Objectives: This study measured the association between socioeconomic status and the eight scale scores of the Medical Outcomes Study short form 36 (SF-36) general health survey in the Whitehall II study of British civil servants. It also assessed, for the physical functioning scale, whether this association was independent of disease.
Methods: A questionnaire containing the SF-36 was administered at the third phase of the study to 5766 men and 2589 women aged 39 through 63 years. Socioeconomic status was measured by means of six levels of employment grades.
Results: There were significant improvements with age in general mental health, role-emotional, vitality, and social functioning scale scores. In men, all the scales except vitality showed significant age-adjusted gradients across the employment grades (lower grades, worse health). Among women, a similar relationship was found for the physical functioning, pain, and social functioning scales. For physical functioning, the effect of grade was found in those with and without disease.
Conclusions: Low socioeconomic status was associated with poor health functioning, and the effect sizes were comparable to those for some clinical conditions. For physical functioning, this association may act both via and independently of disease. (Am J Public Health. 1997;87:1484-1490).
Introduction The goal of medical care has been described as the restoration and preservation of function and well-being. [1] Functional outcomes are now often measured by self-reported health status questionnaires in routine, [2] as well as research, clinical settings. [3,4] The Medical Outcomes Study short form 36 (SF-36) general health survey is among the most widely used validated [5] measures of health status, being administered internationally in more than 260 clinical trials. [6] Interventions aimed primarily at improving functional status have been tested in randomized trials. [7]
However, strong inverse associations between socioeconomic status (SES) and measures of health functioning have been reported in general population studies, [8,9] largely in those concentrating on physical functioning in older age groups. [10-12] In functional assessments of patient populations, social factors are usually measured as potential confounders. [13,14] However, since SES clearly influences the risk of both developing and dying from certain diseases [15] -an effect that is independent of established risk and prognostic factors [16-18] -effects on health functioning are to be expected. It has further been proposed that certain risk factors may predict physical disability via mechanisms independent of disease. [19]
The SF-36 consists of 36 items scored in eight scales covering physical and mental health functioning. Existing population studies using the SF-36 have patchily reported group differences by age, sex, social class, and region [20-23] in the context of population "norms" rather than as variables of interest. Such norms are relatively crude because of the wide range of age and social class of the subjects in these studies. By contrast, the participants in the Whitehall II study of British civil servants [24] were aged 35 to 55 years at baseline in 1985 through 1988, and all were employed in offices in and around Whitehall, London. We hypothesized that the SF-36 would demonstrate SES differences in health mirroring those seen in morbidity and mortality studies. For the physical functioning scale, we further hypothesized that the relationship with SES would be found in those both with and without disease.
Methods Study Population The Whitehall II Study was set up to investigate the potential psychosocial, behavioral, and biological explanations of the inverse social gradient in coronary and other diseases. All nonindustrial civil servants aged 35 through 55 years working in the London offices of 20 departments were invited to participate in the study. The overall response rate at recruitment (Phase 1) was 73%. The true response rate is likely to be higher, however, because around 4% of the civil servants on the lists provided by the civil service had moved before the study and were therefore not eligible for inclusion. In total, 10 308 civil servants participated, of whom 67% (6895) were men and 33% (3413) were women. Enrollment in the study occurred from November 1985 through March 1988. Phase 2 data collection consisted of a postal questionnaire. Phase 3 data collection was carried out from August 1991 through May 1993, when the participants were aged 39 through 63. Participants were sent a questionnaire and attended a screening examination where an interviewer sought missing information. There were 197 deaths by the end of May 1993. At Phase 3, 8355 participants (5786 men and 2589 women) responded to the questionnaire (83% response rate).
Questionnaire The SF-36 consists of 36 items scored in eight scales: general health perceptions (5 items), physical functioning (10 items), role limitations due to physical functioning (4 items), bodily pain (2 items), general mental health (5 items), role limitations due to emotional problems (3 items), vitality (4 items), and social functioning (2 items). The remaining item, relating to change in health, is not scored as a separate dimension. The Phase 3 questionnaire included an Anglicized version of the SF-36 which substituted "walking half a mile," "walking one hundred yards," and "did you feel full of life?" for the US items "walking several blocks," "walking one block," and "did you feel full of pep?" respectively. As an example of scale content, the physical functioning scale comprises items on vigorous activities (e.g., strenuous sports, running), moderate activities (e.g., housework, playing golf), lifting and carrying, climbing stairs, bending, kneeling, and walking. The physical functioning scores ranged from 0, indicating severe limitation in performing all physical activities, including bathing or dressing, to 100, indicating no limitation in performing all types of physical activities, including the most vigorous. The proportion of missing data from each item of the SF-36 was 0.1% or less. If 50% or fewer of items per scale were missing, a person-specific estimate of the missing score was calculated, by substituting the mean score across completed items in the same scale for that respondent. [6] After an interval of 4 weeks, the complete questionnaire was readministered to a random sample of 289 individuals.
SES was assessed by means of civil service employment grade and access to the use of a car. [25] Information on grade of employment was obtained by asking participants to give their civil service grade. On the basis of salary, the civil service identifies 12 nonindustrial grades. To obtain sufficient numbers for meaningful analysis, we combined the top 6 groups into grade 1 and the bottom 2 groups into grade 6, thus producing 6 grade categories. The salaries ranged from [pound sterling] 6483-[pound sterling] 11 917 (grade 6) through [pound sterling] 28 904-[pound sterling] 87 620 (grade 1) in 1992.
The following diseases were identified (along with numbers of research subjects with disease): angina [26] (n = 450), probable or possible ischaemia on resting electrocardiogram (Minnesota codes 1-1 through 1-3, 4-1 through 4-4, 5-1 through 5-3, and 7-1-1; n = 707), hypertension (blood pressure > 160/90 mmHg or on an antihypertensive medication; n = 1554), claudication [26] (n = 125), diabetes (self-report or score on oral glucose tolerance test [27]; n = 222), chronic bronchitis [28] (n = 914), musculoskeletal disorders (self-report; n = 1257), and cancer (registration and self-report; n = 128). In addition, participants reporting any "limiting long-standing illness, disability, or infirmity" were identified. In order to determine whether effects on physical functioning were independent of disease, subjects with one or more of the above diseases or disabilities at phase 1 or phase 3 (2868 men and 1218 women) were analyzed separately.
Statistical Analysis For each of the eight dimensions, item scores were coded, summed, and transformed to a scale score ranging from 0 (worst health) through 100 (best health). The internal consistency of each scale was assessed by means of Cronbach's alpha. Since internal consistency increases with the number of items in a scale, Cronbach's alpha was also estimated under the assumption of a 10-item scale. The Cronbach's alpha values were calculated by means of the Multitrait Analysis Program. [29] Calculation of the correlation coefficients between each item and each scale (corrected for overlap) revealed no definite scaling errors. All other analyses were performed by means of the statistical package SAS (SAS Institute Inc, Cary, NC).
Test-retest reliability was assessed by calculating the mean difference between the first and repeat measurements for each scale, along with the percentage of differences lying within 1.96 standard deviations of the mean. If the scale and the health it is measuring are stable during this time period, then the mean should be zero. [30]
Ordinary least squares regression was used to estimate the effect of age, civil service employment grade, and access to use of a car, along with the interaction of age and grade on each of the eight scale scores, separately for men and women. Age effects were expressed as the increase in score per 1-year increase in age. Effect sizes (standard differences) were calculated by dividing the difference in two mean scores (e.g., between top and bottom grades) by the sex-specific standard deviation for the scale. The effect of grade was further assessed by calculating the age-adjusted odds of being in the lowest sex-specific quartile of physical functioning. The physical functioning scale was selected for further study because improving physical functioning may be considered a more directly relevant goal of medical care than improving, for example, general health perceptions or vitality. Two-tailed tests were used throughout. None of the distributions of the SF-36 scales are normal; however, nonparametric methods showed the same trends with age and grade.
Results Reliability The Cronbach's alpha was 0.75 or greater for all scales ( Table 1); when estimated for a 10-item scale, it was 0.86 or greater, which is satisfactory. [31] There was an inverse association between Cronbach's alpha and grade, with lower grades having higher alpha coefficients. Women had consistently higher Cronbach's alpha values than men. This is likely to be a reflection of the higher variance of each scale among women and those in lower grades. The proportion of participants at the "ceiling" (i.e., scoring 100) was highest among the high grades. Since there is likely to be true but unmeasured variation in functioning among those at the ceiling, the true magnitude of SES differences in scale scores may be greater than measured. There was no effect of age on Cronbach's alpha.
Age and Sex In these cross-sectional data, there were significant increases (i.e., improvements) with age in general mental health, role-emotional, vitality, and social functioning scale scores in men and women (test for linear trend P < .01 for each) ( Table 2). The mean increase in these scale scores associated with each year's increase in age was from 0.11 through 0.40. Physical functioning scores decreased (P < .0001) with age; the regression coefficient was -0.34 in men and -0.72 in women. These effects were not attenuated when adjusted for grade of employment.
Employment Grade ( Table 3) shows that among men, there were significant (P <or= to .0002) age-adjusted gradients across the six civil service employment grades in all the scales except vitality, with the higher status grades reporting better health functioning. The size of these effects was greater than that of age, with effect sizes greater than 0.3 for the physical functioning, social functioning, and pain scales.
The median physical functioning score in the lowest quartile was 80 for men (range, 0 through 89) and 60 for women (range, 0 through 75). Subjects who were in the lowest quartile of physical functioning reported that their ability to perform vigorous activities and to climb several flights of stairs was limited a lot. Their ability to perform moderate activities, lift and carry, bend and kneel, and walk more than 1 mile tended to be limited a little. Table 4 shows that men and women in the lower grades were at increased risk of poor physical functioning independent of disease status. In men without disease, the age-adjusted odds ratio of being in the lowest quartile of physical functioning was 2.49 (95% confidence interval [CI] = 1.62, 3.85) for grades 5 and 6 compared with grades 1 and 2; in men with disease, the corresponding odds ratio was 1.92 (95% CI = 1.42, 2.58).
Although the Whitehall II study is longitudinal in design, the data presented here are cross-sectional, and this represents a potential limitation. Interpreting the striking increases in mean scores with age in the mental health, role-emotional, vitality, and social functioning scales requires repeated measures of the SF-36 to distinguish a true age-related increase from a cohort effect. Prospective studies suggest that successive generations are more inclined to report their health adversely. [33,34] The increases in mental health scale scores could reflect a decline in prevalence of mental illness with age, although recent population data do not support this. [35] Although physical functioning scores declined overall with age, the effect was more marked among low-grade men. This is consistent with the hypothesis that changes in health function with age are heterogeneous and have environmental determinants; they are not simply the biologically inevitable consequence of the passage of time. [36]
Socioeconomic status may influence health functioning by affecting one or a combination of the following: (1) the risk of disease onset, (2) the risk of disability among people with disease, and (3) mechanisms independent of disease, for example, in the case of physical functioning via cardiorespiratory and locomotor "fitness." Previous studies have tended to concentrate on the first and second effects; we provide some evidence here of the third. The inverse grade gradient in physical functioning remained (indeed, increased) when participants who had disease were excluded. However, before a mechanism relating SES and physical function independent of disease can be confirmed, more sensitive measures of clinical and subclinical disease than were available in the present study are required. [37]
The Whitehall II study population is comparatively young and, as a white-collar occupational cohort, high functioning; the study population does not reflect the extremes of SES. In the general population, therefore, the effect of SES (particularly when accurately measured [38]) on the SF-36 may be larger. There was a tendency toward a monotonic inverse gradient in scale score with employment grade and no evidence of a threshold effect.
Among women, there was an inverse relation between employment grade and the physical functioning, pain, and social functioning scales. The lack of effect of grade on the other scales may reflect the meaning of employment grade among women. Social class based on the woman's own occupation predicts mortality less well than her partner's social class, [39] possibly because the latter is a more accurate reflection of material circumstances. To address this potential confounding, we therefore adjusted for the social class of the women's partner. The associations with grade were unchanged. Age-adjusted scores on each of the eight scales were, however, lower among women without access to the use of a car. Arber [40] has emphasized the distinction between structural (e.g., employment grade) and role (e.g., car access) measures of SES for women; such a distinction may be of particular importance since the SF-36 measures role functioning.
It is well recognized that SES can influence the mortality and other medical outcomes of patients with coronary heart disease [17,18] cancer, [41,42] and HIV infection, [43] an effect that is independent of pathophysiological prognostic factors. It is less clear whether SES also influences the effectiveness of medical interventions. The relationships between SES, health functioning, and traditional medical outcomes should be assessed in clinical trials. Even in SF-36 studies where SES is not a variable of interest, the data presented here strongly suggest the potential for SES to bias or confound any observed effects.
It has been proposed that the SF-36 may provide a measure of health in general populations that will enable comparisons of health status across time, place, and person. [20,21,23] The Whitehall II study is obtaining repeated measures of the SF-36 and will examine psychosocial, lifestyle, biological, and health care predictors of the SF-36, and its change, in order to explain the gradients in health functioning across employment grades. The SF-36 will also be used as a predictor of socioeconomic and gender differences in subsequent sickness absence, health service utilization, morbidity, and mortality.
SES shows a strong inverse relation to the SF-36; the effect is comparable in magnitude to that of some medical conditions. However, the association between low SES and poor physical functioning is found in those both with and without concurrent or preexisting disease. Low SES has been associated with adverse outcomes of medical care, independently of clinical prognostic factors. Since one aim of medical care is to improve the health functioning of patients, further research is justified to ascertain whether specific, potentially modifiable aspects (psychosocial, behavioral, biological) of SES are associated with health functioning.
Acknowledgments This study was supported by grants from the Agency for Health Care Policy and Research (5RO1 HS06516); the New England Medical Center-Division of Health Improvement; the National Heart, Lung, and Blood Institute (2RO1 HL36310); the John D. and Catherine T. MacArthur Foundation Research Network on Successful Midlife Development; the Institute for Work and Health, Ontario, Canada; the Volvo Research Foundation, Sweden; Medical Research Council; Health and Safety Executive; and the British Heart Foundation. Michael Marmot is supported by a Medical Research Council research professorship.
We thank all participating civil service departments and their welfare, personnel, and establishment officers: Dr George Sorrie, Dr Adrian Semmence, and Dr Elizabeth McCloy of the Civil Service Occupational Health Service; the Civil Service Central Monitoring Service and Dr Frank O'Hara; the Council of Civil Service Unions and all participating civil servants. We would like to thank all members of the Whitehall II study team.
REFERENCES 1. Tarlov AR, Ware JE, Greenfield S, Nelson EC, Perrin E, Zubkoff M. The medical outcomes study: an application of methods for monitoring the results of medical care. JAMA. 1989;262:925-930. [Medline Link] [Context Link]
2. McHorney CA, Ware JE, Rogers W, Raczek AE, Rachel Lu JF. The validity and relative precision of MOS short-and-long-form health status scales and Dartmouth COOP charts. Med Care. 1992;30:253-265. [Medline Link] [PsycINFO Link] [Context Link]
3. Katz JN, Larson MG, Phillips CB, Fossel AH, Liang MH. Comparative measurement of sensitivity of short and longer health status instruments. Med Care. 1992;30:917-925. [Medline Link] [PsycINFO Link] [Context Link]
4. Guadagnoli E, Hauptman P, Ayanian J, Pashos C, McNeil B, Cleary P. Variation in the use of cardiac procedures after acute myocardial infarction. N Engl J Med. 1995;333:589-590. [Medline Link] [Context Link]
5. McHorney CA, Ware JE, Raczek AE. The MOS 36-item short-form health status survey (SF-36), II: psychometric and clinical tests of validity in measuring physical and mental health constructs. Med Care. 1993;31:247-263. [Medline Link] [PsycINFO Link] [Context Link]
6. Ware JE, Snow KK, Kosinski M, et al. SF-36 Health Survey Manual and Interpretation Guide. Boston, Mass: New England Medical Center; 1993. [Context Link]
7. Landefeld C, Palmer R, Kresevic D, Fortinsky R, Kowal J. A randomized trial of care in a hospital medical unit especially designed to improve the functional outcomes of acutely ill older patients. N Engl J Med. 1995;332:1338-1344. [Fulltext Link] [Medline Link] [Context Link]
8. Hunt SM, McEwen J, McKenna SP. Social inequalities and perceived health. Effective Health Care. 1985;2(4):151-160. [Context Link]
9. Kuh DJL, Wadsworth MEJ, Yusuf EJ. Burden of disability in a post-war birth cohort in the UK. J Epidemiol Community Health. 1994;48:262-269. [Medline Link] [BIOSIS Previews Link] [Context Link]
10. Pinsky J, Leaverton P, Stokes JI. Predictors of good function: the Framingham study. J Chronic Dis. 1987;40(suppl 1):159S-67S. [Medline Link] [Context Link]
11. Mor V, Murphy J, Masterson-Allen S. Risk of functional decline among well elders. J Clin Epidemiol. 1989;42:895-904. [Medline Link] [Context Link]
12. Seeman TE, Charpentier PA, Berkman LF, et al. Predicting changes in physical performance in a high functioning elderly cohort: MacArthur Studies on Successful Aging. J Gerontol. 1994;49:M97-M108. [Medline Link] [PsycINFO Link] [Context Link]
13. Stewart AL, Greenfield S, Hays RD, et al. Functional status and well-being of patients with chronic conditions. JAMA. 1989;262:907-913. [Medline Link] [Context Link]
14. Spitzer R, Kroenke K, Linze M, et al. Health related quality of life in primary care patients with mental disorders: results from the PRIME-ME 1000 study. JAMA. 1995;274:1511-1517. [Fulltext Link] [Medline Link] [BIOSIS Previews Link] [Context Link]
15. Black D, Morris JN, Smith C, et al. Inequalities in health: the Black report; the health divide. London, England: Penguin Group; 1988. [Context Link]
16. Marmot MG, Rose G, Shipley M, Hamilton PJS. Employment grade and coronary heart disease in British civil servants. J Epidemiol Community Health. 1978;32:244-249. [Medline Link] [Context Link]
17. Ruberman W, Weinblatt E, Goldberg JD, Chaudhary BS. Psychosocial influences on mortality after myocardial infarction. N Engl J Med. 1984;311:552-559. [Medline Link] [PsycINFO Link] [Context Link]
18. Williams RB, Barefoot JC, Califf RM, et al. Prognostic importance of social and economic resources among medically treated patients with angiographically documented coronary artery disease. JAMA. 1992;267:520-524. [Medline Link] [Context Link]
19. Pinsky JL, Branch LG, Jette AM, et al. Framingham disability study: relationship of disability to cardiovascular risk factors among persons free of diagnosed cardiovascular disease. Am J Epidemiol. 1985;122:644-656. [Medline Link] [Context Link]
20. Brazier JE, Harper R, Jones NMB, et al. Validating the SF-36 health survey questionnaire: new outcome measure for primary care. BMJ. 1992;305:160-164. [Medline Link] [Context Link]
21. Jenkinson C, Coulter A, Wright L. Short form 36 (SF-36) health survey questionnaire: normative data for adults of working age. BMJ. 1993;306:1437-1440. [Fulltext Link] [Medline Link] [BIOSIS Previews Link] [Context Link]
22. Garratt AM, Ruta D, Abdall M, Buckingham J, Russell I. The SF-36 health survey questionnaire: an outcome measure suitable for routine use within the NHS? BMJ. 1993;306:1440-1444. [Fulltext Link] [Medline Link] [BIOSIS Previews Link] [Context Link]
23. Lyons RA, Fielder H, Littlepage BN. Measuring health status with the SF-36: the need for regional norms. J Public Health Med. 1995;17:46-50. [Medline Link] [Context Link]
24. Marmot MG, Davey Smith G, Stansfeld S, et al. Health inequalities among British civil servants: the Whitehall II study. Lancet. 1991;337:1387-1393. [Medline Link] [Context Link]
25. Liberatos P, Link BG, Kelsey JL. The measurement of social class in epidemiology. Epidemiol Rev. 1988;10:87-121. [Medline Link] [Context Link]
26. Rose G, Blackburn H, Gillum RF, et al. Cardiovascular Survey Methods. 2nd ed. Geneva, Switzerland: World Health Organization; 1982. [Context Link]
27. WHO Study Group. Diabetes Mellitus: Geneva, Switzerland: World Health Organization; 1985. [Context Link]
28. Questionnaire on Respiratory Symptoms and Instruction for Interviewers. London, England: Medical Research Council; 1976. [Context Link]
29. Hays RD, Hayashi T, Carson C, et al. User's Guide for the Multitrait Analysis Program (MAP). Supplement: Memorandum Revised Multitrait Analysis Program Software (MAP-R). Boston, Mass: New England Medical Center; 1988. [Context Link]
30. Bland JM, Altman DG. Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;i:307-310. [Medline Link] [Context Link]
31. McDowell I, Newell C. Measuring Health-A Guide to Rating Scales and Questionnaires. New York, NY: Oxford University Press; 1987. [Context Link]
32. Berkman LF, Seeman T, Albert M, et al. High, usual and impaired functioning in community dwelling older men and women: findings from the MacArthur Foundation research network on successful aging. J Clin Epidemiol. 1993;46:1129-1140. [Medline Link] [BIOSIS Previews Link] [Context Link]
33. Hagnell O, Lanke J, Rorsman B. Are we entering an age of melancholy? depressive illness in a prospective epidemiological study over 25 years; the Lundby study, Sweden. Psychol Med. 1982;12:279-289. [Medline Link] [PsycINFO Link] [Context Link]
34. Barsky AJ. The paradox of health. N Engl J Med. 1988;318:414-418. [Medline Link] [Context Link]
35. Meltzer H, Gill B, Petticrew M, et al. OPCS Surveys of Psychiatric Morbidity in Great Britain report 1: The Prevalence of Psychiatric Morbidity among Adults Living in Private Households. London, England: Her Majesty's Stationery Office; 1995. [Context Link]
36. Rowe JW, Kahn RL. Human aging: usual and successful. Science. 1987;237:143-149. [Medline Link] [PsycINFO Link] [Context Link]
37. Kuller L, Borhani N, Furberg C, et al. Prevalence of subclinical atherosclerosis and cardiovascular disease and association with risk factors in the cardiovascular health study. Am J Epidemiol. 1994;139:1164-1179. [Medline Link] [BIOSIS Previews Link] [Context Link]
38. Davey Smith G, Shipley MJ, Rose G. Magnitude and causes of socioeconomic differentials in mortality: further evidence from the Whitehall Study. J Epidemiol Community Health. 1990;44:265-270. [Medline Link] [Context Link]
39. Fox AJ, Goldblatt P. Longitudinal study: socio-demographic mortality differentials. London, England: Office for Population Censuses and Surveys; 1982. [Context Link]
40. Arber S. Class, paid employment and family roles: making sense of structural disadvantage, gender and health status. Soc Sci Med. 1991;32:425-436. [Medline Link] [Context Link]
41. Kogevinas M, Marmot MG, Fox AJ, Goldblatt PO. Socioeconomic differences in cancer survival. J Epidemiol Community Health. 1991;45:265-270. [Medline Link] [Context Link]
42. Schrijvers CT, Coebergh JW, van der Heijden LH, Mackenbach JP. Socioeconomic variation in cancer survival in the southeastern Netherlands, 1980-1989. Cancer. 1995;75:2946-2953. [Medline Link] [BIOSIS Previews Link] [Context Link]
43. Hogg RS, Strathdee SA, Craib KJ, O'Shaughnessy MV, Montaner JS, Schechter MT. Lower socioeconomic status and shorter survival following HIV infection. Lancet. 1994;344:1120-1124. [Fulltext Link] [Medline Link] [BIOSIS Previews Link] [Context Link]