Volume 22, No. 3/4
2001

[Table of Contents]

Validity of the US Behavioral Risk Factor Surveillance System's Health Related Quality of Life Survey Tool in a Group of Older Canadians

Stephanie Ôunpuu, Larry W Chambers, Christopher Patterson, David Chan and Salim Yusuf

Investigators at the Centers for Disease Control and Prevention in the US have developed a brief survey tool to measure health-related quality of life (HRQOL-4). In order to support use of such tools in surveillance, it is important to assess their validity in different groups. Subjects were 926 non-institutionalized men and women (age >= 65 years) who completed a health exam and questionnaire. Results indicated that physical and mental health and physical activity limitation were each related to self-perceived health. Compared with subjects who reported excellent health, those with poor self-rated health reported a more than 17-fold increase in the number of unhealthy days in the previous 30. While responses to questions addressing psychosocial factors were most consistently associated with the HRQOL item relating to mental health, responses to health and health behaviour questions were more consistently associated with items related to physical health. This study demonstrated that the HRQOL-4 is not only accepted by older adults in a self-administered format, but also stands up to tests of its validity.

Key Words: health-related quality of life; population health; surveillance

Introduction

Health-related quality of life (HRQOL) is an extraordinarily broad and complex concept that encompasses both physical and mental health. During an era when life expectancy is increasing, the goal is to reduce the number of years lived with poor health (compression of morbidity) despite the cumulative health effects associated with normal aging and pathological disease processes. This makes the measurement of HRQOL particularly relevant to an aging population. The Institute of Medicine in the United States (US) has recently recommended that HRQOL measures be included as "Community Profile Indicators."¹ Information on trends in health status and the identification of high-risk subgroups will guide health policy by tracking the impact of health programs and assist in the allocation of resources among competing programs.

Investigators at the Centers for Disease Control and Prevention (CDC) in the US have developed a brief survey tool to identify health-related quality of life in adult populations.² The four-item "Health Related Quality of Life" core module (HRQOL-4) was developed through expert discussions convened by the CDC, and measures self-perceived health, recent physical and mental health, and recent activity limitation (Figure 1). The conceptual relationship between the four questions on the HRQOL core module is presented elsewhere.² Question 1 of the HRQOL core module focuses on self-rated health, a categorical health item that encapsulates present, past and anticipated health on a scale of excellent, very good, good, fair or poor. Questions 2 and 3 assess the number of days in the past 30 when physical and mental health were not good, and are considered mutually independent. Together they are hypothesized to explain the recent health aspects of question 1. Question 4 is included as a global measure of activity limitation (number of days of activity limitation due to poor health in the past 30 days), and can be interpreted as an indicator of severity for responses to questions 2 and 3. The "unhealthy days index" (unHDI), defined as the number of recent days with reported poor physical or mental health, is calculated by summing the total number of not good days reported for recent physical and mental health (HRQOL items 2 and 3), with 30 days as the highest assigned value.³

The HRQOL core module questions are clear, result in few cognitive difficulties, and when compared with the more lengthy and standardized health measures, such as the SF36, appear to have acceptable construct, criterion and known-groups validity for healthy adults as well as adults with chronic health conditions and disabilities.^4,5 The module is being used in the US Behavioral Risk Factor Surveillance System (BRFSS), a state-based telephone survey that conducts over 100,000 interviews annually across all 50 states. It has also been included in at least four population health surveys in Canada⁶ (A Michalos, personal communication). In order to support its use in surveillance, it is important to assess its validity in different population groups and in direct comparison with a variety of alternative health measures.

In 1998 the Seniors Health Investigation Network (SHINE) study of 926 older adults (>= 65 years of age) was conducted in four family practices located in the central-west region of Ontario. The pilot study documented the prevalence of risk factors, morbidity and disability among older adults and the use of preventive manoeuvres recommended by the Canadian Task Force on Preventive Health Care.⁷ Data were collected through participation in a health examination, completion of a questionnaire, and health record review. The HRQOL core module was included in the study questionnaire. Here we report on selected measurement properties of the HRQOL for this older age group.

Method

Sample

Subjects were recruited from one family practice in Hamilton and three practices in Dunnville. The sampling frame consisted of all non-institutionalized men and women (age >= 65 years) who were ambulatory and had visited one of the participating practices within the last 18 months. Residents of long-term care facilities, those who required a proxy respondent, were being actively treated for cancer or other terminal disease, or who did not provide informed written consent were considered ineligible.

Letters of invitation were sent to all patients who were living in the community and had attended the physician's office within the previous 18 months. The letter included a description of the study, a copy of the SHINE questionnaire, and a consent form. These patients were invited to call their physician's office and register for a SHINE clinic. Reminder postcards were mailed to non-responders within one month, followed by a telephone call to determine their interest and eligibility. Where contact was made and the individual was ineligible, the reason was documented. In order to assess differences between responders and non-responders, some demographic information was collected from those who were eligible but unwilling to participate.

Data Collection

One SHINE clinic was held on-site at the practice, and the other was located at a well-known building on the town hospital property. SHINE clinics were organized as a series of stations: physical measures (blood pressure, heart rate, weight, height, waist and hip circumference); physical performance measures (lower extremity function,⁸ grip strength, cognitive performance measures (Mini-Mental State Examination,⁹ Clock test,¹⁰ and laboratory tests [one 20 mL blood sample]). At each station, a trained research assistant took measurements according to a standardized protocol. The SHINE questionnaire covered demographics, health history, medication use, HRQOL core module (Figure 1) and other psychosocial factors. SHINE participants were asked to complete the questionnaire prior to attending their clinic. All questionnaires were reviewed and queries resolved before each participant left the clinic.

The items on the questionnaire were compiled using previously validated sections from other questionnaires. Depression was measured using the National Centre for Health Statistics short-form question: "During the past 12 months was there ever a time when you felt sad, blue, downhearted or depressed for two weeks or more in a row?"¹¹ Locus of control, defined as the level of perceived control over one's own health and over life, was measured using a series of six scale items (strongly agree - strongly disagree) developed by Bobak and Marmot and validated for a variety of health outcomes in central and eastern Europe.¹² A locus of control score (minimum 6 to maximum 24) was calculated for the six items. Participants were grouped by quartile with the first quartile representing the lowest locus of control level. Level of social integration, which includes items measuring both the quantitative characteristics of the extended social network and its function (i.e. belongingness, practical help and appraisal support), was assessed using a series of six questions tapping these dimensions.¹³ Social integration scores for all participants were grouped into quartiles, with the lowest quartile representing the lowest level of social integration.

We included three other measures of health status in the analysis. First, participants were asked to report their lifetime history of 18 common health problems (e.g. high blood pressure, high blood cholesterol, diabetes, heart attack, cancer by site, etc). Responses were categorized into 0, 1-2, 3-4 or 5+ health problems. Second, an estimate for 10-year coronary heart disease risk was calculated based on gender, age, smoking status, blood pressure, total serum cholesterol, and self-reported history of diabetes using guidelines developed by the Second Joint Task Force of European and other Societies on Coronary Prevention.¹⁴ Finally, we assessed lower extremity function using a method developed by Guralnik et al.⁸ that incorporates static balance, walking speed, leg strength and transfer ability. Scores for lower extremity function were grouped into quartiles for women and men separately, with the lowest quartile representing the lowest level of physical function.

Physical activity was measured using the Habitual Activity Estimation Scale adapted for older adults.¹⁵ A participant was considered active if the usual amount of time expended on moderate and vigorous activity exceeded 150 minutes in a week. Tobacco use was measured as current/former/never smoker.¹⁶

Analysis

We conducted a series of analyses to determine the validity of the HRQOL-4 in this group of Ontario adults. Spearman rank correlation analyses were carried out to study the relationship among the four HRQOL questions, and between these questions and the summary unHDI. We hypothesized that the relationships observed among the four variables would replicate those observed in the USA² and in a recent survey of Ontario adults,⁶ and would reflect the conceptual model described above. Concurrent validity was assessed using five logistic regression models with the five dependent variables being each of the HRQOL core module questions and the unHDI. Responses to self-rated health were dichotomized as excellent/very good/good versus fair/poor. Responses to each of the other three questions and the unHDI were categorized into dichotomous dependent variables (0 and >= 1 days in the past 30).¹⁷ The independent variables were socio-demographic (education, income), psychosocial (locus of control, social integration, depression), physical health (history of illness, risk of coronary heart disease, physical function), and behavioural (smoking, physical activity) factors. Per the conceptual model, we hypothesized that the psychosocial variables would be related to the HRQOL mental health question, the physical health and behavioural variables would be related to the HRQOL physical health and activity limitation questions, and the self-rated health and unHDI would be related to both mental health and physical health variables. Logistic regression analyses were carried out with the 741 subjects who answered all items in the analysis.

Results

A total of 1,952 letters of invitation were distributed. Of these, 337 patients were ineligible, 582 refused to participate, and no contact was made with 107 patients. The final sample size was 926 patients. The response rate, calculated as number of subjects/(total invites - no contact - ineligible) was 61%. Of those who were eligible but did not participate in the study, 66% (n = 385/582) were willing to answer a few brief questions. There were no differences between study participants and non-responders on current smoking status and family history of memory loss. The two groups differed with respect to education and gender (i.e. a greater percentage of participants had at least a secondary school education and were female in comparison with non-responders) (Table 1).

Mean age of study participants was 73.2 years and 60% of the sample was female. Thirty-two percent (n = 292) of subjects reported either excellent or very good health, and 21% (n = 195) of subjects reported fair-to-poor health.

Overall, SHINE participants reported an average of 5.2 unhealthy days during the 30 days preceding the survey. In general, participants reporting low locus of control, low social integration, a recent history of depression, a positive history of health problems, being inactive, and those having poor lower extremity function had a higher unHDI. For each of these variables, a gradient of increased unHDI was observed across each of the quartiles/response options (Table 2).

Spearman rank order correlations indicate that physical health, mental health and activity limitation were all moderately related to self-perceived health. Recent activity limitation was strongly correlated with the unHDI (Table 3). Compared with participants who reported excellent health, those with poor self-rated health reported a more than 17-fold increase in the unHDI (Figure 2).

FIGURE 2
Mean unhealthy days by self-rated health, SHINE study (n = 921)

Results of the logistic regression analyses indicated that the psychosocial variables were associated with each of the dependent variables (Table 4). For example, those indicating a positive history of depression were 1.99 times more likely to report fair/poor health than those with no recent history of depression. The same group was 1.84 times more likely to report at least one day of poor physical health, 3.63 times more likely to report at least one day of poor mental health, and 2.35 times more likely to report at least one day of activity limitation. These relationships are reflected in Model 5 (Table 4), which indicates that those with recent history of depression were 2.82 times more likely to have one or more unhealthy days than were those with no recent depression. Relationships of a similar magnitude were seen for locus of control (odds ratios [ORs] of 1.67 and 3.29 for poor physical and mental health in the low versus high locus of control quartile comparisons), and social integration (ORs of 2.35 and 1.68 for poor mental health and unHDI in the low versus high social integration comparison). As presented here, these odds ratios are simultaneously adjusted for all other variables in the model.

A history of multiple illnesses was associated with increased risk of fair-to-poor self-rated health, at least one day of poor physical health or activity limitation, and the unHDI, but not with recent poor mental health. Inactivity was associated with increased risk of poor-fair health and at least one day of activity limitation, but not with the other HRQOL-4 measures. Tobacco use was not associated with any of the dependent variables included in any of the five models (Table 4).

Discussion

The HRQOL-4 core module used in the U.S. Behavioral Risk Factor Surveillance System is based on subjective evaluations of health and functional status. The four core questions are attractive because of their face validity as shown by the hundreds of thousands of BRFSS respondents who willingly answered these questions by telephone over the last decade. The core HRQOL-4 module is used as a general measure that is broadly applicable across different population groups and diseases. We have demonstrated with the SHINE study that the HRQOL-4 core module is not only accepted by older adults in a self-administered format (922 of 926 participants completed all four questions), but also stands up to tests of its validity.

The direction and magnitude of the relationships between self-perceived health status and recent physical health, mental health and activity limitation in this group of older adults were consistent with those reported for adults of all ages living in the same geographic area,⁶ and with those reported elsewhere for the US population.² The 17-fold difference in unHDI among older adults with self-reported poor versus excellent health is consistent with a 10-fold difference seen among the general adult population,¹⁸ and provides some insight into the explanatory abilities of these brief, simple questions. The magnitude of this relationship supports inclusion of a continuous variable such as the unHDI, which more clearly illustrates the extreme differences in perceived mental and physical health at the ends of the "poor health - excellent health" continuum.

In this study, we quantified the relationships of the HRQOL-4 measures with alternative measures of health status, and other factors considered to influence health status. These analyses provide insights into the aspects of health tapped by the HRQOL, and enable a crude level of calibration for the unHDI. For example, SHINE participants with a positive history of depression reported an average unHDI of 10.6 days, compared with 3.7 days among those with no depression. Participants with five or more health problems reported an average unHDI of 8.7 days compared with 3.1 days among those with no health problems. This general pattern of association (i.e. increased unHDI with increasing levels of compromised health) was consistent across several variables addressing different aspects of self-reported health. While these are crude, unadjusted relationships, the consistent gradient observed for most of the variables analyzed provides some measure of construct validity for the unHDI.

Results of the logistic regression analyses provide estimates of the magnitude of the relationship between the HRQOL core variables while adjusting for all other variables in the analysis. The psychosocial variables (locus of control, depression, social networks) were important in all five models. Further, all three psychosocial variables contributed significantly to the mental health model. These results support the validity of the question on mental health. The poor association observed between self-perceived health and recent mental health limitation indicates that many subjects did not consider their mental health status to be a major component of their general health. However, specific aspects of mental health measured in this survey appear to be encapsulated in responses to the other health measures included in the HRQOL.

While the psychosocial variables were most consistently associated with the HRQOL item relating to mental health, the health and health behaviour variables were more consistently associated with HRQOL items related to physical health. For example, subjects with a history of five or more illnesses had greater likelihood of reporting poor-to-fair self-perceived health, and at least one day of poor physical health or activity limitation. However no relationship was observed between health history and presence of at least one day of poor mental health. A low score on the functional performance measures was associated with at least one day of poor physical health. Inactivity was associated with poor/fair self-perceived health and at least one day of physical activity limitation. No relationship was observed for either of these variables with the mental health question. Another validation study with American adults over 18 years of age similarly found that the HRQOL core items correlated with individual SF-36 scales in a manner consistent with a priori expectations. It was reported that "not good" mental health days correlated most strongly with the mental and the emotional scales, and least strongly with the physical functioning scale. The activity limitation question, which is based on both physical and mental health, correlates with each of the SF-36 scales.³

In another population-based sample of adults over 18 years of age, we found that increased household income, younger age and nonsmoking were positively associated with health status as measured by the HRQOL variables.⁶ The lack of association for the same variables in the SHINE study may be explained by the different age groups studied. The lack of association of smoking with the HRQOL global measures, for example, could be due to the survivor effect (the sicker smokers may have died). Household income may be less relevant in this group of older adults who are mostly retired, although this contrasts with findings from older US adults in the BRFSS.¹⁹ These differences between our findings and those elsewhere may reflect a selection bias in the SHINE study (i.e. those most ill were excluded from attending a clinic) and more uniformly available health and social services in Canada. It is interesting that there is no association observed between any of the HRQOL variables and age in the adjusted models, as one might expect an association between age and health in the over 65 population. It may be that individuals who attended a SHINE study clinic represented a relatively healthier group of older people with more positive attitudes about their health, who enjoy relatively good health status. Indeed disability, which is positively associated with age, precluded 48% of those ineligible from attendance at a SHINE clinic.

The response rate achieved in this survey (64%) is comparable to other population-based surveys of older adults. However caution is advised when interpreting the results, as we are unaware of the ages of those who did not participate, and are therefore unaware of whether the sample is representative of all age groups over 65 years. The large sample size for SHINE has permitted us to demonstrate both the ease of completion of HRQOL-4 core module questions by older adults and their measurement characteristics. Consistent relationships between the four questions and the unHDI have now been demonstrated in several independent studies. The SHINE data have confirmed that the items have construct and concurrent validity.

The proposed Canada Well-being Measurement Act²⁰ calls for the development and regular publication of measures to indicate the well-being of people and communities. The healthy days index will be used in the United States' 2010 Objectives for the Nation²¹ to monitor national progress in achieving health for all. Inclusion of this brief survey instrument in surveillance programs is valuable as it will provide insights into health trends both over time and seasonally, to identify relationships between health and its determinants, and to identify high-risk groups. This information is useful for policy development, evaluation of programs, and to justify more detailed studies of health in specific groups. The accumulating evidence for its validity with Canadian samples support its inclusion in both national and local population health surveys in Canada. Indeed, this broad use would provide additional benefits of community ownership and participation that occur when data are collected nationally and locally and then shared.

Acknowledgements

The SHINE project was facilitated by the following: Geriatric Medicine Research Fund; Specialized Health Care for The Elderly Regional Program; St. Peter's Hospital Foundation; Merck Frosst; The R. Samuel McLaughlin Centre for Research and Education on Aging and Health, Faculty of Health Sciences, McMaster University; Hoechst Marion Roussel; Program of Preventive Cardiology and Therapeutics Research, Hamilton Health Sciences Corporation; and, Public Health, Research, Education and Development Program of Hamilton-Wentworth Social and Public Health Services. We thank our partners, the physicians who took part in this study, Drs. G. Wood, F. Scallan, Smith, Rouf, R Kazemi, B Kazemi and D Chan, and David Moriarty of the Centers for Disease Control and Prevention for his valuable comments on an earlier version of this manuscript. S Ôunpuu is a Canadian Institutes of Health Research Senior Research Fellow.

References

1. Durch J, Bailey LA, Stoto MA, editors. Improving health in the community: A role for performance monitoring. Washington DC: Institute of Medicine; National Academy Press;1997.

2. Hennessy CH, Moriarty DG, Zack, MM, Scherr, PA, Brackbill R. Measuring health-related quality of life for public health surveillance. Public Health Rep 1994;109(5):665-672.

3. Centers for Disease Control and Prevention. Quality of life as a new public health measure - behavioral risk factor surveillance system, 1993. MMWR 1994;43(20): 375-80.

4. Newschaffer CJ. Validation of BRFSS HRQOL measures in a statewide sample. Atlanta (GA): Department of Health and Human Services (US), National Center for Chronic Disease Prevention and Health Promotion;1998.

5. Beatty P, Schechter S, Whitaker K. Evaluating subjective health questions: cognitive and methodological investigations. Proceedings of the American Statistical Association, Section on Survey Research Methods;1996.

6. Ôunpuu S, Kreuger P, Vermeulen M, Chambers LW. Using the U.S. behavior risk factor surveillance system's health related quality of life survey tool in a Canadian city. Can J Public Health. 2000;91:67-72.

7. The Canadian Task Force on the Periodic Health Examination. Clinical preventive health care. Ottawa; 1994.

8. Guralnik JM, Ferrucci L, Simonsick EM, Salive ME, Wallace RB. Lower-extremity function in persons over the age of 70 years as a predictor of subsequent disability. N Engl J Med 1995;332:556-61.

9. Folstein MF, Folstein SE, McHugh PR. Mini-mental state: a practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 1975;12:189-198.

10. Juby A. Correlation between the Folstein Mini-Mental State Examination and three methods of clock drawing scoring. J Geriatr Psychiatry Neurol 1999;12:87-91.

11. Kessler R, Mroczek D, editors. DSM-IV CIDI Short-form Package. United States: National Center for Health Statistics; 1996.

12. Bobak M, Pikhart H, Hertzman C, Rose R, Marmot M. Socioeconomic factors, perceived control and self reported health in Russia. A cross-sectional survey. Soc Sci & Med 1998;47:269-279.

13. Orth-Gomer K, Rosengren A, Wilhelmsen L. Lack of social support and incidence of coronary heart disease in middle-aged Swedish men. Psychosom Med 1993;55:37-43.

14. Wood D, De Backer G, Faergeman O, Graham I, Mancia G, Pyorala K. For the members of the Second Joint Task Force of European and other Societies on Coronary Prevention. Prevention of coronary heart disease in clinical practice. Eur Heart J 1998;19:1434-1503.

15. Hay J. Development and validation of the Habitual Activity Estimation Scale (HAES). Children and Exercise XIX 1997;II:125-9.

16. Anand SS, Yusuf S, Vuksan V et al. The Study of Health Assessment and Risk in Ethnic Groups (SHARE): Rationale and design. Can J Cardiol 1998;14:1349-57.

17. Siegel PZ. Self-reported health status: Public health surveillance and small-area analysis. In: Schechter S, editor. Proceedings of the 1993 NCHS Conference on the Cognitive Aspects of Self-Reported Health Status. Cognitive Methods Staff Working Paper Series (No. 10); 1994 Aug; Hyattsville, Maryland: US Department of Health and Human Services, Public Health Service, National Center for Health Statistics; 1994. p. 11-27.

18. Centers for Disease Control and Prevention. Measuring Healthy Days. Atlanta, Georgia: CDC, November, 2000.

19. Campbell VA, Crews JE, Moriarty DG, Zack MM, Blackman DK. Surveillance for sensory impairment, activity limitation, and health-related quality of life for older adults. MMWR Surveillance Summary, 1999; 48
(SS-8):131-156.

20. Nickerson M. Canada Well-being Measurement Act. Ottawa (ON): Standing Committee of the House of Commons; 2000 Canada. Available from: URL: <http://www.parl.gc.ca/37/1/parlbus/chambus/house/bills/private/c-268/c-268_1/c-268TOCE.html>. Accessed Nov. 14 2001.

21. Department of Health and Human Services. Healthy People 2010. Understanding and improving health. Washington DC: 2000. Available from URL: http://www.health.gov/healthypeople

Author References

Stephanie Ôunpuu, Salim Yusuf, Population Health Institute, Hamilton General Hospital, McMaster Clinic, and Department of Medicine, Faculty of Health Sciences, McMaster University

Larry W. Chambers, University of Ottawa Institute on Health of the Elderly, SCO Health Service

Christopher Patterson, Department of Medicine, Faculty of Health Sciences, McMaster University

David Chan, Department of Family Medicine, McMaster University

Correspondence: Dr. Stephanie Ôunpuu, Population Health Section, Hamilton General Hospital, McMaster Clinic, 237 Barton Street East, Hamilton, Ontario L8L 2X2; Fax: (905) 527-9642; E-mail: stephani@ccc.mcmaster.ca

[Previous] [Table of Contents] [Next]