STAI  ISEL  1.000  -0.165*  -0.367**  0.604**  0.630**  0.601**  -0.522**  -0.377**  -0.352**  0.225**  1.000  0.269**  -0.209**  -0.185**  -0.119  0.063  -0.139  -.080  -0.043  1.000  -0.490**  -0.356**  -0.428**  0.234**  0.242**  0.255**  -0.110  1.000  0.812**  0.783**  -0.714**  -0.505**  -0.419**  0.129  1.000  0.743**  -0.804**  -0.413**  -0.325**  0.102  1.000  -0.646**  -0.572**  -0.498**  0.193**  1.000  0.359**  0.217**  -0.009  1.000  0.580**  -0.377**  1.000  -0.288** 1.000  **p<0.01 level (2-tailed) *p<0.05 (2 tailed)  101  Table 3-7  Multiple Regression of Social Participation (FAI) at 12 Months on Balance Confidence Unadjusted Model (n=148)  Adjusted Model (n=145)  B  Standard Error  β  95%CI  B  Standard Error  β  95%CI  p  ABC  0.135  0.037  0.314  0.062, 0.209  0.119  0.037  0.275  0.045, 0.192  0.002  ABC X BBS  0.006  0.003  0.170  0.000, 0.012  0.007  0.003  0.189  0.001, 0.013  0.023  BBS  0.713  0.139  0.525  0.438, 0.989  0.742  0.137  0.548  0.471, 1.013  <0.001  3.577  1.290  0.169  1.027, 6.128  0.006  0.636  0.295  0.137  0.054, 1.219  0.033  Predictors  Absence of a spousal partner ISEL Adjusted R2  0.453  0.483  Excluded variables: 6 Minute Walk Test, Timed “Up and Go” Test, Center for Epidemiological Studies – Depression scale, number of chronic conditions, stroke status and Trait Anxiety component of the State Trait Anxiety Inventory  102  Figure 3-1 Scatter Plot Illustrating the Interaction Effect of Balance Performance on the Association between Balance Confidence and Social Participation  103  Reference List Andersson, A. G., Kamwendo, K., Seiger, A., & Appelros, P. (2006). How to identify potential fallers in stroke unit: Validity indexes of four test methods. Journal of Rehabilitation Medicine, 38, 186-191. Bandura, A. (1977). Self-efficacy: Toward a unifying theory of behavioral change. Psychological Reveiw, 84, 191-215. Bandura, A. (1997). Self-efficacy. the exercise of control. New York: W.H. Freeman and Company. Barnes, L. L. B., Harp, D., & Jung, W. S. (2002). Reliability generalization of scores on the spielberger state-trait anxiety inventory. Educational and Psychological Measurement, 62, 603-618. Beckley, M. N. (2006). Community participation following cerebrovascular accident: Impact of the buffering model of social support. American Journal of Occupational Therapy, 60(2), 129-135. Berg, K., Wood-Dauphinee, S., Williams, J. I., & Gayton, D. (1989). Measuring balance in the elderly: Preliminary development of an instrument. Physiotherapy Canada, 41, 304311. Berg, K. O., Wood-Dauphinee, S. L., Williams, J. I., & Maki, B. (1992). Measuring balance in the elderly: Validation of an instrument. Canadian Journal of Public Health.Revue Canadienne De Sante Publique, 83(Suppl 2:S7-11), JuAug. Blum, L., & Korner-Bitensky, N. (2008). Usefulness of the berg balance scale in stroke rehabilitation: A systematic review. Physical Therapy, 88(5), 559-566.  104  Botner, E. M., Miller, W. C., & Eng, J. J. (2005). Measurement properties of the activitiesspecific balance confidence scale among individuals with stroke. Disability & Rehabilitation, 27(4), 156-163. Bukov, A., Maas, I., & Lampert, T. (2002). Social participation in very old age: Crosssectional and longitudinal findings from BASE. Journal of Gerontology: Psychological Sciences, 57B(6), 510-517. Cohen, S., & Hoberman, H. M. (1983). Positive events and social supports as buffers of life change stress. Journal of Applied Social Psychology, 12(2), 99-125. Cohen, S., Mermelstein, R., Kamarck, T., & Hoberman, H. M. (1985). Measuring the functional components of social supports. In I. G. Sarason, & B. R. Sarason (Eds.), Social support: Theory, research, and application (pp. 73-94). The Netherlands: Martinus Nijhoff. Cumming, R. G., Salkeld, G., Thomas, M., & Szonyi, G. (2000). Prospective study of the impact of fear of falling on activities of daily living, SF-36 scores, and nursing home admission. Journals of Gerontology Series A-Biological Sciences & Medical Sciences, 55(5), M299-305. Dahan-Oliel, N., Gelinas, I., & Mazer, B. (2008). Social participation in the elderly: What does the literature tell us? Critical Review in Physical & Rehabilitation Medicine, 20(2), 159-176. Desrosiers, J., Demers, L., Robichaud, L., Vincent, C., Belleville, S., Ska, B., et al. (2008). Short-term changes in and predictors of participation of older adults after stroke following acute care or rehabilitation. Neurorehabilitation & Neural Repair, 22(3), 288297.  105  Desrosiers, J., Noreau, L., Rochette, A., Bourbonnais, D., Bravo, G., & Bourget, A. (2006). Predictors of long-term participation after stroke. Disability & Rehabilitation, 28(4), 221-230. Desrosiers, J., Noreau, L., Rochette, A., Bravo, G., & Boutin, C. (2002). Predictors of handicap situations following post-stroke rehabilitation. Disability and Rehabilitation, 24(15), 774-785. Desrosiers, J., Rochette, A., Noreau, L., Bourbonnais, D., Bravo, G., & Bourget, A. (2006). Long-term changes in participation after stroke. Topics in Stroke Rehabilitation, 13(4), 86-96. Dumont, C., Gervais, M., Fougeyrollas, P., & Bertrand, R. (2005). Perceived self-efficacy as a factor associated with the social participation of adults who have sustained a traumatic brain injury. Canadian Journal of Occupational Therapy, 72(4), 222-233. Eng, J. J., Dawson, A. S., & Chu, K. (2004). Submaximal exercise in persons with stroke: Test-retest reliability and concurrent validity with maximal oxygen consumption. Archives of Physical Medicine & Rehabilitation, 85(1), 113-118. Green, J., Forster, A., & Young, J. (2001). A test-retest reliability study of the barthel index, the rivermead mobility index, the nottingham extended activities of daily living scale and the frenchay activities index in stroke patients. Disability & Rehabilitation, 23(15), 670-676. Guyatt, G. H., Pugsley, S. O., Sullivan, M. J., Thompson, P. J., Berman, L., Jones, N. L., et al. (1984). Effect of encouragement on walking test performance. Thorax, 39(11), 818822.  106  Guyatt, G. H., Sullivan, M. J., Thompson, P. J., Fallen, E. L., Pugsley, S. O., Taylor, D. W., et al. (1985). The 6-minute walk : A new measure of exercise capacity in patients with chronic heart failure. Canadian Medical Association Journal, 132, 919-923. Guyatt, G. H., Thompson, P. J., Berman, L. B., Sullivan, M. J., Townsend, M., Jones, N. L., et al. (1985). How should we measure function in patients with chronic heart and lung disease? Journal of Chronic Disease, 38(6), 517-524. Hartman-Maeir, A., Soroker, N., Ring, H., Avni, N., & Katz, N. (2007). Activities, participation and satisfaction one-year post stroke. Disability & Rehabilitation, 29(7), 559-566. Holbrook, M., & Skilbeck, C. E. (1983). An activities index for use with stroke patients. Age and Ageing, 12, 166-170. Horgan, N. F., O'Regan, M., Cunningham, C. J., & Finn, A. M. (2009). Recovery after stroke: A 1-year profile. Disability & Rehabilitation, 31(10), 831-839. Hotchkiss, A., Fisher, A., Robertson, R., Ruttencutter, A., Schuffert, J., & Barker, D. B. (2004). Convergent and predictive validity of three scales related to falls in the elderly. American Journal of Occupational Therapy, 58(1), 100-103. Isaacs, B., Neville, Y., & Rushford, I. (1976). The stricken: The social consequences of stroke. Age & Ageing, 5(3), 188-192. Kleinbaum, D. G., Kupper, L. L., Muller, K. E., & Nizam, A. (2008). Applied regression analysis and other multivariable methods (4th ed.). Belmont, CA: Thomson Brooks/Cole Publishing.  107  Kosak, M., & Smith, T. (2005). Comparison of the 2-, 6-, and 12-minute walk tests in patients with stroke. Journal of Rehabilitation Research and Development, 42(1), 103107. Lawrence, L., & Christie, D. (1979). Quality of life after stroke: A three-year follow-up. Age & Ageing, 8(3), 167-172. Lin, M. R., Hwang, H. F., Hu, M. H., Wu, H. D. I., Wang, Y. W., & Huang, F. C. (2004). Psychometric comparisons of the timed up and go, one-leg stand, functional reach, and tinetti balance measures in community-dwelling older people. Journal of American Geriatrics Society, 52, 1343-1348. Lo, R. S., Cheng, J. O., Wong, E. M., Tang, W. K., Wong, L. K., Woo, J., et al. (2008). Handicap and its determinants of change in stroke survivors: One-year follow-up study. Stroke, 39(1), 148-153. Mathias, S., Nayak, U. S., & Isaacs, B. (1986). Balance in elderly patients: The "get-up and go" test. Archives of Physical Medicine and Rehabilitation, 67(6), 387-389. Mayo, N. E., Wood-Dauphinee, S., Cote, R., Durcan, L., & Carlton, J. (2002). Activity, participation, and quality of life 6 months poststroke. Archives of Physical Medicine & Rehabilitation, 83(8), 1035-1042. Miller, W. C., Deathe, A. B., Speechley, M., & Koval, J. (2001). The influence of falling, fear of falling, and balance confidence on prosthetic mobility and social activity among individuals with a lower extremity amputation. Archives of Physical Medicine & Rehabilitation, 82(9), 1238-1244. Niemi, M. L., Laaksonen, R., Kotila, M., & Waltimo, O. (1988). Quality of life 4 years after stroke. Stroke, 19(9), 1101-1107.  108  Noonan, V. K. (2009). A comparison of participation instruments based on the international classification of functioning, disability and health. Unpublished Doctor of Philosophy, University of British Columbia, Vancouver. Orbon, K. H., Satink, T., de Jong, T., & van der Gulden, J. W. J. (2009). Quality of social participation of older people: A key to happiness? Nederlands Tijdschrift Voor Fysiotherapie, 119(4), 136-143. Podsiadlo, D., & Richardson, S. (1991a). The time "UP & go": A test of basic functional mobility for frail elderly persons. Journal of American Geriatrics Society, 39(2), 142148. Podsiadlo, D., & Richardson, S. (1991b). The timed "up and go": A test of basic functional mobility for frail elderly persons. Journal of the American Geriatrics Society, 39(2), 142-148. Portney, L. G., & Watkins, M. P. (Eds.). (2000). Foundations of clinical research: Applications to practice (second ed.). New Jersey: Prentice-Hall, Inc. Post, M. W., & de Witte, L. P. (2003). Good inter-rater reliability of the frenchay activities index in stroke patients. Clinical Rehabilitation, 17(5), 548-552. Powell, L. E., & Myers, A. M. (1995). The activities-specific balance confidence (ABC) scale. Journals of Gerontology Series A-Biological Sciences & Medical Sciences, 50A(1), M28-34. Radloff, L. S. (1977). The CES-D scale: A self-report depression scale for research in the general population. Applied Psychological Measurement, 1(3), 385-401. Radloff, L. S., & Teri, L. (1986). Use of the center for epidemiological studies-depression scale with older adults. Clinical Gerontologist, 5(1/2), 119-135.  109  Robinson, R. G., Bolduc, P. L., Kubos, K. L., Starr, L. B., & Price, T. R. (1985). Social functioning assessment in stroke patients. Archives of Physical Medicine & Rehabilitation, 66(8), 496-500. Roth, E. J., & Lovell, L. (2007). Community skill performance and its association with the ability to perform everyday tasks by stroke survivors one year following rehabilitation discharge. Topics in Stroke Rehabilitation, 14(1), 48-56. Salbach, N. M., Mayo, N. E., Higgins, J., Ahmed, S., Finch, L. E., & Richards, C. L. (2001). Responsiveness and predictability of gait speed and other disability measures in acute stroke. Archives of Physical Medicine and Rehabilitation, 82, 1204-1212. Salbach, N. M., Mayo, N. E., Hanley, J. A., Richards, C. L., & Wood-Dauphinee, S. (2006). Psychometric evaluation of the original and canadian french version of the activitiesspecific balance confidence scale among people with stroke. Archives of Physical Medicine & Rehabilitation, 87, 1597-1604. Salbach, N. M., Mayo, N. E., Robichaud-Ekstrand, S., Hanley, J. A., Richards, C. L., & Wood-Dauphinee, S. (2006). Balance self-efficacy and its relevance to physical function and perceived health status after stroke. Archives of Physical Medicine & Rehabilitation, 87(3), 364-370. Salter, K. L., Foley, N. C., Jutai, J. W., & Teasell, R. W. (2007). Assessment of participation outcomes in randomized controlled trials of stroke rehabilitation interventions. International Journal of Rehabilitation Research, 30(4), 339-342. Schepers, V. P. M., Ketelaar, M., Visser-Meily, J. M. A., Dekker, J., & Lindeman, E. (2006). Responsiveness of functional health status measures frequently used in stroke research. Disability and Rehabilitation, 28(17), 1035-1040.  110  Schepers, V. P., Visser-Meily, A. M., Ketelaar, M., & Lindeman, E. (2005). Prediction of social activity 1 year poststroke. Archives of Physical Medicine & Rehabilitation, 86(7), 1472-1476. Schuling, J., de Haan, R., Limburg, M., & Groenier, K. H. (1993). The frenchay activities index. assessment of functional status in stroke patients. Stroke, 24(8), 1173-1177. Shinar, D., Gross, C. R., Price, T. R., Banko, M., Bolduc, P. L., & Robinson, R. G. (1986). Screening for depression in stroke patients: The reliability and validity of the center for epidemiologic studies depression scale. Stroke, 17(2), 241-245. Singer, J. D., & Willett, J. B. (2003). Applied longitudinal data analysis. modeling change and event occurrence. New York: Oxford University Press, Inc. Sorensen, L. V., Axelsen, U., & Avlund, K. (2002). Social participation and functional ability from age 75 to age 80. Scandinavian Journal of Occupational Therapy, 9, 71-78. Spielberger, C. D. (1983). Manual for the state-trait anxiety inventory (STAI). PaloAlto, CA: Consulting Psychologists Press. Spielberger, C. D. (1984). In Spielberger C. D., Vagg P. R. (Eds.), Test anxiety. theory, assessment, and treatment. Bristol, PA: Taylor & Francis. Sveen, U., Bautz-Holter, E., Sodring, K. M., Wyller, T. B., & Laake, K. (1999). Association between impairments, self-care ability and social activities 1 year after stroke. Disability & Rehabilitation, 21(8), 372-377. Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics (5th ed.). Boston, MA: Pearson Education, Inc.  111  Talley, K. M., Wyman, J. F., & Gross, C. R. (2008). Psychometric properties of the activities-specific balance confidence scale and the survey of activities and fear of falling in older women. Journal of the American Geriatrics Society, 56(2), 328-333. Teoh, V., Sims, J., & Milgrom, J. (2009). Psychosocial predictors of quality of life in a sample of community-dwelling stroke survivors: A longitudinal study. Topics in Stroke Rehabilitation, 16(2), 157-166. Thorbahn, L. D. B., & Newton, R. A. (1996). Use of the berg balance test to predict falls in elderly persons. Physical Therapy, 76(6), 576-583. Tinetti, M. E., Mendes de Leon, C. F., Doucette, J. T., & Baker, D. I. (1994). Fear of falling and fall-related efficacy in relationship to functioning among community-living elders. Journal of Gerontology, 49(3), M140-7. Turnbull, J. C., Kersten, P., Habib, M., McLellan, L., Mullee, M. A., & George, S. (2000). Validation of the frenchay activities index in a general population aged 16 years and older. Archives of Physical Medicine & Rehabilitation, 81(8), 1034-1038. Wade, D. T., Legh-Smith, J., & Langton Hewer, R. (1985). Social activities after stroke: Measurement and natural history using the frenchay activities index. International Rehabilitation Medicine, 7(4), 176-181. World Health Organization. (2001). International Classification of Functioning, Disability and Health:ICF. Geneva: World Health Organization, 2001.  112  Chapter Four: Discussion, Conclusions and Future Direction 4.1 Overview The overall purpose of this study was to examine the natural history of balance confidence and its relationship with social participation in individuals with stroke over a 1 year period after discharge from rehabilitation. First we compared how balance confidence changed over 1 year in individuals with stroke to matched individuals without stroke. Then we examined whether stroke status was an important predictor of balance confidence and explored other important factors affecting balance confidence in individuals with stroke. Next, we also examined how social participation changed over 1 year in individuals with stroke compared to individuals without stroke and examined the relationship between balance confidence and social participation to determine if balance confidence was an important predictor of social participation. In this chapter we will draw research conclusions, discuss clinical implications of our research findings, propose new ideas for future research and comment on the strengths and limitations of our study. 4.2 Reduced Balance Confidence Is a Persistent and Serious Problem in Individuals with Stroke The findings from chapter 2 confirmed our hypotheses; balance confidence remained significantly lower in individuals with stroke over 1 year. Moreover, stroke status remained as a significant risk factor for reduced balance confidence even after controlling for covariates, and the association between stroke status and balance confidence was dependent on the level of depression symptoms and walking capacity. There was a slow natural improvement in balance confidence over 1 year in individuals with stroke; however, this natural improvement was not clinically meaningful, and reduced balance confidence proves to be a persistent and serious problem in individuals with stroke. Therefore, it is suggested 113  that assessment of balance confidence be incorporated into inpatient rehabilitation. Since balance confidence is a remedial condition (Li et al., 2005; Salbach et al., 2005), it is suggested that interventions be provided to those with reduced balance confidence. Our chapter 2 findings also have important implications for designing interventions for reduced balance confidence in individuals with stroke. Since we found that the association between stroke status and balance confidence is modified by the level of walking capacity, and since walking capacity is a modifiable factor, interventions for reduced balance confidence for individuals with stroke may involve providing experience of personal mastery in activities that demand walking capacity (distance walked or walking speed). Previously, Salbach et al. (2005) ’s study found that change in balance confidence was associated with change in walking capacity in individuals with stroke (within 1 year of first stroke). Moreover, we found that the effect of stroke on an individual’s balance confidence is also dependent on the frequency and amount of depression symptoms. It is therefore suggested that interventions for reduced balance confidence be used as an adjunct to treatment of depression in individuals with stroke. It may be that improvement in balance confidence would be associated with decreased depression symptoms. Future research should investigate the effectiveness of specific interventions for reduced balance confidence in individuals with stroke in the post rehabilitation phase of stroke recovery (like our study sample) using randomized controlled trials. A current systematic review (Rand, Miller, Yiu, & Eng, in progress) of randomized controlled trials for improving balance confidence found that Tai Chi has the largest effect size; other mixed exercises and multi-factorial treatment have a small effect size. Some of these treatments included a walking program (e.g. going for a walk 2 to 3 times a week for 1 year) (Campbell  114  et al., 1997; Robertson, Devlin, Gardner, & Campbell, 2001). Although these trials did not focus on treating individuals with stroke, it would be interesting to find out how Tai Chi works in comparison to program that involves improving walking capacity in individuals with stroke. 4.3 Having Better Balance Confidence May Enhance an Individual’s Social Participation The findings from chapter 3 revealed that there was little change in the level of overall social participation over 1 year in both individuals with stroke and individuals without stroke. Our hypotheses were supported: individuals with stroke had consistently lower level of social participation over 1 year than individuals without stroke; baseline balance confidence remained an independent risk factor for restrictions in social participation at 1 year even after controlling for covariates. We found that baseline balance confidence was the second most important predictor of social participation at 1 year. Moreover balance confidence interacted with balance performance in predicting social participation. For individuals with low balance confidence, the level of balance performance does not appear to influence their level of social participation; whereas, in individuals with high balance confidence, those who had high balance performance showed a much higher level of social participation as compared to those with lower balance performance. These findings suggest that for individuals who have low balance confidence, improving balance confidence may be equally important as improving balance performance in order to optimize social participation. Since reduced balance confidence is a persistent and serious problem in individuals with stroke (findings from chapter 2), it is suggested that interventions for reduced balance confidence be incorporated  115  into rehabilitation program that addresses balance performance issues in individuals with stroke. Future research should investigate the effect of incorporating an intervention for reduced balance confidence in treating individuals with stroke who have balance or mobility problems and observe their level of social participation. An experimental design should be used. The intervention for reduced balance confidence can be implemented as early as when these individuals are in an inpatient rehabilitation program, for example, in physiotherapy sessions. The long term effects of the intervention should be assessed in a longitudinal study in order to fully assess social participation. A reliable and valid measure of social participation like ours should be used. 4.4 Strengths and Limitations The strengths of this study lie in the research design and method of analyses. The study was a longitudinal study with an age- and sex- matched control group. This design allowed us to collect data on growth and change on the same individuals, understand how changes occur naturally and answer the types of research questions we wanted to ask. Since balance confidence, social participation and many variables of interests are complex constructs, using experimental approaches to establish cause-and-effect relationships among them can be expensive. Therefore if there is natural improvement or some kind of investigator influences on the outcome, then an observational method like ours can be of an advantage. The use of multivariate analyses enabled us to systematically investigate the relationships among these variables and to predict the effect of one variable on another. The presence of a control group enhanced the credibility of our findings. In terms of limitations, subjects were tested repeatedly and testing effects may be present. Due to long data collection time (1 year), some data were missing due to attrition. 116  The fact that dropouts had lower balance confidence means that survival bias may be present; therefore, our results are conservative and likely underestimate the degree of the problem of reduced balance confidence in stroke. There is also a possibility that some missing data were not missing at random because some of the reasons for dropping out of the study could include reasons directly related to the dropouts’ level of social participation. If this was the case, our results may be biased. There could be confounding variables that may affect changes over time; however, we did well in trying to control for the effects of confounders. Another limitation is the limited generalizability of our findings. Our sample likely represents a group of higher functioning individuals with stroke who were cognitively intact, able to communicate and walk. So our results might be conservative and under-estimate the problem of reduced balance confidence in the stroke population. Therefore, our findings cannot be generalized to individuals who are wheelchair dependent, who have cognitive or communication deficits. 4.5 Significance of the Study and Recommendations for Practice Individuals with new stroke often require intensive and ongoing rehabilitation. Many factors could affect an individual’s natural recovery and benefits from rehabilitation. Psychological factors such as balance confidence have not been fully explored with the stroke population. Our study contributed significantly to the understanding of how important balance confidence is and its effect on social participation, an important rehabilitation outcome in individuals with stroke. Our study would raise the awareness of clinicians working with individuals with stroke the need to incorporate assessment and treatment for reduced balance confidence into their rehabilitation regime. It is recommended that the assessment of balance confidence for individuals with stroke be initiated in inpatient rehabilitation. This will assist clinicians identify those who have reduced balance confidence 117  or those at risk for developing this problem. Individuals’ level of balance confidence should be monitored throughout the inpatient stay as they receive treatments for balance performance or mobility issues. As individuals with stroke are discharged into the community, specific interventions for reduced balance confidence should be initiated in the form of home-based or community-based program for those who continue to have this problem. Since having better balance confidence may enhance social participation, it is important that follow up in this area be done by therapist working in the community. Today there is a trend in health care to provide interventions that are based on the concept of self-efficacy (Jones, 2006; Jones, Mandy, & Partridge, 2009; Kendall et al., 2007). The Chronic Disease Self-Management Program in British Columbia is one good example; another example would be the Peer Mentorship Program of the Stroke Recovery Association of British Columbia. These programs have received great acceptance in the clinical community. However, many clinicians are not yet familiar with the role that self-efficacy can play in specific aspect of stroke recovery, such as what we have been discussing in this thesis. It is therefore important to introduce the concept of balance confidence to the clinical community so that they will understand its impact on rehabilitation. In addition, education on the assessment of balance confidence and intervention methods should be provided to facilitate knowledge translation.  118  Reference List Campbell, A. J., Robertson, M. C., Gardner, M. M., Norton, R. N., Tilyard, M. W., & Buchner, D. M. (1997). Randomised controlled trial of a general practice programme of home based exercise to prevent falls in elderly women. British Medical Journal, 315(7115), 1065-1069. Jones, F. (2006). Strategies to enhance chronic disease self-management: How can we apply this to stroke? Disability & Rehabilitation, 28(13-14), 841-847. Jones, F., Mandy, A., & Partridge, C. (2009). Changing self-efficacy in individuals following a first time stroke: Preliminary study of a novel self-management intervention. Clinical Rehabilitation, 23(6), 522-533. Kendall, E., Catalano, T., Kuipers, P., Posner, N., Buys, N., & Charker, J. (2007). Recovery following stroke: The role of self-management education. Social Science & Medicine, 64(3), 735-746. Li, F., Harmer, P., Fisher, K. J., McAuley, E., Chaumeton, N., Eckstrom, E., et al. (2005). Tai chi and fall reductions in older adults: A randomized controlled trial. Journal of Gerontology: Medical Sciences, 60A(2), 187-194. Rand, D., Miller, W.C., Yiu, J., & Eng, J.J. (in progress). Interventions for addressing low balance confidence in older adults: a systematic review and meta-analysis. Robertson, M. C., Devlin, N., Gardner, M. M., & Campbell, A. J. (2001). Effectiveness and economic evaluation of a nurse delivered home exercise programme to prevent falls. randomised controlled trial. British Medical Journal, 322, 697-701.  119  Salbach, N. M., Mayo, N. E., Robichaud-Ekstrand, S., Hanley, J. A., Richards, C. L., & Wood-Dauphinee, S. (2005). The effect of a task-oriented walking intervention on improving balance self-efficacy poststroke: A randomized, controlled trial. Journal of the American Geriatrics Society, 53(4), 576-582.  120  APPENDICES Appendix I: Ethics Certificate  121  Appendix II: Psychometric Information of Outcome Measures Used in the Study Note: unless specified all studies involve individuals with stroke Name of measure  Reliability  Validity/Other information  Berg Balance Scale  α = 0.92-0.98;  excellent correlations with Barthel Index,  (BBS) (Berg et al.,  interrater (ICCs =  the Postural Assessment Scale for Stroke  1989; Berg et al.,  0.95-0.98); intra-rater  Patients, Functional Reach Test, the  1992)  (ICC = 0.97), test-  balance subscale of Fugl-Meyer  retest(ICC = 0.98)  Assessment, the Functional Independence  (Blum & Korner-  Measure, the Rivermead Mobility Index  Bitensky, 2008)  and gait speed (Blum & Korner-Bitensky, 2008); SRM=1.04 (Salbach et al., 2001)  Timed Up & Go  inter/intra-rater  correlated with BBS (r=-0.81), gait speed  (TUG) (Mathias et  (ICC=0.99) (geriatric  (r=-0.61) and Barthel Index of ADL (r=-  al., 1986; Podsiadlo  outpatients)  0.78) (geriatric outpatients) (Podsiadlo &  & Richardson, 1991)  (Podsiadlo &  Richardson, 1991); good discriminant  Richardson, 1991)  validity (community dwelling elderly) (Lin et al., 2004); SRM= 0.73 (Salbach et al., 2001)  Six Minutes Walk  Inter-rater (ICC=0.78);  correlated to BBS (r = 0.78), Chedoke-  (Guyatt et al., 1984;  intra-rater (ICC=0.74)  McMaster stroke impairment score  Guyatt et al., 1985)  (Kosak & Smith, 2005) (r=0.76) (Eng et al., 2002); responsiveness to change (SRM=1.52) (Kosak & Smith, 2005)  Activities-specific  Test/retest (ICC=0.85);  correlated with BBS (ρ = 0.36) and gait  Balance Confidence  α=0.94 (Botner et al.,  speed (ρ = 0.48) (Botner et al., 2005); (ρ =  Scale (ABC) (Powell  2005); α= 0.94  0.3 to 0.6) (Salbach et al., 2006)  & Myers, 1995)  122  Appendix II: Psychometric Information of Outcome Measures Used in the Study (continued) Name of measure  Reliability  Validity/Other information  The Center for  α= 0.85 (general  correlated with other depression measures  Epidemiological  population) and α=  (r= 0.57 to r= 0.82) but did not correlate  Studies Depression  0.90 (psychiatric  with measures of cognitive, physical and  (CES-D) Scale  patients); test/retest  social functioning (Shinar et al., 1986)  (Radloff, 1977)  (0.45-0.70) (different groups) (Radloff, 1977); interrater (r=0.76) (Shinar et al., 1986)  The State-Trait  test/retest: 0.97 for  discriminating ability for stressful and  Anxiety Inventory  Trait and 0.45 for State  non-stressful situations (Metzger, 1976);  (STAI) for Adults  (college students)  correlated with other measures (0.75-0.85)  (Spielberger, 1983)  (Metzger, 1976); K-R  for college students and psychiatric  20 (0.83-0.92) for State patients (Hedberg, 1972) (variety samples) (Hedberg, 1972) 6-item Interpersonal  α= 0.45 to 0.75 and  moderately correlated with Inventory of  Support Evaluation  test/retest (ICC= 0.63  Socially Supportive Behaviors (r= 0.46)  List (ISEL) (Cohen  to 0.85) (medical  (college students) {{544 Cohen, S.  & Hoberman, 1983;  students) (Delistamati  1983}}; correlated with psychosomatic  Cohen et al., 1985)  et al., 2006)  symptoms and presence of stressful events (Delistamati et al., 2006)  123  Appendix III: The Berg Balance Scale (adapted from Berg et al., 1989; Berg et al., 1992) Grading: Please mark the lowest category which applies. 1. Sitting to Standing Instruction: Please stand up. Try not to use your hands for support. Grading: 4: Able to stand no hands and stabilize independently 3: Able to stand independently using hands. 2: Able to stand using hands after several tries. 1: Needs minimal assistance to stand or to stabilize. 0: Needs moderate or maximal assistance to stand. 2. Standing Unsupported Instruction: Stand for two minutes without holding. Grading: 4: Able to stand safely 2 minutes. 3: Able to stand 2 minutes with supervision. 2: Able to stand 30 seconds unsupported. 1: Needs several tries to stand 30 seconds unsupported. 0: Unable to stand 30 seconds unassisted. 3. Sitting Unsupported Feet on Floor Instruction: Sit with arms folded for two minutes. Grading: 4: Able to sit safely and securely 2 minutes. 3: Able to sit 2 minutes under supervision. 2: Able to sit 30 seconds. 1: Able to sit 10 seconds. 0: Unable to sit without support 10 seconds. 4. Standing to Sitting Instruction: Please sit down. Grading: 4: Sits safely with minimal use of hands. 3: Controls descent by using hands. 2: Uses back of legs against chair to control descent. 1: Sits independently but has uncontrolled descent. 0: Needs assistance to sit.  124  Appendix III (continued) Berg Balance Scale (continued) 5. Transfers Instruction: Please move from chair to bed and back again. (One way toward a seat with armrests and one way toward a seat without armrests) Grading: 4: Able to transfer safely with minor use of hands. 3: Able to transfer safely definitely need of hands. 2: Able to transfer with verbal cuing and/or supervision. 1: Needs one person to assist. 0: Needs two people to assist or supervise to be safe. 6. Standing Unsupported with Eyes Closed Instruction: Close your eyes and stand still for 10 seconds. Grading: 4: Able to stand 10 seconds safely. 3: Able to stand 10 seconds with supervision. 2: Able to stand 3 seconds. 1: Unable to keep eyes closed 3 seconds but stays steady. 0: Needs help to keep from falling. 7. Standing Unsupported with Feet Together Instruction: Place your feet together and stand without holding. Grading: 4: Able to place feet together independently and stand 1 minute safely. 3: Able to place feet together independently and stand for 1 minute with supervision. 2: Able to place feet together independently but unable to hold for 30 seconds. 1: Needs help to attain position but able to stand 15 seconds with feet together. 0: Needs help to attain position and unable to hold for 15 seconds. 8. Reaching Forward with Outstretched Arm Instruction: Lift arm to 90 degrees. Stretch out your fingers and reach forward as far as you can. (Examiner places a ruler at end of fingertips when arm is at 90 degrees. Fingers should not touch the ruler while reaching forward. The recorded measure is the distance forward that the fingers reach while the subject is in the most forward lean position.) Grading: 4: Can reach forward confidently more than 10 inches. 3: Can reach forward more than 5 inches. 2: Can reach forward more than 2 inches safely. 1: Reaches forward but needs supervision. 0: Needs help to keep from falling.  125  Appendix III (continued) Berg Balance Scale (continued) 9. Pick Up Object From the Floor Instruction: Pick up the shoe/slipper which is placed in front of your feet. Grading: 4: Able to pick up slipper safely and easily. 3: Able to pick up slipper but needs supervision. 2: Unable to pick up but reaches 1 to 2 inches from slipper and keeps balance independently. 1: Unable to pick up and needs supervision while trying. 0: Unable to try/needs assistance to keep from falling. 10. Turning to Look Behind Over Left and Right Shoulders Instruction: Turn to look behind you over toward left shoulder. Repeat to the right. Grading: 4: Looks behind from both sides and weight shifts well. 3: Looks behind one side only; other side shows less weight shift. 2: Turns sideways only but maintains balance. 1: Needs supervision when turning. 0: Needs assistance to keep from falling. 11. Turn 360 Degrees Instruction: Turn completely around in a full circle. Pause. Then turn a full circle in the other direction. Grading: 4: Able to turn 360 degrees safely in less than 4 seconds each side. 3: Able to turn 360 degrees safely one side only-less than 4 seconds. 2: Able to turn 360 degrees safely but slowly. 1: Needs close supervision or verbal cuing. 0: Needs assistance while turning. 12. Count Number of Times Step Touch Measured Stool Instruction: Place each foot alternatively on the stool. Continue until each foot has touched the stool four times. Grading: 4: Able to stand independently and safely and complete 8 steps in 20 seconds. 3: Able to stand independently and complete 8 steps in more than 20 seconds. 2: Able to complete 4 steps without aid with supervision. 1: Able to complete more than 2 steps-needs minimal assistance. 0: needs assistance to keep from falling-unable to try.  126  Appendix III (continued) Berg Balance Scale (continued) 13. Standing Unsupported One Foot in Front Instruction: Place one foot directly in front of the other. If you feel that you cannot place your foot directly in front, try to step as far enough ahead that the heel of your forward foot is ahead of the toes of the other foot. (DEMONSTRATE to subject.) Grading: 4: Able to place foot tandem independently and hold 30 seconds. 3: Able to place foot ahead of the other independently and hold 30 seconds. 2: Able to take small step independently and hold 30 seconds. 1: Needs help to step but can hold 15 seconds. 0: Loses balance while stepping or standing. 14. Standing on One Leg (must be stroke-affected leg) Instruction: Stand on one leg as long as you can without holding. Grading: 4: Able to lift leg independently and hold for more than 10 seconds. 3: Able to lift leg independently and hold 5 to 10 seconds. 2: Able to lift leg independently and hold at least 3 seconds. 1: Tries to lift leg, unable to hold 3 seconds but remains standing independently. 0: Unable to try or needs assistance to prevent fall.  127  Appendix IV: The Timed “Up and Go” Test (adapted from Mathias et al., 1986) Instruction: Sit with your back against the chair and your arms on the arm rests. On the word ‘go’, stand upright, then walk at your normal pace to the line on the floor, go around the foot stool, return to the chair, and sit down. The time from the word ‘go’ to when the subject returns to the starting position is recorded. The following schematic diagram shows the arrangement for the Timed “Up and Go” Test.  Figure 4: Schematic Diagram Showing the Arrangement/Set Up for the Timed “Up and Go” Test  Turn around after crossing line Sit to Stand  Walk  3 metres  128  Appendix V: The Six Minute Walk Test (adapted from Guyatt et al., 1984) Protocol: Set up a course of known distance within the research facility. Instruct subjects to walk along the specifically defined course. Instruction: I would like for you to walk as far as you can in 6 minutes. Results: 6 Minute Walk Test: ______________meters •  Assistive device used during this test: 0 = None; 1 = Cane or  •  Crutch;  2 = Walker  Completed without a break 0 = No; 1 = Yes  129  Appendix VI: The Activities-Specific Balance Confidence Scale (ABC Scale) (adapted from Powell & Myers, 1995) For each of the following activities, please indicate your level of self-confidence by choosing a corresponding number from the following rating scale. Answer all items even if they are activities you would not do or are unsure about.  0% 10% 20% Not confident  30%  40%  50%  60%  70%  80% 90% 100% Completely confident  How confident are you that you will not lose your balance or become unsteady when you.... a)....walk around the house?_____% b)....walk up and down the stairs?_____% c)....pick up a slipper from the floor?____% d)....reach at eye level?____% e)....reach while standing on your tiptoes?____% f)....stand on a chair to reach?____% g)....sweep the floor?____% h)....walk outside to nearby car?____% i)....get in and out of a car?____% j)....walk across a parking lot?____% k)....walk up and down a ramp?____% l)....walk in a crowded mall?____% m)....walk in a crowd or get bumped?____% n)....ride an escalator holding the rail?____% o)....ride an escalator not holding the rail?____% p)....walk on icy sidewalks?____%  130  Appendix VII: The Center for Epidemiologic Studies Depression Scale (adapted from Radloff, 1977) Below is a list of some of the ways you may have felt or behaved. Please indicate how often you have felt this way during the past week: (circle one number on each line) Rarely or none of the time (<1 day)  Some or a little of the time(1-2 days)  Occasionally or a moderate amount of time (3-4days)  All of the time (5-7 days)  0  1  2  3  0  1  2  3  0  1  2  3  0  1  2  3  0  1  2  3  6. I felt depressed....  0  1  2  3  7. I felt that everything I did was an effort...  0  1  2  3  8. I felt hopeful about the future...  0  1  2  3  9. I thought my life had been a failure...  0  1  2  3  10. I felt fearful....  0  1  2  3  11. My sleep was restless....  0  1  2  3  12. I was happy...  0  1  2  3  13. I talked less than usual....  0  1  2  3  14. I felt lonely....  0  1  2  3  15. People were unfriendly....  0  1  2  3  16. I enjoyed life....  0  1  2  3  17. I had crying spells....  0  1  2  3  18. I felt sad....  0  1  2  3  19. I felt that people disliked me....  0  1  2  3  20. I could not “get going”....  0  1  2  3  During the past week....  1. I was bothered by things that usually don’t bother me.... 2. I did not feel like eating; my appetite was poor.... 3. I felt that I could not shake off the blues even with help from my family... 4. I felt that I was just as good as other people.... 5. I had trouble keeping my mind on what I was doing...  131  Appendix VIII: The State Anxiety Inventory for Adults (adapted from Spielberger, 1983) Instruction: A number of statements which people have used to describe themselves are given below. Read each statement and then circle the appropriate value to the right of the statement to indicate how you feel right now, that is, at this moment. There are no right or wrong answers. Do not spend too much time on any one statement but give the answer which seems to describe your present feelings best. Not at all  Somewhat  Moderately so  Very much so  1. I feel calm.............................  1  2  3  4  2. I feel secure...........................  1  2  3  4  3. I feel tense.............................  1  2  3  4  4. I feel strained........................  1  2  3  4  5. I feel at ease..........................  1  2  3  4  6. I feel upset............................  1  2  3  4  7. I am presently worrying over  1  2  3  4  8. I feel satisfied........................  1  2  3  4  9. I feel frightened....................  1  2  3  4  10. I feel comfortable...............  1  2  3  4  11. I feel self-confident............  1  2  3  4  12. I feel nervous......................  1  2  3  4  13. I am jittery..........................  1  2  3  4  14. I feel indecisive..................  1  2  3  4  15. I am relaxed........................  1  2  3  4  16. I feel content......................  1  2  3  4  17. I am worried.......................  1  2  3  4  18. I feel confused....................  1  2  3  4  19. I feel steady.........................  1  2  3  4  20. I feel pleasant......................  1  2  3  4  possible misfortunes.................  The Trait version STAI has the same set of questions as the State version, instead of asking how the individual feels at the moment, the Trait version asks how the individual generally feel.  132  Appendix IX: Cognitive Capacity Screening Examination (adapted from Jacobs et al., 1977) Instruction: Check items answered correctly. Write incorrect or unusual answers in space provided. If necessary, urge subject once to complete task. Introduction to subject: “I would like to ask you a few questions. Some you will find very easy and others may be very hard. Just do your best.” 1) What day of the week is this?  ___  16) The opposite of large is  ___  2) What month?  ___  17) The opposite of hard is  ___  3) What day of month?  ___  18) An orange and a banana are both fruits. Red and blue are both  ___  ___  19) A penny and a dime are both  ___  ___  20) What are those words I asked you to remember? (HAT)  ___  6) Repeat the numbers 8 7 2.  ___  21) (CAR)  ___  7) Say them backwards.  ___  22) (TREE)  ___  8) Repeat the numbers 6 3 7 1.  ___  23) (TWENTY-SIX)  ___  9) Listen to these numbers 6 9 4. Count 1 through 10 out loud, then repeat 6 9 4. (Help if needed. Then use numbers 5 7 3.)  ___  24) Take away 7 from 100, then take away 7 from what is left and keep going: 100-7 is  ___  10) Listen to these numbers 8 1 4 3. Count 1 through 10 out loud, then repeat 8 1 4 3.  25) Minus 7  ___  ___ ___  4) What year? 5) What place is this?  11) Beginning with Sunday, say the days of the week backwards.  ___  26) Minus 7 (write down answers; check correct subtraction of 7)  12) 9 + 3 is  ___  27) Minus 7  ___  13) Add 6 (to the previous answer or to “12”).  28) Minus 7  ___  ___ 29) Minus 7  ___  30) Minus 7  ___  14) Take away 5 (“from 18”). Repeat these words after me and remember them. I will ask for them later: HAT, CAR, TREE, TWENTY-SIX.  ___  15) The opposite of fast is slow. The opposite of up is  ___ TOTAL CORRECT (maximum score = 30)  ___  133  Appendix X: The 6-item Interpersonal Support Evaluation List (ISEL) (adapted from (Cohen & Hoberman, 1983; Cohen et al., 1985) This scale is made up of a list of statements each of which may or may not be true about you. For each statement check “definitely true” if you are sure it is true about you and “probably true” if you think it is true but are not absolutely certain. Similarly, you should check “definitely false” if you are sure the statement is false and “probably false” if you think it is false but are not absolutely certain. definitely  probably  probably  definitely  true  true  false  false  1. If I wanted to go on a trip for a day (for example, to the country or mountains), I would have a hard time finding someone to go with me. 2. I feel that there is no one I can share my most private worries and fears with. 3. I don’t often get invited to do things with others. 4. If I had to go out of town for a few weeks, it would be difficult to find someone who would look after my house or apartment (the plants, pets, garden, etc.). 5. If a family crisis arose, it would be difficult to find someone who could give me good advice about how to handle it. 6. If I needed some help in moving into a new house or apartment, I would have a hard time finding someone to help me.  134  Appendix XI: Chronic Condition Questionnaire (adapted from Canadian Community Health Survey Cycle 1.1(2000-2001): Statistics Canada (2003) I would like to ask about certain chronic health conditions which you may have. We are interested in ‘long term conditions’ that have lasted or are expected to last 6 months or more and that have been diagnosed by a health professional. Do you currently have:  Please tick √  1. Food allergies?  □  2. Other allergies?  □  3. Asthma?  □  4. Asthma symptoms or asthma attacks in past 12 months?  □  5. Fibromyalgia?  □  6. Arthritis or rheumatism? (excluding fibromyalgia)  □  What kind of arthritis do you have? a. Rheumatoid arthritis b. Osteoarthritis c. Other-specify  □ □ □  □  7. Back problems? (excluding fibromyalgia and arthritis)  □  8. High blood pressure?  □  9. Migraine headaches?  □  10. Chronic bronchitis?  □  11. Chronic obstructive pulmonary disease? (COPD)  □  12. Diabetes?  □  13. Epilepsy?  □  14. Heart disease?  □  15. Heart attack? (damage to the heart muscle)  □  16. Angina? (chest pain, chest tightness)  □  135  Appendix XI: Chronic Condition Questionnaire (continued) Do you currently have: 17. Congestive heart failure? (inadequate heart beat, fluid build up in the  Please tick √ □  lungs or legs? 18. Cancer?  □  19. Stomach or intestinal ulcers?  □  20. The effects of a stroke?  □  21. Urinary incontinence?  □  22. A bowel disorder such as Crohen’s disease or colitis?  □  23. Alzheimer’s disease or any other dementia?  □  24. Cataracts?  □  25. Glaucoma?  □  26. A thyroid condition?  □  27. Parkinson’s disease?  □  28. Multiple Sclerosis?  □  29. Chronic Fatigue Syndrome?  □  30. Multiple chemical sensitivities?  □  31. Any other long-term condition that has been diagnosed by a health  □  professional  (Specify)_____ _____________  136  Appendix XII: A Person-Period Data Set Figure 5: Part of a “Person-Period” Data Set for Records of 4 Variables (Age, Sex, ABC and BBS) of 4 Selected Participants ID#  Age  Sex  ABC  BBS  01  54  1  66.88  51  01  54  1  78.13  46  01  54  1  70.63  45  01  54  1  77.50  51  02  68  1  93.75  50  02  68  1  85.00  51  02  68  1  89.38  52  02  68  1  96.25  52  03  50  0  83.75  56  03  50  0  92.50  55  03  50  0  93.13  55  03  50  0  04  79  1  97.94  54  04  79  1  99.81  54  04  79  1  99.38  56  04  79  1  99.00  53  Each person has multiple records, one for each testing occasion, in a “Person-Period” data set. Four measurement occasions were shown. 137  Appendix XIII: Centering Variables Variables that have been centered are shown in Table 4. Note: categorical variables and the dependent variable ABC were not re-centered Table 4 Method of Centering Variables in the Study Name of variable  Method of centering  Age  minus 50 (minimum age in sample)  Walking Capacity (6MW)  minus grand mean & divided by ten  Basic Mobility (TUG)  minus grand mean  Depressive symptoms (CES-D)  minus grand mean  Baseline Balance Performance (BBS)  minus baseline mean  Baseline Perceived Social Support (ISEL)  minus baseline mean  Anxiety (STAI)  minus grand mean  138  Appendix XIV: Creating an Unconditional Growth Model To determine whether a predictor is time-varying or time invariant, an unconditional growth model was created for each variable. The following variables were used sequentially as the dependent variable, Y: BBS, TUG, 6MWT, CES-D, STAI and ISEL. Time or “months post baseline” was introduced as a predictor into the level 1 sub-model. The sub-model and composite model equations were written as: Level 1: 𝒀𝒀𝒊𝒊𝒊𝒊 = 𝝅𝝅𝟎𝟎𝟎𝟎 + 𝝅𝝅𝟏𝟏𝟏𝟏 𝑴𝑴𝑴𝑴𝑴𝑴𝒊𝒊𝒊𝒊 + 𝜺𝜺𝒊𝒊𝒊𝒊 Level 2: 𝝅𝝅𝟎𝟎𝟎𝟎 = 𝜸𝜸𝟎𝟎𝟎𝟎 + 𝜻𝜻𝟎𝟎𝟎𝟎 𝝅𝝅𝟏𝟏𝟏𝟏 = 𝜸𝜸𝟏𝟏𝟏𝟏 + 𝜻𝜻𝟏𝟏𝟏𝟏  Composite: 𝒀𝒀𝒊𝒊𝒊𝒊 = 𝜸𝜸𝟎𝟎𝟎𝟎 + 𝜸𝜸𝟏𝟏𝟏𝟏 𝑴𝑴𝑴𝑴𝑴𝑴𝒊𝒊𝒊𝒊 + (𝜺𝜺𝒊𝒊𝒊𝒊 + 𝜻𝜻𝟎𝟎𝟎𝟎 + 𝜻𝜻𝟏𝟏𝟏𝟏 𝑴𝑴𝑴𝑴𝑴𝑴𝒊𝒊𝒊𝒊 )  (where Yij is the value of the dependent variable of individual i at time j; “months post baseline” = MPB; 𝜋𝜋0𝑖𝑖 is the intercept of the true change trajectory for individual i ; 𝜋𝜋1𝑖𝑖 is the slope of the true change trajectory for individual i ; 𝜀𝜀𝑖𝑖𝑖𝑖 is the level 1 residual and 𝜁𝜁0𝑖𝑖 and 𝜁𝜁1𝑖𝑖 are level 2 residuals) For example, using BBS as the dependent variable, an unconditional growth model was created using the following syntax (for SPSS): MIXED BBS WITH Mon_post_b /CRITERIA=CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) /FIXED=Mon_post_b | SSTYPE(3) /METHOD=REML /PRINT=G SOLUTION TESTCOV /RANDOM=INTERCEPT Mon_post_b | SUBJECT(id#) COVTYPE(UN). If the main effect for MPB is significant, the variable is treated as time-varying; otherwise, time invariant. You can see that in figure 7, “months post baseline” is not significant and BBS was treated as a time invariant variable. Figure 6  Sample output table for creating an unconditional growth model Estimates of Fixed Effects  a  95% Confidence Interval Parameter Intercept Mon_post_b  Estimate  Std. Error  df  t  Sig.  Lower Bound  Upper Bound  50.312157  .520096  193.852  96.736  .000  49.286384  51.337930  .028979  .029483  142.201  .983  .327  -.029302  .087260  a. Dependent Variable: Berg Balance (/56).  139  Appendix XV: Pseudo R Squares Calculations (Singer, 2003) Computing pseudo R square from the variance components (𝑅𝑅𝜀𝜀2 ): This is the proportional reduction in residual variance as predictors were added. a. To assess the proportion of within-person variation “explained by time”, the following formula was used: 𝑅𝑅𝜀𝜀2 =  𝜎𝜎𝜀𝜀2 (𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 )− 𝜎𝜎𝜀𝜀2 (𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔 ℎ 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 ) 𝜎𝜎𝜀𝜀2 (𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 )  (1)  where 𝜎𝜎𝜀𝜀2 is the level-1 residual variance across all occasions of measurement, for individual i in the  population;  b. Similarly, level 2 or between-person residual variance can be computed. For any subsequent model, the following formula was used: 𝑅𝑅𝜁𝜁2 =  𝜎𝜎𝜁𝜁2 (𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔 ℎ 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 )− 𝜎𝜎𝜁𝜁2 (𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠𝑠 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 ) 𝜎𝜎𝜁𝜁2 (𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢𝑢 𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔𝑔 ℎ 𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚 )  (2)  where 𝜎𝜎𝜁𝜁2 is the level-2 residual variance  𝑅𝑅𝜀𝜀2 of time-varying predictors can be calculated similarly using formula (2) with level-1 residual variance for subsequent models.  140  Appendix XVI: Histograms Showing Frequency Distributions of Variables Figure 7 Frequency Distributions of Time Varying Predictors- TUG, Transformed TUG, CES-D, 6MW, STAI and Transformed STAI  141  Appendix XVI: Histograms Showing Frequency Distributions of Variables (continued) Figure 8 Frequency Distributions of Time-invariant Predictors- BBS, Transformed BBS, ABC, Transformed ABC and ISEL.  142  Appendix XVII: Reasons for Missing Data Forty-three participants missed at least 1 testing occasions; the reasons for missing a testing occasion were: • unreachable: 7 •  cancelled/missed appointments: 12  •  death: 1  •  declined/dropped out: 7  •  health reasons: 5  •  out of town: 5  •  unknown reasons: 6  Other reasons for missing observations (total 15) were: • participant refused to do the test (6MW at 3 months) •  participant has chosen 2 answers for a question therefore unable to score (STAI at 3 months)  •  unable to read the answers to some of the questions (ABC at 6 months)  •  participant refused because foot was sore (6MW at 3 and 6 months)  •  participant refused to answer one of the question therefore unable to score (CES-D at 12 months)  •  no reason listed for 9 missing observations (1 ISEL and 3 CES-D at baseline; 1 CESD at 6 months; 2 6MW at 6 months; 1 BBS at 6 months; and 1 TUG at 12 months)  143  Appendix XVIII: Differences between Dropouts and Participants Table 5 Mean Difference of Variables at Baseline between Dropouts and Participants in the Stroke Group Variable  Dropouts  Participants  Mean (SD) or n  Mean (SD) or n  66.39(10.40)  67.81(10.07)  Male  21  50  Female  7  20  ABC*  51.85(21.68)  65.88(23.69)  BBS**  42.14(10.14)  46.54(8.11)  TUG(sec)  27.56(26.54)  18.90(12.36)  6MWT(m)  229.95(143.49)  282.73(136.03)  CES-D  16.79(6.83)  14.69(9.86)  STAI  37.21(9.45)  35.60(11.83)  ISEL  12.75(1.96)  13.47(2.50)  Age Sex  *p =0.008; ** p =0.026 by Independent-Sample t Test; categorical variable by chi square p= 0.721  144  Appendix XVIII: Differences between Dropouts and Participants (continued) Table 6 Mean Difference of Variables at Baseline between Dropouts and Participants in the Control Group Variable  Dropouts  Participants  Mean (SD) or n  Mean (SD) or n  65.67(10.42)  68.09(9.81)  Male  14  57  Female  4  23  ABC  94.66(5.65)  94.00(8.70)  BBS  55.17(1.38)  54.30(3.34)  TUG(sec)  8.36(2.03)  8.14(1.90)  6MWT(m)  524.50(104.61)  532.72(88.01)  CES-D  7.59(7.77)  6.05(5.15)  STAI*  29.94(7.45)  25.25(5.85)  ISEL  13.78(1.22)  13.37(2.19)  Age Sex  *p =0.004 by Independent-Sample t Test; categorical variable by chi square p= 0.575  145  Appendix XIX: Selected SPSS Output for Regression Analysis with All Predictors on Collinearity Diagnostics (Tolerance, VIF and Condition Index) Coefficients  a  Standardized Unstandardized Coefficients Model 1  B (Constant)  Std. Error 86.320  1.485  .185  .058  c_TUG  -.163  c_CES_D c_state  Coefficients Beta  Collinearity Statistics t  Sig.  Tolerance  VIF  58.146  .000  .153  3.163  .002  .226  4.423  .065  -.094  -2.512  .012  .376  2.662  -.263  .079  -.107  -3.338  .001  .517  1.935  -.388  .064  -.187  -6.107  .000  .566  1.767  c_BBS_b  .880  .118  .327  7.437  .000  .274  3.654  c_social_b  .377  .232  .041  1.624  .105  .842  1.187  -.061  .054  -.029  -1.121  .263  .801  1.248  Gender  -4.057  1.105  -.088  -3.673  .000  .920  1.087  Stroke or controls  -6.049  1.486  -.146  -4.072  .000  .415  2.410  6 min walk every 10m  c_age  a. Dependent Variable: ABC  146  Appendix XIX: Selected SPSS Output for Regression Analysis with All Predictors on Collinearity Diagnostics (Tolerance, VIF and Condition Index (continued) Collinearity Diagnostics  a  Variance Proportions Condition Model Dimension 1  Eigenvalue  Index  6 min walk (Constant)  every 10m  Stroke or c_TUG c_CES_D c_state c_BBS_b c_social_b  c_age  Gender  controls  1  3.419  1.000  .00  .01  .01  .01  .01  .01  .00  .00  .01  .01  2  2.500  1.170  .01  .01  .01  .01  .01  .01  .01  .02  .03  .00  3  1.413  1.556  .00  .01  .04  .09  .09  .02  .13  .00  .00  .00  4  .896  1.954  .00  .00  .01  .02  .08  .00  .51  .00  .13  .02  5  .598  2.392  .00  .01  .01  .02  .05  .00  .24  .01  .65  .02  6  .356  3.099  .00  .01  .03  .77  .63  .01  .04  .00  .00  .00  7  .333  3.204  .02  .17  .41  .00  .11  .00  .00  .04  .12  .09  8  .249  3.706  .01  .03  .24  .07  .00  .14  .06  .24  .00  .25  9  .178  4.382  .01  .37  .16  .00  .01  .78  .00  .06  .00  .06  10  .059  7.640  .95  .38  .09  .01  .01  .02  .01  .62  .05  .54  a. Dependent Variable: ABC  147  Appendix XX: Scatter Plots of Individual Changes in Balance Confidence over Time Figure 9: Scatter Plots of Individual Changes in ABC over Time in the First 49 Stroke Subjects (those with only one data point were excluded in the analyses)  148  Appendix XX (continued) Figure 10: Scatter plots of individual changes in ABC over time in the second 49 stroke subjects (those with only one data point were excluded in the analyses)  149  Appendix XX (continued) Figure 11: Scatter plots of individual changes in ABC over time in the first 49 controls (those with only one data point were excluded in the analyses)  150  Appendix XX (continued) Figure 12: Scatter plots of individual changes in ABC over time in the second 49 controls (those with only one data point were excluded in the analyses)  151  Appendix XXI: Written Models, SPSS Syntax and Selected Output for (a): Unconditional Mean Model, (b) Unconditional Growth Model and (c) Model A. (a) Unconditional Mean Model 1. Written model There are no predictors in the unconditional mean model. It only describes the outcome variations. The sub-model and composite model equations were written as: Level 1: 𝑌𝑌𝑖𝑖𝑖𝑖 = 𝜋𝜋0𝑖𝑖 + 𝜀𝜀𝑖𝑖𝑖𝑖  Level 2: 𝜋𝜋0𝑖𝑖 = 𝛾𝛾00 + 𝜁𝜁0𝑖𝑖 2. Syntax The following syntax was used for SPSS: MIXED ABC /CRITERIA=CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) /FIXED=| SSTYPE(3) /METHOD=REML /PRINT=G SOLUTION TESTCOV /RANDOM=INTERCEPT | SUBJECT(id#) COVTYPE(UN).  152  Appendix XXI (a): Unconditional Mean Model (continued)  3. Output Estimates of fixed effects (unconditional mean model) Estimates of Fixed Effects  a  95% Confidence Interval Parameter Intercept  Estimate  Std. Error  81.256256  1.505684  df 178.894  t  Sig.  53.966  Lower Bound  .000  78.285071  Upper Bound 84.227442  a. Dependent Variable: ABC.  Estimates of covariance parameters (unconditional mean model) Estimates of Covariance Parameters  a  95% Confidence Interval Parameter  Estimate  Residual Intercept [subject = id#]  Variance  Std. Error  Wald Z  Sig.  Lower Bound  Upper Bound  68.028202  4.298422  15.826  .000  60.104235  76.996843  391.647735  43.400966  9.024  .000  315.187210  486.656639  a. Dependent Variable: ABC.  153  Appendix XXI (b): Unconditional Growth Model  1. Written model Time or “months post baseline” was introduced as a predictor into the level 1 sub-model. The sub-model and composite model equations were written as: Level 1: 𝑌𝑌𝑖𝑖𝑖𝑖 = 𝜋𝜋0𝑖𝑖 + 𝜋𝜋1𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 + 𝜀𝜀𝑖𝑖𝑖𝑖 Level 2: 𝜋𝜋0𝑖𝑖 = 𝛾𝛾00 + 𝜁𝜁0𝑖𝑖 𝜋𝜋1𝑖𝑖 = 𝛾𝛾10 + 𝜁𝜁1𝑖𝑖  Composite: 𝑌𝑌𝑖𝑖𝑖𝑖 = 𝛾𝛾00 + 𝛾𝛾10 𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 + (𝜀𝜀𝑖𝑖𝑗𝑗 + 𝜁𝜁0𝑖𝑖 + 𝜁𝜁1𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 )  (where Yij is the value of the dependent variable, ABC, of individual i at time j; “months post baseline” = MPB; 𝜋𝜋0𝑖𝑖 is the intercept of the true change trajectory for individual i ; 𝜋𝜋1𝑖𝑖 is the slope of the true change trajectory for individual i ; 𝜀𝜀𝑖𝑖𝑖𝑖 is the level 1 residual and 𝜁𝜁0𝑖𝑖 and 𝜁𝜁1𝑖𝑖 are level 2 residuals) 2. Syntax The following syntax was used for SPSS: MIXED ABC WITH Mon_post_b /CRITERIA=CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) /FIXED=Mon_post_b | SSTYPE(3) /METHOD=REML /PRINT=G SOLUTION TESTCOV /RANDOM=INTERCEPT Mon_post_b | SUBJECT(id#) COVTYPE(UN).  154  Appendix XXI (b): Unconditional Growth Model (continued) 3. Output Estimates of fixed effects (unconditional growth model) Estimates of Fixed Effects  a  95% Confidence Interval Parameter Intercept Mon_post_b  Estimate  Std. Error  df  t  Sig.  Lower Bound  Upper Bound  80.130009  1.645609  179.687  48.693  .000  76.882805  83.377213  .240913  .081217  150.813  2.966  .004  .080443  .401382  a. Dependent Variable: ABC.  Estimates of covariance parameters (unconditional growth model) Estimates of Covariance Parameters  a  95% Confidence Interval Parameter  Estimate  Residual  Std. Error  Wald Z  Sig.  Lower Bound  Upper Bound  59.145005  4.598312  12.862  .000  50.785556  68.880443  Intercept + Mon_post_b  UN (1,1)  453.512616  51.796012  8.756  .000  362.554262  567.290789  [subject = id#]  UN (2,1)  -7.461664  2.116105  -3.526  .000  -11.609155  -3.314174  UN (2,2)  .333165  .143135  2.328  .020  .143537  .773313  a. Dependent Variable: ABC.  155  Appendix XXI (c): Model A 1. Written model Level 1: 𝑌𝑌𝑖𝑖𝑖𝑖 = 𝜋𝜋0𝑖𝑖 + 𝜋𝜋1𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 + 𝜀𝜀𝑖𝑖𝑖𝑖 Level 2: 𝜋𝜋0𝑖𝑖 = 𝛾𝛾00 + 𝛾𝛾01 𝑆𝑆𝑆𝑆𝑖𝑖 + 𝜁𝜁0𝑖𝑖 𝜋𝜋1𝑖𝑖 = 𝛾𝛾10 + 𝛾𝛾11 𝑆𝑆𝑆𝑆𝑖𝑖 + 𝜁𝜁1𝑖𝑖  Composite: 𝑌𝑌𝑖𝑖𝑖𝑖 = 𝛾𝛾00 + 𝛾𝛾01 𝑆𝑆𝑆𝑆𝑖𝑖 + 𝛾𝛾10 𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 + 𝛾𝛾11 𝑆𝑆𝑆𝑆𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 + (𝜀𝜀𝑖𝑖𝑖𝑖 + 𝜁𝜁0𝑖𝑖 + 𝜁𝜁1𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 ) where SS is stroke status, a time-invariant  level 2 predictor.  2. Syntax The following syntax was used for SPSS: MIXED ABC WITH Mon_post_b stroke_or_not /CRITERIA=CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) /FIXED=Mon_post_b stroke_or_not Mon_post_b*stroke_or_not | SSTYPE(3) /METHOD=REML /PRINT=G SOLUTION TESTCOV /RANDOM=INTERCEPT Mon_post_b | SUBJECT(id#) COVTYPE(UN).  156  Appendix XXI (c): Model A (continued) 3. Output Estimates of fixed effects (model A) Estimates of Fixed Effects  a  95% Confidence Interval Parameter Intercept Mon_post_b stroke_or_not Mon_post_b * stroke_or_not  Estimate  Std. Error  df  t  Sig.  Lower Bound  Upper Bound  94.030816  1.731902  178.773  54.293  .000  90.613215  97.448418  -.119174  .105872  151.586  -1.126  .262  -.328350  .090002  -28.903342  2.498089  178.777  -11.570  .000  -33.832877  -23.973808  .740872  .153259  151.482  4.834  .000  .438071  1.043673  a. Dependent Variable: ABC.  Estimates of covariance parameters (model A) Estimates of Covariance Parameters  a  95% Confidence Interval Parameter  Estimate  Residual  Std. Error  Wald Z  Sig.  Lower Bound  Upper Bound  59.117567  4.593684  12.869  .000  50.766195  68.842796  Intercept + Mon_post_b  UN (1,1)  245.346554  29.960010  8.189  .000  193.124407  311.689921  [subject = id#]  UN (2,1)  -2.106799  1.450938  -1.452  .146  -4.950585  .736988  UN (2,2)  .201052  .127764  1.574  .116  .057861  .698598  a. Dependent Variable: ABC.  157  Appendix XXII: Written Model, Syntax and Selected Output for Model B 1. Written model Level 1: 𝑌𝑌𝑖𝑖𝑖𝑖 = 𝜋𝜋0𝑖𝑖 + 𝜋𝜋1𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 + 𝜋𝜋3𝑖𝑖 𝑇𝑇𝑇𝑇𝑇𝑇𝑖𝑖𝑖𝑖 + 𝜋𝜋4𝑖𝑖 6𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 + 𝜋𝜋5𝑖𝑖 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖 + 𝜋𝜋6𝑖𝑖 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖𝑖𝑖 + 𝜀𝜀𝑖𝑖𝑖𝑖 Level 2: 𝜋𝜋0𝑖𝑖 = 𝛾𝛾00 + 𝛾𝛾01 𝑆𝑆𝑆𝑆𝑖𝑖 + 𝛾𝛾02 𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖 + 𝛾𝛾03 𝐴𝐴𝐴𝐴𝐴𝐴𝑖𝑖 + 𝛾𝛾04 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑖𝑖 + 𝛾𝛾05 𝐵𝐵𝐵𝐵𝐵𝐵𝑖𝑖 + 𝜁𝜁0𝑖𝑖 𝜋𝜋1𝑖𝑖 = 𝛾𝛾10 + 𝛾𝛾11 𝑆𝑆𝑆𝑆𝑖𝑖 + 𝜁𝜁1𝑖𝑖 𝜋𝜋3𝑖𝑖 = 𝛾𝛾30 + 𝜁𝜁3𝑖𝑖 𝜋𝜋4𝑖𝑖 = 𝛾𝛾40 + 𝜁𝜁4𝑖𝑖 𝜋𝜋5𝑖𝑖 = 𝛾𝛾50 +𝜁𝜁5𝑖𝑖 𝜋𝜋6𝑖𝑖 = 𝛾𝛾60 + 𝜁𝜁6𝑖𝑖 Composite: 𝑌𝑌𝑖𝑖𝑖𝑖 = 𝛾𝛾00 + 𝛾𝛾01 𝑆𝑆𝑆𝑆𝑖𝑖 + 𝛾𝛾02 𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖 + 𝛾𝛾03 𝐴𝐴𝐴𝐴𝐴𝐴𝑖𝑖 + 𝛾𝛾04 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑖𝑖 + 𝛾𝛾05 𝐵𝐵𝐵𝐵𝐵𝐵𝑖𝑖 + 𝛾𝛾10 𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 + 𝛾𝛾11 𝑆𝑆𝑆𝑆𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 + 𝛾𝛾30 𝑇𝑇𝑇𝑇𝑇𝑇𝑖𝑖𝑖𝑖 + 𝛾𝛾40 6𝑀𝑀𝑀𝑀𝑖𝑖𝑗𝑗 + 𝛾𝛾50 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖 + 𝛾𝛾60 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖𝑖𝑖 + (𝜁𝜁0𝑖𝑖 + 𝜁𝜁1𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 + 𝜁𝜁3𝑖𝑖 𝑇𝑇𝑇𝑇𝑇𝑇𝑖𝑖𝑖𝑖 + 𝜁𝜁4𝑖𝑖 6𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 + 𝜁𝜁5𝑖𝑖 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖 + 𝜁𝜁6𝑖𝑖 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖𝑖𝑖 + 𝜀𝜀𝑖𝑖𝑖𝑖 )  2. Syntax  The following syntax was used for SPSS: MIXED ABC WITH Mon_post_b stroke_or_not gender c_age c_TUG c_CES_D c_state c_BBS_b c_social_b c_sixMW_tenM /CRITERIA=CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) /FIXED=Mon_post_b stroke_or_not Mon_post_b*stroke_or_not c_age gender c_TUG c_CES_D c_state c_BBS_b c_social_b c_sixMW_tenM | SSTYPE(3) /METHOD=REML /PRINT=G SOLUTION TESTCOV /RANDOM=INTERCEPT Mon_post_b c_TUG c_state c_sixMW_tenM c_CES_D | SUBJECT(id#) COVTYPE(DIAG).  158  Appendix XXII: Model B (continued) 3. Output Estimates of fixed effects (model B) Estimates of Fixed Effects  a  95% Confidence Interval Parameter Intercept  Estimate  Std. Error  df  t  Sig.  Lower Bound  Upper Bound  86.453340  2.415541  163.615  35.790  .000  81.683688  91.222992  -.156817  .095335  138.120  -1.645  .102  -.345321  .031687  -8.298721  2.286155  192.947  -3.630  .000  -12.807785  -3.789656  .417975  .145439  166.552  2.874  .005  .130833  .705117  c_age  -.001242  .090043  156.835  -.014  .989  -.179095  .176612  gender  -3.892394  1.928870  147.189  -2.018  .045  -7.704250  -.080538  c_TUG  -.402255  .130120  47.071  -3.091  .003  -.664013  -.140496  c_CES_D  -.273205  .071416  130.066  -3.826  .000  -.414492  -.131918  c_state  -.125362  .058532  112.208  -2.142  .034  -.241333  -.009392  c_BBS_b  .643670  .184790  228.814  3.483  .001  .279564  1.007777  c_social_b  .639203  .423674  158.776  1.509  .133  -.197560  1.475967  c_sixMW_tenM  .269130  .074579  220.486  3.609  .000  .122152  .416108  Mon_post_b stroke_or_not Mon_post_b * stroke_or_not  a. Dependent Variable: ABC.  159  Appendix XXII: Model B (continued) 3. Output (continued) Estimates of covariance parameters (model B) Estimates of Covariance Parameters  a  95% Confidence Interval Parameter  Estimate  Residual  Std. Error  Wald Z  Sig.  Lower Bound  Upper Bound  39.902900  3.798262  10.506  .000  33.111634  48.087069  72.932360  13.284405  5.490  .000  51.035988  104.223105  Intercept + Mon_post_b +  Var: Intercept  c_TUG + c_state +  Var: Mon_post_b  .169755  .092021  1.845  .065  .058668  .491184  Var: c_TUG  .145426  .093132  1.561  .118  .041450  .510223  Var: c_state  .080715  .044913  1.797  .072  .027122  .240211  Var: c_sixMW_tenM  .121463  .053811  2.257  .024  .050973  .289430  Var: c_CES_D  .049539  .052642  .941  .347  .006172  .397609  c_sixMW_tenM + c_CES_D [subject = id#]  a. Dependent Variable: ABC.  160  Appendix XXIII: Written Model, Syntax and Selected Output for Model C 1. Written model (same as Model B but with interaction terms added to the composite model) Level 1: 𝑌𝑌𝑖𝑖𝑖𝑖 = 𝜋𝜋0𝑖𝑖 + 𝜋𝜋1𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 + 𝜋𝜋3𝑖𝑖 𝑇𝑇𝑇𝑇𝑇𝑇𝑖𝑖𝑖𝑖 + 𝜋𝜋4𝑖𝑖 6𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 + 𝜋𝜋5𝑖𝑖 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖 + 𝜋𝜋6𝑖𝑖 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖𝑖𝑖 + 𝜀𝜀𝑖𝑖𝑖𝑖 Level 2: 𝜋𝜋0𝑖𝑖 = 𝛾𝛾00 + 𝛾𝛾01 𝑆𝑆𝑆𝑆𝑖𝑖 + 𝛾𝛾02 𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖 + 𝛾𝛾03 𝐴𝐴𝐴𝐴𝐴𝐴𝑖𝑖 + 𝛾𝛾04 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑖𝑖 + 𝛾𝛾05 𝐵𝐵𝐵𝐵𝑆𝑆𝑖𝑖 + 𝜁𝜁0𝑖𝑖 𝜋𝜋1𝑖𝑖 = 𝛾𝛾10 + 𝛾𝛾11 𝑆𝑆𝑆𝑆𝑖𝑖 + 𝜁𝜁1𝑖𝑖 𝜋𝜋3𝑖𝑖 = 𝛾𝛾30 + 𝜁𝜁3𝑖𝑖  𝜋𝜋4𝑖𝑖 = 𝛾𝛾40 + 𝜁𝜁4𝑖𝑖  𝜋𝜋5𝑖𝑖 = 𝛾𝛾50 +𝜁𝜁5𝑖𝑖  𝜋𝜋6𝑖𝑖 = 𝛾𝛾60 + 𝜁𝜁6𝑖𝑖  Composite:  𝑌𝑌𝑖𝑖𝑖𝑖 = 𝛾𝛾00 + 𝛾𝛾01 𝑆𝑆𝑆𝑆𝑖𝑖 + 𝛾𝛾02 𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖 + 𝛾𝛾03 𝐴𝐴𝐴𝐴𝐴𝐴𝑖𝑖 + 𝛾𝛾04 𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝐼𝑖𝑖 + 𝛾𝛾05 𝐵𝐵𝐵𝐵𝐵𝐵𝑖𝑖 + 𝛾𝛾10 𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 + 𝛾𝛾11 𝑆𝑆𝑆𝑆𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 + 𝛾𝛾30 𝑇𝑇𝑇𝑇𝑇𝑇𝑖𝑖𝑖𝑖 + 𝛾𝛾40 6𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖  +𝛾𝛾50 𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝐶𝑖𝑖𝑖𝑖 + 𝛾𝛾60 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖𝑖𝑖 + 𝜸𝜸𝟕𝟕𝟕𝟕 𝑺𝑺𝑺𝑺𝒊𝒊 𝑪𝑪𝑪𝑪𝑪𝑪𝑪𝑪𝒊𝒊𝒊𝒊 + 𝜸𝜸𝟖𝟖𝟖𝟖 𝑺𝑺𝑺𝑺𝒊𝒊 𝟔𝟔𝟔𝟔𝟔𝟔𝒊𝒊𝒊𝒊 + (𝜁𝜁0𝑖𝑖 + 𝜁𝜁1𝑖𝑖 𝑀𝑀𝑀𝑀𝑀𝑀𝑖𝑖𝑖𝑖 + 𝜁𝜁3𝑖𝑖 𝑇𝑇𝑇𝑇𝑇𝑇𝑖𝑖𝑖𝑖 + 𝜻𝜻𝟕𝟕𝟕𝟕 𝑺𝑺𝑺𝑺𝒊𝒊 𝟔𝟔𝟔𝟔𝟔𝟔𝒊𝒊𝒊𝒊 + 𝜻𝜻𝟖𝟖𝟖𝟖 𝑺𝑺𝑺𝑺𝒊𝒊 𝑪𝑪𝑪𝑪𝑪𝑪𝑪𝑪𝒊𝒊𝒊𝒊 + 𝜁𝜁6𝑖𝑖 𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑆𝑖𝑖𝑖𝑖 + 𝜀𝜀𝑖𝑖𝑖𝑖 )  161  Appendix XXIII: Model C (continued) 2. Syntax The following SPSS syntax was used:  MIXED ABC WITH Mon_post_b stroke_or_not gender c_age c_TUG c_CES_D c_state c_BBS_b c_social_b c_sixMW_tenM /CRITERIA=CIN(95) MXITER(100) MXSTEP(5) SCORING(1) SINGULAR(0.000000000001) HCONVERGE(0, ABSOLUTE) LCONVERGE(0, ABSOLUTE) PCONVERGE(0.000001, ABSOLUTE) /FIXED=Mon_post_b stroke_or_not Mon_post_b*stroke_or_not c_age gender c_TUG c_CES_D c_state c_BBS_b c_social_b c_sixMW_tenM stroke_or_not*c_CES_D stroke_or_not*c_sixMW_tenM | SSTYPE(3) /METHOD=REML /PRINT=G SOLUTION TESTCOV /RANDOM=INTERCEPT Mon_post_b c_TUG c_state stroke_or_not*c_CES_D stroke_or_not*c_sixMW_tenM | SUBJECT(id#) COVTYPE(DIAG).  162  Appendix XXIII: Model C (continued) 3. Output Estimates of fixed effects (model C) Estimates of Fixed Effects  a  95% Confidence Interval Parameter Intercept  Estimate  Std. Error  df  t  Sig.  Lower Bound  Upper Bound  92.089414  2.454941  181.803  37.512  .000  87.245574  96.933253  -.119648  .089841  123.221  -1.332  .185  -.297479  .058184  -9.836193  2.089515  175.023  -4.707  .000  -13.960081  -5.712305  .254962  .143827  174.755  1.773  .078  -.028899  .538824  c_age  -.088098  .082081  138.525  -1.073  .285  -.250391  .074195  gender  -3.044169  1.681984  115.669  -1.810  .073  -6.375651  .287314  c_TUG  -.304908  .140049  68.226  -2.177  .033  -.584354  -.025461  c_CES_D  -.036809  .088946  357.803  -.414  .679  -.211732  .138114  c_state  -.110711  .052798  87.332  -2.097  .039  -.215647  -.005775  c_BBS_b  .494350  .195602  250.680  2.527  .012  .109118  .879582  c_social_b  .509080  .377915  135.872  1.347  .180  -.238276  1.256435  c_sixMW_tenM  .018621  .074953  380.010  .248  .804  -.128754  .165995  -.407915  .129955  132.727  -3.139  .002  -.664966  -.150864  .530828  .137272  181.489  3.867  .000  .259974  .801681  Mon_post_b stroke_or_not Mon_post_b * stroke_or_not  stroke_or_not * c_CES_D stroke_or_not * c_sixMW_tenM a. Dependent Variable: ABC.  163  Appendix XXIII: Model C (continued) 3. Output (continued) Estimates of covariance parameters (model C)  Estimates of Covariance Parameters  a  95% Confidence Interval Parameter  Estimate  Residual  Std. Error  Wald Z  Sig.  Lower Bound  Upper Bound  34.546532  3.167303  10.907  .000  28.864523  41.347049  50.441926  10.516097  4.797  .000  33.522297  75.901359  Intercept + Mon_post_b +  Var: Intercept  c_TUG + c_state +  Var: Mon_post_b  .193809  .085128  2.277  .023  .081940  .458407  Var: c_TUG  .163421  .093392  1.750  .080  .053316  .500906  Var: c_state  .043065  .039585  1.088  .277  .007107  .260940  Var: stroke_or_not *  .196645  .126519  1.554  .120  .055723  .693959  .509020  .145402  3.501  .000  .290796  .891008  stroke_or_not * c_CES_D + stroke_or_not * c_sixMW_tenM [subject = id#]  c_CES_D Var: stroke_or_not * c_sixMW_tenM a. Dependent Variable: ABC.  164  Appendix XXIV: The Frenchay Activities Index (FAI) (adapted from Holbrook & Skilbeck, 1983) We are interested in finding out how often you carry out some common activities. Please check which best describes how often you perform each activity. In the last 3 months how often have you carried out these activities? Never  Less than once per week  1 or 2 times per week  Most days  a) Prepare a main meal (not just a snack) b) Wash up (cleaning up after a meal, by yourself or sharing the task) c) Wash clothes (e.g. loading and unloading the washing machine) d) Light housework (e.g. dusting or tidying) e) Heavy housework (e.g. vacuuming or making beds) f) Local shopping (e.g. groceries or clothes) g) Attend social activities (e.g. go out to play cards, see movie or go to church) h) Walk outside for more than 15 minutes i) Take part in a hobby activity j) Drive a car or take the bus In the last 6 months how often have you carried out the following activities? Never  Once or twice only  Once or twice per month  Once or twice per week  a) Travel, go on outings or for car rides (traveling for pleasure or vacation, not just routine trips) b) Gardening or yard work c) Household or car maintenance d) Reading books (not just magazines or newspapers) e) Paid work Sub-domains: Domestic: questions a) to e) in first section Work/Leisure: questions g), i) in first section; a), c), e) in second section Outdoors: questions h), j) in first section; b), d) in second section  165  Appendix XXV: Selected SPSS Output for (a) the Unconditional Means Model (UMM), (b) the Unconditional Growth Model (UGM) and (c) the Final Model with Total FAI as the Dependent Variable (a) Results of UMM: Estimates of Fixed Effects  a  95% Confidence Interval Parameter Intercept  Estimate  Std. Error  27.030440  .704089  df 176.953  t 38.391  Sig.  Lower Bound  .000  Upper Bound  25.640947  28.419932  a. Dependent Variable: FAI_15_total.  Estimates of Covariance Parameters  a  95% Confidence Interval Parameter  Estimate  Residual Intercept [subject = ID]  Variance  Std. Error  Wald Z  Sig.  Lower Bound  Upper Bound  13.822239  1.133515  12.194  .000  11.769939  16.232394  82.754964  9.436337  8.770  .000  66.181059  103.479517  a. Dependent Variable: FAI_15_total.  166  Appendix XXV: Selected SPSS Output for (a) the Unconditional Means Model (UMM), (b) Unconditional Growth Model (UGM) and (c) the Final Model with Total FAI as the Dependent Variable (continued) (b) UGM: Estimates of Fixed Effects  a  95% Confidence Interval Parameter Intercept months_post_baseline  Estimate  Std. Error  df  t  Sig.  Lower Bound  Upper Bound  26.582323  .768459  177.615  34.592  .000  25.065838  28.098807  .069008  .049818  148.419  1.385  .168  -.029437  .167453  a. Dependent Variable: FAI_15_total.  Estimates of Covariance Parameters  a  95% Confidence Interval Parameter  Estimate  Residual  Std. Error  Wald Z  Sig.  Lower Bound  Upper Bound  11.905099  1.392881  8.547  .000  9.465499  14.973471  Intercept +  UN (1,1)  86.378170  11.418425  7.565  .000  66.662674  111.924527  months_post_baseline  UN (2,1)  -.532253  .607207  -.877  .381  -1.722356  .657850  UN (2,2)  .089983  .055494  1.621  .105  .026866  .301379  [subject = ID]  a. Dependent Variable: FAI_15_total.  167  Appendix XXV: Selected SPSS Output for (a) the Unconditional Means Model (UMM), (b) Unconditional Growth Model (UGM) and (c) the Final Model with Total FAI as the Dependent Variable (continued) (c) Final model: Estimates of Fixed Effects  a  95% Confidence Interval Parameter Intercept months_post_baseline stroke_or_not months_post_baseline *  Estimate  Std. Error  df  t  Sig.  Lower Bound  Upper Bound  32.619939  .835875  175.365  39.025  .000  30.970270  34.269608  -.020671  .067729  147.843  -.305  .761  -.154513  .113171  -12.678270  1.212122  175.162  -10.460  .000  -15.070514  -10.286026  .185699  .098788  147.208  1.880  .062  -.009528  .380925  stroke_or_not a. Dependent Variable: FAI_15_total.  Estimates of Covariance Parameters  a  95% Confidence Interval Parameter  Estimate  Residual  Std. Error  Wald Z  Sig.  Lower Bound  Upper Bound  11.989822  1.406173  8.527  .000  9.527596  15.088363  Intercept +  UN (1,1)  46.161547  7.347623  6.283  .000  33.790371  63.062002  months_post_baseline  UN (2,1)  .083548  .498897  .167  .867  -.894272  1.061369  UN (2,2)  .079038  .054942  1.439  .150  .020236  .308702  [subject = ID]  a. Dependent Variable: FAI_15_total.  168  Appendix XXVI: Selected SPSS Output for (a) the Unconditional Means Model (UMM), (b) Unconditional Growth Model (UGM) and (c) the Final Model with Domestic Sub-domain of FAI as the Dependent Variable (a) UMM: Estimates of Fixed Effects  a  95% Confidence Interval Parameter Intercept  Estimate 9.697750  Std. Error .317885  df 176.652  t  Sig.  30.507  Lower Bound  .000  Upper Bound  9.070409  10.325092  a. Dependent Variable: domestic.  Estimates of Covariance Parameters  a  95% Confidence Interval Parameter  Estimate  Residual Intercept [subject = ID]  Variance  Std. Error  Wald Z  Sig.  Lower Bound  Upper Bound  3.070645  .247445  12.409  .000  2.622022  3.596027  16.845815  1.925855  8.747  .000  13.464200  21.076741  a. Dependent Variable: domestic.  169  Appendix XXVI: Selected SPSS Output for (a) the Unconditional Means Model (UMM), (b) Unconditional Growth Model (UGM) and (c) the Final Model with Domestic Sub-domain of FAI as the Dependent Variable (continued) (b) UGM: Estimates of Fixed Effects  a  95% Confidence Interval Parameter Intercept months_post_baseline  Estimate  Std. Error  df  t  Sig.  Lower Bound  Upper Bound  9.474256  .354290  177.962  26.742  .000  8.775107  10.173406  .034475  .022023  155.082  1.565  .120  -.009028  .077979  a. Dependent Variable: domestic.  Estimates of Covariance Parameters  a  95% Confidence Interval Parameter  Estimate  Residual  Std. Error  Wald Z  Sig.  Lower Bound  Upper Bound  3.016640  .339509  8.885  .000  2.419494  3.761165  Intercept +  UN (1,1)  17.638456  2.435165  7.243  .000  13.456859  23.119446  months_post_baseline  UN (2,1)  -.070428  .127669  -.552  .581  -.320654  .179798  UN (2,2)  .002203  .011780  .187  .852  6.201709E-8  78.287401  [subject = ID]  a. Dependent Variable: domestic.  170  Appendix XXVI: Selected SPSS Output for (a) the Unconditional Means Model (UMM), (b) Unconditional Growth Model (UGM) and (c) the Final Model with Domestic Sub-domain of FAI as the Dependent Variable (continued) (c) Final model: Estimates of Fixed Effects  a  95% Confidence Interval Parameter  Estimate  Intercept months_post_baseline stroke_or_not months_post_baseline *  Std. Error  df  t  Sig.  Lower Bound  Upper Bound  11.777594  .431564  230.756  27.290  .000  10.927284  12.627904  -.032846  .030106  194.159  -1.091  .277  -.092223  .026531  -4.843608  .625856  230.015  -7.739  .000  -6.076751  -3.610464  .140621  .043845  190.691  3.207  .002  .054137  .227105  stroke_or_not a. Dependent Variable: domestic.  Estimates of Covariance Parameters  a  95% Confidence Interval Parameter  Estimate  Residual Intercept +  Var: Intercept  months_post_baseline  Var: months_post_baseline  Std. Error  Wald Z  Sig.  Lower Bound  Upper Bound  2.891498  .278969  10.365  .000  2.393315  3.493382  12.851188  1.546989  8.307  .000  10.150282  16.270783  .004075  .008192  .497  .619  7.923208E-5  .209583  [subject = ID] a. Dependent Variable: domestic.  171  Appendix XXVII: Selected SPSS Output for (a) the Unconditional Means Model (UMM), (b) Unconditional Growth Model (UGM) and (c) the Final Model with Leisure/Work Sub-domain of FAI as the Dependent Variable (a) UMM Estimates of Fixed Effects  a  95% Confidence Interval Parameter Intercept  Estimate 7.285673  Std. Error .241424  df 176.780  t 30.178  Sig.  Lower Bound  .000  Upper Bound  6.809229  7.762117  a. Dependent Variable: leisure_work.  Estimates of Covariance Parameters  a  95% Confidence Interval Parameter  Estimate  Residual Intercept [subject = ID]  Variance  Std. Error  Wald Z  Sig.  Lower Bound  Upper Bound  2.987384  .242051  12.342  .000  2.548725  3.501539  9.224276  1.111487  8.299  .000  7.283935  11.681497  a. Dependent Variable: leisure_work.  172  Appendix XXVII: Selected SPSS Output for (a) the Unconditional Means Model (UMM), (b) Unconditional Growth Model (UGM) and (c) the Final Model with Leisure/Work Sub-domain of FAI as the Dependent Variable (continued) (b) UGM Estimates of Fixed Effects  a  95% Confidence Interval Parameter Intercept months_post_baseline  Estimate  Std. Error  df  t  Sig.  Lower Bound  Upper Bound  7.174235  .279749  177.950  25.645  .000  6.622183  7.726288  .016900  .024115  153.816  .701  .484  -.030739  .064539  a. Dependent Variable: leisure_work.  Estimates of Covariance Parameters  a  95% Confidence Interval Parameter  Estimate  Residual  Std. Error  Wald Z  Sig.  Lower Bound  Upper Bound  2.365892  .269241  8.787  .000  1.892898  2.957076  Intercept +  UN (1,1)  10.118729  1.536600  6.585  .000  7.513894  13.626580  months_post_baseline  UN (2,1)  -.152672  .109496  -1.394  .163  -.367280  .061936  UN (2,2)  .030858  .012081  2.554  .011  .014325  .066470  [subject = ID]  a. Dependent Variable: leisure_work.  173  Appendix XXVII: Selected SPSS Output for (a) the Unconditional Means Model (UMM), (b) Unconditional Growth Model (UGM) and (c) the Final Model with Leisure/Work Sub-domain of FAI as the Dependent Variable (continued) (c) Final model: Estimates of Fixed Effects  a  95% Confidence Interval Parameter  Estimate  Std. Error  df  t  Sig.  Lower Bound  Upper Bound  Intercept  8.897598  .338955  176.909  26.250  .000  8.228682  9.566514  months_post_baseline  -.005186  .033223  156.539  -.156  .876  -.070809  .060437  -3.620944  .490826  175.977  -7.377  .000  -4.589607  -2.652282  .043407  .048206  153.421  .900  .369  -.051828  .138641  stroke_or_not months_post_baseline * stroke_or_not  a. Dependent Variable: leisure_work.  Estimates of Covariance Parameters  a  95% Confidence Interval Parameter  Estimate  Residual  Std. Error  Wald Z  Sig.  Lower Bound  Upper Bound  2.371761  .269800  8.791  .000  1.897765  2.964145  Intercept +  UN (1,1)  6.860569  1.213407  5.654  .000  4.850772  9.703076  months_post_baseline  UN (2,1)  -.110916  .098964  -1.121  .262  -.304882  .083050  UN (2,2)  .030397  .012052  2.522  .012  .013975  .066117  [subject = ID]  a. Dependent Variable: leisure_work.  174  Appendix XXVIII: Selected SPSS Output for (a) the Unconditional Means Model (UMM), (b) Unconditional Growth Model (UGM) and (c) the Final Model with Outdoors Sub-domain of FAI as the Dependent Variable (a) UMM: Estimates of Fixed Effects  a  95% Confidence Interval Parameter Intercept  Estimate  Std. Error  10.172852  .241829  df 175.560  t 42.066  Sig.  Lower Bound  .000  9.695585  Upper Bound 10.650119  a. Dependent Variable: outdoors.  Estimates of Covariance Parameters  a  95% Confidence Interval Parameter  Estimate  Residual Intercept [subject = ID]  Variance  Std. Error  Wald Z  Sig.  Lower Bound  Upper Bound  2.370816  .192886  12.291  .000  2.021368  2.780675  9.498810  1.118826  8.490  .000  7.540666  11.965441  a. Dependent Variable: outdoors.  175  Appendix XXVIII: Selected SPSS Output for (a) the Unconditional Means Model (UMM), (b) Unconditional Growth Model (UGM) and (c) the Final Model with Outdoors Sub-domain of FAI as the Dependent Variable (continued) (b) UGM: Estimates of Fixed Effects  a  95% Confidence Interval Parameter Intercept  Estimate  Std. Error  df  t  Sig.  Lower Bound  Upper Bound  10.015480  .270428  176.751  37.036  .000  9.481797  10.549164  .023781  .020853  150.446  1.140  .256  -.017422  .064984  months_post_baseline a. Dependent Variable: outdoors.  Estimates of Covariance Parameters  a  95% Confidence Interval Parameter  Estimate  Residual  Std. Error  Wald Z  Sig.  Lower Bound  Upper Bound  2.035730  .232196  8.767  .000  1.627915  2.545708  Intercept +  UN (1,1)  9.725673  1.434701  6.779  .000  7.283717  12.986325  months_post_baseline  UN (2,1)  -.063856  .092817  -.688  .491  -.245773  .118061  UN (2,2)  .016419  .009504  1.728  .084  .005280  .051058  [subject = ID]  a. Dependent Variable: outdoors.  176  Appendix XXVIII: Selected SPSS Output for (a) the Unconditional Means Model (UMM), (b) Unconditional Growth Model (UGM) and (c) the Final Model with Outdoors Sub-domain of FAI as the Dependent Variable (continued) (c) Final model: Estimates of Fixed Effects  a  95% Confidence Interval Parameter  Estimate  Intercept months_post_baseline stroke_or_not months_post_baseline *  Std. Error  df  t  Sig.  Lower Bound  Upper Bound  12.072462  .297524  172.751  40.576  .000  11.485211  12.659712  .013226  .028688  151.479  .461  .645  -.043455  .069906  -4.351257  .432489  172.868  -10.061  .000  -5.204896  -3.497618  .021953  .041824  149.735  .525  .600  -.060688  .104594  stroke_or_not a. Dependent Variable: outdoors.  Estimates of Covariance Parameters  a  95% Confidence Interval Parameter  Estimate  Residual  Std. Error  Wald Z  Sig.  Lower Bound  Upper Bound  2.047541  .234409  8.735  .000  1.636004  2.562601  Intercept +  UN (1,1)  4.972624  .965282  5.151  .000  3.398999  7.274787  months_post_baseline  UN (2,1)  -.041274  .079918  -.516  .606  -.197911  .115363  UN (2,2)  .016504  .009599  1.719  .086  .005279  .051599  [subject = ID]  a. Dependent Variable: outdoors.  177  Appendix XXIX: Test of Assumptions of Normality, Linearity and Homoscedasticity of Residuals Figure 13 Plot of Predicted Values of the FAI against Residuals Showing Assumptions Met  178  Appendix XXX: Outliers Analyses (Regression Analysis of FAI on ABC and Covariates) Examination of extreme values and histograms of all independent and dependent variables revealed a few extreme scores for BBS, TUG and ISEL that appeared to be disconnected from the rest of their distributions. Figure 14 presents all the histograms and Table 7 presents the extreme values of all the variables. Since ISEL has a fairly normal distribution form and the disconnected scores were not far-disconnected, this variable was not transformed. Transformation of BBS did not change distribution form so this variable was also not transformed. Logarithmic transformation of TUG appeared to improve distribution form so the transformed variable was used for the regression analyses; however, the results were not different from using the untransformed TUG.  179  Appendix XXX: Outliers Analyses (Regression Analysis of FAI on ABC and Covariates) (continued) Figure 14 Histograms Showing Distribution of Predictors (a) 6MW, (b) BBS, (c) TUG & Transformed TUG, (d) CES-D, (e) STAI, (f) ABC, (g) ISEL at Baseline and (h) FAI at 12 Months  (a)  (b)  (c)  (c)  (d)  (e)  180  Appendix XXX: Outliers Analyses (Regression Analysis of FAI on ABC and Covariates) (continued) Figure 14 (continued) Histograms Showing Distribution of Predictors (a) 6MW, (b) BBS, (c) TUG & Transformed TUG, (d) CES-D, (e) STAI, (f) ABC, (g) ISEL at Baseline and (h) FAI at 12 Months  (f)  (g)  (h)  181  Appendix XXX: Outliers Analyses (Regression analysis of FAI on ABC and covariates) (continued) Table 7: Extreme Values of Baseline 6MWT, BBS, TUG, CES-D, STAI, ABC, ISEL and FAI Extreme Values Case Number Baseline 6 Min Walk  Highest  Distance (m)  Lowest  Baseline Berg Balance (/56)  Highest  Lowest  Baseline TUG Time  Highest  Lowest  Value  1  195  733  2  131  732  3  164  680  4  122  669  5  125  668  1  89  46  2  60  56  3  93  66  4  18  84  5  70  1  12  56  2  21  56  3  95  56  4  99  56  5  100  1  69  14  2  93  23  3  72  27  4  146  36  5  40  36  1  60  70.7  2  89  60.0  3  69  46.2  4  18  42.6  5  40  41.2  1  126  5.3  92  56  a  b  c  182  Extreme Values Case Number  Baseline CES-D total score  Highest  Lowest  Baseline State  Highest  Lowest  Basline ABC  Highest  Lowest  Value  2  128  5.4  3  124  5.6  4  104  5.7  5  163  6.2  1  97  43  2  65  39  3  88  37  4  44  34  5  41  32  1  176  0  2  174  0  3  173  0  4  172  0  5  170  0  1  41  67  2  97  67  3  11  57  4  37  57  5  47  56  1  195  20  2  186  20  3  181  20  4  178  20  5  174  1  109  100.00  2  125  100.00  3  160  100.00  4  162  100.00  5  165  100.00  1  44  9.38  d  20  e  f  183  Extreme Values Case Number  Baseline 6 item ISEL  Highest  Lowest  Fai15_12_total  Highest  Lowest  Value  2  89  10.00  3  69  13.13  4  93  21.25  5  40  23.75  1  136  18  2  9  17  3  54  17  4  81  17  5  145  17  1  97  3  2  146  5  3  181  8  4  65  8  5  42  8  1  126  44  2  168  44  3  131  43  4  182  42  5  138  1  93  0  2  96  2  3  28  6  4  74  7  5  71  7  g  41  h  i  184  Appendix XXXI: Collinearity Diagnostics: Regression Analysis of FAI on ABC and Covariates Tolerance and VIF values of the unadjusted and adjusted models are listed in Table 8 and 10. None of the tolerances approaches zero and none of the VIF is greater than 10. Table 9 and 11 presents the condition index and the variance proportions of all the predictors in the 2 models. None of the values of either the condition index or the variance proportions indicated that collinearity was a concern.  Table 8: Coefficients, Tolerance and VIF Values of the Unadjusted Model Coefficients  a  Standardized Unstandardized Coefficients Model 1  B (Constant)  Std. Error -1.633  .736  c_abc  .135  .037  c_BBS  .713  c_abc_bbs  .006  Coefficients Beta  95.0% Confidence Interval for B t  Sig.  Lower Bound  Upper Bound  Collinearity Statistics Tolerance  VIF  -2.219  .028  -3.088  -.178  .314  3.624  .000  .062  .209  .497  2.010  .139  .525  5.115  .000  .438  .989  .353  2.833  .003  .170  2.052  .042  .000  .012  .541  1.848  a. Dependent Variable: c_FAI_12  Table 9: Condition Index and Variance Proportions of the Unadjusted Model Collinearity Diagnostics  a  Variance Proportions Model  Dimension  Eigenvalue  Condition Index  (Constant)  c_abc  c_BBS  c_abc_bbs  1  1  2.138  1.000  .00  .07  .06  .06  2  1.263  1.301  .37  .03  .01  .06  3  .391  2.338  .33  .59  .04  .33  4  .208  3.208  .30  .30  .88  .54  a. Dependent Variable: c_FAI_12  185  Appendix XXXI: Collinearity Diagnostics: Regression Analysis of FAI on ABC and Covariates (continued) Table 10: Coefficients, Tolerance and VIF Values of the Adjusted Model Coefficients  a  Standardized Unstandardized Coefficients Model  B (Constant)  Std. Error -2.905  .837  c_BBS  .742  .137  c_abc  .119  c_abc_bbs partnering_status c_ISEL  Coefficients Beta  95.0% Confidence Interval for B t  Sig.  Lower Bound  Upper Bound  Collinearity Statistics Tolerance  VIF  -3.470  .001  -4.560  -1.249  .548  5.422  .000  .471  1.013  .352  2.844  .037  .275  3.188  .002  .045  .192  .483  2.071  .007  .003  .189  2.293  .023  .001  .013  .529  1.890  3.577  1.290  .169  2.774  .006  1.027  6.128  .967  1.034  .636  .295  .137  2.159  .033  .054  1.219  .898  1.113  a. Dependent Variable: c_FAI_12  186  Appendix XXXI: Collinearity Diagnostics: Regression Analysis of FAI on ABC and Covariates (continued) Table 11: Condition Index and Variance Proportions of the Adjusted Model Collinearity Diagnostics  a  Variance Proportions Model  Dimension Eigenvalue  Condition Index  (Constant)  c_BBS  c_abc  c_abc_bbs  partnering_status  c_ISEL  1  2.264  1.000  .01  .05  .06  .06  .01  .03  2  1.725  1.146  .13  .01  .02  .01  .14  .00  3  .940  1.552  .01  .02  .00  .02  .03  .82  4  .508  2.112  .07  .01  .14  .24  .58  .10  5  .366  2.488  .41  .10  .49  .14  .19  .00  6  .197  3.392  .38  .82  .29  .54  .06  .05  a. Dependent Variable: c_FAI_12  187  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0069229/manifest

Comment

Related Items