Understanding the Determinants of Health for People With Type 2 ...

 RESEARCH AND PRACTICE 

Understanding the Determinants of Health for People With Type 2 Diabetes | Sheri L. Maddigan, PhD, David H. Feeny, PhD, Sumit R. Majumdar, MD, MPH, FRCPC, Karen B. Farris, PhD, and Jeffrey A. Johnson, PhD

The health of a population is determined by a large number of factors. Specific medical conditions, such as diabetes, have a significant impact on health status. Diabetes affects approximately 5% of Canadians aged 20 years and older, and the prevalence increases with age.1 Individuals aged 65 years and older account for 50% of diabetes cases but represent only 15% of the population.1 Ninety percent of all diagnosed cases of diabetes are type 2 diabetes, which is associated with a substantial physical and emotional burden for individuals who have the disease and their families.2–4 The health-related quality of life (HRQL) deficits reported by people with type 2 diabetes are generally attributed to the disease itself, its restrictive treatment regimens, and its associated comorbidities.2–4 However, HRQL deficits associated with type 2 diabetes may be better explained in the context of a more holistic determinants of health framework, because population health is not solely associated with disease and treatment.5,6 The general approach to studying factors associated with HRQL and type 2 diabetes, however, has tended to focus on demographic and clinical factors.7–11 There has been less emphasis on individual lifestyle factors (e.g., stress), the social environment (e.g., social integration), and access to health care. Previous research has shown that demographic characteristics and clinical factors (e.g., complications and comorbidities, duration of diabetes, and insulin use) have an impact on HRQL associated with type 2 diabetes, and some heterogeneities of HRQL and the disease can be explained by these factors.10 This is not surprising because a number of these variables are determinants of population health.6 There are many frameworks that conceptualize the determinants of health and their causal associations, but because of the complexity of the specified associations, it is often difficult to use such frameworks analytically. Hertzman et al.12 proposed a relatively simple

Objective. We assessed which of a broad range of determinants of health are most strongly associated with health-related quality of life (HRQL) among people with type 2 diabetes. Methods. Our analysis included respondents from the Canadian Community Health Survey Cycle 1.1 (2000–2001) who were aged 18 years and older and who were identified as having type 2 diabetes. We used regression analyses to assess the associations between the Health Utilities Index Mark 3 and determinants of health. Results. Comorbidities had the largest impact on HRQL, with stroke (–0.11; 95% confidence interval [CI] = –0.17, −0.06) and depression (–0.11; 95% CI = –0.15, −0.06) being associated with the largest deficits. Large differences in HRQL were observed for 2 markers of socioeconomic status: social assistance (–0.07; 95% CI = –0.12, −0.03) and food insecurity (–0.07; 95% CI = –0.10, −0.04). Stress, physical activity, and sense of belonging also were important determinants. Overall, 36% of the variance in the Health Utilities Index Mark 3 was explained. Conclusion. Social and environmental factors are important, but comorbidities have the largest impact on HRQL among people with type 2 diabetes. (Am J Public Health. 2006;96:1649–1655. doi:10.2105/AJPH.2005.067728)

conceptual scheme for organizing and analyzing the relative importance of individual-level determinants of health on the basis of a commonly used population health framework.6 They categorized the determinants of health into 3 domains for analysis: stage of life cycle, subpopulation partitions, and sources of heterogeneity. The objectives of our analysis were to (1) assess the magnitude of HRQL deficits associated with determinants of health and type 2 diabetes, and (2) assess the contribution of Hertzman et al.’s domains to explaining the variance in HRQL associated with type 2 diabetes.

forces, and some remote areas of the country; however, it still represents approximately 98% of the Canadian population aged 12 years and older.13 We used a multistage stratified cluster design combined with random sampling methods to select the sample.13 Proxy reporting was permitted; however, certain components of the interview were only appropriate for self-response.14 Data for Cycle 1.1 were collected between September 2000 and November 2001; 131 535 respondents were surveyed, and the overall response rate was 84.7%.14

Sample

METHODS Survey Design We used data from the Canadian Community Health Survey (CCHS) Cycle 1.1, which is a cross-sectional survey of the Canadian population aged 12 years and older. Data about use of health services, determinants of health, and health status are collected.13 The survey excludes individuals who live on crown and reserve land or in institutions, individuals who are members of the Canadian armed

September 2006, Vol 96, No. 9 | American Journal of Public Health

Approximately 6361 respondents reported a diagnosis of diabetes, which represented a weighted percentage of 4.1% of the Canadian population. An algorithm was developed on the basis of age, treatment regimen, duration of time from initial diagnosis to initiation of insulin therapy, and age at diagnosis to categorize respondents as having type 1 (9.9%) or type 2 diabetes (90.1%). Fifty-four respondents could not be categorized because of missing data. The study was restricted to respondents aged 18 years and older who were

Maddigan et al. | Peer Reviewed | Research and Practice | 1649


categorized as having type 2 diabetes (n = 5497); 83.1% had complete data and were included in our analysis (n = 4678). The majority of excluded respondents were missing data on determinants of health.

Health Utilities Index Mark 3 HRQL was assessed with the HUI3, a preference-based measure of HRQL.15 The HUI3 was administered as a 31-item questionnaire that assessed usual health status. HUI3 health states are defined by a classification system that includes a set of HRQL attributes, with 5 to 6 levels of functioning for each attribute. Eight attributes are included in the Health Utilities Index Mark 3 (HUI3) system: vision, hearing, speech, ambulation, dexterity, emotion, cognition, and pain and discomfort. A utility function that ranges from −0.36 to 1.0 is used to obtain overall scores for health states (–0.36 = worst possible health, 0.0 = dead, and 1.0 = perfect health).15 Differences of greater than 0.03 for HUI3 overall scores are considered to be clinically important.16 This value was derived from a variety of types of evidence; in part, the clinically important difference is based on the premise that a change in 1 level of functioning for any of the 8 attributes is qualitatively important. As such, 0.03 represents the smallest change in an overall score that results from a 1-level change in functioning for 1 attribute (e.g., the difference in overall score between having level 1 and level 2 functioning for the vision attribute). Cross-sectional comparisons between groups of individuals who have diabetes and who are known to differ in their levels of HRQL also support this value as clinically important.3,17,18

Determinants of Health Determinants of health were selected in accordance with the 3 domains of the conceptual scheme.12 The stage of life cycle domain reflects the fact that age, in part, determines vulnerabilities or susceptibility to disease. For example, individuals are most susceptible to chronic disease between the ages of 45 to 75 years.12 With type 2 diabetes, however, the lower age boundary for the chronic disease stage becomes less relevant. All individuals with type 2 diabetes

have already developed a chronic disease, and they frequently have additional comorbidities and complications.7,19 For our analysis, stage of life cycle was represented by age, chronic medical conditions, and duration of diabetes. Of the 25 chronic conditions included in the CCHS, stroke, heart disease, and osteoarthritis were of particular interest because they are consistent with the chronic disease stage.12 Moreover, heart disease and stroke are 2 common macrovascular comorbidities associated with diabetes7,19 that also are associated with large HRQL deficits.7,20 Furthermore, it is relatively common for individuals with type 2 diabetes to also have osteoarthritis, because both conditions are associated with aging and obesity.21 Depression is a relevant comorbidity because depression has been associated with additional HRQL deficits that are associated with diabetes.22 Diagnoses of heart disease, stroke, and osteoarthritis were self-reported. Respondents whose predicted probability for major depression exceeded 0.90 on the Composite International Diagnostic Interview Short Form for Major Depression were considered to have depression.23 Subpopulation partitions are segments of the population across which heterogeneities in health status are observed, including gender, socioeconomic status, geographic location, race/ethnicity, and special populations12 (groups not defined by 1 of the other subpopulation partitions but who share a particular characteristic that is associated with patterns in health status). Individuals with type 2 diabetes who use insulin can be considered a special population because they generally have HRQL deficits compared with those who do not use insulin.8,24 The CCHS provided data on socioeconomic status, gender, race (Aboriginal or non-Aboriginal), geographic location (rural or urban), and insulin use (users or nonusers). Because data on income was missing, we used other markers of socioeconomic status: education, social assistance, and food insecurity. Food insecurity was determined by 3 questions about financial access to a sufficient quantity and quality of food; respondents were considered to have food insecurity if, for financial reasons, they did not have enough food

1650 | Research and Practice | Peer Reviewed | Maddigan et al.

to eat or did not eat the quality or variety of foods desired. Sources of heterogeneity are mechanisms that operate across subpopulation partitions and stage of life cycle. Sources of heterogeneity include individual lifestyle factors, physical and social environment, genetic endowment, and differential access to health care. Individual lifestyle factors include body mass index (BMI), smoking status, alcohol consumption, physical activity, and ability to cope. Level of physical activity was determined by energy expenditure, a variable derived from 47 questions about participation in activities. Energy expenditure was derived from metabolic equivalent level, frequency, and time per session of each physical activity.14 Life stress was used to determine ability to cope and was measured with a 5-point Likert scale. Exposure to environmental tobacco smoke by a family member who smoked inside the home was used to assess the physical environment, and sense of belonging to the community (measured with a 4-point Likert scale) and marital status were used to assess the social environment. Access to medical care was determined by selfperceived unmet healthcare needs and whether respondents had a regular medical doctor.

Analysis We used multiple regression analysis to assess the clinical importance (i.e., magnitude of the unstandardized regression coefficients) and the statistical significance of each determinant of health in a single overall model. Normalized sampling weights14 and bootstrap variance estimates25 were used to account for the multistage stratified cluster design. The statistical significance of each determinant of health was derived from the 95% confidence intervals (on the basis of the bootstrap variance estimate) of the regression coefficients. Collinearity in the regression model was assessed from the tolerance of each independent variable and was not problematic.26 All analyses were conducted with SPSS Version 12.0 software (SPSS Inc, Chicago, Ill) and with a bootstrap algorithm designed by Statistics Canada. Previous research that used the HUI3 detected a nonlinear association between age

American Journal of Public Health | September 2006, Vol 96, No. 9


and HRQL among the Canadian population3,27; thus, age was modeled in a meancentered quadratic form28 (i.e., age was analyzed as a deviation from its mean and the square of this variable), which reduced collinearity between age and its square.28 The total number of self-reported medical conditions other than heart disease, stroke, depression, and osteoarthritis also were included. Duration of diabetes was analyzed in quartiles. For life stress, response options were collapsed to create 3 categories: not at all stressful, not very stressful/a bit stressful, and quite a bit stressful/extremely stressful. For BMI, respondents were categorized as obese (BMI ≥ 30) or not obese (BMI < 30).29 For marital status, respondents were categorized as married/partnership or not married. Alcohol consumption was analyzed as a dichotomous variable (heavy drinker vs not). Respondents who consumed 5 or more drinks during 1 occasion more than once a month were considered heavy drinkers.30 Respondents’ physical activity level was categorized as inactive, moderately active, or active in accordance with guidelines used in previous health surveys.14,31 The proportion of explained HRQL variance (R 2 ) within each domain was then assessed with 3 regression models that each contained the determinants for their respective domains (stage of life cycle, subpopulation partitions, and Sources of Heterogeneities). To determine the unique contribution to the explained variance of a particular domain, the R 2 change was calculated between models that had the 2 domains and models that had all 3 domains. We performed analyses of the cases with complete data; however, some of this data had been imputed by Statistics Canada during data processing.13 The demographic characteristics of respondents who were excluded from the analysis because of missing HUI3 scores (n = 76) were compared with respondents included in the analysis sample (n = 4678). The overall HUI3 scores of the analysis sample (n = 4678) were compared with those of respondents who were excluded from the analysis because of missing data on determinants of health (n = 819). Chi-square and t tests were used, when appropriate, for these analyses.

RESULTS The average age of respondents included in our analysis was 61.6 years (SD = 13.3), and the average duration of diabetes was 9.3 years (SD = 9.8) (Table 1). Attaining less than high-school education was relatively common (42.4%), and physical inactivity (64.6%) and

obesity (36.4%) were prevalent among this sample. The HUI3 scores of respondents who were excluded from the analyses because of missing data on determinants of health (n = 819) were significantly lower than the overall HUI3 scores of respondents who had complete data (n = 4678) (mean difference = –0.14; 95% confidence interval [CI] = –0.17,

TABLE 1—Demographic Characteristics of Sample, by Domain (N = 4678) Determinants Stage of life cycle Age, mean (SD) Duration of diabetes, m (SD) Median (interquartile range) Number of medical conditions,a mean (SD) Has osteoarthritis, % yes Suffers the effects of stroke, % yes Has heart disease, % yes Predicted probability of depression > 0.90, %

61.6 (13.3) 9.3 (9.8) 6.0 (2.0, 13.0) 2.7 (1.7) 19.4 4.8 20.6 7.2 Subpopulation partitions

Gender, % male Level of education, % Less than secondary Secondary graduation Some postsecondary, college, trade school University degree Aboriginal status, % yes Some food insecurity, % yes Social assistance, % yes Rural geographic location, % rural Insulin use, % yes

51.7 42.4 16.3 29.3 12.0 1.8 15.5 7.5 19.3 15.5 Sources of heterogeneity

Current smoking status, % current smoker Heavy drinkers, % yes Body mass index > 30.0, % yes Physical activity index Active Moderately active Inactive Life stress, % Not at all stressful Not very stressful/a bit stressful Quite a bit/extremely stressful Family member smokes inside house, % yes Marital status, % married Sense of belonging to the community, % Strong Somewhat strong Somewhat weak Weak Regular medical doctor, % yes Self-perceived unmet healthcare needs, % yes Health Utilities Index Mark 3, mean (SD)

19.0 6.9 36.4 14.5 20.9 64.6 21.1 55.9 23.0 24.5 67.7 22.5 38.1 24.1 15.3 96.0 12.7 0.78 (0.26)

Note. Domains refer to the categories of determinants of health proposed by Hertzman et al.12 a Other than stroke, heart disease, osteoarthritis, or depression.




TABLE 2—Unstandardized Regression Coefficients for Each Domain Alone and Overall Model (N = 4678) Model 1: Stage of Life Cycle

Model 2: Subpopulation Partitions

Model 3: Sources of Heterogeneity

Model 4: Overall Model

Stage of life cycle Agea Age Age2 Osteoarthritis Stroke Heart disease Depression Number of medical conditionsb Duration of diabetes 13 y Level of education Less than secondary Secondary graduation Some postsecondary, college, trade school University degree Food insecurity Social assistance Aboriginal status Rural geographic location Male Insulin use

–0.001* –0.0001* –0.08* –0.14* –0.07* –0.16* –0.05*

–0.003* –0.0001* –0.06* –0.11* –0.05* –0.11* –0.04*

0.05* 0.07* 0.02* Reference Subpopulation partitions

0.03* 0.04* 0.01 Reference

–0.09* –0.03 –0.04* Reference –0.13* –0.12* 0.02 0.00 0.03* –0.08* Sources of heterogeneity

–0.04* –0.03 –0.02 Reference –0.08* –0.07* –0.01 0.00 –0.01 –0.04*

Current smoker Heavy drinker Body mass index > 30 Physical activity index Active Moderately active Inactivec Life stress Not at all stressful Not very stressful/a bit stressful Quite a bit/extremely stressful Family member smokes inside house Married Sense of belonging to the community Strong Somewhat strong Somewhat weak Weakd

–0.01 0.04* –0.04* 0.08* 0.09* Reference

–0.02 –0.11* –0.01 0.06* 0.11* 0.11* 0.07* Reference

–0.02 –0.01 –0.03* 0.06* 0.06* Reference

–0.02 –0.08* 0.01 0.00 0.09* 0.08* 0.05* Reference Continued


−0.10; P < .05). There were some demographic differences between the respondents excluded because of missing HUI3 scores (n = 76) and the analysis sample (n = 4678). Across all determinants of health included in the model, stroke and depression were associated with the largest HRQL deficits and were nearly 4 times the clinically important difference of 0.03 (Table 2, Model 4). Each additional medical condition was associated with a clinically important deficit (–0.04). As hypothesized, a nonlinear association between age and HRQL was detected. Socioeconomic status (food insecurity, social assistance, and failure to complete a secondary education) and insulin use also were associated with clinically important HRQL deficits (Table 2, Model 4). Within the sources of heterogeneity, the largest differences between respondents were sense of belonging to the community, life stress, and self-perceived unmet healthcare needs. Approximately 36% of the variance in HRQL was explained by the determinants of health included in the model. The stage of life cycle variables explained the most variance (27%) among the 3 domains individually (Table 2, Models 1–3). Taken together as a full model (Table 2, Model 4), the largest unique contribution to the explained variance in HRQL was that of the stage of life cycle variables (15%), followed by sources of heterogeneity (6%) and subpopulation partitions (2%) (Table 2).

DISCUSSION Type 2 diabetes is a chronic medical condition in which many factors potentially influence HRQL and health status. Some of these factors are disease related, but many are associated with demographic characteristics, social factors, and health behaviors. We used population-based data from the CCHS to construct a model of multiple determinants of health associated with type 2 diabetes. It was evident that 2 of the comorbidities of interest— stroke and depression—were associated with the largest HRQL deficits (–0.11 for each), even after socioeconomic and behavioral determinants of health were included in the model. The clinically important deficits associated with comorbidities suggest that prevention and management of comorbidities could



TABLE 2—Continued Regular medical doctor Self-perceived unmet healthcare needs Variance explained by domain alone Unique variance explained by domaine

R 2adj = 0.27 R 2 = 0.15

R 2adj = 0.11 R 2 = 0.02

–0.06* –0.14* R 2adj = 0.16 R 2 = 0.06

–0.01 –0.08* R 2adj = 0.36

a

Age was modeled as a mean-centered quadratic (b1[age – 61.55] + b2[age – 61.55]2). Other than stroke, heart disease, osteoarthritis, or depression. c Inactive was chosen as the reference category for physical activity to allow for comparisons between inactive and moderately active and active. d Weak sense of belonging was chosen as the reference category for sense of belonging so that the gradient between categories could be observed. e For example, R 2 change when stage of life cycle variables were added to a model that included subpopulation partitions and sources of heterogeneity. *P < .05 based on bootstrap variance estimate. b

be vital to preserving or improving HRQL among people with type 2 diabetes. From both the clinical and broader health policy perspectives, efforts at primary and secondary prevention of heart disease and stroke could have a significant impact on HRQL among people with type 2 diabetes. Identifying individuals with depression and providing appropriate treatment is important because of the magnitude of the deficit associated with this comorbidity and diabetes. Across the subpopulation partitions domain, the largest deficits were associated with the 2 markers of income: social assistance (–0.07) and food insecurity (–0.08). Income and social status have been recognized as 2 of the most important determinants of health for the Canadian population.5,6 It was interesting to note, however, that the 2 markers appeared to capture somewhat different phenomena, because they were both independently associated with clinically important deficits that reached statistical significance. Although social assistance may have captured respondents with low income, as intended, it is possible that food insecurity reflected, in addition to low socioeconomic status, the impact of poor nutrition on both diabetes and overall health. Education also was associated with health status, which confirmed that it is an important determinant independent of its association with income.6 It has been suggested that the association between education and HRQL is in part attributable to the association between higher levels of education and healthier lifestyles, including not smoking, higher levels

of physical activity, and better access to healthy foods.5 Despite controls for a number of these factors in our analysis, the association between education and HRQL persisted. It has been suggested that education may also influence diabetes-related knowledge, the ability to communicate with healthcare providers, treatment choices, and the ability to adhere to complex self-care regimens, which in turn affect HRQL.32 One of the largest differences in HRQL was observed between respondents who reported a weak sense of belonging to the local community and respondents who reported a strong or somewhat strong sense of belonging. Sense of belonging was previously found to be associated with self-rated health among the Canadian population,33 when Ross considered sense of belonging to be a measure of social capital, despite the fact that it was evaluated as an individual-level variable (as it was in our study). Perhaps when analyzed at the individual level, sense of belonging may better reflect social integration (i.e., a measure of the degree to which individuals are socially isolated).34 Life stress was another Source of Heterogeneity where clinically important differences between respondents were observed. Respondents who felt their lives were quite a bit stressful or extremely stressful reported HRQL deficits nearly 3 times the clinically important difference compared with those who felt their lives were not at all stressful. Stress is recognized as a determinant of health for the general population, but high levels of stress may be particularly problematic for individuals with diabetes because


stress is associated with poor glycemic control.35,36 This might explain the magnitude of the deficit associated with high stress levels that we observed. Of the 3 domains in the conceptual scheme,12 stage of life cycle variables accounted for the largest proportion of the variance in HRQL. We represented this domain in a manner somewhat specific to diabetes, because we included the comorbidities that are more frequently associated with diabetes and the duration of diabetes. Thus, it is not clear if stage of life cycle would be the dominant domain for the general population or for other chronic diseases. Furthermore, the use of cross-sectional data precluded us from definitively stating that stage of life cycle variables were the most important determinants of health associated with diabetes, because we were unable to assess causal associations among variables. For example, we could not capture the fact that Aboriginals, individuals who smoke, and individuals who are sedentary may be more likely to develop comorbidities, such as heart disease.37,38 Thus, the full explanatory power of the subpopulation partitions and sources of heterogeneity domains may not have been captured in our analysis. A number of limitations should be noted in this analysis. The algorithm used to distinguish between respondents with type 1 and type 2 diabetes has not been previously validated. A number of the criteria in the algorithm have been used and validated previously,39–41 but the algorithm as a whole has not. Misclassification of individuals’ type of diabetes could affect both the internal and external validity of the results if determinants of health differ between the 2 diseases. However, we are reasonably confident that the algorithm accurately classified respondents, because categorization of type 1 and type 2 diabetes produced by the algorithm was 10% and 90%, which is generally recognized as the distribution of type 1 and type 2 diabetes in Canada.42 Another potential limitation is the accuracy of self-reported data on a number of determinants of health, including medical conditions, alcohol use, BMI, duration of diabetes, and level of physical activity. Although the questions about medical conditions specified that the condition must have been diagnosed by a



health professional, there remained potential for individuals to overreport or underreport any of the medical conditions. BMI may have been subjected to bias because it was derived from self-reported height and weight. As with any health behavior, questions about alcohol consumption and physical activity may have been answered in a socially desirable manner. Approximately 17% of the respondents were missing data, which lead to their exclusion from our analysis. Certain covariates (sense of belonging, physical activity, and depression) accounted for the majority of missing data. Respondents who were missing data on covariates tended to have a worse HRQL. Demographic differences also existed between respondents who did and did not have complete HUI3 scores. Regardless, we chose to include all of the variables we had originally proposed, because the intent of our analysis was to represent Hertzman et al.’s conceptual scheme as fully as possible.12 We used imputed data when it was supplied by Statistics Canada, but we did not impute other missing elements. It is not possible to determine the impact missing data had on our analysis, but it may limit the generalizability of the results. The strengths of our analysis also should be noted. We used a large population-based sample of Canadians to model the determinants of health associated with type 2 diabetes. Furthermore, we used a broad framework rather than limit our analysis to medical-related and disease-related factors. Without this framework, we may not have included a number of variables that were associated with clinically important HRQL deficits, particularly sense of belonging and life stress. Moreover, by including multiple domains of determinants of health, we were able to confirm that previously detected associations, such as those seen between HRQL and comorbidities and HRQL and insulin use, were not merely confounded by socioeconomic or behavioral factors. A final strength of our analysis was the use of the HUI3 as the measure of HRQL, a measure that we have previously shown to have construct validity for this population.18 Overall, our analysis confirmed that many of the determinants of health specified in Hertzman et al.’s conceptual scheme12 were indeed

important for type 2 diabetes. Clinically important heterogeneities in HRQL were associated with stroke and depression, which emphasizes the importance of preventing and managing comorbidities and complications associated with type 2 diabetes. Social and behavioral determinants of health (socioeconomic status, life stress, and sense of belonging) also were important and showed that factors other than medical factors have an impact on the health of individuals with type 2 diabetes. A better understanding of the broader determinants of health is the necessary first step that will permit the development of interventions and policies for improving HRQL among this vulnerable population.

About the Authors Sheri L. Maddigan, David H. Feeny, Sumit R. Majumdar, and Jeffrey A. Johnson are with the Institute of Health Economics, Edmonton, Alberta. David H. Feeny and Jeffrey A. Johnson are also with the Department of Public Health Sciences, University of Alberta, Edmonton. David H. Feeny also is with the Department of Economics, University of Alberta, and Health Utilities Inc, Edmonton. Sumit R. Majumdar is with the Department of Medicine, University of Alberta. Karen B. Farris is with the College of Pharmacy, University of Iowa, Iowa City. Requests for reprints should be sent Jeffrey A. Johnson, PhD, Institute of Health Economics, #1200 – 10 405 Jasper Ave, Edmonton, Alberta, Canada T5J 3N4 (e-mail: [email protected]). This article was accepted October 21, 2005.

Contributors All authors originated the study, interpreted findings, and reviewed drafts of the article. S. L. Maddigan completed the analyses and led the writing.

Acknowledgments We would like to thank Statistics Canada for allowing us access to data through the Research Data Center at the University of Alberta. We thank the University of Alberta for providing facilities to access the data and Irene Wong for her assistance with this project. Research and analysis were based on data from Statistics Canada. Note. The opinions expressed here do not necessarily represent the views of Statistics Canada.

Human Participant Protection This study was approved by the health research ethics board of the University of Alberta.

References 1. Health Canada. Responding to the Challenge of Diabetes in Canada. First Report of the National Diabetes Surveillence System (NDSS). Ottawa, Ontario: Health Canada; 2003. 2. Ahroni JH, Boyko EJ. Responsiveness of the SF36 among veterans with diabetes mellitus. J Diabetes Complications. 2000;14:31–39.


3. Maddigan SL, Feeny DH, Johnson JA. Healthrelated quality of life deficits associated with diabetes and comorbidities in a Canadian National Population Health Survey. Qual Life Res. 2005;14:1311–1320. 4. Anderson RJ, Freedland KE, Clouse RE, Lustman PJ. The prevalence of comorbid depression in adults with diabetes: a meta-analysis. Diabetes Care. 2001; 24:1069–1078. 5. Health Canada. Toward a Healthy Future: Second Report on the Health of Canadians. Ontario, Ottawa: Health Canada; 1999. 6. Evans RG, Stoddart GL. Producing health, consuming health care. Soc Sci Med. 1990;31:1347–1363. 7. Lloyd A, Sawyer W, Hopkinson P. Impact of long-term complications on quality of life in patients with type 2 diabetes not using insulin. Value Health. 2001;4:392–400. 8. Petterson T, Lee P, Hollis S, Young B, Newton P, Dornan T. Well-being and treatment satisfaction in older people with diabetes. Diabetes Care. 1998;21:930–935. 9. Redekop WK, Koopmanschap MA, Stolk RP, Rutten GE, Wolffenbuttel BH, Niessen LW. Healthrelated quality of life and treatment satisfaction in Dutch patients with type 2 diabetes. Diabetes Care. 2002;25:458–463. 10. Rubin RR, Peyrot M. Quality of life and diabetes. Diabetes Metab Res Rev. 1999;15:205–218. 11. Coffey JT, Brandle M, Zhou H et al. Valuing health-related quality of life in diabetes. Diabetes Care. 2002;25:2238–2243. 12. Hertzman C, Frank J, Evans R. Heterogeneities in health status and the determinants of population health. In: Evans R, Barer M, Marmore T, eds. Why Are Some People Healthy and Others Not? The Determinants of Health of Populations. New York, NY: Aldine De Gruyter, 1994:93–132. 13. Beland Y. Canadian community health survey– methodological overview. Health Rep. 2002;13:9–14. 14. Statistics Canada. CCHS Cycle 1.1, Public Use Microdata File Documentation. 2004. Available at: http:// prod.library.utoronto.ca:8090/datalib/codebooks/cstdli/ cchs/2001/engdoc.pdf. Accessed June 4, 2006. 15. Feeny D, Furlong W, Torrance GW, et al. Multiattribute and single-attribute utility functions for the Health Utilities Index Mark 3 system. Med Care. 2002; 40:113–128. 16. Horsman J, Furlong W, Feeny D, Torrance G. The Health Utilities Index (HUI®): concepts, measurement properties and applications. Health Qual Life Outcomes. 2003;1:54. 17. Maddigan SL, Feeny DH, Johnson JA. A comparison of the Health Utilities Indices Mark 2 and Mark 3 in type 2 diabetes. Med Decis Making. 2003;23:489–501. 18. Maddigan SL, Feeny DH, Johnson JA. Construct validity of the RAND-12 and Health Utilities Index Mark 2 and Mark 3 in type 2 diabetes. Qual Life Res. 2004;13:435–448. 19. de Visser CL, Bibo HJ, Groenier KH, de Visser W, Meyboom-de Jong B. The influence of cardiovascular disease on quality of life in type 2 diabetics. Qual Life Res. 2002;11:249–261. 20. Grootendorst P, Feeny D, Furlong W. Health Utilities Index Mark 3: evidence of construct validity for



stroke and arthritis in a population health survey. Med Care. 2000;38:290–299. 21. Sturmer T, Brenner H, Brenner RE, Gunther KP. Non-insulin dependent diabetes mellitus (NIDDM) and patterns of osteoarthritis. The Ulm osteoarthritis study. Scand J Rheumatol. 2001;30:169–171. 22. Goldney RD, Phillips PJ, Fisher LJ, Wilson DH. Diabetes, depression, and quality of life: a population study. Diabetes Care. 2004;27:1066–1070. 23. Patten SB, Brandon-Christie J, Devji J, Sedmak B. Performance of the composite international diagnostic interview short form for major depression in a community sample. Chronic Dis Can. 2000;21:68–72.

40. Johnson JA, Majumdar SR, Simpson SH, Toth EL. Decreased mortality associated with the use of metformin compared with sulfonylurea monotherapy in type 2 diabetes. Diabetes Care. 2002;25:2244–2248. 41. Hahl J, Hamalainen H, Sintonen H, Simell T, Arinen S, Simell O. Health-related quality of life in type 1 diabetes without or with symptoms of long-term complications. Qual Life Res. 2002;11:427–436. 42. Center for Chronic Disease Prevention and Control. Diabetes in Canada. 2nd ed. Ottawa, Ontario: Health Canada; 2002.

24. Glasgow RE, Ruggiero L, Eakin EG, Dryfoos J, Chobanian L. Quality of life and associated characteristics in a large national sample of adults with diabetes. Diabetes Care. 1997;20:562–567. 25. Rust KF, Rao JN. Variance estimation for complex surveys using replication techniques. Stat Methods Med Res. 1996;5:283–310. 26. Menard S. Applied logistic regression analysis. Thousand Oaks, Calif: Sage; 1995.

Fighting Global Blindness Improving World Vision Through Cataract Elimination By Sanduk Ruit, MD, Charles C. Wykoff, MD, D.Phil., MD, Geoffrey C. Tabin, MD

27. Austin PC. A Comparison of methods for analyzing health-related quality-of-life measures. Value Health. 2002;5:329–337.

noperated cataract is the cause of millions of cases of visual impairment and blindness in poor populations throughout both the developing and the developed world. This wonderfully written volume shares the experiences of a team of surgeons who have demonstrated how the surgical procedures can be simplified and made more efficient, accessible, and far less expensive. It is a step–by–step manual to solving the problem where adequate surgeons can be trained to follow suit. Subject matter ranges from ways to increase demand among those who need surgery to the organization of surgical services, responsibilities among different personnel, efficient layout of clinical facilities, and how to sustain services at the least cost to those who need it most.

U

28. Pedhazur EJ. Multiple Regression in Behavioral Research. 3rd ed. Toronto, Ontario: Harcourt Brace & Company Canada, Ltd; 1997. 29. Health Canada. Health risk classification according to body mass index (BMI). Available at: http://www. hc-sc.gc.ca/hpfb-dgpsa/onpp-bppn/cg_quick-reference_ table1_e.html. Accessed: September 15, 2003. 30. Shields M, Tremblay S. The health of Canada’s communities. Health Rep. 2002;13(Suppl):1–23. 31. Campbell Survey on Well-Being in Canada. Available at: http://www.cflri.ca/cflri/pa/surveys/88survey. html. Accessed August 10, 2005. 32. Brown AF, Ettner SL, Piette J et al. Socioeconomic position and health among persons with diabetes mellitus: a conceptual framework and review of the literature. Epidemiol Rev. 2004;26:63–77. 33. Ross N. Community belonging and health. Health Rep. 2002;13:33–39. 34. House JS, Landis KR, Umberson D. Social relationships and health. Science. 1998;241:540–544.

ISBN 0-87553-067-2 • spiral bound • 2006 $31.50 APHA Members • $45.00 Nonmembers American Public Health Association 800 I Street, NW, Washington, DC 20001 www.apha.org

TO ORDER: web www.aphabookstore.org • email [email protected] • fax 888.361.APHA • phone 888.320.APHA M-F 8am-5pm EST

35. Kramer JR, Ledolter J, Manos GN, Bayless ML. Stress and metabolic control in diabetes mellitus: methodological issues and an illustrative analysis. Ann Behav Med. 2000;22:17–28. 36. Jaber LA, Lewis NJ, Slaughter RL, Neale AV. The effect of stress on glycemic control in patients with type II diabetes during glyburide and glipizide therapy. J Clin Pharmacol. 1993;33:239–245. 37. Health Canada. Diabetes among Aboriginal people in Canada: the evidence. Available at:http://www. hc-sc.gc.ca/fnihb-dgspni/fnihb/cp/adi/publications/ the_evidence.pdf. Accessed September 15, 2003. 38. Resnick HE, Howard BV. Diabetes and cardiovascular disease. Annu Rev Med. 2002;53:245–267. 39. Eurich DT, Majumdar SR, Tsuyuki RT, Johnson JA. Reduced mortality associated with the use of ACE inhibitors in patients with type 2 diabetes. Diabetes Care. 2004;27:1330–1334.

