UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The relationship of self-efficacy with depression, pain, and health status in the arthritis self-management… McGowan, Patrick Thomas 1996

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

Item Metadata

Download

Media
831-ubc_1996-147967.pdf [ 10.5MB ]
Metadata
JSON: 831-1.0076940.json
JSON-LD: 831-1.0076940-ld.json
RDF/XML (Pretty): 831-1.0076940-rdf.xml
RDF/JSON: 831-1.0076940-rdf.json
Turtle: 831-1.0076940-turtle.txt
N-Triples: 831-1.0076940-rdf-ntriples.txt
Original Record: 831-1.0076940-source.json
Full Text
831-1.0076940-fulltext.txt
Citation
831-1.0076940.ris

Full Text

T H E RELATIONSHIP OF S E L F - E F F I C A C Y W I T H DEPRESSION, PAIN, A N D H E A L T H STATUS IN T H E ARTHRITIS S E L F - M A N A G E M E N T P R O G R A M  by PATRICK THOMAS McGOWAN  B.A., The University of British Columbia, 1969 M . S . W . , The University of British Columbia, 1975 A THESIS S U B M I T T E D I N P A R T I A L F U L F I L L M E N T O F T H E REQUIREMENTS FOR T H E D E G R E E OF DOCTOR OF PHILOSOPHY in T H E F A C U L T Y O F G R A D U A T E STUDIES (Interdisciplinary Studies) We accept this thesis as conforming to the required standard y  T H E UNIVERSITY OF BRITISH C O L U M B I A October 1996 © Patrick Thomas McGowan, 1996  In  presenting this  degree at the  thesis in  University of  partial  fulfilment  of  the  requirements  British Columbia, I agree that the  for  an advanced  Library shall make it  freely available for reference and study. I further agree that permission for extensive copying  of  department  this thesis for scholarly purposes may be granted or  by  his  or  her  representatives.  It  is  by the  understood  that  head of copying  my or  publication of this thesis for financial gain shall not be allowed without my written permission.  Department The University of British Columbia Vancouver, Canada  DE-6 (2/88)  ABSTRACT The Relationship of Self-efficacy with Depression, Pain and Health Status in the Arthritis Self-Management Program  Over the past decade results from a series of research studies have contributed to the development and evaluation of the Arthritis Self-Management Program (ASMP), a volunteer-led patient education program for persons with arthritis. To date, these studies have primarily focussed on program effectiveness, process, implementation, and dissemination. In these studies self-efficacy was identified as an important construct contributing to the program's effectiveness, however, the exact relationship between self-efficacy and health outcomes has not been determined. In this dissertation research I investigate the evidence of a causal relationship between self-efficacy and three program outcomes (a decrease in depression, less pain, and a higher self-rating of overall health status), and attempt to determine the nature of that relationship.  The research methodology involved the use of structural equation modeling (SEM) with two longitudinal samples, one (n=122) of 1991 A S M P participants in British Columbia, the other (n=189) of 1992 A S M P participants in Ontario. In the analysis self-efficacy was paired separately with depression, pain and perceived health status.  The results of the S E M failed to confirm a dominant causal relationship from selfefficacy to depression, or to pain. This may indicate that these variables have a reciprocal or "spiral" relationship or that both sets of variables may be caused by factors not considered in the ii  analysis. The results of the S E M between self-efficacy and perceived health status did, however, show that higher self-rated health status leads to higher self-efficacy at a later time. The data did not show statistical significance for other causal patterns among these variables.  The findings suggest that self-efficacy may play a moderator role in the complex relationship involving individuals with arthritis, their behaviors, and health outcomes. As well, the findings have implications for health promotion planning and research in that they reinforce the complex interplay of psychological and behavioral variables (probably influenced by social variables) in programs which attempt to give individuals greater control over their health. The efficacy and effectiveness of the A S M P has been established in previous studies. This study in no way calls these into question. It does, however, suggest that the mechanism by which these effective outcomes are achieved warrants further investigation.  iii  T A B L E OF CONTENTS Page Abstract  ii  Table of Contents  iv  List of Tables  viii  List of Figures  x  Acknowledgements  xi  Chapter One The Arthritis Self-Management Program  1  The Problem  2  A Solution: The Arthritis Self-Management Program  6  Evaluation of the A S M P in British Columbia First B C Study - February to May, 1989 Second B C Study - January to April, 1991  10 12 15  Evaluation of A S M P Nationally Study Design Outcome Measures Analysis Plan . . Subjects Overview of Statistical Analyses Findings - A S M P Participants Findings - A S M P Leaders Subjects Overview of Statistical Analyses Effect of A S M P on Leaders in English-Speaking provinces Effect of A S M P on Leaders in French-Speaking populations Summary Impact of A S M P on Participants Impact of A S M P on Leaders Study Limitations  18 19 21 22 23 25 26 28 28 29 29 31 32 32 33 34  The Relationship Between Self-Efficacy and Health Status Background Canadian Experience with Self-Efficacy in the A S M P  36 36 37  iv  Theoretical Orientation of the Proposed Research Chapter Two Social Learning Theory - Self-Efficacy Theory  42 45  Nineteenth Century Psychology  45  Twentieth Century Contributions  46  Social Learning Theory Today  50  Self-Efficacy Theory  . . . 53  Performance Accomplishment Vicarious Experience Verbal Persuasion Emotional Arousal  58 59 59 59  Review of Applied Social Learning Theory Social Learning Theory in Health Education Programs . Self-Regulation and Self-Reinforcement Cognitive Restructuring Emotional Coping Conclusion Self-Efficacy Theory in Health Education Programs Self-Efficacy, Depression and Arthritis Depression As A Cause of Low Self-efficacy Reciprocal Relationship Between Self-efficacy, Performance Accomplishments and Depression Association of Self-efficacy and Depression Association Between Depression and Performance Self-efficacy, Performance and Depression Effect of Other Variables on the Relationship Between Self-efficacy and Depression Self-Efficacy and Pain Management in Arthritis Arthritis Pain Cold-Pressor pain Summary - Arthritis pain Childbirth Pain Tension Headache Pain Self-Efficacy and Perceived Health Status Self-Efficacy and Smoking Cessation Self-Efficacy and Cardiac Rehabilitation Conclusion Critical Appraisal of Self-Efficacy Theory Internal consistency Parsimony of the Theory Plausibility in Integrating with Prevailing Theories v  61 61 61 63 68 69 70 70 75 78 79 79 80 81 86 87 93 95 96 99 101 103 106 106 107 107 107 108  Ecological Validity Weaknesses of Self-Efficacy Theory  108 109  Chapter Three Measurement of Self-Efficacy and Health Status Variables Used in this Study  .113  Self-Efficacy Scales  113  Development of the Arthritis Self-Efficacy Scale  115  Procedure Employed in the Development of the Arthritis Self-Efficacy Scale . . . .  116  Comments  118  Visual Analogue Scale to Measure Pain  118  Causal Modeling of Self-Efficacy and Health States  120  Statement of Research Question  120  Method  120  Chapter Four Results  125  Procedures  125  British Columbia Sample Ontario Sample Subjects B C Sample Ontario Data Measures Self-Efficacy Depression Other measures Relationship between self-efficacy and depression  125 127 128 128 129 130 130 134 136 138  Modeling strategy  138  B C Results  143  Ontario Results  151  Relationship Between Self-Efficacy and Pain and Between Self-Efficacy and Health  157  Self-Efficacy and Pain Self-Efficacy and Health Chapter Five Discussion  158 160 162 vi  Implications for Health Promotion Planning Correlation Designs Behaviours Included in the Research Measurement of Behaviours  164 169 173 174  Implications Relative Effectiveness of Self-Efficacy in Bringing about Changes in Health Status  179  Sample Selection and Nature of the Data  184  180  References  192  List of Appendices Appendix A Appendix B Appendix C Appendix D Appendix E  225  A S M P Program Overview B C A S M P Study Questionnaire National A S M P Study Questionnaire Behaviours Identified in Nominal Group Process Precede-Proceed Model of Health Promotion Planning  vii  LIST OF TABLES  page Table 1.  Function of A S M P Personnel  9  Table 2.  A S M P Scales for Outcome Measures Employed During the First B C Study . . . 12  Table 3.  Comparison of 1991 B C Results and 1990 Stanford Results  Table 4.  Planned Distribution of Participants by Province and Language Group for the  17  National Study  20  Table 5.  A S M P Scales for Outcome Measures Employed During the National Study . . .  22  Table 6.  Questionnaires Completed by Participants, by Province and Language Group . . 23  Table 7.  Profile of Respondents in Treatment and Comparison Groups by Province During the National Evaluation of A S M P  25  Table 8.  Analysis of Post-treatment Results in Manitoba and Ontario  27  Table 9.  Statistics on Questionnaire Completion of Leaders by Province and Language Group  29  Table 10.  Analyses of Leaders' Results From English-Speaking Provinces  31  Table 11.  Analyses of Leaders' Results From French-Speaking Quebec  30  Table 12.  Correlations Between Changes in Arthritis Self-Management Knowledge and Health Behaviours and Changes in Health Status in 501 Subjects Who Took the Arthritis Self-Management Course, From Lorig et al., 1989 Correlations of Changes Between Self-Efficacy (SE) and Health Status in A Sample of 204 A S M P Participants Who Completed Pre-program and 4 Month Post-Program Questionnaires, From Lorig et al., 1989 Correlations of Changes in Arthritis Self-Efficacy with Changes in Three Health Status Indicators in A S M P Participants in British Columbia During 1990  Table 13.  Table 14.  Table 15.  38  39 40  Correlations of Changes in Arthritis Self-Efficacy with Changes in Health Status in A S M P in Provinces of East of Saskatchewan During 1992-1993 41  viii  Table 16.  Methods of Focusing Interventions to Enhance Elements of Social Cognitive Theory  52 I  Table 17.  Sources of Self-Efficacy  Table 18.  Gender Distributions, Type of Arthritis, and Mean Education Attainment of British Columbia and Ontario Samples  58 129  Table 19.  Factor Analysis of the Eleven Items Used in the B.C. and Ontario Evaluations 132  Table 20.  Parameter Estimates of the Items Comprising Depressive Effect Subscale, Based on the Four Factor Model, Three Time Points 135  Table 21.  Scale Reliability of Self-Efficacy and Depression Scales Used in the BC and  Table 22.  Ontario Studies  T" 138  Sequence of Nested Models of the Self-Efficacy and Depression Relationship  . 140  Table 23.  Mean Scores of British Columbia Sample on Self-efficacy, Depression and VAS Pain 143  Table 24.  Correlations Between Self-Efficacy and Depression Items for British Columbia Sample  146  Table 25.  BC Sample Data: Summary of Model-fitting Approach Using Ml Estimator . 148  Table 26.  BC Sample Data: Measurement Model Mm2 Using Ml Estimator (Indicator Loadings and Correlations Among Factors)  149  Table 27.  Mean Scores of Ontario Sample on Self-efficacy, Depression and VAS Pain . 152  Table 28.  Correlations Between Self-Efficacy and Depression Items for Ontario Sample  153  Table 29.  Ontario Sample Data: Summary of Model Fitting Approach  155  Table 30.  Ontario Sample Data: Measurement Model Mm2 (Indicator Loadings and Correlations Among Factors)  156  Table 31.  BC Data: Summary of Regression Model-fitting Approach  159  Table 32.  Ontario Data: Summary of Regression Model-fitting Approach  160  ix  LIST OF FIGURES Page Figure 1.  Influence of Efficacy and Outcome Expectations  Figure 2.  Roles of Locus of Control and Self-Efficacy  54  Figure 3.  Social Learning Theory: Components and Processes  57  Figure 4.  Theorized Relationship Between Self-Efficacy, Performance Accomplishments and Depression. From Yusaf and Kavanagh (1990)  Figure 5.  Figure 6.  Figure 7.  Figure 8.  Figure 9.  Figure 10.  Figure 11.  .54  78  The Saturated Model of Relationship Between Self-Efficacy and Depression  139  A Decision Tree for Selecting the Most Parsimonious Model Consistent with the Data  142  The Saturated Model of the Relationship Between Depression and Self-Efficacy, for the B C Sample Data  150  The Saturated Model of the Relationship Between Depression and Self-Efficacy, for the Ontario Data  157  The Saturated Regression Model of the Relationship Between Self-Efficacy and Pain for B C Sample Data  158  The Saturated Regression Model of the Relationship Between Self-Efficacy and Pain for Ontario Data  159  The Saturated Model of the Relationship Between Self-Efficacy and Health for Ontario Sample Data  161  x  ACKNOWLEDGEMENTS  I would like to thank Dr. Lawrence Green and Dr. Kate Lorig for their assistance in this research. Dr. Green, Director of the U B C Institute of Health Promotion Research, served as my supervisor during the doctoral program and his encouragement, inspiration, and support are greatly appreciated. Dr. Lorig, Senior Researcher at the Stanford Arthritis Research Centre, has provided encouragement, support and resources. I consider them as mentors and am grateful for the opportunities that were provided through our association. I would like to express special thanks to the members of my dissertation committee: Dr. Nand Kishor (Educational Psychology and Special Education), Dr. Nancy Waxier-Morrison (Anthropology and Sociology), and Dr. Patricia Vertinsky (Educational Studies). Their expertise and attention to detail is appreciated. As well, I am indebted to The Arthritis Society (British Columbia & Yukon) for the opportunity and support they provided to enable me to pursue doctoral studies.  xi  1  CHAPTER ONE  THE ARTHRITIS SELF-MANAGEMENT P R O G R A M  This chapter begins by assessing the importance of the arthritis problem, introducing the concept of self-management, and describing the Arthritis Self-Management Program (ASMP). A review of the methodology is provided, as well as results of three evaluations of the A S M P conducted by the author: two evaluation studies conducted in British Columbia, and an evaluation study conducted during the national implementation of the program. These evaluation studies provided the background for the study undertaken for this thesis, in that they revealed promising outcomes, but could not provide information regarding the relationship between selfefficacy and the outcome measures. Lastly, this chapter describes process evaluations which examined the relationships between self-efficacy and the program outcomes of pain, depression, and disability.  This doctoral thesis will focus on the causal relationship between self-efficacv and health status. The question this research seeks to probe beyond the analyses reported in the foregoing studies is:  Is there evidence of a causal relationship between self-efficacy and health status in the Arthritis Self-Management Program when the multiple covariates of both are controlled; and, if so, what is the nature of that relationship?  To address this question, the analysis here will use structural equational modeling (SEM) to assess alternative causal explanations for the relationships. Two data sets will be used in the analysis: a data set from a 1991 B C study, and an Ontario data set from a 1994 study. The variables that will be included in this analysis are self-efficacy, depression, pain, and perceived health status.  The Problem  In population studies estimating the prevalence of arthritis, the term "arthritis" is usually lumped under the broader category of Musculoskeletal Disorders (Activity Limitation Survey, Statistics Canada, 1989 & 1990), of which arthritis is the largest subcategory, consisting of about 60% of reported conditions. The term arthritis is used as an umbrella name for more than 120 different conditions that can cause aching and pain in the body's joints and connective tissues. Some forms are serious, and, if left untreated, can result in substantial or complete disability. Others may cause only a mild discomfort, which may be totally controlled with proper care and treatment.  Musculoskeletal disorders are among the most common of medical conditions and affect nearly one million Canadians (Reynolds et al., 1992). This is more than double the prevalence rate for all cancers combined in Canada (National Cancer Institute of Canada, 1989). In Canada these disorders are a leading cause of long-term disability (Badley, Rasooly, & Webster, 1994; Lee, Helewa, Smythe, Bombardier, & Goldsmith, 1985; Cunningham & Kelsey, 1984; Badley & Tennant, 1992; LaPlante, 1988).  3 The total cost ensuing from musculoskeletal disorders in 1986 was estimated to be $8.2 billion in Canada, which represents almost 2% of the GNP (Badley, 1995). Results from the 1986-1987 Health and Activity Limitation Survey (Statistics Canada, 1989 & 1990) showed that the prevalence of disabling musculoskeletal disorders was 47.6/1000 Canadian population, and that the most common type of disabling musculoskeletal disorder was arthritis, at 27.2/1000 population (Reynolds et al., 1992).  In population surveys conducted in several countries, the combined rates of arthritis and rheumatism reported are usually around 15% (Adams & Benson, 1992; Miles, Flegal & Harris, 1993; Lee, Helewa, Smythe, Bombardier & Goldsmith, 1985; Pincus, Callahan & Burkhauser, 1987; L a Vecchia, Negri, Pagano & Decarli, 1987; LaPlante, 1988; Collins, 1988; Cunningham & Kelsey, 1984; Badley & Tennant, 1992; Jacobson, Lingarde & Manthorpe, 1989). In the United States, arthritis affects approximately 38 million persons (Centre for Disease Control, 1990) . It is the most common self-reported chronic condition affecting women (NHIS, 19891991) , ranked ahead of self-reported hypertension, ischemic heart disease, and other chronic conditions; it is the most common cause of activity limitation among elderly women, and second most common among elderly men (Verbrugge, 1984; LaPlante, 1988).  The prevalence of disorders such as arthritis increases with age (Badley, 1992, Badley et al., 1994; Badley & Tennant, 1992, Cunningham & Kelsey, 1984; Jacobson et al. 1989; Lee et al., 1985; Reynolds et al., 1992), and therefore the number of Canadians who will be living with arthritis will increase, because the proportion of older persons in the population is increasing (Statistics Canada, 1990). The National Health Interview Study (1989-1991) reports that the  4 prevalence of self-reported arthritis and of associated activity limitation increases directly with age. Unfortunately, the ordinary senior must cope with more than one chronic health condition (US Public Health Service, D H E W Publication, 1979), and it has been shown that this comorbidity has a dramatically negative impact on functional capability (Yelin & Felts, 1990). The US Bureau of Census estimates that between 1985 and 2000, the number of persons over age 65 will increase by more than 9%, from 12% of the total population to 13% (US Bureau of the Census, 1986). If this were to occur, and if the age-adjusted prevalence rate of arthritis and associated disability did not change, the number of people with arthritis would increase by 15%, and the number with activity limitation associated with arthritis would increase by 45% (Yelin & Katz, 1990). For women over the age of 15 years living in the US, it has been projected that the prevalence of self-reported arthritis will increase from 22.8 million (22.7%) to 35.9 million (26.7%) between 1989-1991 and 2020 (Day, 1993).  No other group of diseases causes so much pain and disability over so long a period of time (Burckhardt, 1985). The impact arthritis has on the general population can not be based on measures of mortality (e.g., number of life years lost), because people do not generally die of arthritis but of another condition. A "Quality Adjusted Life Years" model (Reynolds et al., 1993) has been developed to calculate the impact of arthritis on Population Health Expectancy (PHE). This model: (a) examines the quantity of life issues (i.e., prevalence estimations and life table progressions) for individuals with arthritis; (b) examines the quality of life issues (i.e., the adjusted factors for changes in the quality of life) for individuals with arthritis; and (c) combines the quantitative and qualitative data to produce the quality adjusted life years lost among persons with arthritis, expressed as a reduction in PHE. Using Canadian data from The Health and  Activity Limitation Survey of Canada (Chambers, Reynolds & Badley, 1991), Reynolds et al. (1993) reported: Women who develop arthritis can expect to lose 3.3 years, and men 1.6 years of healthy life owing to arthritis and its comorbid conditions. Alternatively, the difference in P H E indicates that arthritis reduces 3.3 years of healthy life from a woman and 1.6 years from a man that she/he would otherwise have lived if he/she did not have arthritis, (p. 1045). The impact, calculated using this model, is substantial; the values of 3.3 and 1.6 healthy years of life lost (for women and men respectively) translate into about 50.9 million potentially healthy years of life lost among the Canadian population over fifteen years of age.  Since the expectation of cure is not a realistic one for most people with arthritis, and since medical interventions have only limited benefits for the most part, more emphasis needs to be placed on bettering the patients' quality of life and degree of independence, by developing their coping and self-management strategies. Holman and Lorig (1992) strongly argue for the feasibility and benefits of the self-management approaches taken by patients and health professionals in tackling chronic illnesses. They propose that seven broad skills are central to self-management: (a) minimizing or overcoming physical debility; (b) forming realistic expectations and emotional responses to the vicissitudes of the illness; (c) understanding and managing the symptoms; (d) learning how to judge the effects of medications and how to manage their use; (e) becoming proficient in techniques of problem solving; (f) communicating with health professionals; and (g) using community resources. To engage in self-management the individual must acquire new knowledge and master new skills, both of which appear to be dependent upon the individual's self-efficacy.  6 Canada harbours several health care philosophies that shape and guide its health care delivery systems. The most influential of these are the so-called medical model and public health perspective. The limitations of the medical model in dealing with chronic health conditions is well documented, and over the past decade we have witnessed increasing emphasis placed instead on public health and self-management strategies (Ottawa Charter, 1986; Achieving Health For A l l : A Framework for Health Promotion, 1986). The public health perspective is limited in this case, however, in that the scientific community has been unable to discover the causes of and important risk factors for arthritis. Nevertheless, a public health approach is justified when dealing with a chronic health condition as prevalent as arthritis, because implicit is the conceptualization of the entire population afflicted with arthritis as the "patient". Within this perspective, clinically important interventions that affect as little as five percent of the population are enormously important. A minimal intervention is defined by Hovell and Black (1989) as a technique that yields therapeutic effects with little expense of time or money and that has no side-effects. Specifically, these are therapeutic or preventative services and programs which: (a) result in either a small effect on a large portion of the population or large effects on a small portion of the population; (b) do not require much money, personnel, technology, or time to provide; and (c) involve little or no risk of side-effects.  A Solution: The Arthritis Self-Management Program  The Arthritis Self-Management Program was developed, piloted, implemented, and evaluated by the Stanford Arthritis Centre at the School of Medicine in Stanford, California. Kate Lorig, R N , Dr P H , has been the principal developer of the A S M P course, which teaches  people to become "health self-managers" (Lorig, 1990). After a series of research studies which examined both the impact and process of the A S M P , the course was re-designed to strengthen and enhance self-efficacy. Albert Bandura (1991) defines self-efficacy as "beliefs in one's capabilities to mobilize the motivation, cognitive resources, and courses of action needed to meet given situational demands" (Bandura, 1991, p. 229).  Two fundamental assumptions lie at the foundation of this program: (a) that individuals can learn the general principles of managing their own infirmity; and (b) that knowledgeable individuals practising self-care will experience less pain and an improved physical function, and will have reduced health care costs. There are also several ancillary goals that the A S M P works towards: (a) that health education be easily accessible and affordable; (b) that trained lay people be able to deliver a structured program effectively; and (c) that the lay instructors be acceptable both to the people with the health condition and to the professional community (Lorig, Lubeck, Kraines, Seleznick, & Holman, 1985).  The A S M P course emphasizes self-help and developing the skills necessary to ease problems caused by arthritis. Instead of simply making suggestions, as other programs do, the course actually teaches the methods of self-management. Participants are presented with a large number of self-management skills, offered practice in these skills, and encouraged to choose the techniques that will work best for them. Because the course emphasizes individual skill development, it benefits people with all different forms of arthritis. Each course has between 10 and 15 participants, and family members are permitted to join if they wish. The course is led by a pair of trained lay instructors and comprises six two-hour sessions spread over a period of two  8 or three months. It costs each individual about $25, is advertised by public announcements, and is offered in a community setting (e.g., senior centres, libraries, church basements). The,course is taught in groups, and an emphasis is placed on group discussion, practice, and the use of contracts and diaries for weekly feedback. The activities of each of the six sessions of the A S M P are shown in Appendix A .  The A S M P course is simple and inexpensive to run. The use of audio-visual materials and the demonstration of devices is not permitted, because such activities would shift the emphasis away from the curriculum, because they would allow for no quality control, and, most importantly, because information-giving would be substituted for skill-building and confidencebuilding.  The course was developed with lay people functioning as group leaders. The main reason for the use of lay leaders is that, in light of the considerable prevalence of arthritis and the high cost of using health care professionals as leaders, the course would otherwise be too expensive to be widely accessible. In fact, some of the most successful patient educators in the most successful programs have been lay people (Cox, 1979; Zapka & Mazur, 1977; Thompson, Gallagher & Nies, 1983; Lorig, Feigenbaum, Regan, Ung, & Holman, 1986). Also, acting as a group leader in the delivery of a self-help course may have several benefits for the leaders themselves, who are often drawn from an older and difficult-to-reach population. Each group leader receives 18 hours of training on how to use a standardized protocol. The three course materials are an instructor's manual, a 20-page pre-printed flip chart, and "The Arthritis Helpbook" for participants. These three adjuncts fit together and are used as the curriculum.  9 There were three major steps in implementing the program: providing training opportunities to enable selected volunteers to become trainers; providing workshops to teach volunteers how to become A S M P leaders; and encouraging and supporting the certified leaders to organize and offer the course in their own communities.  The personnel involved in the program include a provincial coordinator, trainers, leaders, and course participants. Table 1 provides a description of the function of the A S M P personnel.  Table 1 Function of A S M P Personnel FUNCTION  ASMP PERSONNEL Coordinator  Advertises and promotes the program Recruits volunteers Trains, supervises, and supports trainers and leaders  Trainers  Conduct three-day workshops to train leaders how to deliver the ASMP Use standardized A S M P Trainers' Manual  Leaders  Take the three-day workshop to learn how to lead the A S M P Use the A S M P Leader's Manual and The Arthritis Helpbook Deliver the A S M P in pairs  Participants  Take the six-session A S M P (two hours per week for six weeks) Receive a copy of The Arthritis Helpbook  10 Evaluation of the A S M P in British Columbia  Prior to the Arthritis Branch Community Support Project ( A B C Project) from 1988 to 1991, the A S M P course had been implemented successfully in the United States, Australia, and New Zealand (Lorig & Holman, 1993). At the time, there was no Canadian experience with the program, and therefore data describing its reception by, interest for, and relevance within our health care system was not available. It was not possible to tell whether the program would have the same positive effects on participants in Canada as it did elsewhere. The Canadian health care system differs from that of the United States in that it is a public rather than a private system. In our different system we did not know whether a shift from health care professionals to volunteer leaders as the source of patient education for persons with arthritis would be accepted. As well, characteristics of the patient population in previous studies of A S M P were expected to be different from those of the Canadian population.  The purpose of the British Columbia evaluations was to determine whether the program would bring about the same beneficial impacts here. The A B C Project was taken to communities where branch volunteers had expressed an interest in directing the program, and where individuals expressed an interest in participating in it. Because funding was received to implement the A S M P as a health promotion demonstration project, the program had to be accessible to all interested persons. Therefore, a one-group pre- and post-test design was chosen for the evaluation. Although such a design has several threats to both internal and external validity, it is still appropriate for the primary purpose of assessing replicability (i.e., whether similar program effects would occur).  11 During the first 20 months of the A B C Project, several research methodologies were employed to capture data describing implementation, thereby addressing the issue of replicability. Feasibility, viability, popularity and program costs were determined by the following methods: observations from key personnel, record keeping, documentation, cost accounting, the use of leader surveys to develop a "Leader Profile".  In this time period, it was observed that the A S M P was accepted both by the medical community and by patients themselves. Persons of all age groups, education levels, background work experience, and geographic areas of British Columbia volunteered to become leaders for this program in their communities. A survey of the costs of implementing and monitoring the A S M P reveals that the largest portion of costs was up-front training expenses, and that the cost per participant decreased as the number of participants increased (McGowan, 1990).  To address the major question of whether the same positive results that had been recorded elsewhere would be reproduced here, a pre- and post-program quantitative study was conducted with 96 A S M P participants. During the first of the six sessions, the participants were given an 11-page questionnaire by the course leaders, and were asked to complete it at home and return it at the following session. They were told that if they did not wish to complete the questionnaire they should simply return it uncompleted to the course leader. Either way, they were welcome to participate in the course.  Four months after the first questionnaire, when the participants had finished the course, they were mailed a second questionnaire and asked to complete and mail it to the coordinator in  12 the enclosed, addressed, and stamped envelope.  The Outcome Measures used in the pre- and-post experimental questionnaires were the same as those used by Dr. Kate Lorig and her Stanford colleagues in their research (Lorig et al., 1985). They include the measures shown in Table 2.  Table 2 A S M P Scales for Outcome Measures Employed During the First B C Study Outcome Measure  Variable Pain  Visual Analogue Scale (Dixon & Bird, 1981)  Disability  Stanford Health Assessment Questionnaire: Disability Scale  Depression  CES Depression Scale  Perceived Self-Efficacy  Arthritis Self-Efficacy Scale (Lorig, Chastain, Ung, Shoor, & Holman, 1989)  Quality of Life  Cantril Quality of Life Scale (Cantril, 1965)  Medication Usage  Medication Count  Doctor Visits  Frequency Count  Health-related Behaviours  Frequency Count  First B C Studv - February to Mav. 1989  A sample of 96 individuals in British Columbia, or 64% of the participant population, completed both the pre- and post-experimental questionnaires between February and May 1989. The data analysis consisted primarily of a series of paired t-tests which examined whether or not  13 the mean change from the baseline pre-test to the four-month post-test was zero for each of a number of outcome variables. These tests were done with matched pre- and post questionnaires from a sample of 96 subjects, and then for sub-samples divided on the basis of the subjects' diagnoses (e.g., Rheumatoid Arthritis or Osteoarthritis).  The analysis showed some positive and encouraging results. Although many of the changes were not statistically significant, the direction of the change was toward improvement for almost all of the response variables. Furthermore, there was certainly no evidence of any harmful effects; indeed, the almost unanimous agreement in the direction of the changes demonstrated the positive effects of the intervention.  The number of visits to the doctor for a routine check-up (i.e., doctor-suggested) was significantly lower at post-test (p_ =.028). The total number of visits, including those initiated by the patient for a specific problem, also showed a significant drop (p. =.038). Although the number of patient-initiated visits dropped, the decline was not statistically significant (p_ =.22), but since the number of this type of visit was so small at baseline, a very large sample size would have been needed to register a statistically significant decline. Several of the comparisons showed changes in a positive direction, but the sample size was too small to be statistically significant.  The Disability rating (as measured by the Stanford Health Assessment Questionnaire) exhibited no significant change (p_=095), but there was a slight trend towards improvement. A s with the number of patient-initiated visits, the level of disability was low (mean score of 1.0) in the pre-experimental test.  14 There were no statistically significant changes on any of the three self-efficacy subscales, but changes on the Pain and Other Symptoms subscales were toward improvement, and the Function subscale recorded virtually no change.  There was no significant change on the CES-Depression scale (p. =.43). The amount of physical activity and the use of medications, which were also measured, similarly did not manifest significant changes.  The subgroup with a diagnosis of Rheumatoid Arthritis (46 of 96 cases) yielded results in line with those of the whole group. There were significant declines in the total number of visits to the doctor (p_ =.043), and number of doctor-suggested visits (p =.027). Although no other response variables showed a statistically significant change from baseline to post-test, the direction of the change, however slight, was always toward improvement. The results are similar for the subgroup with a diagnosis of Osteoarthritis (37 cases), but the small sample size resulted in generally higher p_ values.  The research team concluded that for future research endeavours examining the impact of the A S M P , much larger sample sizes would be needed to obtain statistically meaningful results to compare with those observed by Dr. Lorig in her studies. The small sample size in this study notwithstanding, the team did observe an improvement in the quality of life, a decrease in the number of doctor visits, and a direction of change in all the outcome variables towards improvement, all of which are indications for the positive effects of the A S M P . With respect to the disability measure, there is no recent evidence anywhere showing that the A S M P improves  15 the state of the disability, but in a four-year follow-up study, it was found that A S M P participants had a slower rate of disability degeneration than those in comparison groups (Lorig & Holman, 1989).  Second B C Study - January to April. 1991  This sample was comprised of the 149 persons who had completed the course ending April, 1991 as well as the two questionnaires. As with the first sample, data analysis consisted primarily of a series of paired t-tests which examined the mean change from baseline pre-test to four month post-test scores on a number of outcome variables. Using the mean changes from baseline removes the variability due to different baseline scores (i.e., each subject is used as his/her own control). Paired t-tests is a repeated measure analysis when there are only two time points. In this second analysis, the research team did not find the same change in doctor visits that was observed with the first sample. This is most likely because of the relatively low number at baseline. A still larger sample size might detect meaningful change. Pain levels did decline significantly (p_ = .013). Quality of Life produced a result at variance with the first sample, for it was not significantly reduced. Disability level did not change significantly (p. = .811), although, again, there was a small improvement. Two of the three self-efficacy subscales, Pain and Other Symptoms, evidenced marked improvement (p_ =.008; p_ =.013). As well, a significant decrease in depression levels (p =.041) was found.  Analysis of this second sample is consistent with the findings of the first sample in that, despite not reaching significant levels of change with certain variables, the direction of change in  16 all variables was towards improvement. However, it should be noted that with multiple t-tests, as with most tests of significance, when individual t-tests are performed at a significance level of .05, the overall probability of a Type I error is considerably higher than .05. Here one should not be too hasty in bestowing "statistical significance" when the p-value is near .05. We have examined these patterns in the context of all the other p-values above, and we have noted the consistent direction of change from pre-test to post-test.  Information gathered from all methods provides a consistent testimony of expressed and measured positive impacts of the program. With respect to the quantitative group analysis, in spite of the small sample size, we were able to show statistically significant improvements in Quality of Life, fewer doctor visits, and a pattern of consistent improvements in all the outcome measures, indicating evidence for positive effects of the A S M P .  Table 3 compares the results obtained in this second B C study with those obtained by Dr. Lorig of the Stanford Arthritis Centre in a study she conducted between February and June, 1990.  17 Table 3 Comparison o f 1991 B C Results and 1990 Stanford Results  BRITISH C O L U M B I A Baseline Change  STANFORD Baseline Change V A S Pain  5.7  - 18%  5.2  -9%  C E S Depression  15.2  - 16%  14.7.  -9%  Self-Efficacv Pain  5.8  14%  5.8  7%  Other Symptoms  6.4  11%  6.3  6%  Function  7.1  3%  6.6  -  |  N - 97  N = 148  Note. One Group Pre- and Four Months Post-Design. K. Lorig (personal communication, 1991).  The A S M P had a positive impact in both studies, even though the effect was not so pronounced in the B C study as in the Stanford study. These results provided encouragement to proceed with further development o f A S M P i n Canada, and to try the program on a national scale.  18 Evaluation of A S M P Nationally  In 1992, the National Office of The Arthritis Society obtained a Seniors Independence Program (SIP) grant of $508,000 to cover the costs of implementing the Arthritis SelfManagement Program in the provinces of Saskatchewan, Manitoba, Ontario, Quebec, New Brunswick, Nova Scotia, Prince Edward Island, and Newfoundland. The A S M P had already been implemented in British Columbia, Alberta, and the Yukon Territory between 1988 and 1991, through assistance in the form of grants from the SIP to the British Columbia and Yukon Division of The Arthritis Society. O f the $508,000 grant awarded in 1992, a budget of $25,000 (5% of the total) was set aside to fund the evaluation.  In preparing the grant application, it was fully understood that $25,000 would not be sufficient to conduct a comprehensive evaluation of the program's implementation and impact; it is the usual practice in similar research and development projects to allocate at least 25% of the total budget to evaluation. However, the A S M P had already undergone research and evaluation extensively in the United States and gone through the pilot projects in British Columbia, Alberta, and the Yukon. Therefore it was decided that the focus of this limited evaluation would be on how the program affected the health status and three specific behaviours of the participants, interpreted via a set of standardized measures.  The evaluation of the national Arthritis Self-Management Program was conducted by an Evaluation Team coordinated by Patrick McGowan and formed in early 1992 to design and oversee research activity. The seven-member team consisted of Elizabeth Badley, PhD, of the  19 Arthritis Community Research & Evaluation Unit and The Wellesley Hospital Research Institute in Toronto, Ontario; Dennis Choquette, M D , of the Hopital Notre-Dame de Montreal in Montreal, Quebec; Gerry Devins, PhD, of The Clarke Institute of Psychiatry at The University of Toronto, Toronto, Ontario; Steven Edworthy, M D , of the Joint Injury and Arthritis Research Group at The University of Calgary, Calgary, Alberta; Lawrence Green, DrPH, at the Institute of Health Promotion Research of The University of British Columbia, Vancouver, British Columbia; Kate Lorig, DrPH, of the Stanford Arthritis Centre, at The Stanford University School of Medicine, Stanford, California; and Patrick McGowan, M S W , of The Arthritis Society (BC & Yukon Division) in Vancouver, British Columbia.  Study Design  In February of 1992, the Evaluation Team met in Vancouver, B C to plan the study design and to determine the outcome measures, and again in January of 1993 to review the preliminary results. Ethical approval to conduct the study was obtained by Dr. Edworthy through The University of Calgary.  The study design involved 900 individuals in the treatment (experimental) group (i.e., persons who would participate in the A S M P ) and 450 individuals who would serve as a comparison (control) group. Previous experience in B C had shown that it would not be possible to use a randomization process to allocate interested people into either experimental or control groups. It was hoped that persons in the comparison group would be from communities where the program had not been implemented yet, but where plans were in progress for implementation  20 in four month's time. Table 4 shows the intended number of individuals in each group for each province involved. The task of identifying persons to be included in each group was given to the Divisional A S M P Coordinators.  Table 4 Planned Distribution of Participants by Province and Language Group for the National Study Comparison Group  Treatment Group  Province Saskatchewan  100 (from Regina)  50 (from Saskatoon)  Manitoba  100 (Winnipeg)  50 (Brandon)  Ontario  200 (Toronto, Thunder Bay, & Sudbury)  100 (Ottawa & London)  Quebec  200 French-speaking (Montreal & Sherbrook) 100 English-speaking (Montreal & Sherbrook)  100 French-speaking (Hull, Trois Rivieres, Montreal, & Quebec City) 50 English-speaking (Hull, Trois Rivieres, Montreal, & Quebec City)  New Brunswick  70 English-speaking (Fredericton) 30 French-speaking (Fredericton)  30 English-speaking (Caraquet) 20 French-speaking (Caraquet)  Nova Scotia, Newfoundland, & Prince Edward Island  100 (Halifax & Sydney)  50 (Cornerbrook & Truro)  Sub-Total  670 English-speaking 230 French-speaking  330 English-speaking 120 French-speaking  TOTAL  900  450  The evaluation consisted of using questionnaires to assess A S M P participants once before they started the program, and again four months later. For each participant the change from preto post-program was computed. However, in the application submitted for SIP funding, it was  21 clearly understood that the A S M P would not be withheld from persons so that control groups could be formed. Therefore, the comparison group would consist of individuals living in communities where the program had not been implemented and to whom the program had not been administered. The comparison group would be requested to complete the questionnaire twice, four months apart, and the results would be used to ascertain whether changes would occur even without participation in the program.  The Evaluation Team estimated that a total of 900 program participants would be required for the experimental group, and 450 participants required for the comparison group. W i t h the limited evaluation budget (i.e., $25,000), research personnel could not be hired in every province, and, therefore, the task of insuring that a specified number of course participants in each province completed the questionnaires was given to each of the Divisional A S M P coordinators.  Outcome Measures  The pre- and post-course questionnaires contained outcome measures similar to those used by Dr. Kate L o r i g and her colleagues at Stanford, and by Patrick M c G o w a n in the two previous studies in British Columbia. They consist of the measures shown in Table 5.  22 Table 5 A S M P Scales for Outcome Measures Employed During the National Study  Variable  Outcome Measure  General Health  M O S Scale (Stewart & Ware, Eds., 1992)  Pain  Visual Analogue Pain Scale (Dixon, 1981)  Depression  MOS-Depression Scale (Stewart & Ware, Eds., 1992)  Perceived Self-Efficacy  Arthritis Self-Efficacy Scale (Lorig, Chastain et al., 1989)  Quality of L i f e  Cantril Quality of L i f e Scale (Cantril, 1965)  Illness Intrusiveness  Illness Intrusiveness Scale (Devins et al., 1983)  Mental Health  Mental Health Scale (Stewart & Ware, Eds., 1992)  Energy/Fatigue  Energy/Fatigue Scale (Stewart & Ware, Eds., 1992)  Positive Affect  Positive Affect Scale (Stewart & Ware, Eds., 1992)  Physical Limitations  Stanford Health Assessment Disability Scale (Fries, Spitz, & Young, 1982)  Medication Usage  Medication Count  Doctor Visits  Frequency Count  Health-related Behaviours  Frequency Count  Analysis Plan  The main goal of this evaluation was to assess the impact of A S M P on participants who were involved in the program implemented across Canada. Because there were two distinct language groups, the English and the French, and because the A S M P had not been translated, implemented, or evaluated i n French-speaking populations, the evaluation was carried out separately for the English-speaking provinces and for Quebec.  23 Subjects  In total, 1489 participants completed the pre-program questionnaires (839 individuals from the treatment groups, 346 individuals from the comparison groups, and 304 course leaders ) 1  from various provinces. Both the pre-program questionnaire and the four-month-later postprogram questionnaire were received from 632 treatment group members, of whom 16 did not complete the scales, and 168 comparison group members. Thus, 616 experimental subjects and 168 control subjects were included in the analyses. Table 6 divides this figure province by province.  Table 6 Questionnaires Completed by Participants, by Province and Language Group Four Months  Pre  Later  Saskatchewan Manitoba Ontario Quebec English French New Brunswick English French NS, N F L D , & PEI Total  Treat. 82 108 229  Comp. 13 47 92  Treat. 51 80 200  Comp.  Comp. 100 40 38  28 57  32 50  24 58  10 17  80 50 72  27  51  —  34 99  21 101  26 42  84  76  105  30 2 29  87  6 1 8  839  346  616  168  —  (%)  Treat. 38 26 13  31 200  —  Attrition Rate  —  Leaders participated in a different course and thus were not included in the A S M P evaluation analyses. However, their pre-treatment scores were included in other analyses, including analyses of yearly variation in arthritis symptoms. 1  24  It is apparent that the desired participation in the treatment and comparison groups was attained only in Manitoba, Ontario, and French-speaking Quebec. In Saskatchewan, New Brunswick, and the Atlantic provinces, only the treatment groups were adequately numerous; the comparison group was missing completely for Saskatchewan at post-test, and was inadequate in all four Atlantic provinces. Thus, the small number of persons in the treatment and especially in the comparison groups in some provinces limited the nature and detail of possible analyses. However, sample sizes in Manitoba, Ontario, and French-speaking Quebec provided adequate treatment and comparison groups to ascertain the effectiveness of the ASMP as it was being implemented across Canada.  Table 7 presents demographic data on subjects, separately for Manitoba, Ontario, French Quebec, and the provinces without comparison groups.  25 Table 7 Profile of Respondents in Treatment and Comparison Groups by Province During the National Evaluation of A S M P  Manitoba/Ontario  Quebec Comp. 60.9 13.9  Treat.  75  89  15.5 13.8  14.5 11.9  13.1 11.9  42.4 32.9 24.7  45.1 30.1 24.6  26.5 54.4 19.1  39.3 39.3 21.5  85  122  68  214  Treat. 60.2 13.9  Comp. 59.2 15.0  Treat. 61.9 13.8  89  80  81  13.5 12.5  16.0 15.8  O A (%) R A (%) Other (%)  35.4 32.5 32.1  #of Participants  280  Age (years)  M SD  Females (%) Arthritis (yrs.)  M SD  SK/NB/NS PEI/NFLD  Overview of statistical analyses  There were 14 critical outcome measures, which were grouped into three sets: quality of life measures (QLM), behavioural measures (BM), and arthritis-related doctor visits. The quality of life set consisted of the following measures: illness intrusiveness, pain, quality of life, health status, self-efficacy in relation to pain, self-efficacy in relation to other symptoms, depression, mental status, energy/fatigue, and positive affect. The behaviour measures included stretching, relaxing, and walking. The analyses of the Q L M and B M sets were undertaken using multivariate methods: multivariate analysis of covariance with pre-program scores as covariates, in cases where provinces had a comparison group (Manitoba, Ontario, and Quebec); and  26  multivariate analysis of variance, in cases where provinces did not have adequate numbers in their comparison groups (Saskatchewan, New Brunswick, Nova Scotia, Newfoundland, and Prince Edward Island). The multivariate approach minimizes experimentwise alpha (i.e., a Type 1 Error) and takes advantage of relatedness among the measures in each set. To identify measures that contributed most to the overall effects, univariate analysis of covariance or variance was performed on each outcome measure, as a follow-up to those multivariate analyses that revealed significant relationships.  The analyses evaluating the impact of the A S M P in this study focused on the Manitoba and Ontario samples, because only these two provinces had adequate numbers in both treatment and comparison groups (see Table 7). However, additional analyses were also conducted to determine whether the effectiveness of A S M P in provinces which lacked adequate comparison groups compared favourably to the findings in Manitoba and Ontario. Evaluation of the impact of A S M P in Quebec was carried out separately, because the majority of participants were French-speaking.  Findings - A S M P Participants  The effect of the A S M P as recorded in the post-treatment measures was evaluated for the Q L M and B M sets separately by multivariate analysis of covariance ( M A N C O V A s ) with pretreatment scores on all relevant dependent variables (i.e., measures) simultaneously covaried. The treatment group was entered as the independent variable, a between-subjects factor. Separate M A N C O V A s showed a significant overall effect in the treatment group for both the  27 Q L M set, F (10,344) = 2.13, p < .05, and for the B M set, F (3,358) = 5.18, p. < .01. A n  A N C O V A on doctor visits did not yield significant effects: F (1,362) = 1.89.  Following significant M A N C O V A s , univariate analyses of covariance (ANCOVAs) were performed on each dependent measure, with the single corresponding pre-treatment measures as covariates. Table 8 shows post-treatment scores, adjusted scores (post-treatment scores adjusted for pre-treatment scores), and the results of the univariate A N C O V A s . The adjusted scores were computed by first estimating the regression equation between the pre- and post-treatment scores, and then using the post-treatment scores that would have been obtained if all the pre-treatment scores had been the same. That is, the linear effects of the pre-treatment scores were removed.  Table 8 Analyses of Post-Treatment Results in Manitoba and Ontario Post-treatment scores Comparison Treatment M SD M SD QLM Illness Intrusiveness Pain Quality of life Health status SE Symptoms SE Pain Depression Mental status Energy Positive affect BM Stretching Relaxing Walking Doctor visits  Adjusted scores Treat. Comp.  F (1,362)  3.31  1.37  3.15  1.19  3.23  3.23  0.00  5.35  2.00  5.49  2.14  5.31  5.53  1,10  5.91  1.73  5.49  1.74  5.92  5.48  4.86*  2.21  0.86  2.25  0.87  2.17  2.28  1.82  59.80  19.52  53.33  22.47  59.42  53.71  7.87**  56.16  21.26  50.89  22.09  56.55  50.50  7.48**  1.40  0.89  1.49  0.92  1.41  1.48  0.75  3.36  0.91  3.21  0.97  3.36  3.20  4.03*  2.12  0.91  2.08  0.95  2.18  2.03  3.49x  2.76  0.98  2.63  1.06  2.78  2.61  3.17x  2.00  1.23  1.69  1.22  1.98  1.71  4.56*  2.60  2.71  1.17  2.05  2.40  1.36  12.53**  2.10  1.27  2.12  1.17  2.07  2.15  0.30  1.35  1.76  1.04  1.48  1.33  1.06  1.89  Note: SE = Self-Efficacy. **p < .01. *rj<.05, x <.10. E  28 Inspection of Table 8 reveals that participation in the A S M P produced improved measures of quality of life and self-efficacy, reduced pain and other symptoms, and improvements in the mental state of the participant. Increases in the behavioural measures of relaxation and stretching were also seen. Marginal improvements were found in energy levels and positive affect, but differences in disability levels between the treatment and comparison groups on post-test scores were not significant: F (1,362)=. 18. The A S M P resulted in overall improvement in the participants' quality of life and increased their engagement in beneficial physical activities. The greatest improvements occurred for those variables that were the immediate target of A S M P intervention: self-efficacy in dealing with arthritis symptoms and pain.  Findings - A S M P Leaders  Subjects. In total, 304 leaders in various provinces completed the pre-test questionnaire. Only 193 leaders, however, returned both the pre- and post-program questionnaires, and 11 of these questionnaires were incomplete. Thus, the analyses were conducted on data from 182 leaders. Table 9 shows by province, the number of leaders who completed the baseline questionnaire, and who completed both pre- and post-program questionnaires.  29 Table 9 Statistics on Questionnaire Completion by Leaders, by Province and Language Group  Saskatchewan Manitoba Ontario Quebec English French New Brunswick English French NS, N F L D , PEI Total  Pre-program 13 46 114  Post-program 4 31 73  Attrition (%) 69.2 32.6 35.9  15 34  12 18  20.0 47.1  16 11 55  9 7 28  17.31 12.11 49.1  304  182  40.1  Overview of Statistical Analyses The evaluation of the effects of A S M P on leaders was carried out separately for the English-speaking provinces and for Quebec. A n initial review of baseline survey scores showed no differences between the leaders from Manitoba and Ontario and the leaders from the other English-speaking provinces. Because there was no comparison group for leaders, the effect of A S M P was evaluated by comparing pre-program with post-program results only in terms of the difference in scores.  Effect of A S M P on Leaders in English-Speaking Provinces A separate M A N O V A conducted on each set of different scores showed significant effect of treatment for both the Q L M set, F (10,142) = 5.18, p < .001, and for the B M set, F (3,149) = 9.29, p_ < .001. A univariate A N O V A on doctor visits yielded no significant effects, F (1,151) = .50. The follow-up univariate A N O V A s were performed to identify measures most affected by  treatment. The results of these analyses (see Table 10) indicate improvements on the illness intrusiveness scale, the pain level scale, the self-efficacy pain and other symptoms subscales, and the frequency of relaxation and walking behaviours.  Table 10 Analyses of Leaders' Results from English-Speaking Provinces  Pre-treatment  Post-treatment  Illness Pain Quality of life Health status S E symptoms S E pain Depression Mental status Energy Positive affect  M 3.21 5.65 6.35 1.94 68.35 64.48 1.13 3.64 2.53 3.14  SD 1.33 2.67 1.98 0.89 19.11 20.96 0.70 0.77 0.97 0.99  M 2.99 4.55 6.13 2.01 72.51 69.22 1.02 3.72 2.59 3.25  SD 1.32 2.39 1.83 0.91 17.13 19.44 0.62 0.73 0.95 0.90  F(l,151) 9.14** 30.90** 1.33 1.19 8.43** 8.02** 3.78x 1.59 0.61 1.65  Stretching Relaxing Walking  1.80 1.94 1.98  1.20 2.38 1.22  1.97 3.10 2.22  1.22 2.77 1.20  3.20x 23.95** 4.56*  Doctor  1.27  1.45  1.17  1.40  .50  Note: **rj<.01, *rj<.05, xrj<.10, n=152.  31 Effect of the A S M P on Leaders in French-Speaking Populations Separate M A N O V A s were conducted on the Q L M and behavioural sets of outcome scores. The M A N O V A on the Q L M set yielded no significant effect at the .05 level: F(10,20)=1.04. A M A N O V A on the behavioural set also fell short of significance: F (3, 27) = 2.80, p_ < . 06. A univariate A N O V A on doctor visits did not yield a significant effect: F (1,29) = .13. The follow-up univariate A N O V A s were performed on the behavioural set to identify measures most affected by treatment. The results of these analyses (see Table 11) indicate a negative effect for the frequency of stretching behaviour: F (1,29) = 7.50, p < .01.  Table 11 Analyses of Leaders' Results from French-Speaking Quebec Post-treatment M SD 2.43 1.30 3.83 2.52 6.51 1.92 1.56 1.13 75.96 17.67 69.83 20.30 1.04 0.66 3.70 0.85 3.03 1.12 3.49 1.05  Illness Pain Quality of life Health status SE symptoms SE pain Depression Mental status Energy Positive affect  Pre-treatment M SD 2.59 1.30 4.30 2.91 7.50 2.52 1.33 1.18 75.77 20.60 70.66 22.70 0.92 0.61 3.81 0.64 0.98 3.29 3.63 0.89  Stretching Relaxing Walking  2.29 3.37 2.67  1.44 3.10 1.37  1.63 3.84 2.37  1.47 2.80 1.35  7.50** .51 1.30  Doctor  0.86  1.18  0.78  1.39  .13  Note: **p < .01. *p < .05, xp<.10, n=27.  F(l,29) .53 .68 2.36 2.71 .00 .03 1.49 .67 1.56 .70  32 Summary  Impact of A S M P on Participants In the English-speaking populations of Saskatchewan, Manitoba, Ontario, New Brunswick, Nova Scotia, Prince Edward Island, and Newfoundland, individuals who participated in the Arthritis Self-Management Program experienced health benefits. The data obtained from Manitoba and Ontario offered adequately populous comparison groups, and therefore facilitated the most rigorous statistical analysis.  In these two provinces, participants experienced a significant overall effect of treatment on the Quality of Life set of measures. Specifically, participation in A S M P resulted in improved quality of life, higher self-efficacy on both subscales (i.e., pain and other symptoms), better mental status, and an improvement in the individual's reserves of energy. Also, participants performed relaxation and stretching activities more frequently than before. Marginal improvement was also found for positive affect.  In the other English-speaking provinces - Saskatchewan, New Brunswick, Nova Scotia, Prince Edward Island, and Newfoundland^where comparison groups were not available, analyses showed that participants experienced less pain, improvement in both self-efficacy scales, less depression, and better mental status. They also engaged in more relaxation activities.  The results observed in the provinces where comparison groups were not available are consistent with the results obtained in the previous studies conducted in British Columbia, which  33 used a one-group pre- and post-test design. In spite of the limitation that the absence of comparison groups poses for all these provinces, analysis suggests that the effectiveness of the A S M P across the English-speaking provinces was fairly homogeneous.  A n analysis of the data obtained from the French population in Quebec showed only a few differences between the treatment and comparison group - namely that the participants in the treatment group engaged in more stretching and relaxation activities and visited their doctor more often. This was the first test of the A S M P in French, and it should be noted that the French Quebec attrition rate was higher than that of any of the other provinces and higher than the English Quebec rate. For the implementation, all program materials (i.e., The Arthritis Helpbook, The Trainer's Manual, The Leader's Manual) were translated into French, the leader training workshops were provided in English and simultaneously translated into French, the program was delivered in French, and, lastly, the evaluation questionnaires were translated into French. Cultural idiosyncrasies, and diminished sensitivity of the instruments in detecting and measuring change, may explain why positive results were not obtained.  Impact of A S M P on Leaders In the English-speaking provinces an analysis was completed using information obtained from 152 volunteer leaders who completed both the baseline and follow-up questionnaire. On average, leaders were much younger than participants (49 yrs. vs. 60 yrs.), and a larger percentage of leaders than of participants had Rheumatoid Arthritis (47% vs. 33%). The analysis of scores on the outcome measures showed that there were no significant differences between leaders from Manitoba and Ontario and those from the other English-speaking provinces.  34 The scores from the pre-program questionnaire of the leaders (n=152) were compared with the participants' scores (n=730), and it was found that leaders had higher pre-program scores on each scale of the Q L M set, but not on the behavioural set. A second analysis was conducted to determine whether participation in leading the A S M P had beneficial effects on the leaders. The results of these analyses indicate that leaders made improvements on the illnessintrusiveness scale, on the pain-level scale, on both self-efficacy scales, and in their frequency of walking (Figure 6).  In Quebec the analysis was conducted with the information obtained from 30 leaders. On average, leaders were 15 years younger than participants, and were affected by Rheumatoid Arthritis more than members of the participant group. A comparison of pre-treatment scores from the 30 leaders with scores from 180 Quebec participants showed that leaders had higher (i.e., better) scores on all the outcome measures except the number of doctor visits. A n analysis of baseline and follow-up scores of the 30 leaders showed that no significant changes in the outcome measures took place.  Study Limitations  The evaluation design had to accommodate the planned implementation activity (i.e., the main purpose of the grant), which produced inevitable research limitations. Additionally, because the study had a restricted budget, research personnel could not be hired, and therefore all recruitment and the majority of data gathering activities fell on the Divisional A S M P Coordinators as additional job responsibilities. The major limitations in the conducted national  35  study are: 1.  Adequate sample sizes for comparison groups were obtained only in Manitoba, Ontario and French-speaking Quebec.  2.  Participants were not randomly assigned to treatment and comparison groups. Despite this limitation, the statistical analyses showed that treatment and comparison groups in Manitoba and Ontario were similar in all pre-treatment measures.  3.  Subjects in the treatment groups completed the questionnaires at different times of the year than their counterparts in the comparison groups. Analyses that examined yearly variations in scores on outcome measures suggested that, had treatment and comparison groups been tested at the same time, the effects of ASMP would have been more pronounced in Manitoba and Ontario, if not elsewhere.  4.  French populations had had no previous encounter with the ASMP, and a psychometric analysis for the French questionnaire had not yet been conducted.  Despite these limitations, adequate sample sizes for both treatment and comparison groups in both pre-test and post-test rounds were obtained in the provinces of Manitoba and Ontario, and these permitted a rigorous analysis to be conducted.  36 The Relationship Between Self-Efficacy and Health Status  Background  Since the inception of the Arthritis Self-Management Program (ASMP) in the early 1980's, a series of linked research studies has been conducted. These studies demonstrate that: 1.  Lay leaders are able to lead the course, and obtain results similar to those in courses led by health professionals (Lorig et al., 1986);  2.  The program produces benefits such as increased physical activity, increased levels of self-efficacy, and improved health status (Lorig et al., 1985; Lorig & Holman, 1993; McGowan & Green, 1995);  3.  The key mechanism responsible for bringing about the improvements in health status appeared in early analyses to be self-efficacy (Lenker, Lorig, & Gallagher, 1984);  4.  Reinforcement (other than the series of self-management sessions and the improved sense of control and health status, which may be intrinsically reinforcing) is not needed to maintain these improvements which last as long as four years without further extrinsic reinforcement (Lorig & Holman, 1989; Holman, Mazonson, & Lorig, 1989);  5.  Significant and needed cost-savings may be accomplished through this program (Holman etal., 1989).  37 Canadian Experience with Self-Efficacy in the A S M P  Under the guidance of Patrick McGowan, a series of applications of the A S M P has been implemented in Canada, first in British Columbia (1988-1992), then in Alberta and the Yukon (1990-1992), then nationally in both official languages (1992-1994), and then in First Nations communities in B C (1991-1994). In 1994 the Osteoporosis Society of B C (OSTOP) received a federal grant to implement and evaluate The Osteoporosis Prevention and Self-Management Program. And, more recently, The Multiple Sclerosis Society of B C has received a grant from the British Columbia Health Research Foundation to develop a Multiple Sclerosis SelfManagement Program based on the work of the A S M P .  Published evaluation studies have shown that the A S M P is able to bring about improvements in self-efficacy levels (i.e., self-efficacy in managing pain and self-efficacy in handling arthritis symptoms) and in self-rated health status (i.e., pain level, quality of life, depression, and overall health). In a series of studies in 1993, Dr. Lorig demonstrated that selfefficacy could be enhanced by an intervention consisting of patient education, and that a patient education intervention designed to enhance self-efficacy yielded greater benefits to patients than similar interventions that did not emphasize self-efficacy. In researching the mechanisms through which the A S M P brings about benefits to patients, Lorig, Seleznick, et al. (1989) found that changes in behaviours did not strongly correlate with changes in health status (pain, disability, and depression). Table 12 (Lorig, Seleznick, et al., 1989) depicts the correlations of three health status outcomes (pain, disability, and depression) with changes in knowledge and behaviours. Only exercise showed significant correlations with two out of three of the health  38 outcomes in the hypothesized direction.  Table 12 Correlations Between Changes in Arthritis Self-Management Knowledge and Health Behaviours and Changes in Health Status in 501 Subjects Who Took the Arthritis Self-Management Course, from Lorig et al.. 1989 Pain Knowledge Correlation P 95% confidence interval Exercise Correlation P 95% confidence interval Relaxation Correlation P 95%) confidence interval Self-management activity index Correlation P 95% confidence interval  Disability  Depression  -0.05 0.26 0.04/-0.15  -0.01 0.85 0.08/-0.10  -0.05 0.45 0.07/-0.17  -0.14 0.004 -0.04/-0.23  -0.10 0.03 -0.01/-0.19  -0.02 0.75 0.09/-0.14  -0.04 0.37 0.05/-0.13  0.01 0.99 0.09/-0.09  -0.01 0.90 0.11/-0.13  -0.11 0.02 -0.01/-0.20  -0.07 0.13 0.02/-0.16  -0.03 0.67 0.09/-0.15  Note: Associations analyzed using Pearson correlations. Depression data were collected from 276 subjects only during the last 2 study years. P values calculated by Student's 2-tailed Mest; 95% confidence interval based on Fisher's transformation. The self-management activity index was derived by summing the scores for exercise, relaxation, and walking 4 blocks (scored as -1 for decreased speed; 0 for no change; +1 for increased speed). n=501.  Dr. Lorig found that changes in self-efficacy had a correlation with changes in health status higher than the correlations between changes in knowledge and behaviours and changes in health status. Table 13 illustrates associations of changes between self-efficacy and present and future health status related to arthritis.  39 Table 13 Correlations of Changes Between Self-efficacy (SE) and Health Status in a Sample of 204 A S M P Participants Who Completed Pre-Program and Four-Month Post-Program Questionnaires. From Lorig* Pain  Disability  Depression  Baseline SE with baseline HS Functional SE Other-Symptom SE PainSE  -0.24 -0.30 -0.23  -0.61 -0.33 -0.25  -0.29 -0.55 -0.38  Baseline SE with 4-month HS Functional SE Other-Symptom SE PainSE  -0.30 -0.30 -0.24  -0.64 -0.31 -0.22  -0.28 -0.34 -0.24  Baseline HS with 4-month SE Functional SE Other-Symptom SE PainSE  -0.22 -0.27 -0.22  -0.67 -0.38 -0.35  -0.14 -0.31 -0.25  4-month SE with 4-month HS Functional SE Other-Symptom SE PainSE  -0.39 -0.45 -0.42  -0.75 -0.43 -0.38  -0.30 -0.41 -0.35  Note: SE = Self-Efficacy; HS = Health Status. Associations analyzed using Pearson correlations. Data was obtained from a sample of 204 persons who took the Arthritis Self-Management Program. All correlations were significantly different from zero (E<0.01). * Source - K. Lorig (personal communication, November, 1995).  The correlation coefficients between baseline self-efficacy and four-month health status are of the same magnitude as the coefficients between baseline health status and four-month selfefficacy, and therefore the direction of change cannot be ascertained.  McGowan et al. (1994) investigated the correlation of changes in self-efficacy with changes in behaviours and the measures of health status. The correlation of changes in behaviours with  40 the changes in health status were extremely low and not significant, but the correlations of changes in self-efficacy with changes in health status were statistically significant. Table 14 shows the correlations between changes in self-efficacy and the health status measures (i.e., pain level, depression, and quality of life) for a sample of 149 persons who participated in the A S M P in British Columbia during 1990.  Table 14 Correlations of Changes in Arthritis Self-Efficacy with Changes in Three Health Status Indicators in A S M P Participants in British Columbia During 1990 Pain Functional SE Other Symptom SE PainSE  -.19* -.27** -.18*  Depression "  -.15 -.42** -.35**  Quality of Life -.07 -.07 -.16*  Note: SE = Self-Efficacy. Associations analyzed using Pearson correlations. Pain and Quality of Life were measured using visual analogue scale; Depression measured by CES-D. n = 149.( * g < .05, **D<.001).  Similarly, in a sample of 600 A S M P participants from the 1992-94 Canadian study (McGowan et al., 1994), no statistically significant correlations between changes in behaviours and health status were found. However, there were significant correlations between changes in self-efficacy and changes in health status. These relationships were found in spite of the seasonal variation in weather that influenced the outcomes (McGowan et al., 1994). Table 15 gives the correlations of changes in two of the three self-efficacy subscales (other symptom self-efficacy and pain self-efficacy) with changes in the health status set of outcome measures (pain, quality of life, health status, and depression).  41 Table 15 Correlations of Changes in Arthritis Self-Efficacy with Changes in Health Status in A S M P in Provinces East of Saskatchewan During 1992-1993  Pain  Quality of Life  Health Status  Pain  -.27  .15  -.19  -.25  Other Symptoms  -.21  .20  -.22  -.35  Self-efficacy in relation to  Depression  N o t e : Associations analyzed using Pearson correlations. S E = Self-Efficacy. P a i n and Quality o f L i f e were measured by visual analogue scales; Health Status and Depression measured by scales used in M O S study, n = 600. A l l o f the correlations are significant at rj <.01 i n one-tailed tests.  In summary, analyses have shown that changes in self-efficacy correlate better than behavioural change with health status changes. However, the mechanism explaining the relationship between self-efficacy and health status has not been satisfactorily established empirically. None of these correlations, for example, rules out the possibility that self-efficacy is produced or dampened by improvements or deteriorations in health status. The low correlations between specific behaviours and health outcomes also do not rule out the possibility that different individuals change different behaviours among the several recommended behaviours, thus suppressing the specific correlations. To date, there is only evidence showing that selfefficacy and health status measures are interrelated, and correlational data showing that participation in the A S M P results in benefits to both.  The structural equation modelling that will be undertaken in this thesis represents a secondary analysis of the data collected as part of the evaluation of the A S M P in Canada. In order to understand the development of the latent variables used in the structural equation  42  modelling and the limitations of the primary analysis that was carried out, a detailed description was needed. As well, the success of the structural equation modelling depends strongly on a sound theoretical framework. The framework for this structural equation model was developed from findings of the evaluation which show that changes in self-efficacy were associated with changes in the health outcomes.  Theoretical Orientation of the Proposed Research  The major focus of A S M P research to date has been on impact evaluation. These evaluations have been concerned with internal validity and have emphasized the use of randomization controlled experiments. When it had been established that the A S M P could bring about the desired changes in the outcome measures, and that the changes were attributable to the intervention (i.e., the ASMP), a number of "process" studies were conducted to gain an understanding of how and under what circumstances the intervention was able to bring about the desired changes.  The theoretical orientation of my doctoral research is based on "Theory-Driven Evaluation" (Chen & Rossi, 1983, 1987). The major emphasis of this type of evaluation is to introduce more substantive theory into the evaluation. In the evaluations of A S M P to date, the major concern was to obtain unbiased treatment effects. Chen and Rossi (1987) note that: If the only concern in a program evaluation is to obtain unbiased treatment effects, then a randomized controlled experiment that maintains its integrity need not be designed with  43 any knowledge of the relationships among exogenous, treatment, intervening, and outcome variables. Hence randomized experiments can be designed as "black box" researches in which how treatments affect outcomes is unknown. (Chen & Rossi, p. 291).  As noted by Peter H . Rossi, there are major disadvantages with this approach in program evaluation: The domination of the experimental paradigm in the program evaluation literature has unfortunately drawn attention away from a more important task in gaining understanding of social programs, namely, developing theoretical models of social interventions... A n unfortunate consequence of this lack of attention to theory is that the outcomes of evaluation research often provide narrow and sometimes distorted understandings of programs. (Chen & Rossi, 1983, p. 284).  Another major disadvantage to this type of evaluation relates to external validity. The end result of a black box evaluation is to know whether or not a given treatment-as-unit is effective and to what extent it is so. A transfer into a different administrative environment and subsequent modifications to fit the requirements of that environment may drastically alter the treatment's effectiveness, if the elements changed are among the most important within the treatment-as-unit. Therefore, a major benefit of departing from the black box treatment-as-unit approach to evaluation is an enhanced ability to generalize from the researches in question to other circumstances.  The theory-driven evaluator constructs a theoretical model of program inputs, mediating  44 processes, and outputs; devises measures of all of these parts; and gathers and analyses these data. In addition to studying the existence of cause-effect relationships the evaluator also examines program inputs and causal mediating processes.  Chen and Rossi (1981, 1983, 1987) have written extensively on theory-driven evaluation, tying it closely to the method of linear structural equation modeling, a methodology that models correlational and causal relationships among latent variables inferred from multiple observed variables. Using the theory-driven evaluation approach one gains strong inference about the existence of a cause-effect relationship, and from structural modeling one gets more theoretical information about the program inputs and causal mediating processes.  This chapter has summarized the evaluations on A S M P and it has pointed out that the relationship between self-efficacy and the outcome variables is unknown. In spite of the limited understanding of the relationship between self-efficacy and the outcome measures, it seems that both theorists and program planners have assumed that the self-efficacy variable is responsible for bringing about the positive changes in health status. To address this current state of affairs, I will use structural equation modeling to investigate whether there is a causal relationship between self-efficacy and the outcome variables of depression, pain, and perceived health status.  45 CHAPTER TWO  SOCIAL LEARNING THEORY - SELF-EFFICACY THEORY  Self-efficacy theory was developed within the framework of social learning theory, and social learning theory was itself developed within the larger context of the imitation theory of the nineteenth century. This chapter: (a) provides a precis of the origin of imitation theory; (b) describes the major contributions made in the development of social learning theory during the twentieth century, up to the present state of the theory; (c) describes self-efficacy theory (its function within the framework of social learning theory, its components, and how it is operationalized); and (d) reviews the literature on the relationship of self-efficacy with depression, pain, and perceived health status, with a special focus on the area of arthritis. As well, it provides an overview of instances where self-efficacy theory has been applied in smoking cessation, cardiac rehabilitation, and adherence to medical regimen programs. The chapter concludes with a brief explanation of the major weakness of self-efficacy theory.  Nineteenth Century Psychology  Nineteenth century psychologists tended to postulate unitary theories intended to account for all human behaviour. According to Allport (1985), these theories, termed "Simple and Sovereign," were basically six: hedonism, egoism, sympathy, imitation, gregariousness, and satisfaction. Social learning theory has developed within, or evolved from, the structure of imitation theory, from William Bagehot (1875) through Gabriel Tarde (1903), James Mark  Baldwin (1895), and Charles Horton Cooley (1902). Ideomotor theory was developed by Charles Horton Cooley, who in 1902 wrote: "...all ideas press toward [sic] expression. Since many of our own ideas come about while we are observing other people, it follows that we tend to act out our perceptions of the stimulus pattern".  William James (1890) and Claude Bernard (1926) viewed imitation as a basic human instinct. Their line of thinking was gradually displaced or elaborated by twentieth-century learning theories in which instinct played a lesser role.  Twentieth Century Contributions  In 1921, Freud developed the concept of identification, which can be described as affective imitation. He believed that a strong emotional attachment could explain why the child imitates the mother. Freud's concept was favoured by Blanton & Blanton (1927), who focused their studies on the process of empathy and termed imitation theory "postural tensions." They believed this could explain why tensions in the mother are often transferred to the child.  In the 1920's and 1930's, the idea of classical conditioning, based on the original Pavlovian theory, became a popular explanation of the mechanism by which individuals learned, and with it the echo principle of learning prevailed (Allport, 1924; Holt, 1931; Humphrey, 1921). Most psychologists up to this point thought imitation had a motivational force, for only then could it account for conformity.  47 George Herbert Mead (1934) elaborated upon Baldwin's concepts. He thought that people perceive what others are doing, and also perceive their own response to it, and thus an interweaving process takes place that results in mutual understanding and continual accommodation and readjustment. This idea heavily influenced Mead's theory of socialization, bringing a more sociological perspective to bear on the social psychology of imitation and learning.  Instrumental learning was first conceived of and formalized by Miller and Dollard (1941). Gestalt psychologists (e.g., Kohler & Asch) believed that comprehension of the purpose of an act is needed along with the context in which it would be productive.  Lastly, cultural anthropologists (Wissler, 1923) viewed imitation as the agent for the perpetuation of culture (i.e., the younger imitating the older). This group was known as "the diffusionists."  It is clear that imitation is a protean concept, and several distinct mechanisms may be involved: motor mimicry, classical conditioning, instrumental conditioning, cognitive structuring, and identification. Near the middle of the twentieth century the emphasis shifted from trying to explain why to trying to understand how (Lott & Lott, 1985).  Neal Edgar Miller and John Dollard (1941) appropriated the stimulus-response mechanism from classical conditioning in their work with human and animal subjects. They explained imitation as a response learned through experiencing situations in which positive rewards are received for copying (matching) a model's behaviour and negative reinforcements are  48  received for not copying the model's behaviour. Their work can be viewed as a particular foundation for social learning theory and self-efficacy, insofar as its emphasis was on social reinforcement and the reinforcement that comes from mastery.  Orval Hobart Mowrer (1960) suggested instead that not following a model's behaviour causes anxiety, and that the individual will follow the model's behaviour to avoid anxiety. Thus, for Mowrer, the cues associated with the model's behaviours acquire a reinforcing value that is more psychodynamic than social or instrumental in its reinforcing value.  Bandura and Walters (1963) introduced the concept of vicarious reinforcement They studied the behaviour of children as they were playing, and noted that a child did not have to perform a specific behaviour in order to learn it. They observed that a child could learn by watching as another child performed the behaviour and received a positive or negative reward for doing so. This formed the platform from which Bandura launched a series of studies that formed his "social learning theory," later called social cognitive theory.  Julian B . Rotter (1966) applied social learning theory to clinical psychology, and from this marriage issued his idea of generalized expectancies of reinforcement. He posited that learning is the sum of one's history of receiving positive or negative reinforcement for various behaviours. He originated the idea of the "locus of control", by which he referred to whether the control over reinforcement or reward was perceived to be internal or external to oneself.  Jacob L . Gewirtz and Karewn G. Stingle (1968) regarded imitative responses as instrumental learning which may be either purposefully guided or incidental. They proposed that  49  one must "learn to learn" through exposure to models, and either positive or negative responses will follow the behaviour. In their view, identification is generalized imitation, which is behaviour that is changeable in accordance with the many laws of behaviours.  In 1969, Albert Bandura articulated a unique conceptual foundation. He suggested that imitation and identification were synonymous terms, referring to behavioural modification resulting from exposure to modeling stimuli. His analysis was distinctly different from Gewirtz & Stingle's, in that he believed the acquisition of a new behaviour was independent of reinforcement. According to Bandura, reinforcement theories fail in situations where the observer does not carry out the behaviour, where the model and observer do not receive a reinforcement, and where there is a delay in the time it takes for the observer to perform the behaviour. Bandura accounted for observational learning by the mediation of imaginal and verbal codes that become coupled with external stimuli.  In 1973, Walter Mischel first proposed several constructs that provided a cognitive basis for social learning theory.  In 1977, Bandura provided the first theoretical treatment of his cognitive concept of selfefficacy with the publication of "Self-Efficacy: Toward a Unifying Theory of Behavioural Change." In 1978 he introduced the concept of reciprocal determinism. This notion sits in stark contrast to traditional operant conditioning, in as much as it suggests that the environment may not only influence the individual, but also be influenced by him/her. Elaborating upon this concept, he suggested that learning is dynamic and multi-directional. A n individual's behaviour,  50 his/her c o g n i t i v e processes, a n d his/her e n v i r o n m e n t all operate as interacting determinants o f e a c h other.  T h e e n v i r o n m e n t i n question was p r i m a r i l y the social e n v i r o n m e n t . T h i s , together  w i t h the concepts o f m o d e l i n g a n d v i c a r i o u s learning, gave s o c i a l learning theory its s o c i a l dimensions.  In 1986, B a n d u r a r e n a m e d social learning theory as s o c i a l c o g n i t i v e theory, reflecting his g r o w i n g p r e o c c u p a t i o n w i t h the c o g n i t i v e processes at w o r k i n social learning.  In 1981, P a r c e l a n d B a r a n o w s k i delineated the stages that transpire i n a directed b e h a v i o u r change process i n health education, a n d s p e c i f i e d s o c i a l learning theory constructs that c a n be a p p l i e d at each stage i n a health education p r o g r a m . N o r e e n C l a r k (1987) further elaborated o n w a y s i n w h i c h social learning theory c o u l d be a p p l i e d i n health e d u c a t i o n a n d health p r o m o t i o n .  Social Learning Theory Today  F i v e c r u c i a l assumptions are i m p l i c i t i n social learning theory: 1.  that i n d i v i d u a l s are capable o f s y m b o l i z i n g the m e a n i n g o f their b e h a v i o u r ;  2.  that i n d i v i d u a l s c a n foresee the consequences o f their b e h a v i o u r a l patterns;  3.  that i n d i v i d u a l s c a n learn b y o b s e r v i n g others;  4.  that i n d i v i d u a l s are able to self-regulate a n d self-reinforce their b e h a v i o u r ; a n d  5.  that i n d i v i d u a l s are capable o f reflecting u p o n a n d a n a l y s i n g their o w n b e h a v i o u r .  51  According to Perry, Baranowski, and Parcel (1990), social learning theory is a sound theory on which to build health education and health promotion programs because: (a) it synthesizes previously dissociated cognitive, emotional and behaviourist explanations of behaviour change; (b) its constructs provide ample opportunities for future research; and (c) it purports to prognosticate how the individual's behaviour will change. It has gained increasing attention in health promotion since the 1986 Ottawa Charter for Health Promotion defined health promotion as "a process of enabling people to increase control over, and to improve, their health," making the self-efficacy concept in social learning theory attractive.  The requisites of a good theory, as set out by Glanz, Lewis, and Rimer (1990), are that it tells us the what (i.e., the elements to be considered), the why (i.e., the process by which changes occur in the dependent variable), the when (i.e., the timing and sequences), and the how (i.e., the methods or ways to focus interventions to bring about a change).  Perry et al. (1990) summarize the several concepts operating in social learning theory: 1.  environment  the factors physically external to the individual;  2.  situation  the individual's perception of the environment;  3.  behavioural capacity  the knowledge and skill to perform a behaviour;  4.  expectation  anticipated outcome of a behaviour;  5.  expectancy  the values an individual places on an outcome;  6.  self-control  personal regulation;  7.  observational learning  acquisition of a behaviour by watching it being performed by another;  52 8.  reinforcement  response to a behaviour that will either encourage or discourage repetition;  9.  self-efficacy  confidence in one's ability to perform a behaviour;  10.  emotional coping responses strategies to deal with emotional coping;  11.  reciprocal determinism  dynamic interaction of an individual's cognitive processes, behaviour, and environment.  Each of these concepts may be manipulated - for example, the encouragement of mastery learning may be a good way to increase behavioural capacity. Table 16 describes program strategies that can be used to influence these elements. Table 16 Methods of Focusing Interventions to Enhance Elements of Social Cognitive Theory ELEMENTS  P R O G R A M STRATEGIES  Performance Accomplishments  Personal experience - skills mastery Mastery experiences Contracting & concurrent feedback  Vicarious Experience  Group members to help each other in problem-solving Modelling perseverant successes Group work  Social Persuasion  Group work Contracting (short-term goals) Structuring success situations  Physiological State  Reinterpretation of physiological signs and symptoms, e.g., fatigue Stress management (relaxation techniques)  Appraisal of Efficacy Information  Efficacy validation Dealing with preconceptions Credibility of others  53  Social learning theory (especially in its latest incarnation as social cognitive theory) places great importance on cognitive events. Whereas operant conditioning implies a mechanical association among stimulus, response, and reinforcement, social learning theorists believe that learning employs cognitive aids, and also that cognition has a causal influence on learning. Social learning theorists differ from behaviourists in their belief that there is judgement in addition to thought, and that reinforcement works because the individual integrates data regarding the consequences of each behaviour over time, and regulates behaviour based on the aggregate consequences.  Self-Efficacy Theory  In 1977 Bandura published "Self-Efficacy: Toward a Unifying Theory of Behavioural Change." Besides being a significant milestone in the ongoing development of social learning theory, this was the first declaration of self-efficacy as a theory. The theory states that psychological procedures alter the level of strength of self-efficacy, and is concerned with the effects of self-referent thought on psychosocial functioning. Specifically, it states that behaviour change is mediated through cognitive processes (e.g., thinking, perceiving, and believing), and that cognitions (e.g., attitudes and beliefs) about a behaviour are altered most easily through actual or observed performances of it. According to Bandura (1977a), efficacy expectations influence whether coping behaviour will be initiated, how much effort will be expended, and how long effort will be sustained. The notion of control and mastery provides the basis for selfefficacy theory. The fundamental assumption upon which Bandura's 1977 theory rests is that "psychological procedures serve as ways of creating and strengthening expectations of personal efficacy."  54 There is a distinction between outcome expectations (as used in the health belief model, and in the locus of control) and efficacy expectancies as illustrated in Figure 1. Figure 1 Influence of efficacy and outcome expectations. PERSON = = > • BEHAVIOUR = >  EFFICACY EXPECTATIONS  OUTCOME  OUTCOME EXPECTATIONS  The health belief model and the notion of locus of control are concerned with outcome expectations, whereas self-efficacy is concerned with efficacy expectations. Locus of control has been interpreted by latter-day users (e.g., Wallston, Wallston, & DeVellis, 1978) as a generalized concept about the self, whereas self-efficacy is situation specific - focused on beliefs about one's personal abilities in specific settings. A combination of internality-externality and self-efficacy may influence behaviours.  Figure 2 Roles of locus of control and self-efficacy. Locus of Control Internal External Self-efficacy  High  A  B  Low  C  D  In Figure 2 , " A " would be much more likely to perform the given behaviour than would "D." Those in "B" believe themselves to be capable but will not comply because they are not convinced that compliance will attain some desired effect. Those in " C " believe that outcomes are personally determined, but that they lack the skills to execute the action. This analysis  55  suggests that both an internal locus of control (i.e., outcome expectations) and high efficacy expectations are necessary for a given behaviour to be performed. Several studies which have focused only on locus of control without considering self-efficacy have yielded inconsistent findings (Becker, 1987). Similarly, those (including Becker) who focused on outcome expectancies using the Health Belief Model also produced inconsistent findings (Harrison, Mullen & Green, 1992).  Self-efficacy and locus of control represent different phenomena, and are founded on entirely different conceptual schemes. Self-efficacy is concerned with people's beliefs about their abilities to produce certain performances; locus of control refers to people's beliefs about whether the outcomes they experience are dependent on their actions or are the result of chance, fate, luck, or the fiat of those who control rewards. Research shows that locus of control and self-efficacy bear little or no relation to each other (Manning & Wright, 1983; Taylor & Pompa, 1990). As well, self-efficacy predicts such diverse events as academic performance, proneness to anxiety, level of pain tolerance, career decision-making, and political activism, whereas locus of control is a weak predictor or nonpredictor of these events (Bandura, 1995).  Rotter regards locus of control as a generalized trait, whereas Bandura considers selfefficacy "to be conceptualized and assessed in terms of particularized judgements of capability that vary across realms of activity, different levels of task demands within a given activity domain, and under different situational circumstances" (Bandura, 1995, p. 186).  The majority of models of health behaviour change have been deductively derived from  56  traditional cognitive theory within psychology. Cognitive theorists emphasize the role occupied by the individual's expectations in predicting behaviour. According to this perspective, individual behaviour is a function of a combination of the subjective value of the outcome, and the subjective probability (or expectation) that a particular action will achieve that outcome. These formulations are generally termed value-expectancy theories (e.g., Health Belief Model, Theory of Reasoned Action, Stages of Change). Self-efficacy theory is based on the interactive model of human behaviour within Bandura's social cognitive theory (Bandura, 1986).  The view held by social cognitive theorists (Bandura, 1986) is that behaviour is governed by expectancies and incentives. Expectancies can be subdivided into three categories: 1.  environmental expectancies (i.e., the individual's perception of the way various elements in the world are connected);  2.  outcome expectancies (i.e., the individual's beliefs about how his/her behaviour will influence the outcome); and  3.  self-efficacy expectancies (i.e., the individual's expectancies about his/her competence to perform the behaviour in question).  A n incentive pertains to the subjective value of a particular outcome; behaviour is regulated by reinforcements which result from the individual's perception of the consequences.  Figure 3 illustrates the main elements of social learning theory. Self-efficacy alone does not explain behaviour. Rather, the self-efficacy construct falls within a broader theory of motivation based on social learning theory (Bandura 1977b, 1978).  57  Figure 3 Social Learning theory: components and processes.* INFLUENCES UPON EXPLANATORY VARIABLES  Direct Experience Modelling  Performance Accomplishments Vicarious Persuasion Verbal Persuasion Emotional Arousal  Direct Reward Vicarious Reward Self-Management  EXPLANATORY VARIABLES  OUTCOMES  Behavioural Capability (having the skills necessary for the performance of the desired behaviour)  Efficacy Expectations (beliefs regarding one's ability to successfully carry out a course of action or to perform a behaviour)  Acquisition and Maintenance of New Behaviours  Outcome Expectancies (beliefs that the performance of a behaviour will have desired effects or consequences)  * The shaded areas reflect self-efficacy theory's place within the more inclusive social learning theory.  58  Table 17 depicts self-efficacy theory as described by Bandura (1977), presenting the four sources of self-efficacy information and listing inputs which may influence the sources.  Table 17 Sources of Self-Efficacy SOURCES Performance Accomplishment  Vicarious Experience  |  | | | |  INPUTS participant modeling performance desensitization performance exposure self-initiated performance  live modeling | symbolic modeling  Verbal Persuasion  | | | |  exhortation interpretative treatments self-instruction suggestion  Emotional Arousal  | | | |  attribution relaxation symbolic desensitization symbolic exposure  The changes in self-efficacy vary in proportion to the trustworthiness of the experiential source.  Performance Accomplishment  As stated by Bandura (1977a), this is the major source of perceived self-efficacy, and is based on personal mastery experiences. A success will raise mastery expectations, a failure will lower them. A n intervention would typically consist of having the individual select a realistic but simple task to execute, and having him/her learn the skills required for its execution through  59  instruction and watching others perform it. The individual would set goals (i.e. make contracts) to fulfill, and with successive accomplishments the tasks would become progressively more difficult.  Vicarious Experience  This concept is based on the earlier work of Bandura and Walters (1963) and consists of experience procured watching others perform a behaviour to achieve a set goal. Results are better when the model demonstrates persistence and problem-solving techniques.  Verbal Persuasion  Wherein individuals are convinced that they can perform a behaviour by others telling them they can do it. This source of self-efficacy is not as effective as either performance accomplishments or vicarious experiences.  Emotional Arousal  One's physiology can be a source of self-efficacy. In research with people experiencing phobias, Bandura, Taylor, Williams, Mefford, & Barchas (1985) were able to demonstrate that levels of perceived self-efficacy correlated with catecholamine levels in the blood. Through strategies designed to enhance self-efficacy, they were not only able to raise their subjects' selfefficacy in responding to the phobia, but also succeeded in lowering catecholamine levels. These  60  findings were not only significant in the field that investigates the mind-body connection, but also uncovered a physiological correlation between self-efficacy and the ratio of certain chemicals in the blood.  Self-efficacy theory is based on the interactive model of human behaviour set out in Bandura's (1986) social cognitive theory. It posits the individual as a central processor of efficacy information who processes, weighs, and synthesizes diverse pieces of information concerning his/her capability, and then determines his/her choice of behaviours and effort expenditure accordingly. Beliefs about self-efficacy affect the intention to change a behaviour, the effort expanded to attain this goal, and the level of perseverance to continue striving in spite of setbacks that may undermine motivation (Bandura, 1991).  The emphasis on cognition, anticipated consequences, and reinforcement leads to the classification of this theory as a cognitive behaviourist perspective. The theory postulates that when one is faced with a potentially threatening situation, thought patterns and emotional reactions will affect behaviour and the level of motivation. These personal perceptions of selfefficacy may vary over time and in different situations. Likewise, informational sources and their content, interpretation of efficacy information, and the impact of the interpretation of perceptions of self-efficacy may similarly vary (Jenkins, 1987). As early as 1970, these concepts were suggested as an alternative to the prevailing attitude-change theories in health education (Green, 1970).  61  Review of Applied Social Learning Theory  Social Learning Theory in Health Education Programs  Noreen Clark (1987) has conducted a thorough review and credible evaluation of health education programs that have incorporated elements of social learning theory (declaredly or implicitly). The three components of social learning considered in her review were: 1.  Self-Regulation and Self-Reinforcement  2.  Cognitive Restructuring  3.  Emotional Coping  Self-Regulation and Self-Reinforcement Self-regulation is the process by which individuals arrange environmental inducements, generate cognitive supports, and produce consequences for their own actions. Self-reinforcement is the process by which individuals buttress and encourage their own behaviours by rewarding themselves when they reach self-determined targets.  A review of smoking cessation health education programs that employed self-regulation and self-reinforcement strategies included studies conducted by Davis, Faust, & Ordentlich (1984); Glasgow, Schafer, & O'Neill (1981); and Strecher (1983). In these studies, selfregulation and self-reinforcement were operationalized by programs in which participants identified triggers and cues, monitored themselves (i.e., kept records of smoking behaviour), signed contracts to refrain from smoking, practised breathing and relaxation exercises, avoided  62  familiar environmental cues to smoking, and rewarded themselves upon attaining a selfprescribed goal. In the studies reviewed, the experimental subjects initially showed better results than the control groups, but these positive results were not sustained at follow-up.  In programs focusing on weight loss for obese children (Wheeler & Hess, 1976; Kingsley & Shapiro, 1977), the strategies included self-monitoring and self-regulation through governance of triggers to eating apparent in the environment, establishment of self-inducements to increase physical exercise, and rewards for attained weight loss. These self-regulation and selfreinforcement programs produced weight loss, but, again, these improvements had not been maintained at follow-up.  Clark concluded that self-regulation and self-reinforcement are effective in dealing with problems of smoking and obesity because arresting these conditions involves making specific behaviour changes for specific situations where the behavioural context is clear. But selfregulation and self-reinforcement may not succeed into health education programs where the causal relationships are not clear. Clark was not able to confirm that programs using social learning theory in their intervention(s) are better than programs using other theories, but she did conclude that "implementing such programs was better than having no program at all" (p. 272). Significantly, there were no long-term effects in any of the programs. She suggested that "conscious self-regulation is not inherently self-reinforcing, and will last only as long as the conscious efforts are induced or intrinsically rewarded" (p. 272).  Cognitive Restructuring This term was not used by Bandura, but has been used by others who have accepted his formulation of social learning theory. According to Bandura (1977b), symbols that represent events, cognitive operations, and cognitive relationships serve as vehicles of thought. Cognitive restructuring refers to the manipulation of symbols that convey relevant information. By such manipulation one can gain understanding of causal relationships, create new patterns of knowledge and solve problems, the last without actually undertaking any overt activities. Therefore, theoretically, behaviour changes rely heavily upon cognitive representations of contingencies, and may result from recognizing the association of environmental events and consequences. The capacity to represent future consequences mentally to oneself provides a cognitively-based source of motivation, that is, an incentive to act in order to alter an expected consequence.  The aim of health education programs that use cognitive restructuring is to enable individuals to manipulate symbols in order to articulate the different consequences of different behaviours, and therefore to increase outcome expectancy and self-efficacy. Program strategies have included contingency planning, anticipatory problem solving, performance of skills, role modeling, and rehearsal.  Cognitive restructuring through role modeling and rehearsal was used in a child health promotion program with 100 four-year olds. Puppets served as role models, and the behaviours that the puppets theatricalised were reinforced through songs, poems and materials that the children took home. The intervention was evaluated by Parcel, Bruhn, & Murray (1983), who  found no differences between the verbalized behaviours of children in the experimental group and those of children in the control group except that children in the experimental group were more likely to declare that they would not smoke cigarettes, pipes, or cigars.  Bowler and Morisky (1983) have reported on various components of a program designed to influence behaviour and health outcomes related to blood pressure control. With low-income black outpatients as subjects, an initial diagnostic baseline survey of 305 ambulatory hypertensive patients was conducted at Johns Hopkins Hospital in Baltimore to determine patients' educational needs (Green, Levine, & Deeds, 1975). Respondents reported confusion about the specifics of their therapeutic regimens, felt a lack of family understanding and general . support, and experienced feelings of powerlessness in controlling the contingencies surrounding the management of the regimes. Following the needs assessments, a two-year patient education study was initiated which involved four hundred patients. A randomized factorial design distributed the patients into experimental and control groups representing all 8 combinations of three interventions. Study participants were mostly black (91%), and female (70%), and the median age was 54 years.  The program, developed and implemented by Green et al. (1975), consisted of experimentally varied combinations of an exit interview, a home visit, and small group sessions. The exit interview consisted of a 15-minute interview by a graduate health education student who also had a nursing background. During the interview, the patient's medications were reviewed, and personalized through adaptation to his/her individual daily schedule; diet restrictions were reviewed and explained in clear language; and problems in adhering to the clinic appointment  65  schedule were discussed. In the home visit an instructional reinforcement session was given to an adult family member. This was designed to increase family or peer support and reinforcement for the patient.  The small group sessions, which were defined as "internality training", consisted of three two-hour sessions in which a social worker and a nurse co-led groups of eight to ten persons. The main objective of the sessions was to increase patients' feelings of personal control over their medical regimen (Bowler and Morisky, 1983). This third intervention, based on Rotter's notion of "locus of control," used two main methods. One was to help the patient see him/herself as having the power to effect change. This was done by: (a) challenging or confronting "external statements" with "internal" questions in an attempt to get the patient to examine his/her reasons for choosing one option among several available; (b) rewarding "internal" statements; and (c) helping the patient to recognize his/her focus on the results of the behaviour. The second method used was behavioural reinterpretation, also known as cognitive restructuring, which attempts to get the patient to see that the way he/she looks at or responds to a situation may affect his/her behaviour. This is done by obtaining other group members' perspectives on difficult and problematic situations.  Evaluations were conducted at the end of the two-year project (Levine et al., 1979) and at five years (Morisky et al., 1983). Outcome measures included self-reported medication compliance, appointment keeping, and blood pressure control status, as well as mortality at five years. Post-intervention assessments were made six months after completion of the interventions. With respect to medication compliance and appointment keeping, the home visit intervention had  66  the biggest impact: 53% reported compliance as compared to 40% in the control group, and 76% kept appointments as compared to 63% in the control group. The largest improvement in blood pressure control, a 28% increase, was achieved by those assigned to all three interventions, followed by those assigned to various combinations of two interventions, followed by two of the single interventions, demonstrating an education dose-response relationship. The small-group education demonstrated the strongest impact among single interventions on blood pressure control (18%i), the family support an intermediate effect (11%). The effect of the exit interview alone was indistinguishable from that of the control group.  At five years the group which received all three interventions continued to demonstrate a positive impact on appointment-keeping behaviour, with a group mean of 86%. The greatest improvement was found in the group assigned to either the three interventions and in the group assigned to a combination of family support and small group interventions. Patients who had been assigned to any one of the experimental groups also displayed a statistically significant 65% increase in blood pressure control (from 40% to 66%), as compared to 22% in the control group (from 41%o to 50%). The family support intervention accounted for the largest variation in blood pressure control. Interestingly, improvements in blood pressure control in the control group took place between year two and year five with patients from the clinic where the positive aspects of the educational program had been introduced into routine procedure. Morisky et al. (1983) also examined mortality rates over the five-year follow-up period. Results showed that the education group (those assigned to any combination of the three interventions) were 57% lower in their allcause mortality rate, and 53% lower in their hypertension-related mortality rate, than the control group.  67  Contingency planning and anticipatory problem-solving strategies were used in a smoking prevention program with 1,526 seventh grade students (Hurd et al.,1980). The strategies were implemented through the agency of films, videos, discussion groups, and selected peers and college students who functioned as models. During the study period, the number of occasional smokers declined in all five schools. The investigators were able to conclude that the programs that contained each of three elements of intervention (social pressures, curriculum, personalization, and commitment) effectively reduced smoking in this population.  Condiotte and Lichtenstein (1981) conducted a study using mastery learning to increase participants' belief that they could give up smoking. The investigators were interested in efficacy enhancement that might result from the treatment. The treatment included nicotine fading, regular pace vs. rapid pace smoking, and individualized and group discussion designed to enhance self-efficacy. They found that those who had benefited from the program through achieving mastery had significantly higher levels of self-efficacy when followed up. They found that a higher level of perceived self-efficacy at the completion of treatment indicated a greater probability that the subject would remain abstinent through the entire three-month course of the experiment. In another smoking cessation study, Coelho (1984) found that adult smokers who had quit during the program had higher levels of self-efficacy than those who had not.  Lenker et al. (1984) investigated why persons who are better able to self-manage their arthritis do not improve in their health. They interviewed people who made positive and negative health changes, and they inferred three recurrent cognitive states: perceived self-control over the disease, emotional status, and perceived social support. Subjects whose health had  68  changed for the better had more self-control, had more social support, and elicited more positive emotional responses. Two different conclusions are possible: (a) those who judge themselves able to have an impact on their disease are the ones who self-manage and attain better health; or (b) that the more positive the self-management, the stronger the feeling of having greater impact on the disease.  In the studies reviewed, increased self-efficacy was associated with improved health status, abstinence from smoking, and functional ability in arthritis patients. Long-term blood pressure control remained after the investigators' efforts at internality training and enhancement of belief in self-control. This suggests self-efficacy and health behaviour change are strongly related.  Emotional Coping Through direct and vicarious experiences, a vast array of stimuli acquire the ability to activate and guide behaviour. People not only respond to stimuli, but interpret them as well (Bandura, 1977). Until an individual develops effective coping behaviour, threats produce high emotional arousal and various defensive manoeuvres. Cognitive activity can reduce fear by producing an awareness of other contingencies or by rearranging the existing contingencies. The threat is reduced when one arranges contingencies so that different consequences can be expected. Cognitive activities can also directly produce physiological effects (Bandura, 1977).  Anticipatory discussions and relaxation techniques have been used to reduce emotional responses associated with health problems, so as to increase efficacy expectations and the  69  individual's intention to take needed action. Fassler (1980) conducted a study of six- to twelveyear olds admitted to hospital for tonsillectomy. The children were encouraged to describe anticipated problems. The study staff were able to talk over these anticipated problems, provide reassurance, and thus to reduce anxiety. Results showed that children in the experimental group had lower levels of anxiety than did children in control groups who received regular hospital intervention. In an earlier study conducted by Skipper and Leonard (1968), both the children having tonsillectomy and their mothers were provided with opportunities to express their concerns about the operation. The researchers found that the mean systolic blood pressure and mean pulse rate of the children in the experimental group were significantly lower, at each of three different times after admission to hospital, than the mean pressure and pulse of the control group children. The investigators concluded that intervention with the mother enabled the parent to cope with associated stress, and subsequently to reduce the stress felt by the child.  Conclusion In her review, Clark noted that there were only a few published studies on the use of social learning theory in program evaluations, that no consistent pattern emerged, and that in the different studies she reviewed, "social learning theory was used across several age groups, ethnic groups, and income levels" (p. 270). There were several problems that she noted in conducting her evaluation: 1.  Design and Method Issues. Sample sizes were small, and non-health issues (such as phobias, lack of assertiveness, etc.) were investigated. Health-related issues with larger sample sizes were needed.  2.  Eclectic Approaches. Evaluations have not been able to distinguish the effects of  70  the different aspects of social learning theory. Programs have been eclectic educational approaches, and evaluations have taken the so-called "black box" approach. 3.  Lack of Comparison. There have been no studies to compare the results of social learning theory with those of other theories.  4.  Measurement Issues. This was a concern especially when self-efficacy was a working concept because self-efficacy is situational different researchers have different instrumentation, however, researchers have already developed a generalized scale (Sakano & Tohijoh, 1986; Sherer, 1982).  Clark (1987) concluded that, all things considered, social learning theory held promise for health education, but that because limited data was available, it was hard to determine the contribution they would make.  Self-Efficacy Theory in Health Education Programs  Self-Efficacv. Depression and Arthritis In studies examining the emotional reactions associated with arthritis, several standardized measures have been used to determine depression levels, including the Minnesota Multiphasic Personality Inventory (MMPI), Beck Depression Inventory, Center for Epidemiologic Studies-Depression scale (CES-D), and anxiety and depression subscales of the Arthritis Impact Measurement Scales. The validity of these scales has been questioned. For example, using the M M P I with arthritis patients has been questioned (Karoly, 1985), since it was originally developed for use with psychiatric populations (Welsh & Dahlstrom, 1956). Also, the  71  CES-D scale contains questions, responses to which are influenced by aspects of the arthritis disease process other than depression (Blalock, DeVellis, Brown, Wallston, 1989).  In this research, the two main types of arthritis experienced by participants in the British Columbia and Ontario samples are osteoarthritis (42% and 52% respectively), and rheumatoid arthritis (19% and 28% respectively). Osteoarthritis is by far the most prevalent type of arthritis (Lawrence, Hochberg, Kelsey, McDuffie, Medsger, Felts, & Shulman, 1989; Yelin & Felts, 1990) affecting approximately 10% of the general population, while rheumatoid arthritis affects approximately 1% of the general population. Rheumatoid arthritis is a chronic disease characterized by lack of control. Persons cope with both physical symptoms (joint inflammation, stiffness, severe pain) and the unpredictability of symptom occurrence (Genest, 1983). The course of R A is similarly unpredictable, with a tendency toward both spontaneous remissions and exacerbations of symptoms. Most of the research on arthritis has focused on rheumatoid arthritis. The rheumatology literature is replete with research studies that examine the prevalence of depression in persons with specific types of arthritis, in comparison both with persons experiencing other chronic conditions, and with the general population. Most investigations have found that persons with rheumatoid arthritis and fibromyalgia, 1% and 3% of the entire Canadian population respectively, show elevated levels of depression relative to healthy controls. Rimon (1974) reported the prevalence of depression in rheumatoid arthritis patients to range between 22% and 80%. More recent studies of adults with rheumatoid arthritis have reported depression is between 35% and 42% (Ahles, Yunus, & Masi, 1987; Frank et al., 1988) and between 21% and 34% (Creed, Murphy, & Jayson, 1990).  72 The literature on depression and rheumatoid arthritis is difficult to evaluate because of the variations in the definition of depression. Contradictory findings can be attributed to the inclusion of heterogeneous subtypes of depression in research studies (Morrow, Parker, & Russell, 1994). As well, a problem with most studies is that scales used to measure depression contain items about somatic symptoms that can result from the arthritis as well as from depression (Blalock et al., 1989; Callahan, Kaplan, & Pincus, 1991; Peck, Smith, Ward, & Milano, 1989; Pincus & Callahan, 1993; Pincus, Callahan, Bradley, Vaughn, & Wolfe, 1986).  Depression among fibromyalgia patients has been found to range from 20% to 86% (Hudson, Goldenberg, & Popo, 1992), and between 22% and 29% (Burckhardt et al., 1994). Between 26% and 31% of persons with ankylosing spondylitis were found to be depressed (Barlow, Macey, & Struthers, 1993). There have been no studies examining depression levels among large samples of persons with osteoarthritis (Bradley, 1993). The prevalence of depression in persons experiencing rheumatoid arthritis, fibromyalgia and ankylosing spondylitis is greater than the 10% to 15% depression reported in the general population (Murphy, Olivier, Monsonth, Sobol, & Leighton, 1988). A recent review of the prevalence of depression in persons with rheumatoid arthritis, using the DSM-III diagnostic criteria for depressive disorders, revealed a range of depression from 21% to 34% (Creed, Murphy, & Jayson, 1990), which is similar to the prevalence found in other chronic illnesses (Frank, Beck, & Parker, 1988).  Using the CES-D as the measure for depression, studies involving patients with rheumatoid arthritis report sample means between 11.6 and 15.8 on a scale of 0 to 60 (Revenson, Schiafinno, Majerovitz, & Gibofsky, 1991; Brown, 1990). This depression mean score is higher  73  than the mean score in healthy community populations which ranges between 7.9 and 9.3 (Radloff, 1977).  In a classic study (Cassileth et al., 1988) depression levels were compared among patient groups in five diagnostic groups. Examining a population of 758 patients in five diagnostic categories (i.e., arthritis, diabetes, cancer, renal disease, and dermatologic disorders) the researchers found that the groups did not differ significantly from one another or from the general public on measures of anxiety, depression, positive affect, emotional ties, loss of control, and global mental health, but all had significantly higher scores for psychological status when compared with the sixth group, patients under treatment for depression. Patients with depression were significantly different from all other groups within each age-matched and sex-matched category. In this study, participants completed a questionnaire based on the Dupuy General Well-Being Schedule (Dupuy, 1972), which was developed further by the Rand Corporation to measure mental health in adult members of the U.S. population. This measure, the Mental Health Index, was field-tested on over 1000 adults in each of six sites chosen from four U.S. census regions (Veit & Ware; 1983; Brook et al., 1979). The reliability and validity of this instrument are well documented (Veit & Ware, 1983; National Center for Health Statistics, 1977; Brook et al.,1983). Its anxiety and depression subscales were found to correlate highly with the Spielberger State Anxiety Scale and the Beck Depression Inventory (Cassileth, Lusk, Hutter, Strouse & Brown, 1984). Also, a test incorporating items from the Mental Health Index displayed excellent reliability and validity in patients with arthritis (Meenan, Gertman & Mason, 1980; Meenan, Gertman, Mason & Dunaif, 1982). In the arthritis group (n=81) differences in depression scores were not found between the types of arthritis (i.e., rheumatoid arthritis, osteoarthritis, lupus, or scleroderma).  74  There was a significant direct relationship between higher mental-health scores and advancing age across all patient populations. Regardless of diagnostic groups, patients with recently diagnosed illnesses had poorer mental-health scores than did patients whose illness had been diagnosed more than four months previously. A direct relation between declining physical status and mental-health scores was observed. These results suggest that psychological adaptation (i.e., as measured by mental health score) among patients with chronic illnesses is remarkably effective and fundamentally independent of specific diagnosis.  The findings from the 1988 Cassileth et al. study are consistent with the findings from the study conducted by Beaumont (1994). In this study he found that patients suffering from depressive symptoms have been shown to have worse physical and social functioning, spend more days in bed, have more bodily pains and worse current health than those with hypertension, diabetes mellitus and arthritis.  In another study (Beekman, Kriesgsman, Deeg & van Tilburg, 1995) findings were somewhat different. These authors examined the relationship of four aspects of physical health and depressive symptoms in a community-based sample of 224 older inhabitants, aged 55-89 years, of a small Dutch town. Depression was measured by the Center for Epidemiological studies Depression Scale. In their sample there was a linear association between the number of depressive symptoms and age, and the number of chronic diseases and subjective health. When chronic diseases were examined separately, mean depression scores were higher than the depression scores in the total population for all types of chronic diseases. The only disease showing a statistically significant association with depression was arthritis.  The findings of studies reported in the literature on depression in persons with arthritis appear to be consistent. Persons experiencing arthritis, particularly rheumatoid arthritis and fibromyalgia, appear to have higher depression levels than persons not experiencing chronic health conditions, and depression level appears to increase with declining physical status (Cassileth et al. 1989) and age (Beekman et al.1995). As well, the prevalence of depression in persons with rheumatoid arthritis appears to be the same as found in other chronic conditions (Cassileth et al., 1989; Frank et al., 1988; Weismann & Myers, 1978). In this research, the average age of study participants was 60 years (McGowan, 1990; McGowan et al., 1994).  Depression as a cause of low self-efficacy. Depressed people often display reductions in self-efficacy (Cane & Gotlib, 1985; Davis & Yates, 1982; Kanfer & Zeiss, 1983; Miller, 1984). Bandura (1986) has indicated that negative emotions and moods affect self-efficacy level, and explains that depressed persons experience dysfunctions in self-evaluation - a source of selfefficacy. He explains that dysfunctions can occur in each of the self-regulatory sub-functions: (a) in how personal experiences are self-monitored and cognitively processed; (b) in the evaluative self-standards that are adopted; and, (c) in the evaluative self-reactions to one's own behaviour.  (a)  Self-Monitoring. People who are prone to depression tend to misperceive their  performance attainments or to distort their memory of them in self-slighting directions. In contrast, non-depressed persons are more inclined to distort personal experiences in a self-enhancing fashion. Thus, despite gaining the same pattern of outcomes as persons who are not depressed, depressed persons underestimate their successes but are very much aware of their failures, whereas non-depressed persons remember their successes  76  well but recall fewer failures than they have actually experienced (DeMonbreun & Craighead, 1977; Nelson & Craighead, 1977; Wener & Rehm, 1975).  In studies relating depressive mood to recollection of past experiences, the more depressed people were, the more they underestimated how well they had performed (Buchwald, 1977). Thus, the correlation at each point in time between measures of selfefficacy and depression in arthritis patients may be attributable to this tendency for depressive states to dampen perceived success, regardless of the changes in behaviours achieved.  (b)  Standard-Setting. Compared to non-depressed persons, depressed persons tend to  set higher standards for themselves relative to their attainments, and to evaluate their performances as poorer for similar accomplishments (Golin & Terrell, 1977; Loeb, Beck, Diggory, & Tuthill, 1967; Schwartz, 1974). Unlike persons who are not depressed, whose standards are realistically tied to accomplishments, depressed persons set higher goals for themselves in the face of declining improvement and were thus more selfdissatisfied even though their actual performances are the same. The people who overaspire and belittle their actual accomplishments are also the people most vulnerable to depression.  Depression is most likely to arise when personal standards of merit are set well above one's perceived efficacy to attain them (Kanfer & Zeiss, 1983). Perceived inefficacy to accomplish desired goals debilitates motivation and creates discontent,  77  whereas unfulfilled goals are motivating when they are regarded as reachable (Bandura & Cervone, 1983, 1986).  (c)  Performance Appraisal. The likelihood of depressive reactions is heightened  when unattainable standards are combined with a tendency for processing performance information in self-belittling ways. Depressed persons are not especially charitable to themselves in how they judge their performance determinants. Persons who are depressed, while not always discounting their contributions to successes, nevertheless are quick to attribute failures to themselves (Kuiper, 1978; Rizley, 1978; Peterson & Seligman, 1984). Self-blame begets self-punishment.  Judgements of adequacy involve social comparison processes and depressionprone individuals tend to use social comparative information in self-depreciating ways. When exposed to the high attainments of others, persons who are depressed judge their own accomplishments as less praiseworthy than do persons who are not depressed (Ciminero & Steingarten, 1978). Self-devaluation for performances that fall below the level achieved by others is especially evident in depressed females (Garber, Hollon, & Silverman, 1979).  In summary, Bandura (1982b) explains that depressed persons experience dysfunctional self-evaluation and therefore: misperceive or distort their performance accomplishments; set standards for themselves that are too high and non-achievable; readily attribute failure to themselves; and use social comparative information in self-  78  depreciating ways. This analysis would imply a relationship between depression and self-efficacy, where depression level predicts self-efficacy; however, this relationship was not supported by the data from the British Columbia and Ontario samples.  Reciprocal relationship between self-efficacy, performance accomplishments, and depression. In the literature there is substantial support for a correlation between depression and reduced judgements of self-efficacy (Yusaf & Kavanagh, 1990; Kavanagh, 1992).  Figure 4 Theorized relationship between self-efficacy, performance accomplishments and depression. From Yusaf and Kavanagh (1990).  According to these authors, depressive feelings have a reciprocal relationship with both self-efficacy and performance. Self-efficacy judgements appear to affect emotional states both directly (when people imagine the future consequences of their inefficacy), and through their effect on later performance. Reductions in self-efficacy and performance appear to be both consequences of depression and the determinants of it (Figure 4).  The model suggests that self-efficacy judgements, performances, and moods have  79  reciprocal influences on each other. A low sense of self-efficacy often deepens the person's sadness, especially when it makes the opportunity for positive outcomes seem remote, or when the performance domain is crucial to the person's self-esteem. When aversive outcomes occur, these too feed the depressive mood. Emotional states can be expected to affect self-efficacy both directly and though impact on performance.  Association of self-efficacy and depression. Several studies have demonstrated an association between depression and self-efficacy. Researchers have demonstrated that depressed people often display reductions in self-efficacy (Cane & Gotlib, 1985; Davis & Yates, 1982; Kanfer & Zeiss, 1983; Miller, 1984). There is evidence showing that changes in self-efficacy are mediated by differences in recalled performances under happy and sad moods (Wright & Mischel, 1982), that depressed people tend to report negative aspects of their experience (Bower, 1983; Lishman, 1972; Lloyd & Lishman, 1975), that depressed people evaluate their own performances more negatively than people who are not depressed (Forgas, Bower, & Krantz, 1984; Lobitz & Post, 1979; Smollen, 1978), and that depressed people reward themselves less readily (Gotlib, 1981). Also, it has been shown that selective retrieval of mood-consistent information occurs and is thought to be mediated by differential evaluative responses to the judgement target under happiness or sadness (Bower, 1991; Schwartz & Bless, 1991).  Association between depression and performance. The effect of depression on performance has also been substantiated in the research literature. Poor performance is often observed in depressed persons (Cohen, Weingartner, Smallberg, Pickar, & Murphy, 1982; Miller, 1975), and depressed people are more self-absorbed (Ingram & Smith, 1984), which  80 decreases their ability to pay attention to others' contributions or their empathic responses. The impact of depression on performance occurs primarily through its effects on task selection, persistence, and effort, although some direct effects from depressive symptoms may be observed. As well, depressed people may feel unwilling to muster the effort required to perform a given task. This was demonstrated in a study involving 726 subjects with rheumatoid arthritis and 192 matched controls (Katz & Yelin, 1994). In this study 14% of persons with rheumatoid arthritis reported depressive symptoms, compared to 4.7% of persons in the control group. The goal of the study was to explore the association between depressive symptoms and the activities in which persons with rheumatoid arthritis participate. Using a measure of life activities developed by Yelin et al. (1987), the researchers found that persons with rheumatoid arthritis who were depressed performed 12% fewer of their valued activities than persons with rheumatoid arthritis who were not depressed.  Self-efficacy, performance, and depression. Evidence is also available supporting the notion that self-efficacy and skills (i.e., performance) are determinants of depression. Depressed persons are not only concerned about the cause of blame for past events, they are also concerned about the implications of the events for the future (Beck, Rush, Shaw, & Emery, 1979). If persons expect that aversive situations may occur, and that they will once again be unable to prevent a negative outcome, this is likely to compound their reactions to an event. As a result, a person's depression levels are related both to aversive past events and to self-efficacy about control of situations he/she will be face in the future. Reductions in self-efficacy achieve greater significance when these expected losses are greater - for example, when the self-efficacy deficit covers a wide range of task domains and the person sees no chance of future improvement.  81  Poorer performances are most likely to affect later depression (a) when they mean that a valued external outcome is not obtained (or an aversive situation is not terminated), (b) when they induce negative self-reactions, or (c) when the task involves control of negative mood. Lastly, anxiety or fear seems to occur when people anticipate that they will be unable to control a potentially aversive situation (Bandura, 1988).  Effects of other variables on the relationship between self-efficacy and depression. The discussion of the relationship between self-efficacy and depression to this point has covered two main areas: how depressed persons use dysfunctional self-evaluation processes, and an explanation how these mechanisms are believed to operate. As well, there is a growing body of evidence suggesting that depression may have a reciprocal relationship with both self-efficacy and performance. Now I will address another possible relationship between self-efficacy and depression - the possibility that self-efficacy and depression covary together because they are responding to the same underlying causes, rather than because one is influencing the other.  In a study which examined mediational versus moderational relationships among control, self-efficacy, and causal attributions in adjustment to chronic illness, Schiaffino & Revenson (1992) obtained data from 64 persons with rheumatoid arthritis. The average age of respondents was 53 years (SD=14) and data was obtained at two intervals four months apart. A mediator was defined as a variable that partly accounts for the relation between a predictor and a criterion variable (Baron & Kenny, 1986), a moderator as a variable that affects the direction and strength of the relation between a predictor and a criterion variable.  Perceived control was measured with two items from the Implicit Models of Illness  82  Questionnaire (Turk, Rudy, & Salovey, 1986); self-efficacy was measured by three self-report items assessing respondents, perceived ability to manage their pain, to deal with physical limitations, and to continue daily activities despite the illness. Causal attributions were measured by three 7-point scales assessing dimensions of internality, stability, and globality (Peterson et al., 1982). Depression was measured by the 20-item CES-D Scale (Radloff, 1977). The investigators used hierarchial multiple regression analysis to examine the relationships among perceived control, self-efficacy, and causal attributions, and their relation to concurrent and later adaptation (i.e., depression and disability) to a stressful illness event, (inflammation in rheumatoid arthritis).  A relationship between causal attributions and self-efficacy would be one of moderation if the relationship of self-efficacy to depression were positive under conditions of internal, stable, and global attributions, and negative or zero under conditions of external, unstable, and specific attributions. The investigators found greater evidence for moderational than mediational models of the relationship. Specifically, when internal, stable, global attributions were made, lower selfefficacy was associated with greater depression. When external, unstable, specific attributions were made, self-efficacy had virtually no relationship to depression. Thus, the relationship between causal attributions and self-efficacy is one of moderation in determining depression.  These findings suggest that self-efficacy and depression covary reflecting underlying causes. In this case, that cause is causal attributions, because self-efficacy is derived in part from causal attributions, for they are one source of appraisal information made about an event (Bandura, 1986; Strecher, DeVellis, Becker, & Rosenstock, 1986).  83  In another study which examined the influence of self-efficacy on depression (Schiaffino, Revensen, and Gibofsky, 1991), the investigators studied 101 recently diagnosed adult patients with rheumatoid arthritis, gathering data twice in a 14-month interval. A s in the 1992 study cited above, self-efficacy was measured by three self-report items assessing respondents' perceived ability to manage their pain, to deal with physical limitations, and to continue daily activities despite the illness. Depression was measured by the 20-item CES-Depression Scale (Radloff, 1977), and pain was measured by the Arthritis Impact Measurement Scales (Meenan, Gertman, & Mason, 1980). The mean age of this sample was 51 years (SD=15). To determine whether pain was responsible for the association between self-efficacy and depression, a hierarchical multiple regression analysis was performed using depression as the criterion variable. The researchers found that at l o w levels of pain, depression did not vary in response to self-efficacy beliefs; but at high levels of pain, individuals with greater self-efficacy beliefs were more depressed one year later than individuals with low self-efficacy beliefs. Self-efficacy appraisals, for patients reporting higher levels of pain, predicted depression one year later, but self-efficacy beliefs were unrelated to concurrent depression either alone or in interaction with severity of pain. Believing in one's ability to handle the situation in the presence of greater pain appears to contribute to greater depression. These results of this study, however, are consistent with self-efficacy theory. Bandura (1986) specifies that reasonably accurate appraisal of one's capabilities is of considerable value i n selecting goals. He emphasizes that an overestimation o f one's personal efficacy - such as a belief that one can handle high levels of arthritis pain - may result in serious undesirable consequences:  People who grossly overestimate their capabilities may undertake activities that are  84 clearly beyond their reach. A s a result, they get themselves into considerable difficulties, undermine their credibility, and suffer needless failures. Some of the missteps, of course, can produce serious, irreparable harm. (p. 393)  Depression, pain, and disability are interrelated among arthritis patients (McFarlane & Brooks, 1988). The causal relationship between pain and depression has been investigated by B r o w n (1990). A s there were contradictory research findings on the relationship between pain and depression, B r o w n conducted structural equation modeling involving 243 patients with rheumatoid arthritis over six time periods six-months apart. The average age of study participants was 53 years (SD=13.4), and the mean duration of their illness since time of diagnosis was 3.33 years (SD=2.21).  B r o w n used structural equation modeling to investigate the direction o f causality i n the relationship between pain and depression. The latent measure of depression was constructed using four subscales from the C E S - D ; the latent variable for pain was constructed using the pain subscale from the Arthritis Impact Measurement Scales (Meenan, Gertman, & M a s o n , 1980) and the V i s u a l Analogue Scale of Pain (Downie, Letham, Rhind, Wright, Branco, & Anderson, 1978). H i s findings were consistent with other studies, in that he found significant associations between the intensity and frequency of pain episodes and the severity of concurrent depressive symptomology i n patients rheumatoid arthritis and other chronic pain syndromes. However, B r o w n found only tentative support for the hypothesis that pain episodes result i n subsequent increases i n the severity of depression in patients with rheumatoid arthritis. During the first twelve months of his study there was no evidence of a causal relationship between the two  85  variables; however, during the last 12 months of the study, the results suggested a predominant causal influence, in which pain predicted depression. Brown suggested that the causal effect of pain on depression may become stronger over 6-month intervals, as rheumatoid arthritis progresses, or as the pain persists and becomes chronic, there is less hope of relief or ability to control it.  The results of Brown's study provide some support for the hypothesis that depression is elicited by chronic pain; however, the explanation of the relationship remains unclear. Further research testing mediational relationships is needed. For example, chronic pain may produce a depressive state as a function of a reduction in one's ability to perform desired and positively reinforcing activities. This explanation would view pain not as the cause of depression, but as a correlate of inability to perform other activities, the effects of which inability are depressing.  In a somewhat similar study conducted by Buescher et al. (1991), the relationship between depression and pain behaviours was examined. Depression was assessed using the Beck Depression Inventory (Beck, Rush, Shaw, et al., 1979), while pain behaviours were assessed by a panel of judges who viewed a 10-minute video of subjects performing a standardized sequence of movements. Hierarchial regression analysis was used to examine the relationship of depression to pain behaviour. After controlling for the effects of age, anatomic stage, and painful joint count, the Beck Depression Inventory score was found to be unrelated to the total pain behaviour score. These results are similar to those of a previous study conducted by Anderson, Keefe, and Bradley (1988).  86  In another study which examined the relationship between depression and pain (Davis, 1989), a purposeful non-random sample of 160 persons with rheumatoid arthritis was gathered from private physicians' offices, a university hospital, and groups associated with The American Arthritis Foundation. The majority were women (76%) and the mean age was 63 years (SD=13). Study participants completed the Chronic Pain Experience Instrument (Davis, 1989) and the CES-D. Analysis consisted of multiple regression analysis using a causal modeling design (Blalock, 1964). Several studies have suggested that depression and pain are closely associated, but there has been much debate about the relationship (Blumer & Heilbronn, 1982; Hendler, 1984; Turk & Salovey, 1984). The results of this study lend support to the hypothesis that depression both influences and is affected by pain.. Being depressed may influence how strongly one perceives the intensity of pain, which then affects one's affective response.  Self-Efficacy and Pain Management in Arthritis When people with arthritis describe their condition, the first and most important descriptor they use is "pain" (Lorig, Cox, Cuevas, Kraines, & Britton, 1984; McGowan & Green, 1995). Sixty-one percent of patients with rheumatoid arthritis and 75% of patients with degenerative joint disease ranked pain as the most important symptom to be treated (McKenna & Wright, 1985). Pain associated with arthritis is the primary reason patients with arthritis seek medical attention (Buckelew & Parker, 1989), and pain management is the primary concern for patients with rheumatoid arthritis (Gibson & Clark, 1985). Gibson and Clark conducted a survey and found that 85% of rheumatologists prescribe simple analgesics for their patients as a supplement to anti-inflammatory medications.  87  The notion that arthritis pain is simply and only the direct result of physiological stimulation has become increasingly untenable, as the research of many investigators has demonstrated the effects on pain of emotional arousal (Bowers, 1968; Craig, 1984); of interpretation and context (Beecher, 1958); of past pain history (Fordyce et al.,1973); of cognitive factors such as attention focus, imagining, and self-encouragement (Chaves & Barber, 1974; Turk, Meichenbaum, & Genest, 1983); of perceived control over painful stimulation (Glass, Reim & Singer, 1971; Miller, 1979); of modeling (Craig, 1983); and of ethnic background (Tursky & Sternbach, 1967; Weisenberg, Kreindler, Schachat & Werboff, 1975; Zborowski, 1969). As well, there is evidence showing that the higher a person's perceived selfefficacy in handling stressors, the lower are his/her simultaneous pulse rate and blood pressure, and the lower his/her blood level of catecholamines (Bandura et al., 1985).  Arthritis pain Rheumatoid arthritis. Flor and Turk (1988) found significant associations between painrelated cognitions and pain level in samples of persons with chronic pain and rheumatoid arthritis. The purposes of their study were: (a) to examine the relationship among two sets of cognitive variables postulated to play a central role in chronic pain by the cognitive-behavioural perspective, namely, pain-related self-statements and convictions of control, with pain and disability levels; and (b) to compare the contribution of these cognitive variables in a sample of chronic pain patients with limited organic findings (chronic back pain), and to a sample for whom the physical basis of the pain was better established (i.e., rheumatoid arthritis). Multiple regressions were computed with disease-related variables as independent, and pain and disability measures as dependent variables. The results demonstrate that both situational pain-related self-  88  statements and more general pain-related perceptions of control are highly related to reports of pain severity and disability, and that the cognitive-related variables are more related to pain and disability than the disease-related variables are. The global convictions are related to selfefficacy expectations that may instigate situation-specific attributions about the inability to control the nociceptive stimulation and, consequently, lack of effort. Conversely, lack of success in controlling the aversive sensations may reinforce the more general sense of helplessness and low self-efficacy. The researchers caution, "the strong association between reports of pain and disability and pain-related cognitions should not be taken as implying a causal relationship. It is quite plausible that the high levels of pain and disability foster the use of castastrophizing selfstatements, inhibit the use of coping self-statements, and produce feelings of helplessness and lack of resourcefulness, as well as the converse. Future research using longitudinal designs and path analyses is required to establish the direction of causality" (Flor & Turk, 1988, pp. 262). The authors also point out methodological problems that make the conclusions of the study tentative: the small sample size, the high percentage of female patients in the samples, and the reliance of all dependent variables on patient self-report.  In another study investigating the effectiveness of cognitive-behavioural treatment for rheumatoid arthritis (O'Leary, Shoor, Lorig & Holman, 1988), fifteen females were administered a cognitive-behavioural treatment designed to enhance perceived self-efficacy to manage pain and disease effects, by providing instruction in skills relevant to self-management of the disease. Self-efficacy theory postulates that people's perceptions of their capabilities affect their behaviour, their thought patterns, and their emotional reactions (Bandura, 1986). In the case of rheumatoid arthritis, stronger self-efficacy beliefs may affect both behaviour and  89  emotional responses to the effects of the disease. Previous research by Shoor and Holman (1984) had found that the higher the patients' perceived self-efficacy to manage pain, the lower their pain four weeks later. In the 1988 O'Leary et al. study, significant interrelationships were obtained between patients' perceptions of their self-efficacy to manage pain and their actual level of pain. Thus, in research studies involving people with rheumatoid arthritis, the results showed correlations between the participants' self-efficacy to manage pain and actual pain level experienced. The sample size in the 1988 O'Leary study was small - only 30 females, with 15 in the experimental group and 15 in the control group. Also, the average age was only 49 years, considerably younger than the average age in my research. The investigators cautiously point out that: Correlations between enhancement of perceived self-efficacy and improvement in arthritis condition cannot be taken as proof of a causal relationship between the two; however, in connection with converging evidence from experimental laboratory studies, such findings in a naturalistic study would be suggestive of such a relationship, (p.537)  The relationship between self-efficacy and pain behaviours has been studied by Buescher et al., (1991). Their study sample was recruited consecutively from a rheumatology practice and consisted of 72 males diagnosed with rheumatoid arthritis, average age 62 years (SD=7). Subjects completed the Beck Depression Inventory (Beck et al.,1979) and the Arthritis SelfEfficacy Scales (Lorig, Chastain et al.,1989). Then each subject was videotaped while performing a 10-minute standardized sequence of movements. A panel of judges viewed the tapes and noted frequencies of specific pain-related behaviours. Data analysis consisted of developing a correlation matrix between self-efficacy and the other measures of pain behaviours,  90  and a series of hierarchial regression analyses to examine the relationship between pain behaviours and selected independent variables.  The major finding of the study was that there were small but statistically significant relationships between all three subscales in the Arthritis Self-Efficacy Scale and pain behaviours. When self-efficacy was high, patients exhibited fewer pain behaviours. The authors note, however, that the variance in pain behaviour accounted for by self-efficacy only ranged from five to seven percent. Therefore, additional variables besides self-efficacy appear to be related to pain behaviour. As well, the authors caution against the inference of a causal relationship regarding the effects of self-efficacy on pain behaviour and vice versa. Subjects' confidence in their ability to manage the disease may affect their physical movements and facial expressions. Conversely, subjects who have difficulty maintaining normal movements may lose confidence in their ability to cope with this disease. Therefore reciprocal relationships are possible.  Osteoarthritis. The role of self-efficacy and knee pain in persons with osteoarthritis was investigated by Rejeski, Craven, Ettinger, McFarlane, and Shumaker (1996). The main purpose of this study was to evaluate the ability of self-efficacy to predict performance during simulated activities of daily living.  Seventy-nine community-based adults with osteoarthritis in the knees were recruited through local advertising and mass mailings. The sample was mostly female (70%) and the mean age was 69 years (SD=6). Task-specific efficacy beliefs were measured using a standardized measurement protocol (Bandura, 1977b), where participants, using a 10-point  91 confidence ladder, rated their confidence in completing a number of specific tasks. The confidence ladder ranged from " 0 " (completely uncertain) to "10" (completely certain). Knee pain was rated on a 10-point ladder with the phrases "no pain at a l l " and "pain as bad as it could be" anchoring the ends of the scale. "Difficulty" was assessed using the same type of scale, with the phrases "easy" and "extremely difficult" at each end of the scale. Performance-related perceptions were rated using the following 5-point bipolar scales: uncomfortable(l)-comfortable(5); uncoordinated(l)~coordinated(5); weak(l)~strong(5);  unstable(l)--stable(5).  Subjects completed self-efficacy scales prior to stair-climbing and lift-and-carry tasks. They then completed the pain, task difficulty, and perceptions of physical ability scales. Pearson-Product Moment correlations showed statistically significant correlations (-.35, p<.01) between self-efficacy and pain intensity. Stepwise linear regression models were used to test for the hierarchial effects of (1) oxygen consumption and strength, (2) most intense pain, and (3) self-efficacy on performance and ratings of disability. The investigators found that: (a) selfefficacy explained a significant portion of the variance in both performance and self-report o f physical disability; (b) pre-task self-efficacy beliefs and knee pain experienced during the performance of physical tasks influenced speed of movement, as well as the post-task difficulty ratings and perceptions of physical ability. A s well, the bivariate correlations and the hierarchial regression analysis suggested that self-efficacy beliefs covaried, in part, with the pain that participants experienced during the performance of the activities. These findings are consistent with self-efficacy theory, which suggests self-efficacy beliefs are due, in part, to physical capabilities and symptoms experienced during performance (Bandura, 1977b, 1986).  \  92  Fibromyalgia. The relationship between self-efficacy and pain in persons with fibromyalgia was investigated by Buckelew, Murray, Hewett, Johnson, and Huyser (1995). The researchers assessed whether self-efficacy predicted self-report pain and physical activity over and above demographic variables, disease severity, and psychological status.  The study sample consisted of 79 female subjects with a mean age of 44 years (SD=10). All subjects met the American College of Rheumatology 1990 classification criteria for fibromyalgia. Assessments included tender point tests (i.e., myalgic score) performed by trained physicians and questionnaires completed by the subjects. Subjects completed several outcomes measures. Pain intensity was measured by the Visual Analogue Scale for pain (Huskisson, 1974, 1983; Revill, Robinson, Rosen, & Hogg, 1976). Overall impact was measured by the Arthritis Impact Measurement Scales (Meenan, Gertman, & Mason, 1980). These scales have been used in samples of persons with mixed types of arthritis (Meenan et al., 1980) and measure the impact of arthritis in various areas of a person's life: mobility, physical activity, activities of daily living, anxiety, depression, and pain level. Self-efficacy was measured by the Arthritis Self-Efficacy Scale (Lorig, Chastain et al., 1989). Multiple regression analyses were conducted with pain and physical activity as the dependent variables.  Self-efficacy predicted self-report pain over and above the impact of demographic variables, disease severity, or psychological distress. Neither demographic variables nor myalgic score significantly predicted pain. The results suggest the importance of self-efficacy, although the authors note that the combination of demographic variables, myalgic scores, psychological status, and self-efficacy only accounted for a range of 21% to 26% of the variance in predicting  93 self-report p a i n . T h i s suggests the importance o f other variables not addressed i n the study, s u c h as s o c i a l support a n d other c o p i n g behaviours. A s w e l l , the authors w a r n that the research w a s essentially a correlational study. It p r o v i d e d e v i d e n c e f o r a n association b e t w e e n s e l f - e f f i c a c y a n d p a i n , but c o u l d not establish the direction o f the relationship.  C o l d - p r e s s o r p a i n . T h e relationship between s e l f - e f f i c a c y a n d p a i n l e v e l has b e e n investigated i n studies i n v o l v i n g undergraduate female university students.  L i f t (1988)  investigated whether s e l f - e f f i c a c y expectancies are causal determinants o f b e h a v i o u r , as B a n d u r a (1977b) asserts, or whether they are s i m p l y correlates o f b e h a v i o u r change.  In research  i n v o l v i n g 62 f e m a l e undergraduates he investigated whether s e l f - e f f i c a c y has v a l i d i t y as a true causal determinant o f b e h a v i o u r change or is a correlate o f change that has already o c c u r r e d . Subjects participated i n a " c o l d - p r e s s o r test" w h i c h consisted o f s u b m e r g i n g their n o n - d o m i n a n t a r m into f r e e z i n g water for as l o n g as p o s s i b l e (i.e., until it b e c a m e too p a i n f u l ) . S e l f - e f f i c a c y scales, the Internal-External L o c u s o f C o n t r o l Scale (Rotter, 1966), R o s e n b a u m ' s (1980) S e l f C o n t r o l S c a l e , a n d a v i s u a l analogue p a i n scale were c o m p l e t e d p r i o r to the first o f three trials. T h e s e l f - e f f i c a c y a n d p a i n scales were c o m p l e t e d at each trial, a n d the length o f t i m e s u b m e r g e d was r e c o r d e d b y the researcher. B e t w e e n the trials, subjects were g i v e n " b o g u s " p e r f o r m a n c e f e e d b a c k i n d i c a t i n g they were d o i n g w e l l (i.e., i n the 90th percentile i n a national s a m p l e o f f e m a l e undergraduates) or not d o i n g w e l l (i.e., i n the 37th percentile).  T h e results s h o w e d that self-efficacy expectations c o u l d be readily m a n i p u l a t e d a n d that these expectations c o u l d strongly predict future c o p i n g p e r f o r m a n c e . C h a n g e s i n s e l f - e f f i c a c y s i g n i f i c a n t l y predicted changes i n tolerance t i m e ; h o w e v e r , a subject's p e r f o r m a n c e s e l f - e f f i c a c y  94 did not predict her perceptions of pain. In another study using cold-pressor pain, Neufeld and Thomas (1977) examined the role of false efficacy feedback on 83 female undergraduates who were using relaxation to alleviate cold-pressor pain. The study demonstrated that perceived efficacy at mastering a coping response significantly elevated subsequent pain tolerance, and that elevated tolerance was not accompanied by increases in objective or subjective measures of the designated coping response of relaxation. It would appear that subject's appraisal of the effectiveness of her experimental coping resources was the critical factor in her increased coping tolerance. The results suggest that perceived self-efficacy, independent of actual performance on the coping task, affect a person's ability to cope with pain. However, it must be noted that despite the greater tolerance exhibited, there were no differences in the subjects' perceptual experience of pain.  In a third study on management of cold-pressor pain, Reese (1982) found that the strength of perceived self-efficacy to tolerate pain was significantly correlated with pain threshold and tolerance. Also, contrary to the previous cold-pressor pain studies, self-efficacy in reducing pain accurately predicted pain tolerance as well as threshold. These studies demonstrate effects of self-efficacy in laboratory-induced pain.  The research of Litt (1988) and Neufeld and Thomas (1977) found that increases in selfefficacy significantly predicted increases in tolerance time, but that the subject's self-efficacy did not predict her perception of pain intensity. The results of the study by Reese (1982), however, are contrary to the previous studies, in that he found self-efficacy in reducing pain accurately predicted pain tolerance as well as pain threshold, the other studies found that self-efficacy was  95  related to coping performance but not to the criterion variable of pain intensity. The relevance and comparison of the studies by Litt and by Neufeld and Thomas with research involving persons with arthritis must be tentative as these were contrived laboratory experiments with undergraduate college females using artificially induced pain, where the arthritis population tends to be older women. As well, clinical pain experienced by persons with arthritis is characterized by higher levels, and because there is little control people experience anxiety, which may further increase the pain.  Summary - Arthritis pain. To summarize, there have been several studies which have examined the relationship of self-efficacy and pain in the area of arthritis. These studies have demonstrated statistically significant negative correlation levels between self-efficacy and pain intensity at one point in time (Lorig, Mazenson, & Holman, 1993; Rejeski et al. 1996; Buckelew et al. 1995). The studies involved samples of persons experiencing different types of arthritis, and did not involve interventions. The study conducted by Flor and Turk (1988) examined cognitive-behavioural models of chronic pain in rheumatoid arthritis, and demonstrated that both situational pain-related self-statements and more general pain-related perceptions of control, which are likely to be related to self-efficacy expectations, were negatively related to reports of pain severity. The study conducted by Buescher et al., (1991) demonstrated a small but significant relationship between self-efficacy and pain behaviours in a sample of 72 senior males. In the study examining the relationship between self-efficacy and pain levels in adults with osteoarthritis, Rejeski et al. (1996) showed a strong negative relationship between self-efficacy levels and pain levels experienced in performing daily activities. With a population of persons with fibromyalgia, Buckelew et al. (1995) found a significant negative relationship between self-  96  efficacy and pain intensity currently being experienced. Lastly, Lorig found statistically significant negative relationships between self-efficacy and pain levels in pre-intervention assessment scores with a sample of mixed types of arthritis (refer to Chapter 1). These studies illustrate the relationship between self-efficacy and perceived pain level and pain behaviours at the same time point, in persons experiencing different types of arthritis.  Another series of studies has illustrated the relationship between self-efficacy and pain levels following cognitive-behavioural treatment interventions where self-efficacy was experimentally manipulated (Shoor and Holman, 1984; Lorig et al., 1985, 1986, 1993; O'Leary etal.,1988; McGowan, 1990; Buescher et al.,1991; Lorig & Holman, 1993; McGowan et al., 1994; Buckelew et al., 1995; McGowan and Green, 1995). In further analysis Lorig, Chastain et al. (1989) investigated the relationship between changes in self-efficacy and changes in pain levels, and found statistically significant negative correlation levels. The same finding was made by McGowan et al. (1994). Statistically significant correlations were observed between changes in self-efficacy and changes in pain level following the A S M P , but only a weak negative relationship was found between changes in exercise and changes in pain level (Lorig, Chastain et al., 1989).  Childbirth pain. In relationship to childbirth, Manning and Wright (1983) investigated the relative roles of self-efficacy expectations, outcome expectations, and importance of not using pain medication as predictors of persistence of pain during labour and delivery. This study was unique in examining both self-efficacy expectancies and outcome expectancies. The Manning and Wright study demonstrates that laboratory-derived procedures for studying self-  97 efficacy can be applied and tested in a field setting, which increases the external validity of the self-efficacy construct by extending its application to a naturalistic situation encountered by millions of women each year. Also, the study replicates a major self-efficacy finding, that selfefficacy predicts behaviour after treatment better than behaviour during treatment.  Self-efficacy expectancies represent an individual's assessment of his/her potential for having the ability to perform behaviour X, whereas outcome expectancies involve the individual's assessment that behaviour X will lead to outcome Y. Study subjects were 52 primiparous women with a mean age of 27 years. The study design involved three phases: Phase 1 was the week after their last childbirth class; Phase 2 was during the early stages of labour; and Phase 3 was the week following delivery. Participants completed questionnaires three times, once during each of the phases. During the labour phase the measure for self-efficacy for ability to control pain included five self-efficacy scores, one for each 5-hour interval in a hypothetical 25-hour labour. After delivery, once the actual length of labour had been determined, the selfefficacy expectancy score for the time block corresponding to the woman's actual endpoint of labour was selected as the woman's relevant self-efficacy score. The scale consisted of a number of statements concerning her anticipated ability to control the pain of labour and delivery without pain medication. A 12-point continuous scale was used. A " N o " with "complete certainty" was scored as a 1, continuing up through " N o " with "complete uncertainty" as a 6. A "Yes" with "complete uncertainty" was scored as a 7, continuing up through a "Yes" with "complete certainty" as a 12. Outcome expectancy (i.e., the expectancy that pain-control techniques would lead to the ability to control pain of labour and delivery without pain) was measured using a 6point Likert item indicating agreement/disagreement with the statement "the techniques for controlling pain which are taught in Prepared Childbirth classes will make it possible for a  woman to go through labour and delivery without pain for X hours." The importance of not using pain medication during labour and delivery was measured by a single 7-point Likert item ranging from 1 "not important to me" to 7 "extremely important to me."  Persistence in pain control was assessed by determining whether the subject used pain medication, and, if so, the percentage of time in labour without medication. Other measures used in the analysis included Rotter's (1966) Internal-External Locus of Control Scale, the MarlowCrowne Social Desirability Scale (Crowne & Marlowe, 1960), the type of pain training received (didactic vs. didactic plus experiential), the length of labour, the type of delivery (vaginal vs. cesarean), and past medication use for pain.  , Analyses showed that self-efficacy expectations, outcome expectations, and importance correlated negatively with use of pain medication, and positively with percentage of time in labour without pain medication. With "the use of pain medication" as the criterion, hierarchial multiple regression analysis found that self-efficacy expectations add to outcome expectations and importance in predicting medication use. Using "percentage of time without medication" as the criterion, self-efficacy expectancies added to outcome expectancies. The investigators also found that of the additional predictor variables, five (type of pain-related training, length of labour, type of delivery, past medication use for pain, and locus of control) were unrelated both to the use of medication and to the percentage of time without medication.  Self-efficacy expectations predicted persistence in pain control without medication better than outcome expectations, importance, and seven other alternate predictors. Thus, the  99  investigators were able to demonstrate that the more efficacious the women judged themselves to be, the less likely they were to request medication during delivery, and if in fact they did request it, the longer they tolerated pain before doing so. These findings were similar to those of previous research in childbirth by Cogan, Henneborn, and Klopfer (1976), who found that cognitive and psychological factors rather than aspects relating to actual methods of training were related to the perception of pain and the use of medication. The authors note two limitations of the study, namely that subjects were motivated to begin with, as they were attending childbirth classes, and that there was a lack of differentiation between outcome and self-efficacy expectations, because there were high correlations between the two constructs.  Tension headache pain. As it bears upon tension headaches, Holroyd et al.,1984, found that an increase in self-efficacy was a sure sign of mitigation and comparative rarity of subsequent headaches. Forty-three college students suffering from recurrent tension headaches participated in a six-session E M G Biofeedback training program. The study was based on the premise that if the subjects perceived biofeedback training as a credible treatment, and perceived themselves as succeeding at the biofeedback task through their own efforts, they would view (a) their headaches as having an internal locus of control, and (b) themselves as self-efficacious. These cognitive changes were expected to lead to new and more persistent efforts to cope with headache-related stresses that, in turn, would alter psychological and physiological stressresponses triggered headaches. Participants received "bogus" feedback on how well they were using the E M G feedback (i.e., high-success or moderate-success feedback). Participants completed hourly ratings of headache intensity, and recorded the frequency, duration, average intensity, and medication intake. They also completed a self-efficacy scale developed by  100  Reynolds, Creer, Holroyd, & Tobin (1982), modified to include only questions associated with the onset of headaches, and the Wallston et al. (1978) Multidimensional Health Locus of Control Scale.  Analysis of covariance was conducted with pre-treatment score as the covariate. Subjects who received feedback for increasing E M G activity showed substantially higher levels of E M G activity than did subjects who received feedback for decreasing E M G activity. The high-success feedback group rated themselves as more successful at the biofeedback task, had lower headache activity scores at post-treatment, and showed larger reductions in headache activity than did subjects who received moderate-success feedback. The A N C O V A ' s conducted on post-treatment headache activity scores revealed a significant main effect for performance feedback. Persons in the high-success feedback group were more confident that they could manage headache-related stresses without incurring a headache, and they viewed the variables affecting their headaches as more within their control than the moderate-feedback group did. Thus, changes in self-efficacy scores occurring during the treatment were significantly correlated with improvements in headache activity at the post-treatment assessment. These results suggest that improvements in headache activity following biofeedback training are mediated by cognitive changes induced by performance feedback, rather than by reductions in E M G activity: in the study, improvements in tension headache were clearly related to performance feedback, but unrelated to the direction of change in E M G activity.  It is noteworthy that the findings of the research conducted in arthritis are similar to the findings of studies relating to cold-pressor pain, childbirth, and tension headaches. In the study  101  investigating the relationship between self-efficacy and pain level experienced during childbirth labour and delivery, Manning and Wright (1983) demonstrated that there were correlations between self-efficacy and coping behaviours but not between self-efficacy and actual pain level; actual pain levels were not assessed. Instead, the investigators considered whether the subject used analgesics to deal with the pain, and/or the amount of time she spent without analgesic. The findings obtained by Holroyd et al. (1984) in their research examining the relationship between self-efficacy level and tension headache pain found a negative relationship between self-efficacy and headache activity, as measured by the mean pain intensity over a one-week period.  Self-Efficacv and Perceived Health Status The relationship between self-efficacy and perceived health status was investigated by Taal, Rasker, and Seydel (1993). This study involved 73 adults with R A who were recruited by four rheumatologists. The study consisted of an in-home interview followed by a mail questionnaire four weeks later. The average age of subjects was 60 years (SD=14 years).  During the in-home interview, four main areas were examined: (a) main disease problems; (b) significant arthritis-related problems (i.e., problem-index); (c) whether the subject had received recommendations regarding medication use, adjustments in daily activities and physical therapy by health professionals, and whether he/she had experienced problems adhering to these recommendations; (d) social support and arthritis self-efficacy levels. The researchers developed a social support scale which consisted of four items - two measuring perceived emotional support, and two measuring instrumental support. To measure arthritis self-efficacy the investigators developed a nine-item, five-point arthritis self-efficacy scale.  102  Four weeks later, subjects were mailed the D U T C H - A I M S questionnaire (Taal, Jacobs, Seydel, Wiegman, & Rasker; 1989). The DUTCH-AIMS measures nine aspects of health status: mobility, physical activity, dexterity, household activity, social activity, activities of daily living, pain, depression, and anxiety. With the DUTCH-AIMS, the higher the score the worse health status. As well, subjects completed a horizontal visual analogue scale for global assessment of arthritis, and two current laboratory measures of disease activity (haemoglobin count [Hb] and erythrocyte sedimentation rate [ESR]) were provided by the participating rheumatologists. Higher ESR signifies more arthritis-related disease activity taking place, and higher Hb represents better health status.  There were significant negative correlations between arthritis self-efficacy and health status, and between self-efficacy and ESR, and a significant positive correlation between selfefficacy and Hb. Because self-efficacy was significantly negatively correlated with ESR and positively correlated with Hb, partial correlations between self-efficacy and self-reported health status were conducted, controlling for the laboratory measures. Significant partial correlations were found, and therefore the negative significant correlations between self-efficacy and selfreported health status could not be explained by the significant correlations between self-efficacy and disease activity. Also, partial correlations between self-efficacy and ESR and Hb were calculated, controlling for self-reported health status, and they were not significant. This suggests that the correlations between self-efficacy and ESR and Hb can be explained by the correlations with self-reported health status.  In this study Taal et al. seemed to subscribe to an underlying assumption that there is a causal relationship between self-efficacy and health status, namely that self-efficacy leads to  103  improvements in health status. The rationale presented employs the following logic: as rheumatoid arthritis is unpredictable and disease activity varies, people perceive that the disease is uncontrollable, and this leads to low self-efficacy expectations about managing the disease and its consequences. The individual acquires the feeling that he/she has no control over the disease, and this leads to anxiety and depression, which lead in turn to increased perceptions of pain, and reduced efforts to cope with the disease's consequences or to engage in daily activities. As a consequence, health status further deteriorates. By this reasoning, strategies aimed at strengthening self-efficacy expectations about managing pain or other physical or psychosocial consequences of R A would lead to better self-management behaviours and eventually to better health status. Unfortunately, as the researchers employed correlation analysis, they can only conclude that there is a relationship between self-efficacy and health status.  Self-Efficacy and Smoking Cessation Self-efficacy for smoking cessation is relevant to individual behaviour change, as well as to health education programs, since most smokers who quit do so without professional assistance. In reviewing the literature, one can find that though most programs aim at getting people to discontinue smoking achieve a certain amount of success at the end of the program, the percentage of cases that relapse is high (Bernstein & Glasgow, 1979; Hunt & Bespalec, 1974).  Marlatt and Gordon (1980) focused on "relapse-situation analysis," and investigating situations in which subjects relapsed into their prior condition of alcohol abuse, drug use, or smoking. They found that 76% of the relapses had occurred under three main conditions:,  104  1  interpersonal negative emotional states (e.g., frustration, anger, anxiety, depression, loneliness) - 37%;  2  social pressures to resume behaviour - 24%;  3  interpersonal conflict (e.g., anger or worry) -15%.  They hypothesized that negative self-referent ideas result from the first lapses which makes the individual prone to relapse. The process seems to progress as follows: the individual feels guilty after the first slip-up, and makes personal attributions for the lapse as a weakness. This leads to negative feelings and decreases confidence in his/her ability to remain abstinent. The individual starts to focus on the positive aspects of the old habit, which sets the stage for full relapse. Effective coping mechanisms are needed in such high-risk situations. Relapse might be avoided if one of two conditions exists: (a) the individual is possessed of enough self-efficacy to mobilize sufficient will to resist (self-efficacy increases as the person successfully resists temptations); or (b) the individual's self-efficacy to recover from a lapse is high enough to reestablish self-control after the lapse.  A number of researchers have investigated the relationship between perceived selfefficacy in maintaining abstinence and outcome measures of successful abstinence. Self-efficacy questionnaires designed to measure an individual's confidence in his/her ability to maintain abstinence from smoking have been developed by several researchers (Colletti, Supnick, & Payne, 1985; DiClemente, 1981; DiClemente, Prochaska, & Gilbertini, 1985; Godding, Glasgow, & Klesges, 1985; Reynolds et al., 1982). In the studies, subjects have typically discontinued smoking with the assistance of a behaviour treatment program. Post-treatment self-  105  efficacy scores were used to predict the likelihood of abstinence at follow-up. Studies (Colletti, Supnick, & Payne, 1985; Mclntyre, Leichtenstein, & Mermelstein,1983) showed that the level of self-efficacy was related to the probability of a relapsed smoker at three and six months after treatment - enhanced self-efficacy correlated with the reduction of smoking. DiClemente (1981) found that the higher the subjects' self-efficacy rating at the end of the treatment, the longer were they able to maintain abstinence, and less the difficulty that they had in maintaining it. Prochaska, Crimi, Lapsanski, Martel, & Reid (1982) found that when compared to those who relapsed, those who continued to abstain relied more on inner-directed, experiential processes of change, and demonstrated higher levels of self-efficacy. The predictive power of self-efficacy was demonstrated by Colletti, Supnick, and Rizzo (1981), who found a telling correlation between an individual's self-efficacy in a particular situation and whether or not he/she did ultimately relapse.  The effects of several moderating and mediating variables in conjunction with selfefficacy have been examined by several researchers: locus of control and confidence in treatment rationale (Walker and Franzini, 1983), expectancies of the positive effects of smoking (Godding et al., 1985), degree of physical dependence (Killen, Maccoby, & Taylor, 1984; Mclntyre et al., 1983), history of coping, and motivation to discontinue smoking (Reynolds et al., 1982). The findings in all of the above studies indicated that the only good predictor of outcome was the level of self-efficacy. On all counts, the findings of research on self-efficacy and smoking provide consistent evidence that perceived self-efficacy is a reliable predictor of who will remain abstinent and who will relapse. The studies also show that self-efficacy to abstain is a better judge of the prospect of relapse than the degree of physiological dependence, past history of  106  coping, motivation to quit, confidence in treatment rationale, or expectancies concerning the rewards of smoking.  Self-Efficacy and Cardiac Rehabilitation With myocardial infarction recovery is expedited tremendously by the enhancement in the patient's and the spouse's estimation of the patient's physical and cardiac capabilities. The higher the prior self-efficacy in one's capabilities, the more probable is recovery (Ewart, Taylor, Reese & DeBusk, 1984; Taylor, Bandura, Ewart, Miller & DeBusk, 1985). Ewart (1992), drawing on her education work with patients following heart attacks, argues that self-efficacy theory provides methods to anticipate and identify patients' uncertainties, and to increase their participation in healthful exercise by raising their self-appraisal. The theory asserts that the most effective way to help patients overcome inappropriate fears of exertion is to have them perform feared activities in gradually increasing intensities (mastery experiences), to permit them to observe other patients like themselves performing the activity (vicarious mastery), to have medical staff provide reassurance and encouragement (persuasion), and to prevent the "pathologizing" of innocuous bodily sensations or states (purging of misinformed physiological phobias).  Conclusion The studies reviewed consistently show that the individual's perception of his/her selfefficacy is related to different health behaviours. A major difficulty with almost all of the studies cited relates to the measurement of self-efficacy. Researchers in each of the studies developed their own scale to measure self-efficacy, but they did not publish articles detailing the process  107  and procedures used, or specific data that would enable one to evaluate the validity of the scale.  It has been suggested by Blanchard (1982) that "there is a need for wide ranging integration focusing on general behavioural-change procedures and how they can be applied across a variety of problem areas." The general strategy in self-efficacy theory, of assessing and enhancing the self s impression of its efficacy to affect health, has considerable general utility.  Critical Appraisal of Self-Efficacy Theory According to McGuire (1983), the adequacy of a theory can be judged by: 1.  its internal consistency in not yielding mutually contradictory derivations;  2.  the extent to which it is parsimonious, i.e., broadly relevant while using a manageable number of concepts;  3.  its plausibility in integrating with prevailing theories in the field; and  4.  its ecological validity.  Internal consistency. The review undertaken in this chapter of self-efficacy studies demonstrates that when the evidence is taken as a whole, it is consistent in showing that increases in the individual's perception of his/her own self-efficacy are associated with several different types of behaviours, as well as with perceptions of depression and pain.  Parsimony of the theory. Self-efficacy uses a manageable number of concepts. It is a broad and integrative theory which seeks to explain a large body of data. It specifies what strategies and methods may be used to increase levels of self-efficacy. Its hypotheses and  108  applications pertain to specific situations and do not provide broad, character-trait definitions. Also, regarding its broad relevance, it is interesting to note that the programs that have employed self-efficacy theory have cut across diverse ages, ethnicities, and income levels.  Plausibility in integrating with prevailing theories. Certainly, the constructs of modeling and reinforcement fit well with diffusion theory (Green, Gottlieb, & Parcel, 1991) in explaining the stages of adoption. Self-efficacy theory syncretizes well with attribution theory (Kok et al., 1992; Weiner 1985, 1986), learned helplessness theory (Seligman, 1975), depression theory (Bandura, 1982a; Kavanagh, 1992), protection motivation theory (Rogers, 1975, 1983, 1985), locus of control theory (Rotter, 1966), the theory of reasoned action (Ajzen & Fishbein, 1980), the theory of planned behaviour (Ajzen, 1988), and the health belief model (Becker & Rosenstock, 1987; Rosenstock, Stretcher & Becker, 1988), as well as other cognitive behavioural theories (Fleury, 1992). However, Bandura is very critical of theories involving Stages of Change (Prochaska & DiClemente, 1982, 1984, 1985, 1992), and considers them a regressive step in the planning and evaluation of health education programs (Bandura, 1995).  Ecological validity. This refers to the extent to which the theory accords with reality when empirically tested in various environments or circumstances. In the last decade or so, selfefficacy theory has been tried extensively in a wide range of health education programs, including smoking cessation, chronic illness, pain management, medical regimen adherence, and cardiac rehabilitation. The reviews of these programs (Clark, 1987; O'Leary, 1985; Schwarzer, 1992; Strecher et al., 1986) have ascertained that self-efficacy reliably predicts who will benefit from the health education program in question. In the area of smoking cessation, for example, self-efficacy to abstain is a better predictor of who abstains and who still smokes at the close of  109  the program and at the time of follow-up than the degree of physiological dependence, past history of coping, motivation to quit, confidence in treatment rationale, and expectancies concerning the rewards of smoking (O'Leary, 1985). In his review of a variety of health education programs, Schwarzer (1992) concludes: Self-efficacy has proven to be a very powerful behavioural determinant in many studies, and its inclusion in theories on health behaviour, therefore, is warranted. By summing up direct and indirect effects, it can be stated that the total effect of self-efficacy on health behaviours exceeds the effects of any single variable. Self-efficacy determines the appraisal of one's personal resources in stressful encounters and contributes to the forming of behavioural intentions. The stronger their self-efficacy beliefs, the higher are the goals people set for themselves, and the firmer their commitment to engage in the intended behaviour, even if failures mount, (p. 223)  A derivative strength of self-efficacy treatment is suggested by evidence that acquired self-efficacy may be transferrable to other situations where the context is similar, and that selfefficacy can be used as a surrogate for behaviour change, in that it has a strong correlation to behaviour.  Weaknesses of Self-Efficacy Theory. It is unclear how self-efficacy is maintained over time. Studies that have investigated health behaviour using self-efficacy theory have found efficacy expectancies to be an important factor in the individual's desire to initiate lifestyle change. However, the role of self-efficacy in maintaining change over time is not clear (Fleury, 1992). Clark (1987) has declared that the programs using self-efficacy theory have been unable to produce long-term effects. More research exposing the link between the intention to change,  110  the actual behavioural change, and its maintenance is needed.  The role of moderating and mediating variables in conjunction with self-efficacy in health education programs has often been investigated. There is still a need to examine to a greater depth individual differences in valuation of the outcome of proposed behavioural changes, as well as to identify related cognitive mediators such as locus of control (Rotter, 1966), role of social support, (Skipper and Leonard, 1968), attributions (Schiaffino & Revenson, 1992), and pain (Schiaffino, Revenson, & Gibofsky, 1991). Attention to self-efficacy alone is not adequate when interventions are aimed towards behaviour change.  Though the individual's perception of efficacy in a given behaviour may be one important determinant in the decision to initiate behaviour, individual differences in the value of the proposed outcome must also be considered (Ewart, 1992; Maddux & Rogers, 1986). Sometimes individuals may attach a greater value to maintaining unhealthy behaviours.  The majority of research on self-efficacy theory has examined the part played by selfefficacy as the sole agent of behaviour. Few studies have examined efficacy expectation in conjunction with outcome expectation. Outcome expectancies are essential components of value-expectancy models of behaviour change, and may provide an explanation for behavioural changes in conditions of uncertain health outcomes (Maddux & Stanley, 1986; Rosenstock et al.,1988).  Self-efficacy interventions are appropriate in situations such as smoking cessation and weight management programs, where arresting the condition is actuated by making specific  Ill  behaviour changes in circumstances that are well defined. But self-efficacy may not aid health education programs where the causal relationship is not clear.  Lastly, it has been argued that efficacy expectancies are simply epiphenomena that only reflect behaviour change, or health status changes, and that are in no sense the cause of change (Hawkins, 1992, 1995; Eysenck, 1978; Borkovec, 1978; Eastman & Marzilier, 1984). To date, all of the research studies conducted have used correlational designs and therefore the causal relationship between self-efficacy and behaviour or between self-efficacy and health status changes is unknown.  Self-efficacy theory comes under the broad designation of value-expectancy theories. In value-expectancy theories, an individual's behaviour is influenced by the subjective value of the end result and the subjective probability that an action will achieve that result. However, three weak points arise in traditional value-expectancy theories.  First, variables involved in predicting behaviour, in their ranking, and in the nature of their interaction are assumed to be constant from the time the need is perceived through to the time behaviour is initiated. Many factors, either individual or environmental, can interfere and moderate between prediction of behaviour and initiation of behaviour. This is particularly true in health research conducted over a period of weeks or months.  Second, deductively derived motivational models are intended to identify cognitive mediators behind an individual's decision to initiate behaviour. These models do not examine  112  the meaning of behaviour change for the individual. Research that affects health are particularly meaningful for participants.  Third, value-expectancy theories fail to weigh the diverse goals and expectancies of the individuals for whom an intervention is designed. Individual goals and expectancies can vary considerably.  113  CHAPTER THREE  M E A S U R E M E N T OF SELF-EFFICACY AND H E A L T H STATUS V A R I A B L E S U S E D I N THIS S T U D Y  Self-Efficacy Scales  Self-efficacy is defined as one's estimate of one's own capacity for coping successfully with a particular situation or task. Therefore it is behaviour-specific, unlike more general concepts like self-esteem and locus of control as these have been used in health-related research (e.g., Wallston, Wallston, & DeVellis, 1978). According to Bandura, no global sense of selfefficacy exists, since self-efficacy is related to a specific behavioural task. However, researchers have ignored the originator's admonitions regarding specificity, and have developed generalized self-efficacy scales (Sakano & Tohijoh, 1986; Sherer, 1982). Measurement has three parameters: level of self-efficacy, which refers to the performance attainment expected by the individual; strength of self-efficacy, which refers to the individual's confidence in attaining each expected level; and generality of self-efficacy, which refers to the several areas of functioning in which the individual judges himself/herself to be efficacious.  The first approaches to developing self-efficacy scales were in the field of phobias, where behaviours toward threats were ordered hierarchically. This was an effective approach to phobias, as the behaviour was linear, and the individual's condition and progress could be placed on a continuum; however, the hierarchical approach is not appropriate for more complex behaviours such as smoking (Yalow & Collins, 1987). Another method of measuring selfefficacy involves a two-step approach wherein the individuals are first asked to rate each  behaviour on the basis of "magnitude" (i.e., how much of the behaviour they can perform) and then, if they respond affirmatively, they are asked to rate the strength of this belief. Typically these scales employ a range from 10, which indicates little confidence, to 100, which signifies a great deal of confidence, in 10-point intervals, though seven-point Likert scales have also been used. A one-step adaptation of the two-step process is the scale in which respondents are asked to rate their own efficacy on an expanded scale on which the magnitude question is not asked. Correlations between the subsequent performance and the strength measure have been consistently higher than correlations between the subsequent performance and magnitude. It appears that the magnitude ratings are of little utility in the assessment of self-efficacy, and therefore respondents are encumbered with an additional complexity (Yalow & Collins, 1987).  When constructing scales, it is essential that an assessment of self-efficacy retain a clear focus on perceptions of ability rather than whether the individual wills or intends to engage in a specified activity, as these wishes or intentions are clearly altogether different constructs. Belief that one can do something (i.e., self-efficacy) is theoretically, operationally, and empirically distinct from whether one desires or intends to do so (i.e., intention). Studies in which both intentions or willingness and self-efficacy were measured verify that self-efficacy is a major determinant of intention and willingness to perform threatening activities (Arch, 1992a, b; deVries & Backbier, 1994; deVries, Dijkstra & Kuhlman, 1988; Dzewantowski, Noble & Shaw, 1990; Kok, deVries, Mudde & Strecher, 1991; Schwarzer, 1992). Also, though predictions about future behaviour are partially based on self-efficacy, they also include considerations such as incentives, values, and beliefs (Bandura, 1977b). These factors should be taken into account and measured in the research design, to ensure that they are not incorporated into the self-efficacy rating.  115  A literature search using "Medline" and "Psyc Lit" databases for articles, in English, on the development and validation of self-efficacy scales between 1980 and 1995, produced sixtyfive articles. Of these, five articles could be classified as addictive behaviour-specific selfefficacy scales (e.g., alcohol and chemical dependency, and abstinence from smoking); nine were about illness-specific self-efficacy scales (e.g., epilepsy, asthma, COPD, and arthritis); eleven were concerned with situation-specific behavioural scales (e.g., self-efficacy in pre-operative patients, nurses' self-efficacy in their pediatric skills, heart-healthy eating, and condom use); and the other forty articles described the development and validation of a vast range of self-efficacy scales (e.g., public health student self-efficacy, prenatal nursing self-efficacy, career decisionmaking self-efficacy, and athletes' physical self-efficacy).  Development of the Arthritis Self-Efficacy Scale  According to Bandura, self-efficacy is behaviour-specific and not generalizable. The Arthritis Self-Efficacy Scale was developed in 1989 by Lorig, Chastain, Ung, Shoor, and Holman at Stanford. This group has been involved in a patient education program called the "Arthritis Self-Management Program" (ASMP) since 1980, and has conducted a series of research studies on it over a 10-year period. Self-efficacy has been shown to be associated with the program participants' improvements in health. In this context, self-efficacy theory states that: (a) perceived self-efficacy will predict future health condition, given that the individual values an outcome of improved health status, and that the individual believes that he/she will achieve an improvement in health status by performing the behaviour; (b) levels of self-efficacy are not immutable traits, but can be altered; and, (c) the improvement in levels of self-efficacy will correlate with improvements in health status. The hypotheses they tested in developing the scale  116 were: (a) that self-efficacy is associated with present and future health statuses; (b) that selfefficacy can be enhanced through educational intervention; (c) that changes in self-efficacy will be associated with the status of health.  Procedure Employed in the Development of the Arthritis Self-Efficacy Scale  The procedure used by Lorig et al. (1988) involved developing and testing an instrument with one population, refining the scale through a second population, and then using the original population to confirm and validate the scale. A procedural description follows. 1.  Generating the items. The research team reviewed a document entitled "Arthritis Outcomes Manual," and reviewed the behaviours involved in managing the symptoms and physical limitations imposed by arthritis. A rheumatologist developed a set of 23 questions which were administered to individuals in three focus groups (6-10 persons in each group). After the individuals in the focus groups completed the questionnaire, they were asked whether they had had any difficulties in understanding and completing it. Then the researcher explained the concept of self-efficacy and invited the participants to generate additional items. A n additional 20 items were developed.  2.  Developing the instrument. The rheumatologist then developed questions from these 43 items and administered the scale to individuals in a group of 97 people who were about to take the A S M P . They completed the scales at that point, and again four months later, after they completed the A S M P . Additional health status scales they completed measured pain levels, depression, quality of life, and functional ability. Factor analyses were conducted and two factors were revealed: self-efficacy in managing symptoms, and selfefficacy in managing functions. Then a correlation analysis assessed whether the theory functioned as hypothesized, that is whether there were strong associations between self-  efficacy and present and future health status. The correlations were strong enough to be affirmative. Thus the construct validity of the instrument had been addressed. 3.  Determining concurrent validity. A "blinded" researcher visited 43 persons who constituted the sample and who had completed the A S M P and both pre- and postprogram questionnaires. The visit took place in the subject's home. Each subject performed the behaviour specified by the scale, and was monitored for how long he/she took to complete the behaviour. These times were analyzed (on a scale from 0, for no difficulty, to 4, for great difficulty) and then correlated with the self-efficacy scores. The result was a correlation coefficient ofr = .61 (p_<.01).  4.  Conducting a replication study. The original scales were then completed by a second sample of 144 persons who had completed the A S M P and both questionnaires. The original 43 items for self-efficacy were used. This second factor analysis brought forth the original two factors and, additionally, the factor of self-efficacy in ability to handle pain. The correlational analysis conducted demonstrated strong associations between self-efficacy and health.  5.  Conducting a confirmatory study. The scales were then administered to the original sample, and the same analysis was conducted. High coefficient alpha estimates of internal reliability were obtained: .90 for Function Self-Efficacy; .87 for Other Symptoms Self-Efficacy; and .75 for Pain Self-Efficacy.  6.  Conducting a test-retest study. A third sample of 91 individuals who had taken the A S M P completed the scales at two points in time with a mean separation period of 9 days. The correlation coefficients for the three self-efficacy scales were r = .85, r = .87, and r = .90 for the areas of function, pain, and other symptoms respectively.  118  Comments  Bandura states that there are three measurable parameters involved in self-efficacy: generality, level, and strength. Generality refers to the number of behaviours or illness areas to which self-efficacy applies. In the arthritis program example, this was addressed through factor analysis, and the three domains were self-efficacy for pain, function, and symptoms. Level measures the individual's expectations for performance attainments. Strength of self-efficacy is concerned with the participants' confidence in his/her abilities. The following excerpt from Lorig, Chastain, et al. (1989) will illustrate: "AS OF N O W , H O W C E R T A I N A R E Y O U T H A T Y O U C A N : 1. Walk 100 feet on flat ground in 20 seconds? 2. Walk 10 steps downstairs in 7 seconds?" The scale here would range from 10 for "very uncertain" to 100 for "very certain" intervals of 10. The participant's reply to each question would constitute the strength of self-efficacy.  The Visual Analogue Scale to Measure Pain  In the majority of the studies reviewed, pain intensity was assessed using pain visual analogue scales. The Pain Visual Analogue Scale (Downie et al., 1978; Revill et al., 1976; Dixon & Bird, 1981, Scott & Huskisson, 1979; Turk et al., 1983) has been used extensively in the field of arthritis to obtain the individual's perception of his/her pain level. Respondents are asked to place a mark along a 10 cm line, where one end reads "No Pain" and the other end "Pain as bad as could be." There are, however, several inherent problems with this assessment method. Adequate measurement of pain may be hindered by individuals' differing perceptions and  119  responses to pain. A number of specific patient characteristics such as educational level, nature of physical illness, presence of affective disorders (Kremer & Atkinson, 1983), age (Eland, 1985; Jeans, 1983; Patterson & Klopovich, 1987), motor coordination, visual ability, and ethnic background (P. McGuire, 1987; Zbrowski, 1969) will influence the measurement of pain in a particular patient population. In spite of these concerns, the V A S has been used to measure pain intensity in a variety of patients, including, women in labour (Revill et al., 1976), people with cancer (Bond & Pilowsky, 1966), and people with arthritis. Appropriate use of the V A S would encompass any group of patients with acute or chronic pain, provided such use was restricted to the measurement of pain intensity.  Difficulties using the V A S also relate to the difficulty of conceiving of a sensory phenomenon, such as pain, in a straight line continuum. Some respondents at some times may place their marks near the anchor words, yielding data of questionable reliability and validity. Scott and Huskisson (1979) believe that patient access to previous scores is important in obtaining accurate subsequent measures when administering the V A S repeatedly, since previous ratings can be used for comparison. In this study they demonstrated that patients reported differently their pain severity when previous scores were not available. They suggest that initial scores be made available when repeated measures are taken.  As it relates to the assessment of self-efficacy, research on data using numerical anchors (Cervone & Peake, 1986) has demonstrated that people do not correct sufficiently for changes in their starting point when they judge their self-efficacy. When people are depressed it may be that cognitions about low generalized capabilities have been rehearsed so frequently in the negative  120  mood that they become very highly available in that mood. Additionally, depressed persons are prone to rehearse generalized negative cognitions about themselves (Beck, 1991), and their judgements may well be subject to anchoring effects.  Pain, like other quality-of-life measures, remains an individual experience that is variable (dependent upon many factors), private, and yet at least somewhat consistent within each individual. It is important in measuring clinical pain that this perspective be acknowledged.  Causal Modelling of Self-Efficacy and Health States  Statement of Research Question  Is self-efficacy, i.e., a belief regarding one's ability to achieve a goal or successfully carry out a course, a cause of improved health status, or is it, as suggested by Hawkins (1992), an epiphenomenon that merely reflects health status changes? Does self-efficacy predict health status, does health status predict self-efficacy, or do self-efficacy and health status covary together and reflect the same underlying causes?  Method  The problem of identifying causal priority of one variable over another is central to social science research. Recently introduced covariance structure or structural equation modeling (SEM) provides a powerful tool to investigate the nature of relationships among the correlated variables/concepts. Although "correlation does not imply causation," and although S E M cannot  121  provide a definitive answer as to whether changes in one variable are the cause of changes in another variable (Baumrind, 1983), S E M can be used to test whether a hypothesized causal model is consistent or inconsistent with the collected data - a pattern of correlations among variables of interest. S E M follows a logic of confirmatory analysis: "If a model is consistent with reality, then the data should be consistent with the model. But, if the data are consistent with a model, this does not imply that the model corresponds to reality" (Bollen, 1989; p. 68). Thus, S E M provides a basis for rejecting models as inconsistent with reality if they are not consistent with the data. But, S E M cannot confirm that a model is a true model of reality just by showing that it is consistent with the data.  To investigate causal predominance of one variable over another, longitudinal data with control groups is most useful because it helps to establish temporal priority of potentially causal variables, and the controls allow us to rule out other explanations for the ordering, such as maturity or history. When applied to longitudinal data, S E M has many advantages over older rival cross-lagged panel correlation methods (cf. Campbell, 1963): (a) it allows simultaneous assessment of more than two dependent variables in a single model; (b) it allows examination of relationships among latent rather than observed variables, providing estimates of relationships among theoretical constructs free of bias that might ensue from random measurement error in observed variables; and (c) it allows explicit formulation and comparison of more than two alternative models for the data.  Historically, three kinds of correlational methods have been used to make inferences about causal influences of one variable or another and to establish causal predominance of one variable over another: cross-lagged correlational methods, path analysis, and structural equation  122  modeling. A l l three methods are based on analysis of correlations and/or covariances and all three use longitudinal design to argue for or against causal predominance of one variable over another. However, there are substantial differences in the approach and usefulness of these methods.  In cross-lagged correlational methods (CLCM), causal inferences are based on a direct comparison of two cross lagged correlations (Campbell, 1963; Cook and Campbell, 1979). If the cross-lagged correlations are equal (i.e., not different statistically), it is usually agreed that neither one influences the other, and instead, the correlations between them are caused by other variables not considered. In contrast, if one cross-lagged correlation is greater than the other, it is inferred that the variable involved in the stronger correlation and preceding the other variable in time has a causal effect on it.  However, in an extensive critique, Rogosa (1980) pointed out that  both equal and unequal cross-lagged correlations can arise from: (a) large and equal causal effects; (b) large and unequal causal effects; (c) no causal effects (i.e., spurious relationship between the two variables). Based on his critique, he concluded that the use of C L C M for identifying causal inferences "is best forgotten" (Rogosa, 1980, p. 257).  In contrast to C L C M ' s focus on the relative magnitude of the two cross-lagged correlations, path analyses (PA) and structural equation modeling S E M require explicit formulation of causal models and they are designed to test whether a hypothesized causal structure is consistent or inconsistent with the data (i.e., correlation or covariance structure). Moreover, both methods allow a researcher to decide whether one causal model fits the observed data better than another alternative model (Anderson & Gerbing, 1988), although neither method will ever be able to prove the causal effects. Even if a given model fit the data, it is not  123  guaranteed that some other alternative model will describe it better.  The feature difference between the P A and the S E M is that P A focuses on modeling the relationship between observed variables whereas S E M focuses on modeling the relationship between latent constructs or variables (unobserved variables). When using PA, the researcher makes the implicit (and untenable) assumption that a given variable measures a construct perfectly. Thus, imperfect measurement may lead to misleading results. In contrast, by using several indicators of each latent construct, the S E M model's only variance is common to all indicators, and thus, is far less susceptible to misleading results caused by modeling errors in measurement. Of particular importance to longitudinal designs, the S E M requires the researcher to explicitly consider a possibility that measurement errors are correlated across time (e.g., due to badly formulated questions). By focusing on a measurement of a latent construct, the S E M also highlights the researcher's attention to the measurement issue — on what the indicators actually measure. Despite all the advantages of the S E M over the P A and especially over the C L C M , the S E M is not an analytical tool capable of proving causality or the absence thereof; the S E M can only show whether or not a given causal model fits the data and/or whether it fits the data better than alternative models under consideration.  Within the limitations of S E M methods, the proposed study will examine the relationship between self-efficacy and health status; it will investigate whether the link between self-efficacy and health status is mediated by conceptual and empirical commonalities between the two constructs, or whether these constructs are causally linked with one another. More specifically, it will examine whether a model postulating causal effect of self-efficacy on health status, or a model specifying causal effect of health status on self-efficacy, is consistent or inconsistent with  124  the data. It will also consider an alternative model that specifies links both from self-efficacy to health status and from health status to self-efficacy, but postulates that one of the links is stronger than the other.  In this research, the concept of health status refers to the individual variables of depression, pain, and health.  125 CHAPTER FOUR  RESULTS  Procedures  Two samples were available with different sets of measures from two domains, one from BC and one from Ontario. Originally, the plan was to pool data from two large samples, each over 1000 people; however, it turned out that the two samples came from studies with slightly different programs and different administration procedures. It did not seem justified to pool them together in one large sample, as the results could be confounded by these differences. Structural equation modeling assumes that the sample is homogeneous and that processes under study function similarly across all individuals within each sample.  British Columbia Sample  Implementation of the ASMP in BC was funded by the Seniors Independence Program of Health Canada as a health promotion program, and, therefore, the research team was not able to use an experimental design with randomized assignment to either a treatment or a control group. The program was implemented in communities where Arthritis Branch volunteers expressed an interest in becoming leaders and where individuals expressed interest in participating in the program. The purpose of the BC evaluation was to determine whether the program brought about the beneficial impacts that previous research had shown in other locations. The evaluation  126  method chosen was a "One Group Pre- and Post-Test" design. There are several threats to both internal and external validity when using such a design; however, the concern was with replicability (i.e., did similar impacts occur?), and therefore this research design was appropriate.  At the first session of the A S M P course, participants were given an 11-page questionnaire (Appendix B) by the course leaders, and asked to complete it at home and return it to the second session. Participants were told that if they did not wish to complete the questionnaire, they should return it to the course leader, and that they were still welcome to participate in the course. Four months later, course participants who had completed the pre-course questionnaire by the second session were mailed a second questionnaire by the program coordinator, and asked to complete and return it to The Arthritis Society using an enclosed addressed stamped envelope.  In total, 375 people participated in the A S M P between January and April of 1991. Of the 375 participants, 149 (40% of participants) voluntarily completed the pre-test by the second session, and completed the four-month post-program questionnaire in time to be included in the analysis. Given the 40% response rate, it is impossible to rule out sampling bias. However, a comparison of this sample with other populations of A S M P participants from B C and across Canada showed that the gender ratio, average age, educational level, and length of time with arthritis did not differ.  127  Ontario Sample  Evaluation questionnaires were sent to the Ontario A S M P coordinator. In the treatment groups, the coordinator asked the leaders to give each class participant the evaluation questionnaire and study consent form (Appendix C) during the first session. Participants were asked to complete the questionnaire and return it to the leader at the beginning of the second session. Participants who declined were nevertheless free to participate. The leaders briefly scanned the returned questionnaires for completeness.  The coordinator obtained participation in the comparison groups by using a variety of volunteer contacts to identify individuals who met the criteria for participation in the program. Once the necessary number of questionnaires had been collected and checked for completeness, those from both groups were sent to the data collection coordinator at the Institute of Health Promotion Research (IHPR) at The University of British Columbia. The "Question and Answer" computer database software program was used to enter the results, to record the participants' precourse questionnaire completion dates, and to signal the date four months later when the postcourse questionnaires were to be mailed out.  Participants were asked to complete this questionnaire and to return it in the preaddressed stamped envelope provided. If the completed questionnaire was not received by the IHPR within two weeks, a follow-up postcard was mailed to the former participant. If there was still no response after 10 days, the former participant was telephoned. When directly called, the participant either requested another questionnaire, or answered the questionnaire over the telephone.  128  Since the samples were smaller than intended, and since it is recommended that samples include 10 times as many cases as free parameters but at minimum 5 times as many cases as free parameters (Bentler & Chou, 1987) only 3 indicators of each construct were used. These indicators were chosen so as to be unidimensional, valid, and reliable, by means that will be described below.  Subjects B C Sample  The B C sample consisted of 149 participants who took the A S M P during 1992 in British Columbia. Only 129 of these participants, however, completed both the pre- and post-program questionnaires. Seven of these participants were excluded because they were multi variate outliers on critical self-efficacy and depression measures (as indicated by Mahalanobis distance greater than 26.22, using p < .01 and corresponding  X (i2)criticai = 2  26.22). Thus, the final sample  consisted of 122 participants.  Table 18 shows demographic characteristics of the B C sample. High percentages of the sample were affected by osteoarthritis (42.1%) or by rheumatoid arthritis (19.0%). The other participants had a diagnosis of one of the many other rheumatic diseases (38.9%). Table 18 also shows that on average the participants had 12.9 years of education.  129  Table 18 Gender Distributions. Type of Arthritis, and Mean Education Attainment of British Columbia and Ontario Samples British Columbia  Ontario  122  189  Men (%)  10.3  10.1  Women (%)  89.7  89.9  O A (%)  42.1  51.6  R A (%)  19.0  Other (%)  38.9  28.0 20.4  12.9  11.8  n Sex  Arthritis  Education (years)  Ontario data  The Ontario sample consisted of 200 participants who took the A S M P between 1992 and 1994 (see McGowan et al., 1994 for a more detailed description) who completed both the preand post-questionnaires. One participant was excluded because of missing data on more then 25% of the critical self-efficacy and depression measures. As well, ten participants were excluded because their scores represented multivariate outliers on critical self-efficacy and depression measures (as indicated by Mahalanobis distance greater than 26.22, using p < .01 and corresponding  X (i2)criticai 2  =  26.22). Thus, the final sample consisted of 189 participants.  Table 18 shows demographic characteristics of the Ontario sample. The majority of the sample consisted of women (89.9%). High percentages of the sample were affected by osteoarthritis (51.6%) or by rheumatoid arthritis (28.0%). Other participants had a diagnosis of  130  one of the other rheumatic diseases (20.4%). The table also shows that on average the participants had 11.8 years of education.  Measures  Self-Efficacy  Eleven out of the twenty items from "The Arthritis Self-Efficacy Questionnaire" (Lorig, Chastain et al., 1989) were used as the measure of self-efficacy used in both studies. To develop the latent variable for self-efficacy, a review of two data sets was conducted. Both data sets included pre- and post-program arthritis self-efficacy questionnaires. The B C study used the entire Arthritis Self-efficacy Scale (Lorig, Chastain et al., 1989), which contains twenty items divided into three subscales: a self-efficacy to control pain subscale (5 items), a self-efficacy function subscale (9 items), and a self-efficacy to handle other symptoms subscale (6 items). However, the Ontario study used only the pain and other symptoms subscales. To measure selfefficacy to carry out functions, the research co-investigators made the decision to use the Stanford Health Assessment Disability Scale (Fries, Spitz, & Young, 1982). This is a valid and reliable scale and it is used extensively in the field of rheumatology to measure a person's perception to carry out activities of daily living.  As the entire Arthritis Self-Efficacy Questionnaire was used in the B C study, a preliminary Maximum Likelihood Factor Analysis followed by Oblique (Oblinin) rotation was conducted. Results showed that the Arthritis Self-Efficacy Scale did not have the three factors as specified by the authors (Lorig, Chastain et a l , 1989). Rather, the analyses identified one factor  131  containing the "pain" and "other symptoms" items, and one or two factors containing the "function" items. Factor analysis of the eleven items which were used in both the B C and Ontario study addressing "pain" and "other symptoms" showed a single factor solution (Table 19).  132  Table 19 Factor Analysis of the Eleven Items used in the B C and Ontario Evaluations. Analysis number 1 Listwise deletion of cases with missing values Extraction 1 for analysis 1, Maximum likelihood (ML) Initial Statistics: Variable  Communality  Factor  Eigenvalue  7.03267 SE1A .63513 1 .68155 2 .89610 SE2A .54534 3 .57544 SE3A .54593 SE4A .67918 4 .40020 .72069 5 SE5A .35269 .69695 6 SE6A .30386 .63867 7 SE7A .25546 .65551 8 SE8A .23001 .46166 9 SE9A .21340 .69671 10 SE10A .19424 .64923 11 SE11A M L extracted 1 factors. 5 iterations required. Test of fit of the 1 -factor model: Chi-square statistic: 536.1206, D.F.: 44, Significance: Factor Matrix: Factor 1 SE1A SE2A SE3A SE4A SE5A SE6A SE7A SE8A SE9A SE10A SE11A  .76264 .80404 .73413 .83770 .86670 .83035 .79232 .81583 .67228 .73077 .67307  Pet ofVar 63.9 8.1 5.2 5.0 3.6 3.2 2.8 2.3 2.1 1.9 1.8  Cum Pet 63.9 72.1 77.3 82.3 85.9 89.1 91.9 94.2 96.3 98.2 100.0  .0000  133  Final Statistics: Variable SE1A SE2A SE3A SE4A SE5A SE6A SE7A SE8A SE9A SE10A SE11A  Communality  Factor  SS Loadings  .58162 .64648 .53895 .70174 .75117 .68949 .62777 .66558 .45196 .53402 .45302  1  6.64180  PctofVar  Cum Pet  60.4  60.4  Three items were selected from the eleven items in the "pain" and "other symptoms" Arthritis Self-Efficacy subscales: 1.  How certain are you that you can continue most of your daily activities? (SE1)  2.  How certain are you that you can manage your arthritis symptoms so that you can do the things you enjoy doing? (SE2)  3.  How certain are you that you can decrease your pain quite a bit? (SE3)  The choice of these three items was based on the following considerations: 1.  The items were used in both the B C and Ontario studies;  2.  The three items had good face validity;  3.  The B C and Ontario sample items showed a reasonable distribution, in that the data was not excessively skewed and did not show excessive kurtosis;  4.  A preliminary factor analysis of the full Arthritis Self-Efficacy Questionnaire (both Maximum Likelihood and Principal Component methods followed by oblique rotation), for both the B C and Ontario data, showed that these 3 items  134  consistently had the highest loadings (.79 to .87) on a single factor formed by the 11 items of the "self-efficacy to control pain" and "self-efficacy to handle other symptoms" subscales.  While Dr. Lorig and her colleagues (Lorig, Chastain et al., 1989) have claimed that the 11 items represent two factors, "self-efficacy to control pain" and "self-efficacy to handle other symptoms," the claim is not consistent with the results of my preliminary factor analyses. The claim seems to be based on an inappropriate use of the orthogonal rotation technique known as V A R I M A X (see Lorig, Chastain et al., 1989; Lomi,1992). The application of the V A R I M A X rotation method assumes that to-be-rotated factors (i.e., self-efficacy to control pain and selfefficacy to handle other symptoms) are orthogonal (i.e., uncorrelated) and this assumption had not been substantiated in the original report. Moreover, this assumption has been shown to be untenable in the present case, and illogical!  Depression  The participants in the A S M P in B C were administered the entire Center for Epidemiologic Studies Depression Scale (CES-D; Radloff, 1977), while the participants in Ontario were administered several depression-related items on the Medical Outcomes Study Questionnaire (MOSQ; Ware, Nelson, Sherbourne, & Stewart, 1992). The CES-D is a selfreport measure which has demonstrated validity for rheumatoid arthritis populations (Blalock et al., 1989). The CES-D is not a unidimensional measure of depression, but rather consists of four factors: Depressive Affect, Positive Affect, Somatic/Vegetative and Retarded Activity, and  135  Interpersonal Relations (Radloff, 1977; Sheehan, Fifield, Reisine, & Tennen, 1995). Three of the items with the highest parameter estimates (.86 to .94, see Sheehan et al., 1995) on the Depressive Affect factor were chosen as effect indicators of depression (see Table 20).  Table 20 Parameter Estimates of the Items Comprising Depressive Effect Subscale. Based on the Four Factor Model. Three Time Points.  Parameter  Time 1  Time 2  Time 3  X Blues X Depressed X Lonely X Crying X Sad  .879 .936 .760 .793 .861  .902 .921 .805 .766 .878  .899 .935 .845 .769 .912  Source: Sheehan et al., 1995, p. 517.  The three Depressive Affect items chosen as effect indicators of depression were:  1.  Ifelt that I could not shake off the blues even with the help of my family (DEI),  2.  Ifeel depressed (DE2),  3.  1feel sad (DE3).  Participants indicated how often they had felt this way during the preceding week on a 4point Likert-type scale ranging from Rarely or none of the time (less than 1 day) to All of the time (5-7 days). The selection of items focused on the Depressive Affect scale of the CES-D, in order to ensure the unidimensionality of Depression construct, as recommended in the literature (Anderson & Gerbing, 1982, 1988; Bollen & Lennox, 1991; Bollen, 1989). Another reason for selecting these and not other items from CES-D is that they are all effect indicators of  136  depression, in the sense that they serve as indices of depression as opposed to indices of depression causes (other CES-D items such as Ifelt lonely may be considered more of a cause for depression rather than an indicator of depression). It has been pointed out elsewhere that cause indicators are inappropriate for measuring unidimensional constructs such as depression (for a full discussion of various items on CES-D in this context, see Bollen & Lennox, 1991).  The A S M P participants in Ontario were not administered the CES-D, and thus alternative indicators of depression were chosen from among the MOSQ's depression-related items. The items chosen as effect indicators were: 1. Have you felt downhearted and blue (DEI), 2. Did you feel depressed (DEI), 3. Have you been in low or very low spirits (DE3).  Participants indicated to what extent they felt this way during the preceding month, on a 6-point Likert-type scale ranging from None of the time to All of the time. These items were chosen because they are, at least on the face of it, closely related to those selected from CES-D in the B C sample, and because they are also effect indicators as opposed to cause indicators of depression. Other Measures  The relationship between self-efficacy and health status was further examined in the context of self-reported pain and general health. Each of these latter two constructs was measured by only one index. The Pain Visual Analogue Scale (Downie et al., 1978; Revill et al., 1976; Dixon & Bird, 1981, Scott & Huskisson, 1979; Turk et al., 1983) has been used  137  extensively in the field of arthritis to ascertain the individual's perception of his/her pain level. Respondents are asked to place a mark along a 10 cm line where one end reads "No Pain" and the other end reads "Pain as bad as could be." This pain scale was used in both the B C and Ontario studies.  The "general health" question was taken from the Medical Outcomes Study (Ware et al., 1992). Respondents are asked: "In general, would you say your health is? " : Excellent (scored as 0), Very Good (scored as 1), Good (scored as 2), Fair (scored as 3), or Poor (scored as 4). The scale has a test retest reliability of .92 (Stewart, Hays, & Ware, 1988), and has a long history of use in general health population surveys such as the 1990 Canadian Health Promotion Survey (Health and Welfare Canada, 1993) and the Violence Against Women Survey (Statistics Canada, 1994). This question was used in the Ontario study, but not the B C study.  These two measures were used in the regression models with observed variables as opposed to latent variables of relationship between self-efficacy and pain and between selfefficacy and general health, respectively. As shown in Table 21, all the scales in this study had high reliability.  138  Table 21 Scale Reliability of Self-Efficacy and Depression Scales Used in the B C and Ontario Studies. BC DATA Scale  PRE-TEST Alpha  POST-TEST Alpha  Self-Efficacy Total Scale Self-Efficacy Pain Self-Efficacy Symptoms Self-Efficacy Function Self-Efficacy Pain and Symptoms CES-D ONTARIO DATA  .95 .90 .92 .95 .95 .86  .94 .87 .93 .91 .94 .86  Self-Efficacy Pain Self-Efficacy Symptoms Self-Efficacy Pain and Symptoms M O S (Depression  .89 .93 .95 .90  .88 .91 .94 .91  Relationship Between Self-Efficacy and Depression  Modeling Strategy  The development of self-efficacy—depression models and their evaluation proceeded in two steps, guided by recommendations on the application of structural equation modeling (SEM) in the literature (Anderson & Gerbing, 1988; Breckler, 1990; Bollen, 1989). First, a confirmatory factor analysis was conducted to establish the measurement model of self-efficacy and depression. The measurement model specifies relationships between the measured (observed) variables and the latent (unobserved) variables. Second, a model of the self-efficacy—depression relationship (Figure 5) was developed by testing the nested sequence of hypothesized structural models reflecting different causal relationships between self-efficacy and depression.  139 Figure 5 The saturated model of relationship between self-efficacy and depression.  SELF-EFFICACY A N D H E A L T H STATUS TIME 1 (initial assessment)  TIME 2 (four months later)  III  111  TT1  FTT  The structural model was developed by testing a series of nested models (Anderson & Gerbing, 1988), each corresponding to one of the hypothesized configurations of causal relations between self-efficacy and health status. This approach allows testing the hypotheses of interest by means of the difference between two x -values for the two models. The sequence of models 2  tested is described in Table 22, where SE refers to self-efficacy, D E to depression.  140 Table 22 Sequence of Nested Models of the Self-efficacy and Depression Relationship Model  M o d e l description  F r e e s t r u c t u r a l parameters  SMI  Autocorrelation model  SM2  SE1 - DE2  SM3  DEI - SE2  SM4  (SE1 - D E 2 ) = (DE1 -  SM5  Saturated model  4>21»+21» Yn» Y22 <j>2b l|>21> Yll> Y225 Y12 <i>2b ^21, Y l l , Y22, Y21 4>2b ^21> Y l l . Y225 Yl2J Y21 4>21, ^21, Y l l . Y22> Y 12> Y21  SE2)  Constraints  Yl2 Y21  The autocorrelated model (SMI) estimates association between exogenous latent variables (e.g., self-efficacy at Time 1 and depression at Time 1) and the effects of each latent variable at Time 1 on itself at Time 2 (DE1-DE2 and SE1-»SE2). For this model, the paths SE1-DE2 and DE1-SE2 are fixed at zero. The self-efficacy-to-depression, SE1-DE2 model (SM2) postulates a causal effect of self-efficacy on depression, and thus the path SE1-DE2 is freely estimated in this model. A significant increment in overall model fit of SM2 over S M I , as indicated by the difference in x -values for the two models, would suggest that self-efficacy at 2  Time 1 predicts depression at Time 2. In contrast, the depression-to-self-efficacy, DE1-*SE2 model (SM3) postulates a causal effect of depression on self-efficacy, and thus the path DE1-»SE2 path is freely estimated in the model. A significant increment in overall model fit of SM3 over S M I , as indicated by the difference in x -values for the two models, would suggest 2  that depression at Time 1 predicts self-efficacy at Time 2. The self-efficacy-to-depression = depression-to-self-efficacy, (SE1-DE2) = (DE1-SE2) model (SM4) postulates that neither selfefficacy nor depression is causally dominant; acceptance of this model is consistent with the possibility that the relationship between self-efficacy and depression is spurious, that is, mediated by some other variable not included in the model. The unrestricted saturated model (SM5), with all cross-lagged parameters freely estimated, postulates cross-lagged paths from one  141 construct to the other, with one path stronger than the other. The acceptance of the saturated model (indicated by significantly better fit of this model than that of the (SE1-HS2) = (HS1-SE2) model) is consistent with the view that self-efficacy and depression influence each other but that one is more predictive than or causally predominant over the other.  Figure 6 shows a decision tree (Anderson & Gerbing, 1988) that was used to select the most parsimonious model consistent with the data. To find such a model, one follows the branches of the decision tree, guided by the result of significance tests of differences between x 2  values of the two models specified in tree branch nodes. To illustrate: one starts by testing whether the difference between the x -value for SM5 and the x -value for SM4 is significant. If 2  2  the difference in x -values reaches critical x (i> = 3.84, one follows the lower branch, while if the 2  2  difference is not significant one follows the upper branch. By stepping through the tree according to the results of the different x tests, one eventually arrives at the most parsimonious 2  model consistent with the data. The hierarchical structure of hypotheses helps reduce the number of significance tests required to arrive at the best-fit model, and thereby conserves the meaning of the probabilities in the significance tests.  142 Figure 6  A decision tree for selecting the most parsimonious model consistent with the B C and Ontario data.  SM5-SM,  • SM,  The method of estimation used in the present study was maximum likelihood applied to the covariance matrices computed on the raw data. The assumptions required for application of the maximum likelihood estimator were examined by screening data for univariate and multivariate outliers, and by checking univariate and multivariate distributions of observed variables for excessive kurtosis. The B C sample data failed to meet the criteria and thus an attempt was made to re-analyze the data using distribution-free methods of estimation (ADF). A l l analysis were conducted using STATISTICA/W (Statistica/W, 1994, 1995; Steiger, 1994,  143 1995), the program for structional equation modeling.  B C Results  Table 23 shows means and standard deviations for the selected item measures of selfefficacy and depression, as well as the measure of pain that was used for the present modeling efforts. In order to allow comparison with the samples described in the previous literature, Table 23 also includes means and standard deviations for the full scale measures of self-efficacy and depression. Table 24 shows the correlations among the indicators of depression and the indicators of self-efficacy. Table 23 M e a n Scores of British Columbia Sample on Self-efficacy. Depression and V A S Pain Time 1 Mean  Time 2 SD  Mean  SD  I t e m level measures  Self-Efficacy Items (SE) Manage pain Manage symptoms  67.01 60.70  24.13 27.18  73.29 68.57  20.44 22.99  55.16  24.77  61.16  22.68  Blues  .39  .67  .43  Depressed  .69  .77  .61  .70 .80  Sad  .60  .77  .51  .70  Pain subscale  58.15  21.30  Function subscale  68.08  23.13  63.29 70.06  20.36 21.69  Other symptoms subscale  62.93  21.29  68.77  18.96  Depression (DE)  14.49  8.21  12.77  7.75  V A S Pain  5.36  2.45  4.70  2.56  Decrease pain Depression items (DE)  Scale level measures  Self-Efficacy (SE)  144 Inspection of the means and standard deviations in Table 23 suggests a possible problem with this sample: the means for depression items are all within one standard deviation of the floor of the scale, suggesting skewed score distributions. In turn, the floor effects may adversely influence the magnitude of correlations between indicators of depression and indicators of selfefficacy. In fact, univariate skewness for indicators of depression reached 1.6 and univariate kurtosis for the same items reached 1.7. In general, for structural equation modeling purposes, the absolute values of univariate skewness and kurtosis should be lower than 1, since greater absolute values indicate a significant departure from a normal distribution (Muthen & Kaplan, 1985). In contrast, measures of self-efficacy showed approximately normal distribution, with the largest value for univariate skewness and kurtosis not exceeding III. Normalized Mardia coefficient of multi variate kurtosis was 6.38, indicating serious departure from multivariate normality. Thus, B C sample data violates the requirement of no excess kurtosis for structural equation modeling using the Maximum Likelihood Estimator (Bollen, 1989). This violation does not bias path estimates, but it does result in incorrect x tests and other statistics for the 2  models computed over such data.  One possible way of dealing with this problem is to use alterative estimation methods specifically, Arbitrary Distribution Free (ADF) methods. However, the application of A D F methods requires large sample sizes, much larger than the application of M L methods demands. To model the structure of a 12 x 12 correlation matrix, as in the present study, it is recommended that the sample size be larger than 236 people. Clearly, the present sample size of 122 people is inadequate. Given these problems arising from inadequate distributions of depression indicators, I have used the M L estimator, because it at least produces consistent unbiased estimates of model  145 parameters (paths), although any associated statistics including % must be viewed with caution 2  because of the violation of the multivariate kurtosis assumption.  146  o o o  o O O  O © o  O O o  o o ©  o\  o  00 VO f - vo  - H T t VO r - © «n Tt «o i n  — 00 00 v o 00 v o O v o vo T in m  o oo i n T t oo oo T t © o vo i n i n r<->  t~-  i >—1 "~> 0\ C N • — m f - t - - r N r N <3\ vo rN rN co m cn <N  o  1  o ©  2  J o U  o r N ^ r o o o o o i n o r ^ r ^ m o O N T t m © i n r s r N C N C N C N C N  C/3  o v o o v r n T t T t ^ n o o o v O O O O O O O O O O V O O O T t © i n v o c N C N t N c n c N t N GO  a  O m m m ° o o o m v o ^ H v o © v o v o c n C N c n o o c ^ o s © © T t < n v o c N C N " c n c n r N  C  .2 °co CO CD >Oi  o r N C N r ^ - i n C T N O r ^ r ^ v o ' - H o o o T t ^ o r n v o i n — i n r N © i n c n v O T t c n c N C N c n c n r N  T3  O  o  ©  vo  <—i  in  oo  < — i O O ( N ' — i  oo  ©  o o o c n o o r N o o m r N v o o o i o m © i n i n c n T t c n c n r N r S c n c n c N  CN  „ 2 m  CO  c  •O to  -  8  T3  <u  ^ S  rt  o.  ^  i/i  ~ Q.  CN £  «  g.  &.  u  & & 8 & & £  O  Is o E- U  ©'i ^ r N r O T t m v o i > o o c ? \  —  CN —  —  147 The confirmatory factor analyses were conducted, one for Time 1 data and one for Time 2 data, to develop a measurement model and to ascertain that indicators of self-efficacy and indicators of depression measure separable constructs. For these CFAs, the indicators of depression were restricted to load on one factor and the indicators of self-efficacy were restricted to load on the second factor. The two factors were freed to correlate with one another. The results showed that the two-factor model fit the data well, both at Time 1 and at Time 2, 7.16, p > .50 and x  2  (8)  =  x (8) 2  =  10.19, p > .25, respectively. In contrast, the more parsimonious one-  factor model, which assumes that the self-efficacy and depression indicators measure the same construct, did not fit the data either at Time 1 or at Time 2,  x <9> 2  =  96.30, p < .001 and %  2  (9)  =  101.69, p < .001. These results show that selected indicators of depression and the indicators of self-efficacy measure two different constructs.  The measurement model was then estimated across the Time 1 and Time 2 data. For the first longitudinal measurement model (MM1 in Table 25), the four factors (depression 1, selfefficacy 1, depression 2, self-efficacy 2) were freed to correlate with one another, and the error of measurement associated with each indicator at Time 1 was allowed to correlate with the same error at Time 2. The M M 1 fit the data well, as indicated by its  x( ) = 2  42  49.45, p > .20. Next, all  non-significantly correlated errors of measurement were trimmed; the resulting model (MM2) also showed excellent fit to the data, x ( 4 5 ) = 53.78,p>.17. The parameters of this final 2  measurement model M M 2 are shown in Table 25, together with correlations among all factors. Table 23 shows that both depression and self-efficacy indicators had high loadings on their respective factors, and that their loadings were similar at Time 1 and Time 2.  148  OO ON  ON  oo  NO ON  T f T f ON  ON  NO  ON  ON  ON  o  ON  00  ON  ON  >n  <o  V-}  p  p  O  p  p  ON ON  o ON  ro  o  p  ro  (N  oo  t-  ro 00  Tf  oo  Ov  ON  oo  t\  ON  ON  ON  ON  00  ON  ON  OO ON  O0 O0 ON  r~  ON  ro  ro  ro  ON  OO ON  00  00  Tf  v©  >/o  p  O  Ci  p  X  o  ©  Tf 1/0  oo  ©  ON  Tf  p  o  ro  NO ON  o  t\  X  ro  X ©  cs  o  ON  ro  O  o  o  X OJ  OH T f  o  (N  ol  w  Tf  o  o oo  a  o  ON  TJ CN  no  r~  T f  9  T f  fi  in  00  OO  T f  ON T f  ro  no  T f NO  T f  Tf  NO T f  oo  IT)  NO  cn  NO T f  00 </o  VO  >  >/-> Tf  -4 I  ON  CD  Ci  SO  vd  ro in  *->  .3  o oo  r-;  .SP '53  ti  co  <u o  >  • 53 00  * g -a  1 ^ a  a)  e  6  o U  -a o  2  CU  o  ^ o  g S. 3 ^ S2 P  2 2  E  <0 C  2 •£  o u  o ON  Q  11 •a o  i  *1  - H  N  T  Q  5S.  n  >  1/3  W  (fl 1^  OX)  ON  '53  OD  '33  t/5 u  00  m  a S ° •—i  _  .„ O  tfl  .3 S) +^  C  O +  O  J  o g a O  t  «-i  tn  ^« TT  fli  « cS  CQ  S  to c. § J  g 2 2 2 ^ 2 (/)  °N  3 -  NVO  t  T f  a <u  9  T3  S ° tj  §  O  o 2 S iS  S -a °» 53 <u  e  co  CD  T3  <D  CQ  X K  ON  —c  <u a o  O PH  U PQ  a CQ s  T  149 Table 26 B C Sample Data: Measurement Model M M 2 Using M L Estimator (Indicator Loadings and Correlations Among Factors) Factor Indicators  A j  SE  t  1. Depressed  .73*  .06  13.03  2. Downhearted & Blue  .74*  .06  13.40  3. Low spirits  .76*  .05  14.39  1. Manage Arthritis Pain  .89*  .03  30.85  2. Manage Arthritis Symptoms  .87*  .03  27.62  3. Decrease Pain  .79*  .04  19.29  1. Depressed  .78*  .05  15.28  2. Downhearted & Blue  .76*  .05  14.75  3. Low spirits  .75*  .05  14.21  1. Manage Arthritis Pain  .91*  .03  31.73  2. Manage Arthritis Symptoms  .88*  .03  28.26  3. Decrease Pain  .72*  .05  14.78  Depression Time 1  Self-Efficacy Time 1  Depression Time 2  Self-Efficacy Time 2  Correlations Among Factors  1.  2.  3.  1. Self-Efficacy Time 1 2. Depression Time 1  -.43*  3. Self-Efficacy Time 2  .65*  -.53*  -Al*  .76*  4. DepressionTime2 *rj< .001  -46*  4.  150 The structural model was developed following the recommendations in the literature (Anderson & Gerbing, 1988) described in the section on modeling strategy. Figure 7 shows a saturated model that indicates the magnitude of all structural paths, and Table 22 shows the fit of each of the five models under consideration. Figure 7  The saturated model of the relationship between depression and self-efficacy, for the B C sample data.  Following the decision tree in Figure 6, the most parsimonious model that fits the data is the SM4 model. This model indicates that the difference in the magnitude of paths from selfefficacy to depression (-.31) and from depression to self-efficacy (-.11) is not significant. Thus, self-efficacy and depression may exercise about equal influence on each other across time, or they both may be caused by some other variable not considered in the model. In any case, this result does not support the primary hypothesis that higher Self-efficacy at Time 1 lowers depression at Time 2, nor the alternative possibility that higher depression at Time 1 lowers selfefficacy at Time 2. Inspection of the paths shows, however, that the effects seem to be in one direction: the path from self-efficacy to depression is stronger than the path from depression to self-efficacy.  151 It is possible that the problems with the distributions of depression indicators discussed above biased x values, and resulted in rejection of the self-efficacy - depression model. It may 2  also be that the statistical power was too low, despite a sample size comparable to that of samples used in the literature (see Bollen, 1989). These problems are avoided in the modeling of the relationship between self-efficacy and depression in Ontario sample.  Ontario Results  The modeling efforts applied to Ontario sample data proceeded along the same lines as with the B C sample data. Table 27 shows means and standard deviations for the selected item measures of self-efficacy and depression, as well as measures of pain and health status that were used for the present modeling efforts. In order to allow comparison with the samples described in the previous literature, Table 27 also includes means and standard deviations for the full scale measures of self-efficacy and depression. Table 28 shows the correlations among the indicators of depression and the indicators of self-efficacy.  The univariate skewness and univariate kurtosis for all indicators of depression and selfefficacy were within acceptable limits - their absolute values did not exceed 1 for any of the indicators. Normalized Mardia coefficient of multi variate kurtosis was 2.17, indicating that the requirement of no excess kurtosis was not likely to be seriously violated. Thus I have used the M L estimator for all modeling efforts.  152 Table 27 Mean Scores of Ontario Sample on Self-efficacy. Depression and V A S Pain Time 2  Time 1 Mean  SD  Mean  SD  Item level measures Self-Efficacy Items (SE) Manage pain  52.01  24.29  64.52  21.49  Manage symptoms  48.18  25.05  57.93  22.58  Decrease pain  43.82  25.64  56.04  22.98  Downhearted and blue  1.74  1.08  1.61  1.10  Depressed  1.88  1.15  1.67  1.11  Low spirits  1.60  1.15  1.38  1.16  Pain subscale  43.69  22.72  55.43  20.79  Other symptoms subscale  48.26  21.96  58.76  19.35  Depression items (DE)  Scale level measures Self-Efficacy (SE)  Depression (MOS)  1.62  .93  1.41  .93  VAS Pain  6.63  2.09  5.49  1.99  Health Status  2.33  .89  2.28  .88  The confirmatory factor analyses conducted separately for Time 1 data and Time 2 data confirmed that indicators of self-efficacy and indicators of depression measure separable constructs; they showed that the two-factor modelfitthe data well, both at Time 1 and at Time 2, X = 10.27, p > .20 and 2  {S)  X (8> 2  - 9.12, p > .25, respectively. In contrast, the one-factor model  which assumes that the self-efficacy and depression indicators measure the same construct did not fit the data, either at Time 1 or at Time 2,, %  2  {9)  = 238.43, p < .001 and x  2  (9)  = 234.39, p <  .001. These results show that selected indicators of depression and the indicators of self-efficacy measure two different constructs in the Ontario sample.  153  O O O  O O  o O O  o  O O  O O O  o o O  ~~  1  t  vo  r~  O  T l  ON  00 vo  O  O  o  r-  r~  ro  CN T f T f  T f  r-C T f  i—  oo t— vo  ON CO  OO CN  V) CN  NO  T f  10  ON T f  o  O IT)  CN  O  T f  •n  00  ON  T f  CO  T f T f  NO  >n  CN  —i  «r>  vo  r~ oo  CN  CN  CO  CO  T f  CO  ON  —  O r ^ C O ^ H O N C N N O O N O c o ^ O N O V O ^ t ^ o o o c N C N c o c o T f r o  o C N »/o ° ° ' — ' — ' T f i n © o o o r o o o r O T f o o i n r N O C ^ r ^ C N r o r o r O T f T f  O o o r o c N ' n O N v o r o i n r o O r ^ c N C N ^ ^ T f r o O - H O T f u o u o r o r o r o c o T f c o  o r ^ T f r o v o N O T f T f T f r N ' o O U O C N r ^ T f N O O N W - l T f O N O  0 0 o « n » n " n r o r o r o r o T f r o  O ^ t ^ C N ' n o o N O T f o o N O C ^ T f O O T f v o o o r o O N T f T f r ^ t n © o o r ^ ' n n « n r o c N c o c o c o c o ,  o CN  o  'cn cn cu  Q. cu  Q ^ C  E  a  a o  CN PQ  o  —1  cu 3  Q CQ  CN g  '3 ca O. cn CL. U <U cu cu cnC o t>o ra ra ra y b ra c e h ra c e ra  ra ra raQ S 2 S S  N C O T f U O N O r ^ O O O N O - H C N  154 The measurement model was next estimated across the Time 1 and Time 2 data. For the first longitudinal measurement model (MM1 in Table 29), the four factors (depression 1, selfefficacy 1, depression 2, self-efficacy 2) were freed to correlate with one another, and the error of measurement associated with each indicator at Time 1 was allowed to correlate with the same error at Time 2.  The M M 1 fit the data well, as indicated by its x (42) 37.97, p > .50. Next, all non2  =  significantly correlated errors of measurement were trimmed (i.e., fixed to zero); the resulting model (MM2) also showed excellent fit to the data, x ( ) = 40.80, p > .17. The parameters of this 2  45  final measurement model M M 2 are shown in Table 29, together with correlations among all factors. Table 30 shows that both depression and self-efficacy indicators had high loadings on their respective factors, and that their loadings were similar at Time 1 and Time 2.  Table 29 shows the fit of each of the five structural models under consideration. Following the decision tree in Figure 6, the most parsimonious model that fits the data is the SM4 model, the same model supported by the B C sample data. The selection of the SM4 model indicates that the difference in the magnitude of paths from self-efficacy to depression (-.10) and from depression to self-efficacy (-.24) is not significant. Thus, consistent with the results of modeling the relationship between self-efficacy and depression in the B C sample data, the Ontario sample data also do not support the primary hypothesis that self-efficacy causes reduction in depression, nor does it support the alternative possibility that depression causes reduction in self-efficacy.  155  El  a  u  o OS  o  TJ-  o o  TJ. Tt in r~ TOt C > O Os OS os os o o o  oo r-~ as os  os — so so so r- r~ r-^ r— os os OS Os os  o o o  o o o  T t  Tt Tt  CN Tt o  Tt  o o  0  o  © o o  o o o  o  3  I n  O u-  A'  Di  D,  oi  C5  • i-H  so >n rs in >n T t  O O O  1 t  o  t  i  TJ. C i  t $ t  SO cn  o  o  o  CN  o  o in A'  rN  CN O  T+  0  3  A  Tt  Tt  oo  fOS  o  OS r- in cn in cn O CN cn in T t in  Tt  00  o Tt  Tt  Tt  X T3  A  Ito  o  o s  cs  +-» c  o U  Vi  CN o  m T3  H OI  O  fc  to t  •a o  -o JS  "3 fc o o  <u  -a  fc  6  I  §  fc  00  C/2  ti c o m _CJ cs  13 >  $5  s  o  fi C  TJ 6  TJ  o O  s  OJ  O  — CN  fc o  *>  t  Q 50  & "  Vi  t  CO  M  Vi  m  Ui  I  Tt ON  .  - C  •U  —•  oo S« - s  "3 fc  " 'S to >-. *i a  •H " T3 1 5  W W Q <«  g  a  -  C  ffl  *J OS —°o e ^—' c OS S  D  JZ £ O  • - u  OS o oo OS  60  cn  J o  O OO  -o s  C\  O  i  w  *"1  <4H  •a Q  CN Tt O  O O O O O  rN so  ro  O  «s o <a o <S O  o SO o OS Os O  o  'a "  oo fc •g 3 & S O - * ' 2 X T3 cn xi T3 • A.  s s ooo os OH S=  T3  2  w  OJ  CD  2  •S  » a  ^ §r  SJ C3 <u fi  « .2  e  1—1  ca  8 < = * E  * •s  S '•*-*  o  t  8  o « -4-* u o m o t CD t» 2 ^ fc pa CQ fc c/l  =! °  ^*  »-  156  Table 30 Ontario Sample Data: Measurement Model M M 2 (Indicator Loadings and Correlations Among Factors^ Factor  Indicators  2y  SE  t  .84* .96* .90*  .03 .01 .02  33.86 69.98 48.65  .85* .89* .76*  .03 .03 .04  29..50 34.88 20.95  .83* .94* .89*  .03 .02 .02  32.50 58.21 . 44.36  .85* .88* .79*  .03 .03 .03  30.69 35.54 24.37  Depression Time 1  1. Depressed 2. Downhearted & Blue 3. Low spirits Self-Efficacy Time 1  1. Manage Arthritis Pain 2. Manage Arthritis Symptoms 3. Decrease Pain Depression Time 2  1. Depressed 2. Downhearted & Blue 3. Low spirits Self-Efficacy Time 2  1. Manage Arthritis Pain 2. Manage Arthritis Symptoms 3. Decrease Pain Correlations Among Factors 1. Self-Efficacy Time 1 2. Depression Time 1 3. Self-Efficacy Time 2 4. Depression Time 2 *rj<.001  1.  2.  3.  -.46* .60* -.38*  -.47* .65*  -.51*  4.  157 Figure 8  The saturated model of the relationship between depression and self-efficacy, for the Ontario data.  Relationship between Self-Efficacy and Pain and Between Self-Efficacy and Health  The relationship between self-efficacy and health status was further explored in the domain of self-reported pain and self-reported health. Each of these latter two constructs was measured by one index only: a response on a pain visual analogue scale, and on a health visual analogue scale, respectively. Thus it was only possible to model their relationship with regression models using observed as opposed to latent variables. However, the logic for choosing the most parsimonious model describing their relationship is analogous to the logic behind selection of the best structural model: it follows the same decision tree to select the best regression model.  158 Figure 9  The saturated regression model of the relationship between self-efficacy and pain for B C sample data.  painl  .55* (.07)  .62* (.06) pain2  -.42*  self-efficacy 1  .62* (.06)  self-efficacy 2  Self-Efficacy and Pain  Figure 9 shows a saturated regression model of the relationship between self-efficacy and pain for the B C sample data. Table 31 shows the fit of each of the five regression models under consideration. The table indicates that the most parsimonious model that fits the data is the SE Pain = Pain - SE model. The SE - Pain path (-.14) is not significantly different from Pain - SE path (-. 17), indicating that neither self-efficacy nor pain is predictively dominant.  159 Table 31 B C Data: Summary of Regression Model-Fitting Approach RMSEA* Commentary  £  df  P.  Point  90%C.I.  NFj*  NNFj>  SMI  Autocorrelated model  9.71  2  <.01  .176  .075^.295  .946  SM2  SE - Pain  5.47  1  .02  .190  .060<->.360  .970  .867 .846  SM3  Pain - SE  2.95  1  .09  .126  .000-<->-.305  .984  .933  SM4  (SE - Pain) = (Pain - SE)  .07  1  >.50  .000  .000++. 157  1.000  1.032  Model Pain  Saturated model .0 0 SM5 •Point estimate and 90% confidence interval for Steiger-Lind R M S E A Index (Steiger & Lind, 1980; Steiger, 1994) 4NFI •NNFI stands for "linked to" or "affects"  Figure 10  The saturated regression model of the relationship between self-efficacy and pain for Ontario data.  Table 32 shows the fit of each of the five regression models under consideration. As with the B C sample data, the table shows that the most parsimonious model that fits the data is again the SE - Pain = Pain - SE model. In combination, these findings fail to support the hypothesis that self-efficacy causes improvement in health status for the pain domain.  160  Table 32 Ontario Data: Summary of Regression Model-Fitting Approach  Model  Commentary  Pain  RMSEA*  t  df  E  Point  90%C.I.  NFI*  NNFI*  SMI  Autocorrelated model  11.29  2  <01  .155  .075++.250  .947  .864  SM2  SE - Pain  .83  1  .36  .0  .000++.186  1.000  1.000  SM3  Pain - SE  9.26  1  <.01  .207  .758  .16  .099++.339 .000+ +.222  .956  .072  .991  .971  SM4  (SE - Pain) = (Pain -> SE)  1.97  1  SM5  Saturated model  .0  0  Health SMI  Autocorrelated model  18.56  2  <.01  .205  .125+ +.297  .916  .769  SM2  SE - Health  17.03  1  <.01  .285  .174+ +.414  .923  .553  SM3  Health - SE  1.01  1  .32  .005  .000+ +.193  .995  1.000  (SE - Health) = (Health - SE)  5.36  1  .02  .151  .047+ +.288  .976  .878  SM4  0 .0 Saturated model SM5 •Point estimate and 90% confidence interval for Steiger-Lind RMSEA Index (Steiger & Lind, 1980; Steiger, 1994) *NFI •NNFI stands for "linked to" or "affects"  Self-efficacv and health  Figure 11 shows the saturated regression model of the self-efficacy and health relationship for the Ontario data. Table 32 shows thefitof each of the five regression models under consideration; it indicates that the most parsimonious model thatfitsthe data is Health Self-Efficacy model. This model is consistent with the view that improved health leads to a higher level of self-efficacy. The notion that self-efficacy causes improved health status was not supported.  161 Figure 11  The saturated model of the relationship between self-efficacy and health for Ontario sample data.  \  162  CHAPTER FIVE  DISCUSSION  The Arthritis Self-Management Program was originally based on a conventional educational concept, namely, that the cause-and-effect chain is largely a linear process: education leads to knowledge, attitudes, and skills which lead to the adoption of particular practices of behaviour, which, in turn, lead to beneficial changes in health. In most program situations, however, the validity of this sequence has not been tested.  In the early research on the A S M P (Lorig, Chastain et al., 1989), where the sequence has in fact been tested, this straight, sequential chain has not been clearly or strongly verified: changes in behaviours were not associated with improvements in health status. The research showed that the associations between the changes in levels of self-efficacy and changes in elements of health status (i.e., pain, depression, and disability) were stronger than the associations of the changes between behavior and knowledge and between behavior and health status. This same relationship was substantiated by the data from studies conducted in British Columbia and across Canada (McGowan, 1990, 1994). These findings suggested that there was a link between a psychological variable (self-efficacy) and health status variables, a link not necessarily mediated through behaviour. Further examination of the Canadian data showed that the strengths of the correlations between self-efficacy and health status at baseline, and at 4 months following the A S M P , were approximately equal, and therefore a directional relationship was not apparent; it was not possible to conclude that a change in self-efficacy following exposure to the A S M P led to a change in behaviour or health status.  163  Using structural equational modeling (SEM) with latent variables of self-efficacy and depression, data from both British Columbia and Ontario did not support the primary hypothesis that self-efficacy causes reduction in depression, nor did it support the alternate possibility that depression causes reduction in self-efficacy. The results showed that self-efficacy and depression may exercise about equal influence on each other across time, or they both may be caused by some other variable not considered in the analysis. In the review of the literature (Chapter 2), research studies which examine the relationship between self-efficacy and depression were discussed and they help explain the inability to find a causal relationship between self-efficacy and depression. There is evidence that perceptions of self-efficacy are affected by depressed mood, and there appears to be a good understanding of how these mechanisms operate. As well, there is a body of literature supporting the idea that depressive feelings have a reciprocal relationship with both self-efficacy and performance. Lastly, there is evidence suggesting that self-efficacy and depression covary, reflecting common underlying causes - probably causal attributions and pain severity or chronicity.  The results of the structural equation modeling with the observed variables for pain also failed to support the hypothesis that self-efficacy causes improvements in the pain domain. A n examination of the data from both the British Columbia (1992) and Ontario (1994) evaluations of the A S M P showed there were moderate correlations between changes in self-efficacy and changes in pain level. However, the results of the structural equation modeling failed to support a hypothesis specifying that either self-efficacy or pain was predictively dominant. Possible explanations are that self-efficacy expectancies did in fact predict behavior, and subsequently the behavior performance was responsible for reducing pain level; or that coping and relaxation are  164  mediators in the relationship between self-efficacy and pain intensity. Lastly, the results showed that improved health status leads to higher levels of self-efficacy, and the notion that self-efficacy causes improved health status was not supported.  This chapter will describe the implications of my research findings for health promotion planning and research. I will reaffirm the importance and role of healthful behaviors in health promotion planning, and the necessity of comprehensive programs to support and develop environments and behaviour conducive to health.  Implications for Health Promotion Planning  This doctoral research represents only one component of a larger program of research undertaken during the past fifteen years in the development and evaluation of the Arthritis SelfManagement Program. The development model presented by Nutbeam, Smith, and Catford (1989) provides a good overview of the types of research activities needed at the various stages of program development. The model posits that at the first stage of program development, outcome evaluation is the main concern. If a program or its component interventions achieve the desired outcomes under optimal circumstances, the second set of evaluations should attempt to determine whether the program can be repeated and bring about the desired outcomes in more natural circumstances. If the program can be repeated, a third stage of program development involves evaluations directed towards supporting program management (i.e., assessing the penetration of the program, testing the dissemination process of widescale implementation, monitoring quality of delivery, and assessing value for money).  165  To assess the outcome of a program, two questions need to be addressed: 1) can change be observed in the outcome variables, and 2) is the observed change attributable to the intervention? If the program can be shown to achieve its intended goals, the next stage of program development involves a series of process evaluations which examine how the program is implemented, what intervention activities are provided under what conditions, to what audience they are provided, and with what level of effort. It is at this stage of program development that the causal linkages between the intervention and the outcome can also be studied. The Nutbeam et al. framework provides a context that enables me to position and describe my doctoral research in relation to the development and evaluation of the Arthritis SelfManagement Program as it has taken place over the past fifteen years.  Over the last fifteen years, the Arthritis Self-Management Program has undergone a series of research studies which have examined impact, process, and dissemination. In the initial stages of program development, needs assessment studies were conducted to determine the impact arthritis had on peoples' lives, and to suggest areas on which patient education programs might focus (Lorig, 1982). With the information elicited through the needs assessments, a selfmanagement program was developed and subjected to a series of pilot studies which tested its viability and feasibility. When the prototype (Suchman, 1970) program was ready, it underwent a multi-site randomized controlled trail (Lorig et al., 1985). The results from these studies demonstrated that participants in the Arthritis Self-Management Program experienced less pain, depression and disability at the completion of the program, and that, without reinforcement, these improvements had lasted four months later (Lorig et al., 1985; McGowan, 1990; McGowan et al. 1994; McGowan & Green, 1995), twenty months later (Lorig & Holman, 1989), and four years later (Lorig et al. 1993).  166  When it had been established that the A S M P could bring about the desired changes in the outcome measures, and that the changes were attributable to the intervention, a number of studies of the social and psychological mechanisms of change were conducted to gain an understanding of how and under what circumstances the intervention was able to bring about the improved health outcomes. A series of studies showed that lay leaders could lead the program and achieve the same magnitude of impact as professionals (Lorig et al., 1986); that the program was acceptable to health professionals and to people with arthritis (Lorig, 1986); that the program could be disseminated in different populations (Lorig & Holman, 1993;  McGowan &  Green, 1995); and that changes in outcome measures were apparently not dependent on changes in the recommended behaviours (Lorig, Chastain et al., 1989).  The apparent failure to find an association between changes in health-related behaviours (i.e., relaxation and exercise) and changes in the outcome measures of pain, depression, and disability, was not expected, because the A S M P had been planned with the conventional assumption that changes in behaviours would lead to improvements in health status. To investigate this lack of association, Lenker et al. (1984) interviewed program participants and found that persons with positive outcomes indicated they believed they had more control over arthritis and had positive emotional status, while persons with negative outcomes indicated a lack of control and generally had a negative emotional status. This study suggested that subjects' sense of ability to effect change (a concept similar to self-efficacy), whether that sense pre-dated or developed during the program, had interacted with the program to create the health outcomes (Lorig, Seleznick et al., 1989). A n "Arthritis Self-Efficacy Scale" was developed and used in the evaluation of subsequent Arthritis Self-Management Programs. The results of these studies  167  illustrated the relationship between self-efficacy and health status indicators, following cognitive-behavioural treatment interventions designed to manipulate self-efficacy experimentally.  The next step undertaken in the series of investigations was to examine the association (i.e., correlations) between changes in self-efficacy and changes in behaviours and health status indicators, following the intervention (Lorig & Holman, 1993). Significant correlations were found between changes in self-efficacy and changes in the health status indicators of pain, depression, and disability; however, significant correlations were not found between changes in the behaviours of exercise and relaxation and changes in these same health status indicators. These findings, which were based on panel correlational designs, suggest a relationship between self-efficacy and the health status indicators that does not involve behaviour. This line of thought does not coincide with the traditional and widely accepted belief that education leads to the adoption of particular practices of behaviour, which, in turn, lead to beneficial changes in health.  As the principal researcher in developing the A S M P , Dr. Lorig has identified several questions regarding the relationship between the self-efficacy and the outcomes (Lorig, Chastain et al., 1989). First, she has asked whether the Arthritis Self-Efficacy Scales measure perceived self-efficacy for behaviour or expectation for outcome, or some combination of the two. Efficacy expectations and outcome expectations are distinct but related concepts (Strecher et al, 1986). A second question concerns the types of behaviors that have been studied: "the principal behaviors that were taught in the A S M P probably play a role, but the role appears to be small  168  even as an aggregate" (Lorig, Chastain et al., 1989). She has suggested that other behaviours need to be included in subsequent studies, including seeking and using social support, reducing fear that an activity will cause physical harm, communication strategies, developing more realistic expectations concerning the prognosis, and developing a greater capacity to tolerate discomfort. Her third question is, "could the strength of the correlations between perceived selfefficacy and health outcomes be the result of a similarity between efficacy and outcome expectations? That is, are we measuring very similar things and, hence, finding them to be correlated?" (Lorig, Chastain et al., 1989, p. 43). Towards resolving this question, she has suggested research identifying specific behaviours about which the individuals perceive themselves to be efficacious. Lastly, she has asked whether perceived self-efficacy was a mediator of the outcomes, and, if so, what the magnitude of the effect was. She has suggested that clarification of the relationship between self-efficacy and the outcomes depends on the identification of the behaviours which are both affected by self-efficacy and distinct from the outcome variables.  In health promotion planning, behaviour is considered to be a critical variable in the relationship between program interventions and quality-of-life outcomes; it is considered "an inescapable influence in most of medicine and all of health promotion" (Green & Kreuter, 1991, p. 127).  The cumulative findings of the research undertaken in health education programs, as  well as in the process of developing the A S M P , have emphasized instead the important role that self-efficacy plays in bringing about positive changes, and this has assisted program planners in developing new programs. A review of the literature illustrates how self-efficacy-enhancing strategies (modeling, mastery learning, reinterpretation, and persuasion) are being incorporated  169  into programs which address several health concerns. A cautioning note, however, is needed: the popularity of self-efficacy theory in program development should not distract us from the role that behaviours play in bringing about health benefits. While one commends the comprehensive and thoughtful planning and research that have been invested over the past fifteen years in developing the A S M P , it may be worthwhile to note a few limitations in the research design of these studies that may limit the prevailing interpretation of their findings. Some specific limitations or weaknesses in the design and methodology of these studies that may mask, or confound our understanding of, the phenomena under study. Specifically, I will comment on the use of correlational designs, on the specific behaviours considered in the research, and on how they have been measured.  Correlation Designs  Studies that have investigated the relationship between self-efficacy and the outcome measures have generally used correlational designs. Unfortunately, these designs allow only limited conclusions about the role of self-efficacy in bringing about changes in these variables. Positive or negative correlations between improvements in self-efficacy and lower depression, lower pain levels, or improved health status cannot be cited as evidence of a causal link between these two variables. In the tradition of behaviourist psychology, several researchers (Hawkins, 1992; Eysenck, 1978; Borkovec, 1978; Eastman & Marzilier, 1984) have raised the possibility that cognitive elements such as self-efficacy expectancies are not causal but merely epiphenomena of behaviour change. Correlational data allows no resolution of this critique. Positive correlations between cognitions and behavioural or health outcomes may result because  170  contingent feedback provides subjects with convincing evidence of successful performance, including changes in expectancies that lead to improvement.  One of the most important and challenging issues of health promotion research is whether a cause-and-effect relationship can be inferred from an association found in an observational or intervention study. That is, in a study of the relationship between two (or more) variables, can it be established that changes in the explanatory variable, whether occurring naturally or as a result of interventions, actually cause changes in the response variable?  Correlational designs have a number of weaknesses in their ability to answer this research question. These weaknesses will be examined, first by discussing the limitations of standard statistical techniques used in the analysis of these designs, and then by reviewing the rules of evidence for establishing causation.  Most measures of association (e.g., correlation coefficients, measures of agreement, concordance, etc.) treat the variables symmetrically; that is, the value of the measure remains the same if the roles of the two variables are exchanged. Hence, these measures tell us nothing about causality. It may also be the case that a seemingly strong association is just spurious: both variables may simply be responding to changes in some unobserved variable(s), or the effect on Y of S may be hopelessly confounded with the effects on Y of other variables.  171  Regression analysis is also used to analyze correlational designs (Flor & Turk, 1988; Buescher et al., 1991; Rejeski et al., 1996; Buckelew et al., 1995). Regression differs from correlational analysis in that one variable is designated as the outcome or response variable and the other(s) as the explanatory or predictor variable(s). Strictly speaking, regression assumes that each explanatory variable is measured without error, and that only the outcome variable exhibits variability. However, many useful applications of regression relax this assumption and allow the X variables to vary without control also. Nonetheless, regression still requires that distinct roles of outcome and predictor be identified at the outset. In fact, completely different models results from exchanging the roles; a regression of Y on X is not the same as a regression of X on Y . Regression analysis cannot be used to judge whether one has made the right choice in designating which variable is dependent and which independent. The regression of Y on X is not the same as the regression of X on Y , and neither analysis can tell which is the more logical direction or which variable is the more influential cause of the other.  Regardless of the study design from which the evidence comes, a number of basic conditions must be met before one can seriously consider a causal relationship between two variables. 1.  There must be an association between variables that cannot be explained as chance variation (i.e., it must be statistically significant), or as measurement error (e.g., selection or recall bias).  2.  There must not be evidence that the "effect" precedes the "cause"; that is, the time-order of cause and effect must be in the right sequence. The incidence of the effect should rise to a peak some time after exposure to the cause, and then decrease.  172  3.  The association must not be explicable as an effect of confounding; that is, the association must remain when the effects of other plausible variables are taken into account.  4.  There must be a dose-response gradient; that is, a longer or more intense exposure to the cause must be accompanied by a stronger effect.  5.  The results must be consistent across different populations, circumstances and study populations.  6.  There must be coherence with current theory and knowledge; that is, the explanation must be consistent with scientific knowledge of possible underlying mechanisms. In other words, the explanation must make sense.  From this list, the most likely limitations in causal inference from correlational analysis, in the series of studies investigating the relationship between self-efficacy and health outcomes, are the role of other confounding variables, and the time-order of the variables. The fifth criterion would also appear to be problematic if cultural and ethnic differences are factors in the development of cognitive and perceptual response tendencies with regard to self-efficacy, pain, and other variables under study here.  A public health example to illustrate the difficulties in causal inference from observational studies is provided by the long scientific process of establishing the link between smoking and lung cancer. Other "explanations" were offered, such as the genetic hypothesis smokers were predisposed to lung cancer just as they were predisposed to smoking — or the sloppy lifestyle hypothesis - smoking is confounded with other unhealthful behaviours. With  173  systematic experimentation to address and rule out each of these "alternative" explanations, the case for a causal link was strengthened.  Behaviours Included in the Research  There are questions about the suitability of how certain variables were included and measured. In particular, one could seriously question whether the "right" behaviours were measured. For many, behaviours may be manifested not only in activities undertaken but also in activities avoided. That is, a non-behaviour could also be considered a behaviour. The behaviours measured in the A S M P correlational studies were frequency and duration of relaxation and exercise. While these behaviours may be considered important to some, they may not be the important self-management behaviours practiced by participants of the A S M P . To elicit information on which types of behaviors are practiced by seniors with arthritis, a focus group meeting was held with twelve A S M P leaders. Members of this group believed they were enjoying a good quality of life, that they were coping well with arthritis pain, and that they were not depressed. Following the Nominal Group Process (Siegel, Attikisson, & Carson, 1978), the group developed a list of twenty behaviours (Appendix E) they engaged in and that they felt were contributing to their quality of life. The group then ranked the behaviors in terms of importance. According to the group, the behaviours contributing most to their quality of life were: 1.  good communication with family, friends and health-care providers;  2.  constantly looking for information that will help them manage the arthritis;  3.  being involved in interesting, enjoyable and fun activities;  4.  avoiding activities that cause pain;  174 5.  exercising;  6.  taking medications the way they were instructed to by their doctor;  7.  asking for help and assistance when needed.  Interestingly, the group indicated that all of these behaviours had been taught in the A S M P . The results of this study are consistent with Dr. Lorig's (1989) observation that behaviours other than relaxation and exercise may be related to the positive outcomes following A S M P . In fact, relaxation was not one of the twenty behaviours mentioned by the group.  Measurement of Behaviours  In the study evaluating the impact of the Arthritis Self-Management Program research (Lorig et al., 1985), behaviour was assessed using a limited scope of activities, and a derived behaviour score was computed by summing individual items and reporting a mean behaviour score. But since some behaviours may increase and others decrease, an overall mean may obscure changes in individual items — a phenomenon observed in multiple risk factor intervention studies in hypertension control (Green, Levine, Wolle, & Deeds, 1979) and other chronic diseases. Also, it is difficult to know how to combine behaviour scores when some are infrequent but intense and others frequent but not intense. For example, among pain- coping strategies, narcotic medication is an intense strategy but is not used frequently; relaxation techniques may not be as effective but can be used very frequently. How can the "potencies" of the two strategies be compared or combined?  M y doctoral research focused on the causal relationship between self-efficacy and  175  outcomes, which is classified as "Process Evaluation" in the Nutbeam et al. (1989) model. The research was an investigation relating to "attributing causality to the program intervention," specifically, determining the relationship and causal direction between self-efficacy and the outcomes.  In the research, I investigated the direction of influence between self-efficacy paired  separately with each of three indicators of health status - depression, pain, and perceived health status. Using structural equational modeling, I specified a model that allowed reciprocal causation, and then tested the statistical significance of each effect.  Over the last decade there has been an increase in research findings relating to the importance of self-efficacy and its role in bringing about positive health outcomes. Research limitations, however — study designs, behaviours included, and the way they were assessed — have impeded a more comprehensive understanding of self-efficacy's role and relationship. Several key questions regarding the relationship between self-efficacy and health outcomes have surfaced and need to be answered. Specifically, research is needed: a) to address the validity of the Arthritis Self-Efficacy Scale; b) to determine which behaviours play a role in bringing about improvements in pain and depression levels; c) to ascertain whether self-efficacy for pain and depression is the same as perceived pain and depression levels; and d) to determine whether selfefficacy is a mediator or a moderator of the outcomes. As well, an important question relates to the causal direction of self-efficacy and the health outcomes. Dr. Lorig (1989) made the observation that "the data reveal associations that are compatible with self-efficacy causing the outcomes, the outcomes causing self-efficacy, an interaction both ways, or a third factor causing both" (Lorig, Chastain et al., 1989, p. 44).  176  Although S E M modeling cannot provide a definite answer as to whether change in one variable is the cause of change in another variable (Baumrind, 1983), it can be used to test whether or not a hypothesized causal model is consistent with the collected data, through the analysis of the pattern of correlations among the variables of interest. Structural equation modeling follows the logic of confirmatory analysis: " i f a model is consistent with reality, then the data should be consistent with the model. But, if the data are consistent with a model, this does not imply that the model corresponds to reality" (Bollen, 1989, p. 68). Thus, S E M provides a basis for rejecting some models as inconsistent with reality, if they are not consistent with the data (Miller, 1983; Popper, 1934) — assuming, of course, that the data is valid. However, S E M cannot confirm that a model is a true model of reality simply by demonstrating consistency with the data. The process followed in this kind of research, then, is to try to eliminate models, not to prove them. Model fit alone cannot confirm that a model is appropriate, only that it has not been ruled out by the evidence analyzed so far. Substantive knowledge plays a vital role in ruling out some of the possibilities. This "falsification" perspective, advocated by Karl Popper (1934, 1959) and others, holds that a field advances by testing and comparing models to data and determining which models are the fittest to survive. Once we have a set of substantively reasonable models, the next step is to devise additional tests with new data that might help rule one or more out. The assumption is that the models not thereby discarded are closest to the true model.  M y research results could not substantiate a causal relationship between self-efficacy and depression. The review of the literature, however, provides three possible explanations.  177 The first is that persons who are depressed develop lower self-efficacy levels - depression affects self-evaluation, and self-evaluation is a source of self-efficacy. However, the results of the S E M did not indicate this relationship. As well, the subjects in both the British Columbia and Ontario samples did not have high depression scores. Only 19% of the B C sample had rheumatoid arthritis compared to 42% with osteoarthritis, and tests of significance between the rheumatoid arthritis and osteoarthritis populations were unable to detect differences between the two groups (McGowan, 1990).  The mean total score on the CES-D (Radloff, 1977) scale for the B C sample was only 14.5 at the pre-program assessment and 12.8 at the four-month post-program assessment. These means are within the range of 11.6 to 15.8 found in other samples of rheumatoid arthritis patients (Revenson et al., 1991). Using the entire CES-D scale, a score greater than 16 would represent clinical depression. In developing the latent variable for depression in the B C sample, only three items from the CES-D were used — items that focused on the Depressive Affect scale of the CES-D and that measure depression, not indices of depression's causes (Sheehan et al., 1995). The mean scores for these items were between .39 and .80 on a four-point scale ranging from 0 to 3, which indicates low depression. Another indication that participants in the B C sample were not severely depressed is that they traveled to the program location and participated in six twohour sessions, something severely depressed persons would have difficulty accomplishing.  The second alternate explanation suggests that self-efficacy, performances (i.e., behaviour), and moods (i.e., depression) have reciprocal influences on each other, and there is a growing body of literature supporting this relationship. Unfortunately, the variables in my model  178  did not include behaviours and therefore my results cannot add to this body of research. The third alternative explanation is that the relationship between self-efficacy and depression is one of "moderation," as there is research evidence suggesting that both selfefficacy and depression covary reflecting underlying causes — causal attributions and pain severity. This explanation is consistent with my results.  No causal relationship between self-efficacy and pain intensity was found. Several studies have demonstrated associations between these variables — at one point in time, both before an intervention and after the intervention; and at different points in time (i.e., preprogram self-efficacy and pain four months later, and vice versa). Also, studies have shown significant correlations between pre- and post-program changes in self-efficacy and changes in pain levels. Research has included studies which examine the relationship between self-efficacy and pain behaviours, pain coping behaviours, and pain intensity.  Analysis of the structural equation modeling indicated a relationship between selfefficacy and health status, and that higher health status leads to higher levels of self-efficacy. In a recent study (1993) Taal et al. investigated the relationship between self-efficacy and health status using the D U T C H - A I M S as a measure of health status. Their results too showed a significant correlation between self-efficacy and health status. Using a correlational design, the results showed that the more the R A patients viewed themselves as capable of controlling their disease, the better they judged their own health status, and this was independent of disease activity. The significant correlations of self-efficacy with ESR and Hb were explained by the correlations with self-reported health status. These findings imply that self-efficacy may be  179  related to the subjective experience of health status, but not to the more objective indices of health status such as laboratory test results. The results also suggest that self-efficacy may be a determinant of experienced health status that operates partly independent of the underlying physical condition, as would be expected according to social learning theory (Bandura, 1986). Bandura states that self-efficacy is a significant determinant of performance and that it operates partly independent of underlying skills.  In my research, health status was measured using the "general health" question taken from the Medical Outcomes Study (Ware et al., 1992). In a recent study (1996) Ratner, Johnson, and Jeffery developed a model that assessed the respective contributions of self-perceptions of emotional, physical, social, and spiritual health to the variance in overall self-rated health status measured by the Medical Outcomes Study question. The study population included 1,444 noninstitutionalized Yukon residents over the age of fifteen years. They found that only physical health was significant and accounted for 55.1% of the variance in the self-rated health status. This finding suggests that when asked about health status, people think mostly in terms of physical circumstances such as functionality and mobility.  Implications  In summary, if one reflects on the research results of the series of studies conducted on the Arthritis Self-Management Program over the last decade, there is little evidence to suggest a causal relationship between self-efficacy and the outcome measures. The limitations imposed by the use of correlational designs, and the inclusion of only two behaviours (exercise and  180  relaxation) allow weak and only speculative evidence for a causal relationship. In my research, data from both British Columbia and Ontario failed to support the primary hypothesis that selfefficacy causes reductions in depression or pain levels, or an improvement in health status; nor did the data support the alternate possibility that depression and pain cause a reduction in selfefficacy. The results did, however, support the notion that improved health status leads to higher levels of self-efficacy.  The models used in the structural equation modeling analysis did not fit the data, and a brief discussion of possible explanations is warranted. Possible explanations will focus on the relative effectiveness of the self-efficacy construct, and on the sample selection and the nature of the data.  Relative Effectiveness of Self-Efficacy in Bringing about Changes in Health Status  Before the A S M P was implemented, there were no community-based patient education programs available to persons with arthritis. With assistance from The Arthritis Society, several communities developed local arthritis groups to promote public awareness of arthritis and to serve as local fundraising bodies for the Society. Participation in these local group activities did provide a small degree of hope and support for persons coping with arthritis, but the groups' main focus was on fundraising.  When the A S M P was introduced in Canada, it was widely promoted by The Arthritis Society as an effective patient-education program. In the promotional materials, results from  181  Dr. Lorig's research were cited to provide credibility for the program, and treatment personnel (family doctors, rheumatologists, and arthritis health professionals) recommended the program to their patients. These promotional activities, identified with The Arthritis Society, an established, reputable health care institution that specializes in arthritis, seemed effective in motivating and encouraging people to become involved in the program.  Persons interested in participating in the A S M P were required to register for the program. For each group session, individuals needed to get dressed and travel to the location where the A S M P was being held. In the program they were given information about arthritis, and learned and practiced new skills (e.g., communication strategies with health care providers, family, and friends) and behaviours (e.g., exercises and relaxation techniques) to manage the i  pain, functional disability, and psychological and emotional problems that often accompany chronic health conditions. Between the sessions they were instructed to practice the new skills and behaviours at home. During the six sessions they participated in a group program with ten to fifteen other persons with arthritis, socialized, made new friends, and often formed a support group after the program had ended.  Last (1988) defines program effectiveness as "the extent to which a specific intervention, procedure, regimen, or service, when deployed in the field, does what it is intended to do for a defined population" (p. 41). Impact evaluations (Lorig et al., 1985; McGowan, 1990; McGowan et al. 1994; McGowan & Green, 1995) in both the United States and Canada have shown that the A S M P is effective in bringing about positive changes in participants' health status, and selfefficacy was identified as an important construct. It is apparent, though, that many variables may  182 have contributed to the positive changes that participants experienced. Looking at the A S M P from the perspective of a comprehensive program planning and evaluation framework, such as the P R E C E D E - P R O C E E D model (Green & Kreuter, 1990), one can recognize the variety of factors that may have contributed to the outcomes. Firstly, participants were motivated to take the A S M P , not only because it was the only community-based program available, but also because The Arthritis Society had endorsed it as an effective program. With reference to the P R E C E D E - P R O C E E D model (Appendix F), during the sessions: 1.  Participants gained new knowledge about arthritis and its treatment, their beliefs about arthritis were discussed and challenged if incorrect, and they gained confidence in their ability to deal with the day-to-day problems of living with arthritis. (Predisposing factors.)  2.  Participants learned new skills (e.g., problem-solving, communication, accessing services) and behaviours (exercises, relaxation techniques). (Enabling factors.)  3.  Participants received support from each other and learned ways to facilitate support from family and friends. (Reinforcing factors.)  4.  Participants, through discussions with other group members, became knowledgeable about the arthritis service delivery system, and about how to become involved in and improve or influence it. They became involved as volunteers in a variety of capacities (e.g., Arthritis Society Board and Committee members, community branch leaders, and service provision assistants). In doing this they began to take more control of the planning and delivery of arthritis treatment and services. (Environmental factors.)  183  Besides these contributing factors, it is probable that participant behaviours also contributed to the health status improvements that participants experienced. When the program is seen in this light, it is both plausible and probable that several factors working separately and in combination with each other were responsible for bringing about improvements in health status. In contrast to "effectiveness," "efficacy" is defined as "the extent to which a specific intervention, procedure, or service produces a beneficial result under ideal conditions. Ideally, the determination of efficacy is based on the results of a randomized controlled trial" (Last, 1988, p. 41). In my research, structural equation modeling was conducted to determine whether there was a causal relationship between self-efficacy and three indicators of health status (depression, pain, and perceived overall health status). Although there is consensus that there are several factors contributing to health status, in the analysis only one contributing factor was considered self-efficacy. Thus, this research examined the "efficacy" of the A S M P ' s self-efficacy intervention, as opposed to its "effectiveness." The results failed to support a hypothesis of a causal relationship between any of the variables, with one exception: a higher perceived health status leads to higher levels of self-efficacy at a later time. A n explanation for these findings may be that effects brought about by self-efficacy are not strong enough on their own to bring about desired changes, and that additional effects from the other contributing variables and behaviours are required.  This explanation underscores the importance of the health promotion principle of incorporating a comprehensive perspective, when planning interventions aimed at bringing about behaviour changes and improvements in health status (Green & Kreuter, 1990). Human behaviour is complex and influenced by a variety of factors and, therefore, interventions should  184  also be multi-faceted and implemented at different levels (individual, organizational, environmental). The effectiveness of the A S M P may be attributed to the comprehensiveness of the intervention. It is improbable that any single influence, in this case self-efficacy, however potent it may be, is capable of single-handedly bringing about behaviour change and improved health status.  Sample Selection and Nature of the Data  There are two aspects of the sampling method and the nature of the data that merit consideration. The first relates to the use of structural equation modeling with the samples from British Columbia and Ontario. These samples were selected because A S M P implementation had taken place in a timely and orderly fashion, and had followed the specified implementation protocols, and because coordination, training and support had been provided by the same personnel. As well, the participants represented a relatively homogeneous group.  When this research was in the planning phase it was anticipated that two large data sets would be used in the analysis. The first data set was from the A S M P national evaluation (McGowan et a l , 1994), which contained data on 800 participants who had completed both preand post-program questionnaires. The second data set was from Kate Lorig and contained data on approximately 1000 A S M P participants. In reviewing these data sets, however, several concerns came to light. For instance, in the Canadian sample there were groups who had received the program in different languages (English and French); in two provinces, program implementation had been coordinated by more than one coordinator; and different  185  implementation and training protocols were followed in some provinces as the implementation process progressed. The sample provided by Kate Lorig contained data collected over several years on approximately 1000 A S M P participants. These data, however, were from several research studies testing various versions of the program, involved different pre- and postprogram timelines, and tested different outcome measures.  Recent developments in statistical theory have made it possible to account for group differences in conducting structural equation modeling (e.g., multi-group SEM); however, an examination of both the national data set and the data set provided by Kate Lorig showed that if I separated the samples into groups that accounted for language and implementation processes, the group sizes would be very small. I decided instead to use data sets from two studies that I had coordinated, that had reasonable sizes (149 and 200 participants), and about which I did not have concerns regarding variations in the intervention or the implementation processes.  As a result, the sample sizes used in the analysis are considered small for the purposes of structural equation modeling; it might be inferred that the models did not fit with the data because of these small sample sizes — particularly with the B C data. But we must consider why an examination of the relationship between self-efficacy and depression in the B C and Ontario samples yielded opposite results. In the B C sample the relationship between self-efficacy at Time 1 and depression at Time 2, even though it was not statistically significant, was stronger than the relationship between depression at Time 1 and self-efficacy at Time 2. Again, it might be argued that with a larger sample size, the relationship would have reached statistical significance. However, an examination of the relationship between self-efficacy and depression in the Ontario  186  sample showed that the relationship between depression at Time 1 and self-efficacy at Time 2 was stronger than the relationship between self-efficacy at Time 1 and depression at Time 2. The strengths of these relationships were compared to see if they were significantly different from each other. Results indicate they were not different (z = 0.657, 2-sided p-value in excess of .48). Given that the two samples showed opposite results, the more probable explanation would be that the data did not fit the model, and that the small sample size did not unduly influence the results.  A second possible explanation for the research findings in this study relates to the relationship between the variables. A non-linear relationship between self-efficacy and depression, pain, and health status would undermine the usefulness of the analysis. But the nature of the relationship between both the variables and the indicators was addressed during the early stages of the analysis. The indices and variables were plotted against each other, and there was no major departure from linearity.  The research literature does provide alternate explanations of the relationships between and among the variables of self-efficacy, depression, pain, and health status. To begin with, the role of behaviour has been under-emphasized in the A S M P research as only two behaviours were considered in the analysis. There is evidence from focus groups that other behaviours are more important than the ones included. Secondly, behaviour (defined as performance) is believed to have a role in the reciprocal relationship between self-efficacy, performance, and depression.  187  The literature also provides evidence that self-efficacy plays the role of a moderator between pain and depression and between attributions and depression. As well, the studies examining the relationship between self-efficacy and pain suggest that self-efficacy expectancies predict behaviour, and that subsequently the behaviour performance is responsible for reducing pain level, and that coping and relaxation are mediators in the relationship between self-efficacy and pain intensity.  The overall findings of the studies conducted during the planning and evaluation of the A S M P , with the inclusion of my research results, clearly suggest two implications for health promotion planning and research. The first and most important relates to the role of behaviours. One must conclude that there is more evidence for the role of behaviours in bringing about the beneficial health outcomes than there is for the self-efficacy variable. Secondly, relationships which consider self-efficacy in a moderating role have merit and need further investigation.  After undergoing structural equation modeling, my data could not support the notion of a causal relationship from self-efficacy to the health outcomes of depression, pain, or health status. These findings do not undermine or negate the value and importance of self-efficacy in health education and health promotion programs. Self-efficacy is still considered to be important in bringing about positive health outcomes, and should continue to be used in planning. As stated by Glymour et al. (1987), "In the natural sciences nearly every exact, quantitative law ever proposed is known to be literally false. Kepler's laws are false, Ohm's law is false... and so on. These theories are still being used in physics and chemistry and in engineering, even though they are known to be false. They are used because, although false, they are approximately correct.  188  Approximation is the soul of science."  In conclusion, several qualifications should be noted with respect to statements of causality in this study. First, the direction of causality is always inferred and "never" proven. Second, causal modeling may only help the researcher choose relevant causal hypotheses by ruling out those hypotheses not supported by the data. (Rogosa, 1979). Third, inferences regarding causal effects between the two constructs are limited in the present research design by the four-month follow-up period (cf. Gollob & Reichardt, 1987). Lastly, any model is an approximation of reality. A "theory" is an abstract set of ideas that links together concepts, and a model is a formal representation of a theory. If the theory at best approximates reality, the model derived from it can do no better. Model building and modification is a process of a successive approximation (Bollen, 1989) — we must realize that the problems of demonstrating isolation, association, and direction of causation are age-old issues.  The Arthritis Self-Management Program is an innovative approach to patient education which successfully involves people in managing their arthritis. Since its inception nearly 15 years ago, it has been given in 15 countries to over 200,000 persons. The A S M P is a good example of a health education program: it is theoretically grounded, and information obtained from a series of research studies has assisted program development at various stages.  Dr. Lorig (1989) had identified several research questions regarding the associations of changes in self-efficacy and the changes in health outcomes. M y research provides preliminary evidence which addresses two of her questions: 1) whether there is a causal relationship between  189  self-efficacy and the health outcomes, whether both influence each other equally, or whether both self-efficacy and the health outcomes were influenced by other variables or factors not considered in the analysis; and 2) whether other behaviours not considered in the research were responsible for bringing about the health outcomes.  The intent of my research was to investigate whether there was a causal relationship between self-efficacy and three indicators of health status. The results of this analysis do not provide evidence to suggest that the A S M P needs to be changed or modified in any manner. The results do suggest, however, that further research is needed to: (a) determine the relative contributions of the variables towards behaviour change and improvements in health status, and (b) determine the relationships between the variables involved in the A S M P . Research is needed in three key areas. First, research needs to be focussed on determining which behaviours are associated with the successful management of arthritis. In previous research Lorig (1985) included specific behaviours in the analysis, but the behaviors used and the way they were measured and aggregated appear to be problematic. In my focus group meeting with A S M P leaders, group members listed 2 0 different behaviours that they engage in to manage arthritis successfully, and they believed that the majority of the behaviours listed had been taught in the ASMP.  Second, research is needed to determine which behaviours are important and to understand more fully the relationship between behaviours and successful management of arthritis. Research studies that use qualitative methods are best suited to gain better knowledge and a more comprehensive understanding of the types of behaviours that people engage in to  190  manage their arthritis condition. This research could take the form of focus group meetings and interviews with persons with arthritis, family members, and significant others. Involving persons with arthritis as partners in the research would increase the effectiveness of the methodology and facilitate a more comprehensive understanding of the behaviours and their meaning to persons with arthritis (Green et al., 1985). Once the important behaviours are identified, they could be examined in relation to the behaviours taught in the A S M P to determine which are taught and emphasized in the A S M P . When this information is known, the second task would be to determine how the behaviours could be measured.  Third, research is needed to evaluate the A S M P which considers all the contributing variables, the behaviours identified in the previous steps, and the development of a theoretical model of program inputs, mediating processes, and outputs. Structural equation modeling should be used to determine relationships between the input variables, the mediating processes and the outcomes. As well, the evaluation methodology would need to use a randomized controlled design with longitudinal data and involve relatively large sample sizes. This theory-driven approach to evaluation would provide the opportunity to examine both the effects of the A S M P as well as the theoretical framework that explains how the variables operate to produce the program effects.  The results of the S E M failed to support the hypothesis indicating causal relationships between self-efficacy and depression, or between self-efficacy and pain, and that these variables may exercise about equal influence on each other across time or they both may be caused by other variables that were not considered in the analysis. The results of the S E M between self-  191  efficacy and perceived health status did, however, show that improved health status leads to higher self-efficacy at a later time. So what are the practical and clinical implications of these findings? First, as no causa direction was shown, refinements and revisions to the A S M P should not sacrifice one module at the expense of the other. Second, in spite of pressure to reduce the size and scope of the A S M P , the evidence presented here suggests the contrary. The findings reinforce the multi disciplinary approach to the treatment of arthritis. Further, this suggests that other treatment modalities should be considered.  The findings suggest that self-efficacy may play a moderator role in the complex relationship involving individuals with arthritis, their behaviors, and health outcomes. As well, the findings have implications for health promotion planning and research in that they reinforce the complex interplay of psychological and behavioral variables (probably influenced by social variables) in programs which attempt to give individuals greater control over their health. The efficacy and effectiveness of the A S M P has been established in previous studies. This study in no way calls these into question. It does, however, suggest that the mechanism by which these effective outcomes are achieved warrants further investigation.  192  REFERENCES  Adams, P. F., & Benson, V . (1992). Current estimates from the National Health Interview Survey, 1991. Vital & Health Statistics - Series 10: Data from the National Health Survey 184:1-232. Ahles, T. A . , Yunus, M . B., & Masi, A . T. (1987). Is chronic pain a variant of depressive disease? The case of primary fibromyalgia. Pain. 29. 105-111.  Ajzen, I. (1988). Attitudes, personality, and behaviour. Milton Keynes: Open University Press. Ajzen, I., & Fishbein, M . (1980). Understanding attitudes and predicting social behaviour. Englewood Cliffs: Prentice-Hall. Allport, F. H . (1924). Social psychology. Boston: Houghton Mifflin.  Allport, G. W. (1985). The historical background of social psychology. In G. Lindzey & E. Aronson (Eds.). Handbook of social psychology (pp. 1-46). New York: Random House.  Anderson, J. C , & Gerbing, D. W. (1982). Some methods for respecifying measurement models to obtain unidimensional construct measurement. Journal of Marketing Research. 19. 453460.  Anderson, J. C , & Gerbing, D. W. (1988). Structural equation modelling in practice: A review and recommended two-step approach. Psychological Bulletin. 103(3). 411-423. Anderson, K . O., Keefe, F. J., Bradley, L . A . (1988). Prediction of pain behaviour and function status of rheumatoid arthritis patients using medical status and psychological variables. Pain. 33, 25-32.  Arch, E. C. (1992a). Sex differences in the effect of self-efficacy on willingness to participate in a performance situation. Psychological Reports. 70. 3-9.  193  Arch, E. C. (1992b). Affective control efficacy as a factor in willingness to participate in a public performance situation. Psychological Reports. 71. 1247-1250.  Badley, E. M . (1992). The impact of musculoskeletal disorders in the Canadian population, (editorial). Journal of Rheumatology. 19. 337-340. Badley, E. M . (1995). The economic burden of musculoskeletal diseases in Canada (is similar to that for cancer and may be higher. Journal of Rheumatology. 22, 204-206.  Badley, E. M . , Rasooly, I., & Webster, G. (1994). Relative importance of musculoskeletal disorders as a cause of chronic health problems, disability and health care utilization. Journal of Rheumatology. 21. 505-514.  Badley, E . M . , & Tennant, A . (1992). Changing profile of joint disorders with age: Findings from a postal survey of the population of Calderdale, West Yorkshire, United Kingdom. Annals of Rheumatic Disease. 51. 366-71.  Bagehot, W. (1875). Physics and politics. New York: D. Appleton. Baldwin, J. M . (1895). Mental development in the child and in the race. New York: Macmillan.  Bandura, A . (1969). Principles of behavior modification. New York: Holt, Rinehart, & Winston.  Bandura, A . (1977a). Self-efficacy: Toward a unifying theory of behaviour change. Psychological Review. 84(2). 191-215.  Bandura, A . (1977b). Social learning theory. Englewood Cliffs, N.J. Prentice-Hall. Bandura, A . (1978). The self system in reciprocal determinism. American Psychologist. April, 344-359.  Bandura, A . (1982a). The self and mechanisms of agency. In J. Suls (Ed.), Psychological Perspectives on the Self. Vol. 1. Hillside, N.J.: Erlbaum.  194  Bandura, A . (1982b). Self-efficacy mechanisms in human agency. American Psychologist. 37(2), 122-147.  Bandura, A . (1986). Social foundations of thought and action: A social cognitive theory. Englewood Cliffs: Prentice-Hall.  Bandura, A . (1988). Self-efficacy conception of anxiety. Anxiety Research. 1. 77-98. Bandura, A . (1991). Self-efficacy mechanism in physiological activation and health-promoting behaviour. In J. Madden (Ed.), Neurobiology of learning, emotion and affect (pp. 229269). New York: Raven Press.  Bandura, A . (1995). Comments on the crusade against the causal efficacy of human thought. Journal of Behaviour Therapy and Experimental Psychiatry. 26(3\ 179-189.  Bandura, A . , & Adams, N . E. (1992). Analysis of self-efficacy theory of behavioural change. Cognitive Therapy and Research. 1, 187-310.  Bandura, A . , Barr Taylor, C , Lloyd Williams S., Mefford, I. N . , & Barchas, J. D. (1985). Catecholamine secretion as a function of perceived coping self-efficacy. Journal of Consulting and Clinical Psychology. 53,(3), 406-414. Bandura, A . , & Cervone, D. (1983). Self-evaluative and self-efficacy mechanisms governing the motivational effects of goal systems. Journal of Personality and Social Psychology. 45. 1017-1028.  Bandura, A . , & Cervone, D. (1986). Differential engagement of self-reactive influences in cognitive motivation. Organizational Behavior and Human Decision Processes. 38, 92113.  Bandura, A . , & Walters, R. H . (1963). Social learning and personality development. New York: Holt, Rinehart & Winston.  Barlow, J., Macey, S., & Struthers, G. (1993). Gender, depression, and ankylosing spondylitis. Arthritis Care and Research. Vol.6(l). 45-51.  195  Baron, R. M . , & Kenny, D. A . (1986). The moderator-mediator variable distinction in social psychology research: Conceptual, strategic and statistical considerations. Journal of Personality and Social Psychology. 51. 1173-1182. Baumrind, D. (1983). Specious causal attributions in the social sciences: The reformulated stepping-stone theory of heroin use as exemplar. Journal of personality and Social Psychology. 45(6). 1289-1298.  Beaumont, G. (1994). Quality of life in primary care. Human Psychopharmacology Clinical and Experimental. Vol. 9 fSuppl. 1), S25-S29.  Beck, A . T. (1991). Cognitive therapy: A 30-year retrospective. American Psychologist. 46. 368-375.  Beck, A . T., Rush, A . J., Shaw, B. F., & Emery, G. (1979). Cognitive therapy of depression. New York: Guilford.  Becker, M . H . (1987). Introduction to theory in health education. In S. K . Simmonds, P. D . Mullen, & M . Becker (Eds.), Advances in health education and promotion (pp. 241-243). London: JAI Press Inc. Becker, M . H . , & Rosenstock, I. M . (1987). Comparing social learning theory and the health belief model. In W. B. Ward (Ed.), Advances in health education and promotion. Vol. 2. pp. 245-249). London: JAI Press Inc.  Beecher, H . K . (1958). Relationship of significance of wound to pain experienced. Journal of American Medical Association. 161. 1609-1613.  Beekman, A . , Kriesgsman, D., Deeg, D., & van Tilburg, W. (1995). The association of physical health and depressive symptoms in the older population: Age and sex differences. Social Psychiatry and Psychiatric Epidemiology. Vol. 30 (1), 32-38.  Bentler, P. M . , & Bonnett, D. P. (1980). Significance tests and goodness of fit in the analysis of covariance structures. Psychological Bulletin. 88. 588-606.  196  Bentler, P. M . , & Chou, C. (1987). Practical issues in structural modelling. Sociological Methods & Research. 16(1). 78-117.  Bernard, C. L . (1926). Instinct: A study in social psychology. New York: Henry Holt.  Bernstein, D. & Glasgow, R. (1979). The modification of smoking behaviour. InO. F. Pomerleau & J. P. Brady (Eds.) Behavioural medicine: Theory and practice (pp. 233253). Baltimore, M d : Williams & Wilkins. Blalock, H . M . (1964). Causal inferences in nonexperimental research. Chapel Hill: University of North Carolina Press.  Blalock, S. J., DeVellis, R. F., Brown, G. K., & Wallston, K. A . (1989). Validity of the Centre for Epidemiological Studies depression scale in arthritis populations. Arthritis and Rheumatism. 32, 991-997.  Blanchard, E. B. (1982). Behavioural medicine: Past, present and future. Journal of Consulting Clinical Psychology. 50, 795-796.  Blanton, S. & Blanton, M . (1927). Child guidance. New York: Century.  Blumer, D., & Heilbronn, M . (1982). Chronic pain as a variant of depressive disease. The painprone disorder. Journal of Nervous and Mental Disease. 170. 381-406. Bollen, K . A . (1989). Structural Equations with Latent Variables. New York: John Wiley & Sons.  Bollen, K . A . , & Lennox, R. (1991). Conventional wisdom on measurement: A structural equation perspective. Psychological Bulletin. 110(2). 305-314.  Bond, M . R., & Pilowsky, I. (1966). Subjective assessment of pain and its relationship to the administration of analgesics in patients with advanced cancer. Journal of Psychosomatic Research. 10(2). 203.  197  Borkovec, T., D., (1978). Self-efficacy: Cause or reflection of behaviour change? Advances in Behaviour Research and Therapy. 1, 231-236.  Bower, G. H . (1991). Mood congruencies and social judgements. In J. P. Forgas (Ed.). Emotions and social judgements (pp. 31-53). Oxford: Pergamon.  Bower, G. H . (1983). Affect and cognition. Philosophical Transactions of the Royal Society of London. Series B . 302. 387-402.  Bowers, K . (1968). Pain, anxiety and perceived control. Journal of Consulting and Clinical Psychology. 32, 596-602.  Bowler, M . H . , & Morisky, D. E. (1983). A small group strategy for improving compliance behavior and blood pressure control. Health Education Quarterly. 10(1). 56-70.  Bradley, L . A . (1993). Psychosocial factors and arthritis. In H . R. Schumacher, J. H . Klippel, and W. J. Koopman (Eds.), Primer on the Rheumatic Diseases. Tenth Edition (pp. 319-321). Atlanta: Arthritis Foundation.  Breckler, S. J. (1990). Applications of covariance structure modelling in psychology: Cause for concern? Psychological Bulletin. 107(2), 260-273.  Brook, R. H . , Ware, J. E., Davies-Averly, A . , Stewart, A . L . , Donald, C.A., Rogers, W. H . , Williams, K . N . , & Johnston, S.A. (1979). Overview of adult health status measures fielded in Rand's health insurance study. Medical Care. Vol. 17. (Suppl. 7), 1-31.  Brook, R. H . , Ware, J. E., & Rogers, W. H . , Keeler E. B., Davies, A . R., Donald, C. A . , Goldberg, G. A . , Lohr, K . N . , Masthay, P. C , & Newhouse (1983). Does free care improve adults' health?: Results from a randomized controlled trial. New England Journal of Medicine. Vol. 309. 1426-1434. Brown, G. K . (1990). A causal analysis of chronic pain and depression. Journal of Abnormal Psychology. 99(2), 127-137.  198  Buchwald, A . M . (1977). Depressive mood and estimates of reinforcement frequency. Journal of Abnormal Psychology. 86, 443-446. Buckelew, S. P., Murray, S. E., Hewett, J. E., Johnson, J., & Huyser, B. (1995). Self-efficacy, pain, and physical activity among fibromyalgia subjects. Arthritis Care and Research. 8(1), 43-50.  Buckelew, S. P., & Parker, J. C. (1989). Coping with arthritis pain. Arthritis Care and Research. 2(4), 136-144.  Buescher, K . L . , Johnston, J. A., Parker, J. C , Smarr, K . L . , Buckelew, S. P., Anderson, S. K . , & Walker, S. E. (1991). Relationship of self-efficacy to pain behaviour. Journal of Rheumatology. 18, 968-972.  Burckhardt, C. S. (1985). The impact of arthritis on quality of life. Nursing Research. 34. no. 1, 11-17.  Burckhardt, C. S., O'Reilly, C. A., Wiens, A . N . , Clark, S. R., Campbell, S. M . , & Bennett, R. M . (1994). Assessing Depression in Fibromyalgia Patients. Arthritis Care and Research. 2,(1), 35-40. Callahan, L . F., Kaplan, M . R., & Pincus, T. (1991). The Beck Depression Inventory, Centre for Epidemiological Studies Depression Scale (CES-D), and General Weil-Being Schedule Depression Subscale in rheumatoid arthritis: Criterion contamination of responses. Arthritis Care and Research. 4. 3-10.  Campbell, D. T. (1963). From description to experimentation: Interpreting trends from quasiexperiments. In C. W. Harris (Ed.), Problems in Measuring Change pp. 212-242. Madison: University of Wisconsin Press.  Cane, D. B., & Gotlib, I. H . (1985). Depression and the effects of positive and negative feedback on expectations, evaluations, and performance. Cognitive Therapy and Research. 9, 145160.  Cantril, H . (1965). The pattern of human concerns. New Brunswick, N J : Rutgers University Press.  199 Cassileth, B . R., Lusk, E. J., Hutter, R., Strouse, T. B., & Brown, L . L . (1984). Concordance of depression and anxiety in persons with cancer. Psychological Reports. V o l . 54. 588-590.  Cassileth, B. R., Lusk, E. J., Strouse, T. B., Miller, D. S., Brown, L . L . , Cross, P. A . , & Tenaglia, B.S. (1984). Psychosocial status in chronic illness: A comparative analysis of six diagnostic groups. The New England Journal of Medicine. 311.(8). 506-511.  Centre for Disease Control (1990). Arthritis prevalence and activity limitations - United States.  Cervone, D., & Peake, P. K . (1986). Anchoring, efficacy and action: The influence of judgemental heuristics on self-efficacy judgements and behavior. Journal of Personality and Social Psychology. 50,492-501.  Chambers, L . W., Reynolds, D. L . , & Badley, E. M . (1991). Physical disability among Canadians reporting musculoskeletal disease. Final report submitted to Statistics Canada, Health Priorities Analysis Unit, McMaster University, Hamilton, O N , March. Chaves, J., & Barber, T. (1974). Cognitive strategies, experimental modelling and expectation in the attention of pain. Journal of Abnormal Psychology. 83, 356-363.  Chen, H . T., & Rossi, P. H . (1983). Evaluating with sense: The theory-driven approach. Evaluation Review. 7(31 283-302.  Chen, H . T., & Rossi, P. H . (1987). The theory-driven approach to validity. Evaluation and Program Planning. 10. 95-103.  Ciminero, A . R., & Steingarten, K. A . (1978). The effects of performance standards on selfevaluation and self-reinforcement in depressed and nondepressed individuals. Cognitive Therapy and Research. 2, 179-182.  Clark, N . M . (1987). Social learning theory in current health education practice. In S.K. Simmonds, P.D. Mullen, & M . Becker (Eds.), Advances in health education and promotion (pp. 251-275V London: JAI Press.  Coelho, R. J. (1984). Self-efficacy and cessation of smoking. Psychology Reports. 54. 309-310.  200  Cogan, R., Henneborn, W., & Klopfer (1976). Predictors of pain during prepared childbirth. Journal of Psychosomatic Research. 20. 523-533.  Cohen, J. (1988). Statistical power analysis for the behavioural sciences. Hillsdale, New Jersey: Lawrence Erlbaum. Cohen, R. M . , Weingartner, H., Smallberg, S. A., Pickar, D., & Murphy, D. L . (1982). Effort and cognition in depression. Archives of General Psychiatry. 39, 593-597. Colletti, G., Supnick, J. A . , & Payne, A . A . (1985). The smoking self-efficacy questionnaire (SSEQ): Preliminary scale development and validation. Behavioural Assessment.7(3). 249-260.  Colletti, G., Supnick, J. A . , & Rizzo, A . A . (1981). A n analysis of relapse determinants for treated smokers. Paper presented at the American Psychological Association, Los Angeles, California. Collins, J. P. (1988). Prevalence of selected chronic conditions. United States, 1983-85. Rockville: National Centre for Health Statistics. Advance Data, 155: 1-16.  Condiotte, M . M . , & Lichtenstein, E. (1981). Self-efficacy and relapse in smoking cessation programs. Journal of Consulting and Clinical Psychology. 49(5). 648-658.  Cooley, C. H . (1902). Human nature and the social order. New York: Scibner's.  Cox, C. A . (1979). A pilot study using the elderly as community health educators. International Journal of Health Education. 22. 49-52.  Craig, K. D. (1983). A social learning perspective on pain experience. In M . Rosembaum, C. M . Franks, & Y . Jaffe (Eds.), Perspectives on Behaviour Therapy in the Eighties. New York: Springer. Craig, K. D. (1984). Emotional aspects of pain. In P. D. Wall & R. Melzack (Eds.), Textbook of Pain. London: Churchill/Livingston.  201 Creed, F., Murphy, S., & Jayson, M . V . (1990). Measurement of psychiatric disorder in rheumatoid arthritis. Journal of Psychosomatic Research. 34. 79-87.  Crowne, D. P., & Marlow, D.A. (1960). A new scale for social desirability independent of psychopathology. Journal of Consulting Psychology. 24. 349-354.  Cunningham, L . S., & Kelsey, J. L . (1984). Epidemiology of musculoskeletal impairments and associated disability. American Journal of Public Health. 75. 574-9. Davis, A . L . , Faust, R., & Ordentlich, M . (1984). Self-help smoking cessation and maintenance programs: A comparative study with 12-month follow-up by American Lung Association. American Journal of Public Health. 74(11). 1212-1217.  Davis, F. W., & Yates, B. T. (1982). Self-efficacy expectancies versus outcome expectancies as determinants of performance deficits and depressive effect. Cognitive Therapy and Research. 6, 23-35.  Davis, G. C. (1989). Measurement of the chronic pain experience: Development of an instrument. Research in Nursing and Health. 12. 221-227. Day, J. C. (1993). Population projections in the United States, by age, sex, race, and Hispanic origin: 1993-2050. Washington, DC: US Department of Commerce, Bureau of Census.  DeMonbreun, B . G., & Craighead, W. E. (1977). Distortion of perception and recall of positive and neutral feedback in depression. Cognitive Therapy and Research. 4, 311-329.  Devins, G. M . , Binik, Y . M . , Hutchinson, T. A., Hollomby, D. J., Barre, P. E., & Guttmann, R. D. (1983). The emotional impact of end-stage renal disease: Importance of patients' perceptions of intrusiveness and control. International Journal of Psychiatry in Medicine. 13, 327-343. deVries, H . , & Backbier, M . P. H . (1994). Self-efficacy as an important determinant of quitting among pregnant women who smoke; the O-pattern. Preventive Medicine. 23, 167-174.  202 deVries, H . , Dijkstra, M . , & Kuhlman, P. (1988). Self-efficacy: The third factor besides attitude and subjective norm as a predictor of behavioural intentions. Health Education Research. 3, 273-282.  DiClemente, C. C. (1981). Self-efficacy and smoking cessation maintenance: A preliminary report. Cognitive Therapy Research. (5). 175-187. DiClemente, C. C , Prochaska, J. O., & Gilbertini, M . (1985). Self-efficacy and the stages of self-change of smoking. Cognitive Therapy Research.9(2\ 181-200.  Dixon, J., & Bird, H . (1981). Reproducibility along a 10 cm vertical analogue scale. Annals of Rheumatic Diseases. 40. 87-89.  Downie, W., Letham, P., Rhind, V . , Wright, V . , Branco, J., & Anderson, J. (1978). Studies with pain rating scales. Annals of Rheumatic Diseases. 37. 378-381. Dupuy, H . J. (1972). The psychological section of the Current Health and Nutrition Examination Survey: Proceedings of the Public Health Conference on Records and Statistics meeting jointly with the National Conference on Health Statistics. Washington, D.C.: National Centre for Health Statistics.  Dzewantowski, D. A . , Noble, J. M . , & Shaw, J. M . (1990). Physical activity participation: Social cognitive theory versus the theories of reasoned action and planned behaviour. Journal of Sport and Exercise Psychology. 12. 388-405.  Eastman, C , & Marzilier, J. S. (1984). Theoretical and methodological difficulties in Bandura's self-efficacy theory. Cognitive Therapy and Research. 8, 213-229.  Eland, J. M . (1985). The child who is hurting. Seminars in Oncology Nursing. 1(2), 116. Epp, J. (1976). Achieving health for all: A framework for health promotion. Ottawa: Health and Welfare Canada.  203  Ewart, C. K . (1992). Role of physical self-efficacy in recovery from heart attack. In R. Schwarzer (Ed.). Self-Efficacy: Thought control of action (pp. 278-304). Washington: Hemisphere Publishing Corporation.  Ewart, C. K . , Taylor, C. B., Reese, L . B., & DeBusk, R. F. (1984). Effects of early postmyocardial infarction exercise testing on self-perception and subsequent physical activity. American Journal of Cardiology. 51. 1076-1080. Eysenck, H . , J. (1978). Expectations as causal elements in behaviour change. Advances in Behaviour Research and Therapy. 1, 171-175.  Fassler, D. (1980). Reducing preoperative anxiety in children: Information versus emotional support. Parent Counselling and Health Education. 2(83). 130-134.  Fleury, J. (1992). The application of motivational theory to cardiovascular risk reduction. I M A G E : Journal of Nursing Scholarship. 24(3), 229-239. Flor, H . , & Turk, D. C. (1988). Chronic back pain and rheumatoid arthritis: Predicting pain and disability from cognitive variables. Journal of Behavioural Medicine. 11(3). 251-265.  Fordyce, W., Fowler, R., Lehmann, J., DeLateur, B., Sand, P., & Treischmann, R. (1973). Operant conditioning in the treatment of chronic pain. Archives of Physical Rehabilitation. 54, 399-408.  Forgas, J. P., Bower, G. H . , & Krantz, S. (1984). The influence of mood on perceptions of social interactions. Journal of Experimental Social Psychology. 20. 497-513.  Frank, R. G., Beck, N . S : , Parker, J. C , Kashani, J. H . , Elliott, T. R., Haut, A . E., Smith, E., Atwood, C , Brownlee-Duffeck, M . , & Kay, D. R. (1988). Depression in rheumatoid arthritis. The Journal of Rheumatology. 15. 920-925.  Freud, S. (1921). Group psychology and the analysis of the ego. (Transl. London: International Psychoanalytical Press, 1922).  204  Fries, J. F., Spitz, P. W., & Young, D. Y . (1982). The dimensions of health outcomes: The health assessment questionnaire, disability and pain scales. Journal of Rheumatology. 9, 789-793. Garber, J., Hollon, S. D., & Silverman, V . (1979, December). Evaluation and reward of self vs. others in depression. Paper presented at the meeting of the Association for the Advancement of Behavior Therapy, San Francisco. Genest, M . (1983). Coping with rheumatoid arthritis. Canadian Journal of Behavioral Science. 14,392-408.  Gewirtz, J. L . , & Stingle, K. G. (1968). Learning of generalized imitation as the basis for identification. Psychological Review. 75. 374-397.  Gibson, T., & Clark, B. (1985). Use of simple analgesics in rheumatoid arthritis. Annals of Rheumatic Disease. 44, 27-29.  Glanz, K . , Lewis, F. M . , & Rimer, B. K . (1990). Theory, research, and practice in health education: Building bridges and forging links. In K . Glanz, F. M . Lewis, B. K . Rimer (Eds.) Health behaviour and health education (pp. 17-37). San Francisco: Jossey-Bass Publishers.  Glasgow, R. E., Schafer, L . , & O'Neill, H . K . (1981). Self-help books and the amount of therapist contact in smoking cessation programs. Journal of Consulting and Clinical Psychology. 49(5), 659-667. Glass, D. C , Reim, B., & Singer, J. (1971). Behavioural consequences of adaptation to controllable and uncontrollable noise. Journal of Experimental Social Psychology. 7. 244-257. Glymour, C , Schemes, R., Spirtes, P., & Kelly, K. (1987). Discovering causal structure: Artificial intelligence, philosophy of science, and statistical modeling. Orlando, F L : Academic Press.  Godding, P. R., Glasgow, R. E., & Klesges, R. C. (1985). Self-efficacy and outcome expectations as predictors of controlled smoking status. Cognitive Therapy Research. 9(5), 583-590.  205 Golin, S., & Terrell, F. (1977). Motivational and associative aspects of mild depression in skill and chance tasks. Journal of Abnormal Psychology. 86. 389-401.  Gollob, H . F., & Reichardt, C. S. (1987). Taking account of time lags in causal models. Child Development. 58, 80-92.  Gorsuch, R. L . (1974). Factor analysis. Philadelphia: W.B.Saunders. Glaser, B., & Strauss, A . (1967). The discovery of grounded theory. Chicago: Aldine. Gotlib, I. H . (1981). Self-reinforcement and recall: Differential deficits in depressed and nondepressed psychiatric patients. Journal of Abnormal Psychology. 90. 521-530.  Green, L . W. (1970). Should health education abandon attitude-change strategies? Perspectives from recent research. Health Education Monographs. L.(30), 25-48.  Green, L . W. (1974). Toward cost-benefit evaluations of health education: Some concepts, methods and examples. Health Education Monographs. 2 (Suppl. 2), 34-64.  Green, L . W., George, M . A . , Daniel, M . , Frankish, C. J., Herbert, C. P., Bowie, W. R., & O'Neill. (1995). Study of health promotion research in health promotion. Ottawa, O N : The Royal Society of Canada.  Green, L . W., Gottlieb, N . H . , & Parcel, G. S. (1991). Diffusion theory extended and applied. In W. B. Ward & F. M . Lewis (Eds.), Advances in Health Education. (Vol. 3). pp. 91-117. London: Jessica Publishers.  Green, L . W. & Kreuter, M . W. (1991). Health promotion planning: an educational and environmental approach. (2nd. ed.). Mountain View, C A : Mayfield Publishing Co. Green, L . W., Levine, D. M . , & Deeds, S. G. (1975). Clinical trials of health education for hypertensive outpatients: Design and baseline data. Preventive Medicine. 4. 417-425.  Green, L . W., Levine, D. M . , Wolle, J., & Deeds, S. (1979). Development of randomized patient education experiments with urban poor hypertensives. Patient Counselling and Health Education. 1, 106-111.  206  Harrison, J. A . , Mullen, P. D., & Green, L. W. (1992). A meta-analysis of studies of the Health Belief Model with adults. Health Education Research. 7(1), 107-116. Hawkins, R. M . F. (1992). Self-efficacy: A predictor but not a cause of behaviour. Journal Of Behaviour Therapy and Experimental Psychiatry. 23. 25L256.  Hawkins, R. M . F. (1995). Self-efficacy: A cause of debate. Journal of Behavioral Therapy and Experimental Psychiatry. 26,(3), 235-240.  Hawley, D. J., & Wolfe, F. (1988). Anxiety and depression in patients with rheumatoid arthritis: A prospective study of 400 patients. Journal of Rheumatology. 15. 932-940.  Health Canada, Statistics Canada. (1981). The Health of Canadians: Report of the Canada health survey. (Catalogue No. 83-538 E). Ottawa.  Hendler, N . (1984). Depression caused by chronic pain. Journal of Clinical Psychiatry. 45. 3036. Holman, H . , & Lorig, K . (1992). Perceived self-efficacy in self-management of chronic disease. In R. Schwarzer (Ed.), Self-Efficacy: Thought control of action (pp. 305-323). Washington: Hemisphere Publishing Corporation.  ;  Holman, H . , Mazonson, P., & Lorig, K. (1989). Health education for self-management has significant early and sustained benefits in chronic arthritis. Trans Association of American Physicians. 102. 204-208.  Holroyd, K . A . , Penzien, D. B., Hursey, K. G., Tobin, D. L . , Rogers, L . , Holm, J. E., Marcille, P. J., Hall, J. R., & Chila, A . G. (1984). Change mechanisms in E M G biofeedback training: Cognitive changes underlying improvements in tension headache. Journal of Consulting Clinical Psychology. 52. 1039-1053.  Holt, E. B. (1931). Animal drive and the learning process. New York: Holt. Hovell M . F., & Black D. R. (1989). Minimal intervention and arthritis treatment: implications for patient and physician compliance. Arthritis Care and Research. 2, 565-570.  207 Hudson, J. I., Goldenberg, D.L., & Popo, H . G. (1992). Comorbidity of fibromyalgia with medical and psychiatric disorders. American Journal of Medicine. 92, 363-367.  Humphrey, G. (1921). Imitation and the conditioned reflex. Pediatric Seminars. 28. 1-21. Hunt, W. A . , & Bespalec D. A . (1974). A n evaluation of current methods of modifying smoking behaviour. Journal of Consulting Psychology. 52. 431-438.  Hurd, P. D., Johnson, C. A . , Pachacek, T., Bast, L . P., Jacobs, D. R., & Juepker, R. V . (1980). Prevention of cigarette smoking in seventh grade students. Journal of Behavioural Medicine. 3(1). 15-28.  Huskisson, E.C. (1974). Measurement of pain. Lancet. 2. (1). 1127-1131.  Huskisson, E. C. (1983). Visual Analogue Scales. In R. Melzack (Ed.) Pain Measurement and Assessment. New York: Raven Press.  Ingram, R. E., & Smith, T. W. (1984). Depression and internal versus external focus of attention. Cognitive Therapy and Research. 8, 139-152. Jacobson, L . , Lingarde, F., & Manthorpe, R. (1989). The commonest rheumatic complaints of over six weeks' duration in a twelve month period in a defined Swedish population. Scandinavian Journal of Rheumatology. 18. 353-60.  James, W. (1890). Principles of Psychology (Vols 1-2). New York: Holt. Jeans, M . E. (1983). The measurement of pain in children. In R. Melzack (Ed.), Pain Measurement and Assessment (pp. 183). New York: Raven.  Jenkins, L . S. (1987). Self-efficacy: New perspectives in caring for patients recovering from myocardial infarction. Progress in Cardiovascular Nursing. 2, 32-35.  Kanfer, R., & Zeiss, A . M . (1983). Depression, interpersonal standard-setting, and judgements of self-efficacy. Journal of Abnormal Psychology. 92. 319-329.  208  Karoly, P. (1985). The logic and character of assessment in health psychology: Perspectives and possibilities. In P. Karoly (Ed.), Measurement strategies in health psychology (pp. 3-45). New York: Wiley. Katz, P., & Yelin, E. (1994). Life activities of persons with rheumatoid arthritis with and without depressive symptoms. Arthritis Care and Research. 7(2), 69-77. Kavanagh, D. J. (1992). Self-efficacy and depression. In R. Schwarzer (Ed.). Self-Efficacv: Thought control of action (pp. 177-1931. Washington: Hemisphere Publishing Corporation.  Killen, J. D., Maccoby, N . , & Taylor, C. B. (1984). Nicotine gum and self-regulation training in smoking relapse prevention. Behavioural Therapy. 15. 234-248.  Kingsley, R. G., & Shapiro, J. (1977). A comparison of three behavioural programs for the control of obesity in children. Behavioural Theory. 8, 30-36.  Kok, G., Den Boer, D. J., De Vries, H . , Gerards, F., Hospers, H . J., & Mudde, A . N . (1992). Self-efficacy and attribution theory in health education. In R. Schwarzer (Ed.). SelfEfficacy: Thought control of action (pp. 245-262). Washington: Hemisphere Publishing Corporation.  Kok, G., deVries, H , Mudde, A . N . , & Strecher, V . J. (1991). Planned health education and the role of self-efficacy: Dutch research. Health Education Research. 6. 231-238.  Kremer, E. F., & Atkinson, J. H . (1983). Pain language as a measure of affect in chronic pain patients. In R. Melzak (Ed.), Pain Measurement and Assessment. New York: Raven.  Kuiper, N . A . (1978). Depression and causal attributions for success and failure. Journal of Personality and Social Psychology. 36, 236-246.  LaPlante, M . (1988). Data on disability from the National Health Interview Survey. 1983-1985. A n InfoUse Report. Washington, D C : U.S. National Institute on Disability and Rehabilitation Research.  209 Last, J. M . (1988). A dictionary of epidemiology. Toronto: Oxford University Press. La Vecchia, C , Negri, E., Pagano, R., & Decarli, A . (1987). Education prevalence of disease, and frequency of health care utilization. The 1983 Italian National Health Survey. Journal of Epidemiology and Community Health. 41. 161-5.  Lawrence, L . , & McLeroy, K. R. (1986). Self-efficacy and health education. Journal of School Heahh, 56,(8), 317-321.  Lawrence, R. C , Hochberg, M . C , Kelsey, J. L., McDuffie, F. C , Medsger, T. A . , Felts, W. R., & Shulman, L . E. (1989). Estimates of the prevalence of selected arthritis and musculoskeletal diseases in the United States. The Journal of Rheumatology. 16.(4). 427-441.  Lee, P., Helewa, A., Smythe, H . A., Bombardier, C , & Goldsmith, C. H . (1985). Epidemiology of musculoskeletal disorders (complaints) and related disability in Canada. Journal of Rheumatology. 12, 1169-73. Lenker, S. L . , Lorig, K., & Gallagher, D. (1984). Reasons for the lack of associations between changes in health behaviour and improved health status: A n explanatory study. Patient Education & Counselling. 6, 69-72.  Levine, D. M . , Green, L . W., Deeds, S. G., Chwalow, A . J., Russell, R. P., & Finlay, J. (1979). Health education for hypertension patients. Journal of the American Medical Association. 241. 1700-1703.  Lishman, W. P. (1972). Selective factors in memory, Part 2: Affective disorders. Psychological Medicine. 2. 248-253.  Litt, M . D. (1988). Self-efficacy and perceived control: Cognitive mediators of pain tolerance. Journal of Personality and Social Psychology. 34(1). 149-160.  Lloyd, G. G., & Lishman, W. R. (1975). Effect of depression on the speed of recall of pleasant and unpleasant experiences. Psychological Medicine. 5. 173-180.  210  Lobitz, W. C , & Post, R. D. (1979). Parameters of self-reinforcement and depression. Journal of Abnormal Psychology. 88. 33-41.  Loeb, A . , Beck, A . T., Diggory, J. C , & Tuthill, R. (1967). Expectancy, level of aspiration, performance, and self-evaluation in depression. Proceedings of the 75th Annual Convention of the American Psychological Association. 2, 193-194.  Lomi, C. (1992). Evaluation of a Swedish version of the arthritis self-efficacy scale. Scandinavian Journal of Caring Science. 6(3). 131-138.  Lorig, K . (1982). Arthritis self-management: A patient education program. Rehabilitation Nursing. 8. 16-20.  Lorig, K. (1986). Development and dissemination of an arthritis patient education program. Family and Community Health.5. 23-36. Lorig, K . (1990). Arthritis self-management leader's manual. Atlanta, G A : Arthritis Foundation. Lorig, K . , Chastain, R., Ung, E., Shoor, S., & Holman, H . (1989). Development and evaluation of a scale to measure perceived self-efficacy in people with arthritis. Arthritis and Rheumatism. 32. 37-44.  Lorig, K . R., Cox, T., Cuevas, Y., Kraines, R. G., & Britton, M . C. (1984). Converging and diverging beliefs: Caucasian patients, Spanish speaking patients, and physicians. Journal of Rheumatology. 11(1), 76-79. Lorig, K., Feigenbaum, P., Regan, C , Ung, E., & Holman, H . R. (1986). A comparison of laytaught and professional-taught arthritis self-management courses. The Journal of Rheumatology. 13(4), 763-767.  Lorig, K . , & Fries, J. (1990). The arthritis helpbook. Reading, M A .  Lorig, K., & Holman, H . (1989). Long-term outcomes of an arthritis self-management study: Effects of reinforcement efforts. Social Science and Medicine. 29. 221-224.  211  Lorig, K., & Holman, H . (1993). Arthritis self-management studies: A twelve-year review. Health Education Quarterly. 20. 17-28.  Lorig, K . , Kraines, G., & Holman, H . (1981). A randomized prospective controlled study of the effects of health education for people with arthritis. Arthritis & Rheumatism. 24. SI 48.  Lorig, K . , Lubeck, D., Kraines, R. G., Seleznick,M., & Holman, H . R. (1985). Outcomes of selfhelp education for patients with arthritis. Arthritis and Rheumatism. 28(6) 680-685.  Lorig, K. R., Mazenson, P. D., & Holman, H . R. (1993). Evidence suggesting that health education for self-management in chronic arthritis has sustained health benefits while reducing health care costs. Arthritis and Rheumatism. 36(4), 439-446.  Lorig, K., Seleznick, M . , Back, D., Ung, E., Chastain, R., & Holman, H . (1989). The beneficial outcomes of the arthritis self-management course are not adequately explained by behaviour change. Arthritis & Rheumatism. 32. 91-95. Lott, B., & Lott, A . J. (1985). Learning theory in contemporary social psychology. In G. Lindzey & E. Aronson (Eds.), Handbook of social psychology (pp. 109-135). New York: Random House.  Maddux, J. E., & Rogers, R. N . (1986). Protection motivation and self-efficacy: A revised theory of fear appeals and attitude change. Journal of Experimental Social Psychology. 19, 469-479.  Maddux, J. E., & Stanley, M . (1986). Self-efficacy theory in contemporary psychology: A review. Journal of Social and Clinical Psychology. 4(3), 249-255.  Manning, M . M . , & Wright, T. L . (1983). Self-efficacy expectancies, outcome expectancies, and the persistence of pain control in childbirth. Journal of Personality and Social Psychology. 45, 421-431.  Marlatt, G. A . , & Gordon, J. R. (1980). Determinants of relapse: Implications for the maintenance of behaviour change. In P. Davidson & S. Davidson (Eds.), Behavioural Medicine: Changing Health Lifestyles. New York: Brunner/Mazel.  212  McDonald, R. P., & Marsh, H . W. (1990). Choosing a multivariate model: Noncentrality and goodness of fit. Psychological Bulletin. 107. 247-255. McFarlane, A . C , & Brooks, P. M . (1988). Determinants of disability in rheumatoid arthritis. British Journal of Rheumatology. 27. 7-14.  McGowan, P. (1990). The A B C Project - Providing active participation in responsible health care. A report to Health & Welfare Canada Health Promotion Directorate. Seniors Independence Program. Vancouver, B C : The Arthritis Society (B.C. and Yukon Division).  McGowan, P., Badley, E., Choquette, D., Devins, G., Edworthy, S., Green, L . , & Lorig, K . (1994). Arthritis self-management program national implementation: Evaluation of impact on participants' health status. Toronto, Ontario: The Arthritis Society (National Office). McGowan, P., & Green, L . W. (1995). Arthritis self-management in native populations of British Columbia: A n application of health promotion and participatory research principles in chronic disease control. Canadian Journal on Aging. 14(1). 201-212.  McGuire, P. (1987). Doing participatory research: A feminist approach. Amherst, Massachusetts: CIE.  McGuire, W. J. (1983). A contextualist theory of knowledge: Its implications for innovation and reform in Psychological research. Advances in Experimental Psychology. 16.1-47.  Mclntyre, K. O., Leichtenstein, E., & Mermelstein, R. J. (1983). Self-efficacy and relapse in smoking cessation: A replication and extension. Journal of Consulting Clinical Psychology. 51. 632-633.  McKenna, F., & Wright, V . (1985). Pain and rheumatoid arthritis. Annals of Rheumatic Disease. 44, 805. Mead, G. H . (1934). Mind, self and society (posthumous; C. D. Morris, Ed.). Chicago: University of Chicago Press.  213  Meenan, R. F., Gertman, P. M . , & Mason, J. H . (1980). Measuring health status in arthritis: The arthritis impact measurement scales. Arthritis and Rheumatism. Vol. 23. 146-152.  Meenan, R. F., Gertman, P. M . , Mason, J. H . , & Dunaif, R. (1982). The arthritis impact measurement scales: Further investigation of a health status measure. Arthritis and Rheumatism. Vol. 25. 1048-1053. Miles, T. P., Flegal, K . , & Harris, T. (1993). Musculoskeletal disorders: Time trends, cormorbid conditions, self-assessed status, and associated activity limitations. Vital & Health Statistics - Series 3: Analytic and Epidemiologic Studies. 27: 275-288.  Miller, G. (1984). A n assessment of self-efficacy, performance, and cognitive distortion following successes and failure, in moderately depressed and nondepressed college students. Unpublished Honours Project, University of Sydney. Miller, N . E., & Dollard, J. (1941). Social learning and imitation. New Haven: Yale University Press.  Miller, R. W. (1983). Fact and method in social sciences. In B. P. Gasper & J. D. Trout (Eds.), The philosophy of science (pp. 743-762). Cambridge, Mass: MIT Press.  Miller, S. (1979). Controllability and human stress: Method, evidence and theory. Behavioural Research Therapy. 17, 287-304.  Miller, W. R. (1975). Psychological deficit in depression. Psychological Bulletin. 82, 238-260.  Mischel, W. (1973). Toward a cognitive social learning reconceptualization of personality. Psychological Review. 80. 252-283.  Morbidity and Mortality Weekly Report, U . S. Department of Health and Human Services Public Health Service (1994) 43: 433-438. Morisky, D. E., Levine, D. M . , Green, L. W., Shapiro, S., Russell, R. P., & Smith, C. R. (1983). Five-year blood-pressure control and mortality following health education for hypertensive patients. American Journal of Public Health. 73. 153-162.  214  Morrow, K . A . , Parker, J. C , & Russell, J. L . (1994). Clinical implications of depression in rheumatoid arthritis. Arthritis Care and Research. 7(2), 58-63.  Mowrer, O. H . (1960). Learning theory and the symbolic processes. New York: Wiley.  Mullen, P., Green, L., & Persinger, G. (1985). Clinical trials of patient education for chronic conditions: A comparative meta-analysis of intervention types. Preventive Medicine. 14, 753-781.  Murphy, J. M . , Olivier, D. D., Monsonth, R. R , Sobol, A . M . , & Leighton, A . H . (1988). Incidence of depression and anxiety: The Sterling County study. American Journal of Public Health. 78, 534-539. Muthen, B., & Kaplan, D. (1985). A comparison of some methodologies for the factor analysis of non-normal Likert variables. British Journal of Mathematical and Statistical Psychology. 38. 171-189. National Cancer Institute of Canada, (1989). Canadian Cancer Statistics 1989. Toronto, Ontario.  Nelson, R. E., & Craighead, W. E. (1977). Selective recall of positive and negative feedback, self-control behaviors, and depression. Journal of Abnormal Psychology. 86. 379-388.  Neufeld, R. W. J., & Thomas, P. (1977). Effects of perceived efficacy of a prophylactic controlling mechanism on self-control under painful stimulation. Canadian Journal of Behavioural Science. 9, 224-232.  Nutbeam, D., Smith, C , & Catford, J. (1989). Evaluation in health education. A review of progress, possibilities, and problems. Journal of Epidemiology and Community Health. 44, 83-89. O'Leary, A . (1985). Self-efficacy and health. Behavioural Research and Theory. 23(4), 437-451.  O'Leary, A . , Shoor, S., Lorig, K., & Holman, H . R. (1988). A cognitive-behavioural treatment for rheumatoid arthritis. Health Psychology. 7(6), 527-544.  215  Parcel, G.'S., & Baranowski, T. (1981). Social learning theory and health education. Health Education. 12. 14-18. Parcel, G. S., Bruhn, J. G., & Murray, J. L . (1983). Preschool health education program (PHEP): Analysis of education and behavioural outcomes. Health Education Quarterly. 10(3.4). 149-172.  Patterson, K . L., & Klopovich, P. M . (1987). Pain in the pediatric oncology patient. In D. B . McGuire & C. H . Yarbro (Eds.), Cancer Pain (pp. 259). Orlando, F L : Grune & Stratton.  Peck, J. R., Smith, T. W., Ward, J. R., & Milano, R. (1989). Disability and depression in rheumatoid arthritis. A multi-trait, multi-method investigation. Arthritis and Rheumatism. 32, 1000-1006. Perry, C. L . , Baranowski, T., & Parcel, G. S. (1990). How individuals, environments, and health behaviour interact: Social learning theory. In K. Glanz, F . M . Lewis, B . K . Rimer (Eds.), Health Behaviour and Health Education (pp. 161-186). San Francisco: Jossey-Bass Publishers.  Peterson, C , & Seligman, M . E. P. (1984). Casual explanations as a risk factor for depression: Theory and evidence. Psychological Review. 91. 347-374.  Peterson, C , Semmel, A., von Baeyer, C , Abramson, L. Y., Metalsky, G. I., & Seligman, M . E. P. (1982). The Attributional Style Questionnaire. Cognitive Therapy and Research. 6, 287-300.  Pincus, T., & Callahan, L . F. (1993). Depression scores in rheumatoid arthritis: Criterion contamination of patient responses. Patient Education and Counselling. 20. 63-76.  Pincus, T., Callahan, L . F., Bradley, L . A., Vaughn, W. K., & Wolfe, F. (1986). Elevated M M P I scores for hypochondriasis, depression, and hysteria in patients with rheumatoid arthritis reflect disease rather than psychological status. Arthritis and Rheumatism. 29. 14561466.  216  Pincus, T., Callahan, L . F., & Burkhauser, R. V . (1987). Most chronic diseases are reported more frequently by individuals with fewer than 12 years of formal education in the age 18-64 United States population. Journal of Chronic Disease. 40. 865-74. Popper, K . (1934). Selections from the logic of scientific discovery. In B . P. Gasper & J.D. Trout (Eds.), The philosophy of science (pp. 99-116). Cambridge, Mass: MIT Press.  Popper, K . (1959). The Logic of Scientific Discovery. London: Hutchinson and Company.  Prochaska, J. O., & DiClemente, C. C. (1982). Transtheoretical therapy: Toward a more integrative model of change. Psychotherapy Theory Research and Practice. 19(3). 276287.  Prochaska, J. O., & DiClemente, C. C. (1984). Self change processes, self-efficacy and decisional balance across five stages of smoking cessation. In A . R. Liss (Ed.), Advances in Cancer Control: Epidemiology and Research. New York: Alan R. Liss Inc.  Prochaska, J. O., & DiClemente, C. C.(1985). Common processes of self-change in smoking, weight control, and psychological distress. In S. Shiftman and T. Wills (Eds.) Coping and Substance Abuse. Orlando, Florida: Academic Press, 345-364.  Prochaska, J. O., & DiClemente, C. C , & Norcross, J. C. (1992). In search how people change: Applications to addictive behaviors. American Psychologist. 47. 1102-1114.  Prochaska, J. O., Crimi, P., Lapsanski, D., Martel, L., & Reid, P. (1982). Self-change processes, self-efficacy and self-concept in relapse and maintenance of cessation of smoking. Psychological Reports. 5J_, 983-990.  Radloff, L . (1977). The CED-D Scale: a self-report depression scale for research in the general population. Applied Psychological Measurement. 1(3), 385-401.  Ratner, P., Johnson, J., & Jeffery, B. (1996, June). Examining emotional, physical, social, and spiritual health as determinants of self-rated health status. Paper presented at the Fourth Canadian Conference on Health Promotion, Montreal, Qb.  217  Reese, L . B. (1982). Pain reduction through cognitive, self-relaxative and placebo means: A self-efficacy analysis. Unpublished Dissertation, Stanford University, Stanford, California.  Rehm, L . P. (1982). Self-management in depression. In P. Karoly & F. H . Kanfer (Eds.), Selfmanagement and behavior change: From theory to practice (pp. 522-567). New York: Pergamon. Rejeski, W. J., Craven, T., Ettinger, W. H . , McFarlane, M . , & Shumaker, S. (1996). Selfefficacy and pain in disability with osteoarthritis of the knee. Journal of Gerontology: Psychological Sciences. 51B(1), 24-29.  Revenson, T. A . , Schiafinno, K . M . , Majerovitz, D., & Gibofsky, A . (1991). Social support as a double-edged sword: The relation of positive and problematic support to depression among rheumatoid arthritis patients. Social Science and Medicine. 33(7). 807-813.  Revill, S. I., Robinson, J. O., Rosen, M . , & Hogg, M . I. J. (1976). The reliability of a linear analogue for evaluating pain. Anaesthesia. 31. 1991-1198.  Reynolds, D. L . , Chambers, L . W., Badley, E. M . , Bennett, K. J., Goldsmith, C. H . , Jamieson, E., Torrance, G.,W., & Tugwell, P. (1992). Physical disability among Canadians reporting musculoskeletal diseases. Journal of Rheumatology. 19(7). 1020-1030.  Reynolds, D. L., Torrance, G. W., Badley, E. M . , Bennett, K . J., Chambers, L . W., Goldsmith, C. H . , Jamieson, E., Tugwell, P., & Wolfson, M . C. (1993). Modelling the population health impact of musculoskeletal diseases: Arthritis. The Journal of Rheumatology. 20.(6). 1037-1047.  Reynolds, R., Creer, T. L., Holroyd, K. A., & Tobin, D. L . (1982). Assessment in the treatment of cigarette smoking: The development of the smokers' self-efficacy scale. Paper presented at the meeting of the Association for Advancement of Behavior Therapy, Los Angeles, C A .  Rimon, R. (1974). Depression on rheumatoid arthritis. Annals of Clinical Research. 6, 171-175.  218  Rizley, R. (1978). Depression and distortion in the attribution of causality. Journal of Abnormal Psychology. 87, 32-48.  Rogers, R. W. (1975). A protection motivation theory of fear appeals and attitude change. Journal of Psychology. 91. 93-114.  Rogers, R. W. (1983). Cognitive and physiological processes in fear appeals and attitude change: A revised theory of protection motivation. In J. R. Cacioppo & R. E. Petty (Eds.). Social psychology: A sourcebook (pp. 153-176). New York: Guilford. Rogers, R. W. (1985). Attitude change and information integration in fear appeals. Psychological Reports. 56. 179-182.  Rogosa, D. (1979). Causal models in longitudinal research: Rationale, formulation, and interpretation. In J. R. Nesselroade and P. B. Bakes (Eds.), Longitudinal research in the study of behaviour and development (pp. 263-302). New York: Academic Press.  Rosenbaum, M . (1980). A schedule for assessing self-control behaviours: Preliminary findings. Behaviour Therapy. 11. 109-121. Rosenstock, I. M . , Stretcher, V . J., & Becker, M . H . (1988). Social learning theory and the health belief model. Health Education Quarterly. 15. 175-183.  Rossi, P. H . (1991). Comprehensive, tailored, theory-driven evaluations - a smorgasbord of options. In W. R. Shadish, T. D. Cook, & L . C. Leviton (Eds.), Foundations of Program Evaluation, (pp. 377-437). California: Sage Publications.  Rotter, J. B. (1966). Generalized expectancies for internal versus external control of reinforcement. Psychological Monographs. 80. 1-28.  Sakano, Y., & Tohijoh, M . (1986). The General Self-Efficacy Scale (GSES): Scale development and validation. Japanese Journal of Behaviour Therapy. 12,(1), 73-82.  219  Schiaffino, K . M . , Revenson, T. A., & Gibofsky, A . (1991). Assessing the impact of selfefficacy beliefs on adaptation to rheumatoid arthritis. Arthritis Care and Research. 4(4), 150-157.  Schiaffino, K . M . , & Revenson, T. A . (1992). The role of perceived self-efficacy, perceived control, and causal attributions in adaptation to rheumatoid arthritis: Distinguishing mediator from moderator effects. Personality and Social Psychology Bulletin. 18(6). 709-718.  Schunk, D. H . , & Carbonari, J. P. (1984). Self-efficacy models. In J. D. Matarazzo, J. A . Herd, N . E. Miller, & S. M . Weiss (Eds.), Behavioural health: A handbook of health enhancement and disease prevention, (pp. 230-247). New York: Wiley. Schwartz, J. L . (1974). Relationship between goal discrepancy and depression. Journal of Consulting and Clinical Psychology. 42. 309.  Schwartz, N . , & Bless, H . (1991). Happy and mindless, but sad and smart? The impact of affective states on analytic reasoning. In J. P. Forgas (Ed.), Emotion and social judgements (pp. 55-71). Oxford: Pergamon.  Schwarzer, R. (1992). Self-efficacy in the adoption and maintenance of health behaviour: Theoretical approaches and a new model. In R. Schwarzer (Ed.), Self-efficacy: Thought, control and action (pp. 217-243). Washington, D . C : Hemisphere.  Scott, J., & Huskisson, E. C. (1979). Accuracy of subjective measurements made with or without previous scores: A n important source of error in serial measurement of subjective states. Annals of the Rheumatic Diseases. 38, 558-559.  Seligman, M . E. P. (1975). Helplessness. San Francisco: Freeman. Sheehan, T. J., Fifield, J., Reisine, S, & Tennen, H . (1995). The measurement structure of the centre for epidemiologic studies depression scale. Journal of Personality Assessment. 64(3), 507-521.  Sherer, M . (1982). The Self-efficacy scale: Construction and validation. Psychological Reports. 51,663-671.  •  220  Shoor, S. M . , & Holman, H . R. (1984). Development of an instrument to explore psychological mediators of outcome in chronic arthritis. Transactions of the Association of American Physicians. 97.325-331.  Siegel, L . M . , Attkisson, C. C. & Carson, L . G. (1978). Need identification and program planning in the community context. In C.C. Attkisson, W. A . Hargreaves, M.J. Horowitz, & J. E. Sorensen (Eds.) Evaluation of Human Service Programs.(pp. 215- 252). New York: Academic Press.  Skipper, J. K., & Leonard, R. C. (1968). Children, stress, and hospitalization: A field experiment. Journal of Health and Social Behaviour. 9, 275-287. Smollen, R.C. (1978). Expectancies, mood and performances of depressed and nondepressed psychiatric inpatients on chance and skill tasks. Journal of Abnormal Psychology. 87, 91-101.  Statistica/W (Computer Software). (1994). Tulsa, O K : StatSoft, Inc.  Statistica/W (Computer Software). (1995). Tulsa, O K : StatSoft, Inc. Statistics Canada (1989). Health and activity limitation survey microdata user's guide adults in households. Ottawa, Ontario: Statistics Canada.  Statistics Canada (1990). Health and activity limitation survey microdata user's guide adults in institutions. Ottawa, Ontario: Statistics Canada.  Statistics Canada (1994). Violence against women survey [Public use microdata file documentation and user's guide]. Ottawa: Statistics Canada.  Steiger, J. H . (1994). Structural equation modeling. Tulsa, O K : StatSoft, Inc.  Steiger, J. H . (\995\ Structural equation modeling. In STATISTICA for Windows (Addendum): New features in STATISTICA 5.0, 283-487.  221  Steiger, J. H . , & Lind, J. C. (1980, May). Scientifically-based tests for the number of common factors. Paper presented at the annual meeting of the Psychometric Society. Iowa City, IA.  Stephens, T., & Graham, D. F. (Eds.), Canada's health promotion survey 1990: Technical report. (Catalogue no. H39-263/2-1990E). Ottawa: Minister of Supply and Services, Health and Welfare Canada. Stewart, A . L . , Hays, R. D., & Ware, J. E. (1988). The MOS short-form general health survey: Reliability and validity in a patient population. Medical Care. 26. 724-735. Stewart, A . L . , Ware, J. E. Jr., Sherbourne, C. D., & Wells, K. B. (1992). Psychological distress/well being and cognitive functioning measures. In A . L . Stewart, & J. E. Ware, (Eds.), Measuring functioning and well being: The medical outcomes study approach. Durham, N C : Duke University Press.  Strecher, V . J. (1983). Effect of a minimal-contact smoking program in a health care setting. Unpublished doctoral dissertation, University of Michigan.  Strecher, V . J., DeVellis, B., Becker, M . H . , & Rosenstock, I. M . (1986). The role of selfefficacy in achieving health behaviour change. Health Education Quarterly. 13. 79-91.  Suchman, E. A . (1970). Action for what? A critique of evaluation research. In C. H . Weiss, (Ed.), Evaluating action programs: readings in social action and education. Boston: Allyn and Bacon Inc.  Taal, E., Jacobs, J. W., Seydel, E. R., Wiegman, O., & Rasker, J. J. (1989). Evaluation of the Dutch arthritis impact measurement scales (DUTCH-AIMS) in patients with rheumatoid arthritis. British Journal of Rheumatology. 28. 487-491.  Taal, E., Rasker, J. J., & Seydel, E. R. (1993). Health status, adherence with health recommendations, self-efficacy and social support in patients with rheumatoid arthritis. Patient Education and Counselling. 20. 63-76.  Tarde, G. (1903). The laws of imitation (transl.). New York: Henry Holt.  222  Taylor, C. B., Bandura, A . , Ewart, C. K., Miller, N . H . , & DeBusk, R. F. (1985). Raising spouse's and patient's perception of his cardiac capabilities following a myocardial infarction. American Journal of Cardiology. 55, 635-638. Taylor, K . M . , & Pompa, J. (1990). A n examination of the relationship among career decision making self-efficacy, career salience, locus of control, and vocational indecision. Journal of Vocational Behaviour. 37. 17-31.  Thompson, L . W., Gallagher, D., & Nies, G. (1983). Evaluation of the effectiveness of professionals and non-professionals as instructors of "Coping with Depression" classes for elders. Gerontologist. 23, 390-396. Tugwell, P., Chambers, L . , Torrance, G., Reynolds, D., Wolfson, M . , Bennett, K . , Badley, E., Jamieson, E., Stock, S., & the P O H E M Workshop Group (1993). The population health impact of arthritis. The Journal of Rheumatology. 20(6). 1048-1051.  Turk, D. C , Rudy, T. E., & Salovey, P. (1986). Implicit models of illness. Journal of Behavioural Medicine. 9, 453-474.  Turk, D. C , Meichenbaum, D., & Genest, M . (1983). Pain and behavioural medicine: A cognitive-behavioural approach. New York: Guilford Press.  Turk, D. C , & Salovey, P. (1984). Chronic pain as a variant of depressive disease: A critical appraisal. Journal of Nervous and Mental Disease. 172. 1-7.  Tursky, B., & Sternbach, R. A . (1967). Further psychological correlates of ethnic differences in response to shock. Psychophysiology. 4, 67-74.  U.S. Public Health Service (1979). Healthy people: The Surgeon General's report on health promotion and disease prevention. D H E W publication no. 79-55071. Washington, D C : U.S. Government Printing Office.  van Ryn, M . , & Heaney, C. A . , (1992). What's the use of theory? Health Education Quarterly. 19(3), 315-330.  223  Veit, C. T., & Ware, J. E. (1983). The structure of psychological distress and well-being in general populations. Journal of Consulting Clinical Psychology. Vol. 51. 730-742.  Verbrugge, L . (1984). Longer life but worsening health? Trends in health and morbidity of middle-aged and older persons. Milbank Quarterly. 62. 475-519.  Walker, W. B., & Franzini, L . R. (1983). Self-efficacy and low-risk aversive group treatments for smoking cessation. Paper presented at the Meeting of Western Psychological Association, San Francisco, California.  Wallston, K . A . , Wallston, B. S., & DeVellis, R. (1978). Development of the multidimensional health locus of control (MHLC) scales. Health Education Monographs. 6, 160-170. Ware, J. E., Nelson, E. C , Sherbourne, C. D., & Stewart, A . L . (1992). Preliminary tests of a 6-item general health survey: A patient application. In A . L . Stewart and J. E. Ware (Eds.). Measuring Functioning and Well Being: The Medical Outcomes Study Approach. Durham, N.C.: Duke University Press. Weiner, B. (1985). A n attributional theory of achievement motivation and emotion. Psychological Review. 92. 548-573.  Weiner, B. (1986). A n attributional theory of motivation and motivation. New York: Springer.  Weisenberg, M . , Kreindler, M . L . , Schachat, R., & Werboff, J. (1975). Pain: Anxiety and attitudes in black, white, and Puerto Rican patients. Psychosominia Medicine. 37. 123135.  Weismann, M . M . , & Myers, J. K . (1978). Affected disorders in a U.S. urban community. Archives of General Psychiatry. 35. 1304-1311. Welsh, G. S., & Dahlstrom, W. G. (Eds.). (1956). Basic readings on the M M P I in psychology and medicine. Minneapolis: University of Minnesota Press.  Wener, A . E., & Rehm, L . P. (1975). Depressive affect: A test of behavioral hypothesis. Journal of Abnormal Psychology. 84. 221-227.  224 Wheeler, M . E., & Hess, K. W. (1976). Treatment of juvenile obesity by successive approximation control of eating. Journal of Behavioural Therapy and Experimental Psychology. 7, 235-241.  Wissler, C. (1923). Man and culture. New York: Thomas Y . Crowell.  World Health Organization. (1986). The Ottawa Charter for Health Promotion. Health Promotion. 1.3-5.  Wright, J., & Mischel, W. (1982). The influence of affect on cognitive social learning person variables. Journal of Personality and Social Psychology. 43. 901-914.  Yalow, E. S.,& Collins, J. L . (1987). Self-efficacy in health behaviour change: Issues in measurement and research design. In S. K . Simonds, P. D. Mullin & M . Becker (Eds.) Advances in Health Education and Promotion, (pp. 181-199). London: JAI Press Inc. Yelin, E. H . , & Felts, W. R. (1990). A summary of the impact of musculoskeletal conditions in the United States. Arthritis and Rheumatism. 33(5), 750-755.  Yelin, E. H . , Lubeck, D., Holman, H . , & Epstein, W. (1987). The impact of rheumatoid arthritis and osteoarthritis: The activities of patients with rheumatoid arthritis and osteoarthritis compared to controls. Journal of Rheumatology. 14. 710-717.  Yelin, E. H . , & Katz, P. P. (1990). Transitions in health status among community dwelling elders with arthritis.. Arthritis and Rheumatism. 33(8), 1205-1215.  Yusaf, S., & Kavanagh, D., J. (1990). The role of self-efficacy in the treatment of depression. Journal of Cognitive Psychotherapy. 4, 51-70.  Zapka, J., & Mazur, R. (1977). Peer sex education training and evaluation. American Journal of Public Health. 67, 450-454.  Zborowski, M . (1969). People in Pain. San Francisco. California: Jossey-Bass.  225  LIST OF APPENDICES  Appendix A  A S M P Program Overview  226  Appendix B  B C A S M P Study Questionnaire  227  Appendix C  National A S M P Study Questionnaire  238  Appendix D  Behaviour Identified in Nominal Group Process  248  Appendix E  Precede-Proceed Model of Health Promotion Planning  249  APPENDIX A  ASMP PROGRAM  226  OVERVIEW S E S S I O N  TOPICS  1  2  SELF-MANAGEMENT PRINCIPLES  •  DISEASE PROCESS OA, RA, OSTEOPOROSIS & FIBROMYALGIA  •  EXERCISE/FITNESS  •  PAIN MANAGEMENT/ RELAXATION  3  4  5  6  • •  •  •  •  FEAR, ANGER AND FRUSTRATION  •  FATIGUE  •  •  •  DEPRESSION NUTRITION  •  PROBLEM-SOLVING  •  COMMUNICATION SKILLS  •  DOCTOR-CLIENT RELATIONSHIPS  •  MEDICATIONS  •  NON-TRADITIONAL TREATMENTS  •  FEEDBACK/CONTRACTING  227.  APPENDIX B  LEAVE THIS AREA BLANK  A r t h r i t i s Branch Ocranunity Support Project A r t h r i t i s Society TORR NAME  CHI) Birthdate  (H3)  Street Address  PI  date  (H14)  City, Province, Postal Code  (H9)  Telephone Number (Home)  Sex  (H4)  (Work)  H2 medrec H3 brthdate H4  Ethnic Origin  (H5)  H5  Please c i r c l e the highest year of school completed. (H6) 12 3 4 5 6 (primary)  7 8 9 10 11 12 (high school)  13 14 15 16 (college)  17 18 19 20 21 22 (graduate school)  above 22  In what month and year did your a r t h r i t i s begin?  (H7)  H6  H7  The last time I saw a doctor for a r t h r i t i s was month Are you currently: (check only one)  year  (H10)  1.  single  4.  divorced  2.  married  5.  widowed  3.  separated  H10  Are you: (check only one) (Hll) 1.  employed f u l l time  5.  retired  2.  ' employed part time  6.  disabled  7.  other (describe)  3.  seeking work  4.  hamemaker  Hll  If employed, what kind of work do you do?  (Hi2)  If retired, what kind of work did you do mostly?  (H12)  What kind of a r t h r i t i s do you have?  (H16)  H12 H16 H13 site H17 ASH  - 2 -  228  LEAVE THIS AREA BLANK  Riysical Acrdyities/Therapies for A r t h r i t i s During the past month, on an average, how many times per week did you do each of the following? Please f i l l i n each space with a zero or other number. Stretching exercise for a r t h r i t i s t o improve joint movement  times per week  Strengthening exercise for a r t h r i t i s to strengthen muscles and joints  times per week  Practice relaxation techniques  times per week  |151  L i s t which ones: Massage  times per week  Walking for exercise  times per week  Each time you walk for exercise, how many minutes do you walk?  minutes  11  Each time you walk for exercise, how many blocks do you walk?  blocks  14  Swimming (i.e., of lap swijnming) Each time that you swim, how many minutes do you swim? Bicycling (regular or stationary) Each time that you bicycle, how many minutes do you b i c y c l e ? . . . . . .  PLEASE GO ON TO THE NEXT PAGE.  times per week minutes  |12  times per week  |10  minutes  |13  -  3  229  -  LEAVE THIS AREA BLANK  Please check the one response which best describes your usual a b i l i t i e s OVER THE PAST WEEK:  DRESSING & GROOMING Are you able to: - Dress yourself, including tying shoelaces and doing buttons? -  Without ANY Difficulty  With SOME Difficulty  With MUCH Difficulty  UNABLE to do  Shampoo your hair?  35  ARISING Are you able to: - Stand up from an armless straight chair? -  Get i n and out of bed?  36  EATING Are you able to: - Cut your meat? -  L i f t a f u l l cup or glass to your mouth?  37  WALKING Are you able to: - Walk outdoors on f l a t ground? -  Climb up five steps?  38  * Please check any AIDS CR DEVICES that you usually use f o r any of these a c t i v i t i e s : Cane  Devices Used for Dressing (button hook, zipper p u l l , long-handled shoe horn, etc.)  Walker  Built Up or Special Utensils  Crutches  Special or Built Up Chair  Wheelchair  Other (Specify:  * Please check any categories f a r which you usually need HELP FROM ANOTHER PERSON: Dressing & Groaning  Eating  Arising  Walking  - 4-  LEAVE THIS AREA BLANK  230  Please check the one response which best describes your usual a b i l i t i e s OVER THE PAST WEEK: Without ANY With SOME With MUCH UNABLE Difficulty Difficulty Difficulty to do HYGIENE Are you able to: - Wash & dry your entire body? -  Take a tub bath? Get on and o f f the t o i l e t ?  39  REACH Are you able to: - Reach & get down a 5 pound object (such as a bag of sugar) from just above your head? -  Bend down to pick up clothing from the floor?  40  GRIP Are you able to: - Open car doors? -  Open jars which have been previously opened?  -  Turn faucets on and off?  41  ACTIVITIES Are you able to: - Run errands and shop? -  Get i n and out of a car?  -  Do chores such as vaaiuming and yardwork?  42'  * Please check any AIDS CR DEVICES that you usually use f o r any of these a c t i v i t i e s : Raised Toilet Seat  Bathtub Bar  Bathtub Seat  Long-Handled Appliances f o r Reach  Jar Opener (for jars previously opened)  Long-Handled Appliances i n Bathroom Other (Specify:  ) 43  * Please check any categories far which you usually need HELP FROM ANOTHER PERSON: Hygiene  Gripping and Opening Things  Reach  Errands and Chores  44  231-  - 5 -  LEAVE THIS AREA BLANK  How many a r t h r i t i s and related v i s i t s did you make for routine check-ups? (That i s , the doctor suggested the v i s i t . ) Do not include v i s i t s while i n the hospital. In the past 4 months  47  How many a r t h r i t i s and related v i s i t s did you make for a specific problem? (That i s , you made the appointment without the suggestion of your doctor.) In the past 4 months  48  We are interested i n learning whether or not you are affected by pain because of your illness. Please mark an X on the line below to describe your a r t h r i t i s pain i n the recent past. Pain as bad as could be SEVERE  MOD  ERATE  SLIGHT  No -J Pain  45  Take a moment and think of the best possible l i f e and the worst possible l i f e . Now, on the line below, place an X to indicate where your l i f e i s now. Worst possible l i f e j_  Best possible life  We would l i k e to know how confident you are i n performing certain daily a c t i v i t i e s . For each of the following questions, please c i r c l e the number which corresponds to your certainty that you can perform the tasks as of now without assistive devices or help from another person. Please consider what you routinely can do, not what would require a single extraordinary effort. Here i s an example of the way someone might answer the question: EXAMPLE AS OF NOW,  HOW CERTAIN ARE YOU THAT YOU  CAN:  Dial a telephone i n 10 seconds:  6>  10 very uncertain  I 30  40  I 50 60 moderately certain  I 70  80  90  100 \very certain  This person i s uncertain that she could d i a l a telephone i n 10 seconds. Now, please answer the following questions using the same format....  153  232  - 6 AS OF NOW, 1.  HOW CERTAIN ARE YOU THAT YOU  I  I  I  10 20 very uncertain  20  10 very uncertain  I  I  I  I  70  80  90  80  90  I 100 very certain  83  I  30  I  40  I  I  50 60 moderately certain  I  70  100 very certain  84  I  30  40  I I I  50 60 moderately certain  70  80  I  90  I 100 very certain  85  I  30  I  40  I  I  50 60 moderately certain  70  80  90  100 very certain  86  Cut 2 bite-size pieces of meat with a knife and fork i n 8 seconds?  I  I  10 20 very uncertain 6.  I  50 60 moderately certain  Button and unbutton three medium-size buttons i n a row i n 12 seconds?  10 20 very uncertain 5.  I  40  Get out of an armless chair quickly without using your hands for support?  I  4.  I  30  Walk 10 steps downstairs i n 7 seconds?  I  3.  CAN:  Walk 100 feet on f l a t ground i n 20 seconds?  10 20 very uncertain 2.  LEAVE AREA BLANK  THIS  30  40  I  I  50 60 moderately certain  70  80  90  100 very certain  87  Turn an outdoor faucet a l l the way on and a l l the way off?  I I  10 20 very uncertain  I  30  I  40  I  I  50 60 moderately certain  70  80  I  90  I 100 very certain  88  233  - 7 AS OF NOW, HOW CERTAIN ARE YOU THAT YOUR CAN...  LEAVE THIS AREA BLANK  7. Scratch your upper back with both your right and l e f t hands?  I  I  10 20 very uncertain  I  30  I  40  I  I  50 60 moderately certain  70  80  I  90  100 very certain  89  Get i n and out of the passenger side of a car without assistance from another person and without physical aids?  I  20  10 very uncertain  I  30  I  40  I  I  50 60 moderately certain  70  I  80  90  100 very certain  90  Put on a long-sleeve front opening s h i r t or blouse (without buttoning) i n 8 seconds?  I  I  10 20 very uncertain  30  40  50 60 moderately certain  70  80  90  100 very certain  91  In the following questions we'd l i k e t o know how you f e e l about your a b i l i t y t o control your a r t h r i t i s . For each of the following questions please c i r c l e the number which corresponds with the certainty that you can now perform the following a c t i v i t i e s or tasks. 1.  How certain are you that you can control your fatigue?  I  I  10 20 very uncertain  30  40  50 60 moderately certain  70  80  90  100 very certain  92  How certain are you that you can regulate your activity so as t o be active without aggravating your arthritis?  10  I 20  very uncertain  30  40  I 50 moderately 60 certain  70  80  90  I 100 very certain  93  5  - 83.  234-'  Hew certain are you that you can do something to help yourself feel better i f you are feeling blue?  I  I  10 20 very uncertain  30  40  50 60 moderately certain  70  80  90  I 100 very certain  94  4. As compared to other people with a r t h r i t i s like yours how certain are you that you can manage a r t h r i t i s pain during your daily activities?  I  20  10 very uncertain 5.  30  40  50 60 moderately certain  70  I  80  90  I 100 very certain  How certain are you that you can manage your a r t h r i t i s symptoms so that you can do the things you enjoy doing?  I  I  10 20 very uncertain  I  30  I  40  I  I  50 60 moderately certain  70  80  90  100 very certain  6. How certain are you that you can deal with the frustration of arthritis?  I  I  10 20 very uncertain  I  30  I  40  I  I  50 60 moderately certain  70  80  90  100 very certain  In the following questions, we'd l i k e t o know how your a r t h r i t i s pain affects you. For each of the following questions please c i r c l e the number which corresponds t o your certainty that you can now perform the following tasks. 1. How certain are you that you can decrease your pain quite a bit?  10 20 very uncertain  30  40  I 50 60 moderately certain  I  70  80  90  100 very certain  95  _ 9 -  2.  235  LEAVE THIS AREA BLANK  How certain are you that you can continue most of your daily activities?  10 20 very uncertain  I  30  I  40  I  I  50 60 moderately certain  I  70  I  80  90  100 very certain  99  3. How certain are you that you can keep a r t h r i t i s pain from interfering with your sleep?  10 20 very uncertain  30  40  50 60 moderately certain  70  80  90  100 very certain  4. How certain are you that you can make a small-to-moderate reduction i n your a r t h r i t i s pain by using methods other than taking extra medication?  I  I  1  i  I  I  I  I I  I  I  I  10 20 30 40 50 60 70 80 90 100 very moderately very uncertain certain certain How certain are you that you can make a large reduction i n your a r t h r i t i s pain by using methods other than taking extra medication?  10 20 very uncertain  i  i  30  40  i  i  50 60 moderately certain  i  i  i  70  80  90  Please go on t o the next page..  i  100 very certain  |101  1102  - 10 Below i s a l i s t of some of the ways you may have f e l t indicate how often you have f e l t t h i s way during the PAST appropriate space. Rarely or Same or a none of l i t t l e of the time the time (less than (1-2 days) 1 day)  236  LEAVE THIS AREA BLANK  or behaved. Please WEEK by checking the Occasionally A l l of or a the time moderate amount of (5-7 days) time (3-4 days)  1.  I was bothered by things that usually don't bother me.  2.  I did not feel l i k e eating; my appetite was poor.  106_  3.  I f e l t that I could not shake off the blues even with the help from my family.  10 7_  4.  I f e l t that I was just as good as other people.  10 8_  5.  I had trouble keeping my mind on what I was doing.  109_  6.  I f e l t depressed.  110_  7.  I f e l t that everything I did was an effort.  111_  I f e l t hopeful about the future.  112_  I thought my l i f e had been a failure.  113_  8. 9.  10 5_  10. I f e l t fearful.  114_  11. My sleep was restless.  115_  12. I was happy.  116_  13. I talked less than usual.  117_  14. I f e l t lonely.  118_  15. People were unfriendly.  119_  16. I enjoyed l i f e .  120_  17. I had crying spells.  121_  18. I f e l t sad.  122_  19. I f e l t that people disliked me.  123  20. I could not get "going".  126  237- 11 -  LEAVE THIS AREA BLANK  What medications are you taking for your arthritis? (Please c i r c l e YES or NO FOR EACH GROUP.) Aspirin/Aspirirt-Like Product  1  0  YES  NO  127  YES  NO  128  Gold injections, Myodhrisine, Solgonal  YES  NO  Penicillamine, Cuprimine, Depen  YES  NO  YES  NO  Aspirin, Bufferin, Ascriptin, Anacin, Excedrin, Ecotrin, Empirin, T r i l i s a t e , Disalcid, other aspirin Tylenol, other acetaminophen Nonsteroidal. Anti-inflammatory Advil, Anaprox, Ansaid, Butazolidin, C l i n o r i l , Dolobid, Feldene, Ibuprofen, Indocin, Meclomen, Motrin, Naifon, Naprosyn, Nuprin, Orudis, Rufin, Tandearil, Tolectin, Tolmetin, Voltaren Tmmurosuppressive Agents Auronofin, Ridaura, (oral gold)  •  Plaquenil, HydroxycMoroquine  129_ 130_ 131  Chemotherapeutic Agents Imuran, Cytoxan, Azathioprine, Cyclophosphamide, Methotrexate  132 YES  NO  Steroids  133  Prednisone, Cortisone, Hydrocortisone, Decadron  YES  NO  Others for A r t h r i t i s Darvon, Darvocet, Codeine, Percodan, Percocet, Talwin, Dilaudid, Vicodin L i s t on the l i n e any others:  134_ YES YES  NO NO  135  THANK YOU ! PLEASE CHECK BACK TO MAKE SURE THAT ALL PAGES ARE COMPLETE coded PLEASE SHARE ANY ADDITIONAL THOUGHTS OR CONCERNS ON THE BACK OF THIS PAGE  checked entered  APPENDIX C  238  NATIONAL ARTHRITIS SELF-MANAGEMENT PROGRAM STUDY QUESTIONNAIRE  Questions 1 through 11 are to obtain demographic information on participants. 1.  Name:  2.  Street Address:  3.  City:  Province:  4. Telephone:  (home)  5. Gender (circle):  Postal Code:  .  Female  (work)  Birthdate:  Male  month  6.  Mother Tongue (circle one): English Other (specify):  7.  Please circle the HIGHEST year of school you have completed:  French  1 2 3 4 5 6 7 8 9 10 11 12 13,14  8.  day  year  German Italian Portuguese Chinese  15 16 17 18 19 20 21 22 above 22  Have you received a degree or diploma? (e.g., nurse, secretary, electrician) (circle):  No  Yes  If so, what is it?  9. What kind of arthritis do you have? fibromyalgia, etc.)  (e.g., osteoarthritis, rheumatoid arthritis,  :  10. Approximately, in what year did your arthritis begin: 11. Do you live (please check  /):  dalone?  • with parents?  dwith spouse?  \3with children?  • with other relatives?  \3with friends or roommates?  239 Question 12 is a single item question to get participants' perception of their health.  12. Please circle the word below which corresponds to how, in general, you would say your health is. Excellent  Very Good  Good  Fair  Poor  Scoring instructions: excellent - 0, very good - 1, good - 2, fair - 3, poor - 4.  Test Retest Reliability = .92. Source: Ware I E , Nelson EC.Sherboume C D , Stewart A L . Preliminary Tests of a 6-Item General Health Survey: A Patient Application. In Stewart A L , Ware JE (eds.). Measuring Functioning and Well Being: The Medical Outcomes Study Approach. Durham N C , Duke University Press, 1992.  Question 13 measures pain level.  13. W e are interested in learning whether or not you are affected by pain because of your illness. Please mark an X on the line below to describe your A R T H R I T I S P A I N i n the recent past. No Pain  0  1  10  8  Pain as bad as could be  Source: Downie WW, Letham PA, Rhind VM, Wright V, Branco JA, Anderson JA. Studies with Pain Rating Scales. Am, Rheum. Pis. 37:378-381, 1978. Dixon JS, Bird HA. Reproducibility Along a 10 cm Vertical Analogue Scale. Am. Rheum. Pis. 40:87-89, 1981.  Question 14 asks participants about their Quality of Life.  14. Take a moment and think of the best possible life and the worst possible life. N o w , on the line below, place an X to indicate where you life is now. Worst Possible Life Source: Cantril, H. (1965). The pattern of human concerns.  8  9  Best Possible 10 Life  New Brunswick, NJ, Rutgers University Press.  240 Question 15 through 22 asks participants about the medications they are taking.  What medications are you taking for your arthritis? (Please circle Yes or No for each group). 15. Oral gold (e.g., Auronfrn, Ridaura) or gold injections (e.g. Solgonal, Mychrysin)  No  Yes  16. Penicillamine (e.g., Cuprirnine, Depen)  No  Yes  17. Chloroquine (e.g., Plaquenil, Aralen)  No  Yes  No  Yes  No  Yes  .  No  Yes  Clinoril, Naprosyn, A S A , 292, Bufferin, Acetaminophen) .  No  Yes  18. Azathioprine (Imuran), Cyclophosphamide (e.g., Cytoxan, Priocytox), or Methotrexate 19. Prednisone (e.g., Deltasone, Winpred, Novoprednisone), Cortisone (Cortone), Dexamethasone (Decadron), or Hydrocortisone (tablets) 20. Darvon, Codeine, Percodan, Percocet, Talwin, Dilaudid 21. Aspirin, NSAIDs, or Tylenol (e.g., Voltaren, Indocid,  22.  Any  Others (please list these on the line below):  Questions 23 through 35 are from The Illness Intrusiveness Scale and asks participants the degree their illness interferes in various areas of their life.  The following items ask about how much your illness and/or its treatment interfere with different aspects of your life. Please circle the O N E number that best describes your current life situation. If an item in not applicable, please circle the number one (1) to indicate that this aspect of your life is not affected very much. Please do not leave any item unanswered. Thank you.  How much does your illness and/or its treatment interfere with your: 23.  HEALTH  (i.e., your physical well being)?  Not very much  1  2  3  4  5  6  7  Very much  241 24.  DUET (i.e., the things you eat and drink)?  25.  W O R K (i.e., the things you do at work or to maintain your home)?  Not very much 1  Not very much 1 26.  27.  2  2  3  4  5  6  7  Very much  3  4  5  6  7  Very much  ACTIVE RECREATION (i.e., sports)? Not very much 1 2 3  4  5  6  7  Very much  PASSIVE RECREATION (i.e., reading, listening to music)?  Not very much 1 28. FINANCIAL SITUATION? Not very much 1  2  3  4  5  6  7  Very much  2  3  4  5  6  7  Very much  29. RELATIONSHIP WITH YOUR SPOUSE (or with your girlfriend or boyfriend)? Not very much 1 2 3 4 5 6 7 Very much  30. SEX LIFE? Not very much 1  2  31. FAMILY RELATIONS? Not very much 1  2  7  3  4  32. OTHER SOCIAL RELATIONS (i.e.. friends)? Not very much 1 2 3 4  5  6  Very much  7 7  5  Very much Very much  33. SELF EXPRESSION/SELF IMPROVEMENT (i.e., hobbies, courses)? Not very much 1 2 3 4 5 6 7 Very much 34.  RELIGIOUS EXPRESSION (i.e., beliefs, going to church)?  Not very much 1  2  3  4  5  6  7  Very much  35. COMMUNITY AND CIVIC INVOLVEMENT (i.e., participation in events)? Not very much 1 2 3 4 5 6 7 Very much  Scoring instructions: This scale has 1 3 separate domains which are calculated separately, and then the total mean score for all 1 3 items is calculated. Source: Devins, G.M., Binik, Y.M., Hutchinson, T.A., Hollomby, D.J., Barre, P.E., 8 i Guttmann, R.D. ( 1 9 8 3 ) . The emotional impact of end-stage renal disease: Importance of patients' perceptions of intrusiveness and control. International Journal of Psychiatry in Medicine. 1 3 , 3 2 7 - 3 4 3 .  242' , . Questions 36-41 represents the "Other Symptoms'' subscale from the Arthritis Self-Efficacy Questionnaire. 42-46 is the "Pain" subscale.  Questions  How certain are you at the present time that you can: Totally  (circle number)  uncertain 36. Control your fatigue?  I  I  I  I  I  I  I  I  Totally  certain I  I  I  0 10 20 30 40 50 60 70 80 90 100 37. Regulate your activity so as to be active | | | | | | | | | | | without aggravating your arthritis? 0 10 20 30 40 50 60 70 80 90 100 38. Do something to help yourself feel better if you are feeling blue?  I I I I I I I I I I I 0 10 20 30 40 50 60 70 80 90 100  39. Manage arthritis pain during your daily activities (compared to other people with arthritis like yours)?  I I I I I I I I I I I 0 10 20 30 40 50 60 70 80 90 100  40. Manage your arthritis symptoms so that | | | j | | | | | | | you can do the things you enjoy doing? 0 10 20 30 40 50 60 70 80 90 100  41. Deal with the frustration of arthritis?  I  I  I  I  I  I  I  I  I  I  |  0 10 20 30 40 50 60 70 80 90 100 42. Decrease your pain quite a bit?  I  I  I  I  I  I  I  I  I  I  I  0 10 20 30 40 50 60 70 80 90 100 43. Continue most of your daily activities?  I  I  I  I  I  I  I  I  j  I  I  0 10 20 30 40 50 60 70 80 90 100 44. Keep arthritis pain from interfering with your sleep?  I I I I I I I I I I I 0 10 20 30 40 50 60 70 80 90 100  45. Make a small-to-moderate reduction I I I I I I I I I I I in your arthritis pain by using 0 10 20 30 40 50 60 70 80 90 100 methods other than taking extra medication? 46. Make a large reduction in your arthritis pain by using methods other than taking extra medication?  I I I I I I I I I I S 0 10 20 30 40 50 60 70 80 90 100  243:Scoring instructions: Each subscale is scored separately by taking the mean of the subscale items. If one-fourth or less of the data are missing, the score is a mean of the completed data. If more than one-fourth of the data are missing, no score is calculated. Source: Lorig, K., Brown, B. W. Ung, E., Chastain, R., Shoor, S., & Holman, H.R. (1989). Development and evaluation of a scale to measure perceived self-efficacy in people with arthritis. Arthritis and Rheumatism, 32, 37-44. (  Questions 47 through 64 measure Depression, Mental Health, Energy/Fatigue, and Positive Affect.  These questions are about how you feel and how tilings have been with you during T H E PAST M O N T H . (For each question, please circle one number for each question that comes clo to the way you have been feeling). How much time during T H E PAST FOUR WEEKS.... None of the time  A little of the time  Some A good Most of the bit of of the time the time time  All of the time  Questions 48-55 measure Depression.  48.  0  1  2  3  4  5  0  1  2  3  4  5  0  1  2  3  4  5  51. Have you felt emotionally stable  0  1  2  3  4  5  52. Have you felt downhearted and blue  0  1  2  3  4  5  things  0  1  2  3  4  5  Have you been in low or very low spirits  0  1  2  3  4  5  0  1  2  3  4  5  Did you feel depressed  49. Have you been in firm control of your behaviour, thoughts, emotions, & feelings 50. Did you feel that you had nothing to look forward to  53. Have you been moody or brooded about  54.  55. Have you felt so down in the dumps that nothing could cheer you up * This is the MOS Depression II Scale. Scoring instructions:  '  Reverse score items 49 and 5 1 , then take the mean of the 8 hems. Lower score is better (Le.» less depression).  .'  {-•':;'  ;  «3P Consistency Reliability - . 9 1 . Test-Retest Reliability - .82.  Item-Scale Correlations - 52 -.82.  244 Questions 47, 52, 55, 61, and 62 measure Mental Health.  47. Have you been a very nervous person  0  1  2  3  4  5  52. Have you felt downhearted and blue  0  1  2  3  4  5  nothing could cheer you up  0  1  2  3  4  5  61. Have you felt calm and peaceful  0  1  2  3  4  5  62. Have you been a happy person  0  1  2  3  4  5  55. Have you felt so down in the dumps that  * This is the MOS Mental Health Index III (MHI5) Scale. Scoring instructions: Reverse scores for questions 47, 52 and 55 then take trie mean of the 5 items. Higher score is better (i.e., better mental health).  Questions 56 through 60 measure Energy/Fatigue. 56. Did you feel worn out 57. Did you have a lot of energy  0  1  2  3  4  5  58. Did you feel tired  0  1  2  3  4  5  0  1  2  3  4  5  0  1  2  3  4  5  59. Did you have enough energy to do the things you wanted to do 60. Did you feel full of pep * This is the MOS Energy/Fatigue Scale. Scoring instructions:  Reverse scores for 56 and 58, then take the mean of the 5 items. Higher score is better (i.e., more energy and/or less fatigue).  Consistency Reliability - .91. Test-Retest Reliability - .85. Item-Scale Correlations .72 - .80.  245, Questions 61 through 64 measure Positive Affect.  61. Have you felt calm and peaceful  0  62. Have you been a happy person  0  63. Has your daily life been full of things that were interesting to you  0  1  2  3  4  5  64. Have you felt cheerful, lighthearted  0  1  2  3  4  5  * This is the MOS Positive Affect II Scale. Scoring instructions:  Scale is the mean of these 4 items. Higher score is better. ts' Consistency Reliability - .86. Test-Retest Reliability - .81. Item-Scale Correlations - .64 -.78. Source:  Ware JE, Nelson ECSherbourne CD, Stewart AL. Preliminary Tests of a 6-ltem General Health Survey: A Patient Application. In Atewart AL, Ware JE (eds.). Measuring Functioning and Well Being: The Medical Outcomes Study Approach. Durham NC, Duke University Press, 1992.  Questions 65 through 72 measure Physical Limitations.  Please check (/) the ONE best answer for the questions below: At THIS M O M E N T are you able to: Dressing and Grooming 65. Dress yourself, including tying shoes and doing buttons? Arising  66. Get in and out of bed? Eating 67. Lift a full cup or glass to you mouth? Walking 68. Walk outdoors on flat ground?  Without ANY With SOME difficulty difficulty  With MUCH UNABLE difficulty to do  246 Hygiene 69. Wash and dry your entire body? Reaching 70. Bend down to pick up clothing from off the floor? Gripping 71. Turn faucets on and off? Activities 72. Get in and out of a car? Source: Fries, J. F., Spitz, P. W., & Young, D. Y. (1982). The dimensions of health outcomes: the health assessment questionnaire, disability and pain scales. Journal of Rheumatology. 9, 789-793. Fries, J. F., Spitz, P. W., Kraines, R. G., & Holman, H. R. (1980). Measurements of patient outcomes in arthritis. Arthritis & Rheumatism, 23(2). 137-145. Scoring instructions: Without Any Difficulty - 0, With Some Difficulty - 1, With Much Difficulty - 2, Unable to do - 4 Each item is scored independently. Then the 8 means can be used to calculate an overall means.  Questions 73 through 76 ask participants about stretching or strengthening exercises (73), walking (74), stress management (75 & 76). DURING T H E PAST W E E K (even if it was not a typical week), how much T O T A L time (for the entire week) did you spend on each of the following? (Please circle one number for each question). less than none 30 min/wk  30-60 1-3 hrs min/wk per week  more than 3 hrs/wk  73. Stretching or strengthening exercises (range of motion, using weights, etc.)  0  1  2  3  4  74. Walk for exercise  0  1  2  3  4  75. IN T H E PAST W E E K (even if it was not a typical week), how many times did you do mental stress management or relaxation techniques? None 76. If "yes" to question # 7 5 , what did you do to relax?  Times  247 ; Questions 77 through 79 ask participants about doctor visits (i.e., routine checkups (77), specific problem visits (78), and visits to Rehabilitation (79).  In the past FOUR MONTHS, how many arthritis and related visits did you make for: 77. Routine check-ups (ones which your doctor suggested you have)!  visits  78. A specific problem which you needed to see a doctor for?  visits  79. Rehabilitation, physiotherapy, or occupational therapy?  visits  Please check back to make sure ALL the questions are answered.  THANK YOU!  APPENDIX D B E H A V I O U R S IDENTIFIED I N N O M I N A L G R O U P P R O C E S S  Avoids behaviours that cause pain Takes rest when needed Organizes house to make things easier Asks for assistance Takes medications as instructed Makes social contact Exercises Eats a healthy diet "Sloughs off" once in a while Seeks out fun/humour Involves self in interesting/enjoyable activities Helps others Gets a good night sleep Good communication with health care professional Seeks information Tries out new strategies suggested by others Participates in groups and activities Receives physiotherapy Uses available transportation Communicates well with family and friends  APPENDIX  1  Z Z  z  t  < Z  o H  O  o cX  r-. f =3 3 .2 =  H  < UJ  K  tu  O UJ  Q O  s Q UJ UJ  U  o a CM  UJ  O UJ  U  UJ  v  2  E  249  

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.831.1-0076940/manifest

Comment

Related Items