Epidemiology

Elizabeth Breeze, MSc, CStat, Astrid E. Fletcher, PhD, David A. Leon, PhD, Michael G. Marmot, PhD, MBBS, Robert J. Clarke, MD, MRCP, and Martin J. Shipley, MSc

Elizabeth Breeze, Astrid E. Fletcher, and David A. Leon are with the Department of Epidemiology and Population Health, London School of Hygiene and Tropical Medicine, London, England. Michael G. Marmot and Martin J. Shipley are with the Interna- tional Centre for Health and Society, Department of Epidemiology and Public Health, University Col- lege Medical School, London. Robert J. Clarke is with the Clinical Trial Service Unit and Epidemio- logical Studies Unit, University of Oxford, Oxford, England.

Requests for reprints should be sent to Eliza- beth Breeze, MSc, CStat, London School of Hy- giene and Tropical Medicine, Keppel Street, Lon- don WC1E 7HT, England (e-mail: elizabeth.breeze@ lshtm.ac.uk).

This article was accepted May 24, 2000.

A B S T R A C T

Objectives. This study examined (1) the relation of employment grade in middle age to self-reported poor health and functional limitations in old age and (2) whether socioeconomic status at ap- proximately the time of retirement mod- ifies health differentials in old age.

Methods. Survivors of the Whitehall Study cohort of men were resurveyed. Respondents were aged 40 to 69 years when they were originally screened in 1967 to 1970.

Results. Compared with senior ad- ministrators, men in clerical or manual (low-grade) jobs in middle age had quadruple the odds of poor physical per- formance in old age, triple the odds of poor general health, and double the odds of poor mental health and disability. At most, 20% of these differences were ex- plained by baseline health or risk fac- tors. Men who moved from low to mid- dle grades before retirement were less likely than those who remained in low grades to have poor mental health.

Conclusions. Socioeconomic status in middle age and at approximately re- tirement age is associated with morbid- ity in old age. (Am J Public Health. 2001; 91:277–283)

February 2001, Vol. 91, No. 2 American Journal of Public Health 277

There is a small but growing body of evi- dence from the United Kingdom that socioeco- nomic differentials in mortality persist into old age1–3 and may even be widening.4,5 Although rate ratios tend to be smaller for older people than for younger people in the United Kingdom and the United States,4–6 absolute differentials can still be large.5

There is little equivalent information on self-reported morbidity. Analyses of cross- sectional studies show that self-reported health and disability, respiratory function, and blood pressure are all worse among older people in disadvantaged socioeconomic groups.7,8 Analy- ses of the Office for National Statistics Lon- gitudinal Study in England and Wales showed that adverse socioeconomic circumstances were associated with self-reported limiting long- term illness after a 20-year follow-up period among survivors.9

The first Whitehall Study, an investiga- tion of male British civil servants that was ini- tiated in the late 1960s, showed an inverse mor- tality gradient (all causes and major causes) across employment grades.10 The Whitehall II Study, following a later cohort, revealed gra- dients in morbidity in middle age across so- cioeconomic groups.11,12 A resurvey of the sur- vivors of the first cohort enabled us to study the long-term effects of employment grade on self- reported illness in old age.

Methods

Data Source

In the Whitehall Study, 19 029 men, most aged 40 to 69 years, were examined between 1967 and 1970 to identify cardiorespiratory disease and its risk factors.13 Participants com- pleted a questionnaire concerning their jobs, their personal and family medical histories, and their smoking habits. Approximately two thirds of the respondents were also asked about

car ownership and physical activity related to work, and one third were asked about leisure activity in general. A clinical examination in- cluded height and weight, blood pressure, elec- trocardiogram, and a blood sample analyzed for cholesterol and blood sugar. Participants were registered with the National Health Ser- vice Central Register for mortality notification (99% were successfully located).

Resurvey

The resurvey took place in 1997–1998 after a successful pilot study of 400 survivors in 1996.14 The National Health Service Cen- tral Register identified the health authority in which the cohort member was registered with a family doctor. Chief executives of the rele- vant health authorities granted permission to the register to provide addresses of survivors (or, failing this, to forward mail to them). In- vitation letters, consent forms, and question- naires were sent to individuals, along with up to 2 reminders. A short version of the ques- tionnaire covering priority information was sent with the second reminder. The resurvey ques- tionnaire included questions on socioeconomic status (SES) and retirement, diseases diagnosed

Do Socioeconomic Disadvantages Persist Into Old Age? Self-Reported Morbidity in a 29-Year Follow-Up of the Whitehall Study

February 2001, Vol. 91, No. 2278 American Journal of Public Health

TABLE 1—Resurvey Responses by Selected Characteristics: Whitehall Study, 1997–1998

Total No. Invited Completed Full Completed Short to Take Part Questionnaire, No. (%) Questionnaire, No. (%) χ2 P

Age at resurvey, y <75 3029 2316 (76) 262 (9) 75–79 2937 2236 (76) 272 (9) ≥80 2571 1616 (63) 339 (13) < .001

Baseline employment grade High 555 443 (80) 23 (4) Middle 6743 5052 (75) 657 (10) Low 1239 673 (54) 193 (16) < .001

Baseline smoking status Never 2078 1588 (76) 186 (9) Ex-smoker 3370 2496 (74) 318 (9) Pipe/cigar smoker 332 249 (75) 28 (8) Cigarette smoker 2753 1832 (67) 341 (12) < .001

Baseline evidence of cardiovascular disease Yes 1114 813 (73) 127 (11) No 7133 5353 (75) 746 (10) .437

Baseline respiratory symptoms No phlegm 6399 4666 (73) 638 (10) Persistent cough/phlegm 1070 748 (70) 110 (10) Increasing cough/phlegm 409 267 (65) 56 (14) Hospital admission in past 647 481 (74) 69 (11) .018

Total 8537 6168 (72) 873 (10)

by a doctor, and ability to carry out everyday activities.

Outcome Measures

We used 4 measures of self-reported morbidity: general poor health, poor mental health, poor physical performance, and dis- ability. Those rating their health as poor or very poor on a 5-point scale ranging from very good to very poor were classified as being in poor general health. Poor mental health was defined as a score below 60% of the maximum on the 5-item mental health section of the Short Form 36 Health Survey (SF-36).15 Poor physical performance was de- fined as a score below 40% of the maximum on the 10-item physical performance section of the SF-36, which asks people to state whether their health limits their activity ex- tensively, a little, or not at all. Finally, dis- ability was classified as an inability to engage in at least 1 of 5 instrumental activities of daily living (cooking a hot meal, cutting toenails, dressing oneself, doing light housework and simple repairs, and going up and down stairs and steps).

Data on mental health, physical perform- ance, and disability were available only for those who completed the full questionnaire. The SF- 36 indexes were scored as recommended.16 As a result of missing data, 4% of those complet- ing the full questionnaire were not assigned a mental health score, 3% were not assigned a physical performance score, and fewer than 1% were excluded from the disability analyses.

Socioeconomic and Risk Factor Measures

The main baseline socioeconomic clas- sification used was employment grade (high, middle, or low). High grades comprised sen- ior managers and administrators; middle grades comprised executives and professionals (e.g., economists, statisticians, and scientists) in less senior positions; and low grades included cler- ical staff, printing room officers, security of- ficers, messengers, and catering staff.

Other socioeconomic indicators were car ownership and, measured retrospectively at the resurvey, housing tenure at baseline (owner vs renter). These variables were found to be clear discriminators of mortality rates among older people in the United Kingdom in the 1970s,1

were incorporated in the Townsend index of deprivation,17 and have subsequently been used as socioeconomic indicators.5,18

Respondents were considered to have preexisting cardiovascular disease if they had at least 1 of the following at baseline: an ab- normal electrocardiogram; self-reported symptoms of angina, claudication, or poten- tial myocardial infarction19; medication for high blood pressure; or a hospital admission for a heart condition. We adjusted for car- diorespiratory disease clinical risk factors that existed at baseline because these risk factors are associated with later disability20–22 and can lead to more general problems in func- tioning and health. The variables used in the analyses were as follows: being in the top quintile in terms of systolic or diastolic blood

pressure or total cholesterol level (assessed with the entire 1960s cohort), body mass index of 30 kg/m2 or greater, blood sugar level above 96 mg/dL, persistent or increasing du- ration of cough or phlegm or hospital admis- sions for respiratory disease, and 4 or more hospital admissions for other reasons.

Statistical Analysis

Chi-square tests for heterogeneity were used to determine univariate associations. Logistic re- gression (Stata 5 for Windows 3.123) was used to estimate odds ratios (ORs) and 95% confidence intervals (CIs) for each outcome.All models in- cluded adjustment for age at resurvey (younger than 75 years, 75–79 years, 80 years or older).

Results

At the time of the resurvey, there were 8537 men from the original screening who, accord- ing to National Health Service Central Register records, were alive and living in Great Britain. Of these individuals, 6168 completed a full questionnaire (72%) and 873 a short one (10%), 209 of the latter by telephone. Seven percent of respondents had been in high employment grades at the initial screening, 12% had been in low grades, and 81% had been in middle grades. The median age of respondents at the resurvey was 77 years (range: 67–97), and the median follow-up interval was 29 years (range: 26–31).

Response rates were lowest among men in low employment grades, older men, smokers,

February 2001, Vol. 91, No. 2 American Journal of Public Health 279

TABLE 2—Distribution (%) of Characteristics of Resurvey Respondents, by Employment Grade at Baseline: Whitehall Study, 1997–1998

Employment Grade at Baseline, % High Middle Low

(n = 466) (n = 5708) (n = 866) �2 P

Resurvey Age, y

<75 37.8 37.5 30.1 75–79 38.2 36.4 29.3 ≥80 24.0 26.1 40.5 < .001

Net income < $16 500 0.9 8.2 47.8 < .001 Had risen 1 grade category . . . 39.7 50.8 < .001 Had paid job after leaving Civil Service 44.9 22.8 18.7 < .001 Cardiovascular disease

Angina 11.4 14.5 16.7 .03 Heart attack 10.5 11.4 15.0 .006 Stroke 7.3 8.4 8.2 .74

Baseline Cardiovascular disease 11.4 13.3 14.5 .26 Top quintile

Systolic blood pressure 7.3 12.7 15.6 < .001 Diastolic blood pressure 10.7 13.3 13.6 .27 Total cholesterola 23.4 19.0 16.8 .15

Body mass index > 30 kg/m2 1.5 2.7 4.2 .01 Blood sugar > 96 mg/dLa 3.9 4.3 5.0 .52 Respiratory symptoms

No respiratory problem 78.1 75.7 71.8 Persistent phlegm 9.2 12.0 15.4 Increasing phlegm 2.8 4.3 7.2 Hospital admission for respiratory disease 9.9 8.0 5.7 < .001

Ever had 4 or more hospital admissions 11.2 8.9 11.0 .05 (not cardiovascular or respiratory)

Smoking status Never smoked 33.3 25.2 20.8 Ex-smoker 36.1 41.5 32.3 Smoked 1–9 cigarettes or pipe/cigar 16.6 12.8 12.3 Smoked 10–19 cigarettes 6.2 11.2 20.9 Smoked 20 or more cigarettes 7.7 9.4 13.8 < .001

Physical activitya

Walked to work, min 0–9 20.6 18.6 19.6 10–19 43.9 43.9 44.6 ≥ 20 35.5 37.5 35.8 .83

Leisure activity None 18.7 19.9 25.5 Inactive 6.0 7.8 13.2 Moderately active 38.7 44.4 36.5 Active 36.7 27.9 24.8 < .001

Other socioeconomic measures Rented accommodation 4.8 9.2 38.9 < .001 No cara 6.6 14.7 51.3 < .001 Not married 4.7 8.0 21.7 < .001

aSample sizes were smaller for this variable.

and those with increasing symptoms of cough or phlegm at baseline (Table 1). Table 2 shows that socioeconomic indicators were strongly correlated with employment grade, as were smoking, leisure activity, and respiratory dis- ease. Men in the lower employment grades were more likely to be in the top quintile in terms of systolic blood pressure and more likely to have a body mass index above 30 kg/m2.

Twenty-one percent of respondents expe- rienced at least 1 of the outcomes, and these individuals differed markedly from the other

participants. Whereas two thirds of men with poor mental health scores had low ratings on at least 1 of the 5 items of the SF-36 scale, only 6% of the remaining cohort did; those with poor physical performance ratings were lim- ited by their health in at least 7 activities, whereas only 30% of the remaining partici- pants were limited in more than 3 activities.

Respondents in low employment grades were at greatest risk of adverse outcomes for nearly all of the component morbidity items (Table 3), the differentials being greatest for

the more severe physical limitations. In com- parison with those in the high employment grades, men in the middle employment grades had a statistically significant excess risk for 8 of the physical performance limitations.

Figure 1 shows that higher percentages of respondents in low employment grades were at risk for each of the morbidity out- comes. These individuals had more than 4 times the odds of poor physical performance relative to men in high employment grades, 3 times the odds of poor general health, and 2.5 times the odds of poor mental health or a dis- ability (Table 4). Staff in middle employment grades had a statistically significant excess risk of poor general health and poor physical performance.

Baseline clinical indicators of cardiores- piratory disease, clinical risk factors, and risk behavior (smoking) reduced the odds ratios for men in the low employment grades by at most 20%. The other baseline indicators of SES nei- ther reduced the estimates of employment ef- fects nor explained a substantially larger por- tion of the outcomes (data not shown). Not being married at the original screening was an additional factor involved in poor mental health status at follow-up, but it only marginally re- duced the excess odds associated with being in the low employment grades.

After adjustment for all baseline charac- teristics that were independent risk factors for the morbidity outcomes, employment grade at baseline remained a significant factor in all of the outcomes (Table 4). Compared with men in the high employment grades, those in the mid- dle grades had statistically significant excess odds of poor physical performance, and those in the low grades had excess risks for all 4 mor- bidity outcomes.

We looked for evidence of additional so- cioeconomic factors measured at resurvey that could ameliorate, or add to, disadvantages ex- perienced in middle age. After adjustment for other factors, having a job after retirement was not associated with any of the outcomes. How- ever, Table 5 shows that low income after leav- ing the Civil Service (less than $16 500 per year in 1997–1998) was associated with an approximate doubling of the risk of 3 of the outcomes among those in the middle em- ployment grades but not those in the low grades. On the other hand, moving up a grade category between screening and retirement was associated with a smaller risk of poor mental health among those in the low em- ployment grades.

Finally, we examined lifetime cardiovas- cular disease reported at the resurvey as a pos- sible factor on the causal pathway between SES and poor health or functional limitations. As can be seen in Table 2, there were inverse as- sociations between a diagnosis of angina or

February 2001, Vol. 91, No. 2280 American Journal of Public Health

TABLE 3—Odds Ratios for Morbidity Outcomes, by Baseline Employment Grade, Adjusted for Age at Resurvey: Whitehall Study, 1997–1998

Baseline Employment Grade Morbidity Measure Sample, No. (%) High, OR Middle, OR (95% CI) Low, OR (95% CI) P

Mental health Nervous most/all of the time 5899 (1.4) 1.00 1.06 (0.4, 2.7) 2.26 (0.8, 6.2) .045 Down in dumps most/all of the time 5902 (0.7) 1.00 1.57 (0.4, 6.6) 1.72 (0.3, 8.9) .78 Calm none/little of the time 5958 (7.1) 1.00 1.02 (0.7, 1.5) 1.52 (1.0, 2.4) .027 Downhearted most/all of the time 5929 (1.5) 1.00 3.10 (0.8, 12.7) 5.82 (1.3, 25.4) .011 Happy none/little of the time 6022 (5.3) 1.00 1.41 (0.8, 2.4) 2.16 (1.2, 3.9) .010

Physical performance limited extensively by health in: Vigorous activities 6005 (31.2) 1.00 1.11 (0.9, 1.4) 1.49 (1.1, 2.0) .002 Moderate activities 6031 (8.4) 1.00 1.31 (0.9, 2.0) 2.54 (1.6, 4.0) < .001 Lifting or carrying groceries 6019 (5.9) 1.00 2.63 (1.3, 5.2) 6.17 (3.1,12.5) < .001 Climbing several flights of stairs 6027 (16.9) 1.00 2.11 (1.5, 3.0) 3.63 (2.5, 5.3) < .001 Climbing 1 flight of stairs 6003 (4.8) 1.00 3.19 (1.4, 7.2) 8.16 (3.5,19.0) < .001 Bending, kneeling, stooping 6039 (10.9) 1.00 2.05 (1.3, 3.2) 3.94 (2.5, 6.3) < .001 Walking more than half a mile 6029 (14.8) 1.00 1.64 (1.2, 2.3) 2.86 (1.9, 4.2) < .001 Walking half a mile 5954 (10.0) 1.00 1.92 (1.2, 3.0) 3.63 (2.2, 5.9) < .001 Walking 100 yards 5960 (4.1) 1.00 4.23 (1.6, 11.5) 9.05 (3.2, 25.2) < .001 Bathing and dressing oneself 6052 (3.4) 1.00 3.22 (1.2, 8.8) 9.00 (3.2, 25.1) < .001

Activities of daily living Unable to do:

Cutting toenails 6111 (8.6) 1.00 1.60 (1.0, 2.5) 3.21 (2.0, 5.2) < .001 Cooking a hot meal 6078 (4.6) 1.00 1.15 (0.7, 1.9) 1.86 (1.0, 3.3) .015 Light housework, simple repairs 6098 (3.5) 1.00 1.50 (0.8, 3.0) 3.17 (1.5, 6.6) < .001

Unable to do or difficulty witha: Dressing self 6106 (6.3) 1.00 1.69 (1.0, 2.8) 2.64 (1.5, 4.6) .001 Going up and down stairs/steps 6104 (17.4) 1.00 1.81 (1.3, 2.5) 3.08 (2.2, 4.4) < .001

Note. OR = odds ratio; CI = confidence interval. aToo few were unable to do the task to allow the outcome to be modeled.

FIGURE 1—Prevalence of poor outcomes (%) at resurvey, by employment grade at baseline: Whitehall Study, 1997–1998.

heart attack and baseline employment grade. These 2 conditions were also associated with poor health (odds ratios of 2.5 and 3.5, re- spectively, after adjustment for all baseline health indicators, age, and employment grade) and poor physical performance (ORs of 1.9 and 2.1, respectively). Heart attack was also associated with disability (OR = 1.5) and poor mental health (OR = 1.3). However, the asso- ciations between employment grade and these outcomes were essentially unchanged when

experience of angina and heart attack was taken into account.

Discussion

The survivors of the 1960s Whitehall co- hort were mostly in good health, with only 21% having any of the morbidity outcomes. Each of 4 self-reported morbidity outcomes was more prevalent among men in lower Civil Ser-

vice employment grades than among men in high grades nearly 30 years after screening. Men in the low employment grades had a 4- fold risk of physical performance limited by health, a 3-fold risk of poor health, and more than a 2-fold risk of poor mental health and disability.

Previous research has shown that com- bining socioeconomic indicators yields stronger gradients in mortality than using a single mea- sure.1,18 In the present analysis, neither car own- ership nor housing tenure in middle age added to the predictive power of employment grade with regard to the 4 outcomes.

Before it is concluded that SES in mid- dle age is responsible for the associations found, possible biases should be considered. First, the response rate was lower among those in the low employment grades. Because baseline data were available for nonrespondents, we assessed the implications of this difference. We ran mod- els assuming that nonrespondents who had any other risk factors (e.g., heavy smoking or high body mass index) would have experienced the adverse morbidity outcomes. Under these as- sumptions, those in the low grades still had more than 3 times the risk of poor physical per- formance and twice the risk of the other mor- bidity outcomes. Although the assumptions were crude, they suggest that nonresponse dif- ferentials did not substantially bias the esti- mated effects of employment grade.

February 2001, Vol. 91, No. 2 American Journal of Public Health 281

TABLE 4—Odds Ratios for Outcomes, by Baseline Employment Grade, Adjusted for Age at Resurvey and Other Independent Baseline Risk Factors: Whitehall Study, 1997–1998

Adjusted for Age, Fully Adjusted,a

Outcome Baseline Grade OR (95% CI) OR (95% CI)

Rated health as poor/very poor* (n = 6951) High 1.00 1.00 Middle 1.75 (1.0, 3.0) 1.62 (0.9, 2.8) Low 3.06 (1.7, 5.5) 2.50 (1.4, 4.5)

Poor mental health score* (n = 5921) High 1.00 1.00 Middle 1.10 (0.7, 1.6) 1.05 (0.7, 1.5) Low 2.19 (1.4, 3.4) 1.88 (1.2, 2.9)

Poor physical performance score* (n = 5965) High 1.00 1.00 Middle 2.04 (1.3, 3.3) 1.93 (1.2, 3.1) Low 4.32 (2.6, 7.2) 3.67 (2.2, 6.2)

Unable to do at least 1 activity of daily living* (n = 6080) High 1.00 1.00 Middle 1.22 (0.8, 1.7) 1.15 (0.8, 1.6) Low 2.36 (1.6, 3.5) 2.05 (1.4, 3.1)

Note. OR = odds ratio; CI = confidence interval. aThe models included adjustment for the following baseline factors found to be independently associated with outcomes: self-rated health

(age, clinical signs of cardiovascular disease, top quintile diastolic blood pressure, body mass index > 30 kg/m2, respiratory symptoms, ever hospitalised at least 4 times for reasons other than cardio-respiratory disease, smoking habit) ; mental health score (age, married or not, smoking habit); physical performance score (age, high body mass index, respiratory symptoms, hospitalised for non cardio-respiratory disease, smoking habit); disability (age, high body mass index, high blood sugar level/diabetic, respiratory symptoms, smoking habit).

*P < .001.

TABLE 5—Association of Selected Health Outcomes With Characteristics After Retirement, by Employment Grade: Whitehall Study Resurvey, 1997–1998

Employment Gradea

Middle, OR Low, OR Interaction Outcome Characteristic at Resurvey (95% CI) (95% CI) P

Poor mental health score Income < $16 500 (vs higher) 1.95 (1.4, 2.8) 0.81 (0.5, 1.4) .012 Higher grade category at retirement (vs same/lower) 0.82 (0.6, 1.0) 0.44 (0.3, 0.8) .033

Poor physical performance score Income < $16 500 (vs higher) 2.05 (1.5, 2.9) 1.05 (0.6, 1.7) .020 Higher grade category at retirement (vs same/lower) 1.00 (0.8, 1.3) 0.64 (0.4, 1.0) .19

Unable to do at least 1 activity Income < $16 500 (vs higher) 1.79 (1.3, 2.4) 0.98 (0.6, 1.6) .027 of daily living Higher grade category at retirement (vs same/lower) 0.92 (0.7, 1.1) 0.71 (0.4, 1.1) .46

Note. Odds ratios were adjusted for age and independent risk factors. OR = odds ratio; CI = confidence interval. aToo few of those in high grades had low incomes to allow separate analyses, and, by definition, they could not rise a category.

Second, we considered the possibility that men in the lower employment grades might have a more negative outlook generally. Ex- cluding those who reported being “nervous most of the time” or “happy little of the time” did not substantially alter the results (data not shown).

While self-reported measures are subjec- tive, they are predictive of mortality independ- ently of clinical health.24,25 McCallum et al.26

attributed their finding of a contrary effect to individuals’ basing their subjective ratings on objective comorbidities and disability. Self- reported functional status has also been associ- ated with mortality in old age.27,28 Methodo- logical studies of the SF-36 suggest that it is reasonably sensitive to lower levels of morbid- ity,29 that it is reliable and internally consistent,30

and that it is suitable for use with older people.31

There are several possible explanations for an employment grade differential in old age. First,

ill health could precede low socioeconomic sta- tus. However, health disadvantages in middle age seem to be an unlikely explanation of differentials in old age. After adjustment for baseline health, behavior, and marital status, the odds ratios for re- spondents in low vs high employment grades were 3.7 (95% CI=2.2, 6.2) for poor physical performance, 2.5 (95% CI = 1.4, 4.5) for poor health, 2.0 (95% CI=1.4, 3.1) for disability, and 1.9 (95% CI=1.2, 2.9) for poor mental health. Participants in the resurvey had already survived nearly 30 years. Only 18% of the original low- grade cohort members could take part, most hav- ing died. By definition, the survivors must have been less vulnerable to fatal disease than their deceased colleagues, yet those in the low em- ployment grades were still more likely to have severe morbidity in old age than those who had been in the higher grades in middle age.

Second, there could have been a cumulation of psychologic stress affecting biological coping

mechanisms (e.g., cortisol production, decrease in parasympathetic activity).32 In a later cohort of civil servants (Whitehall II), degree of control in one’s job explained a substantial proportion of differences in coronary heart disease incidence among the different grades33,34 and was associ- ated with psychiatric disorders.35 This could not be tested with the Whitehall I cohort.

Third, there could have been cumulating disadvantages in regard to material resources, opportunities to promote health, and lifestyle between the baseline and resurvey. We have information on the cohort at only 2 points in time. There was some evidence that circum- stances arising in later life could add to or ame- liorate disadvantages. Having a low income exacerbated health problems for middle-grade staff, whereas rising a grade category amelio- rated risk of poor mental health among staff who had been in the low employment grades. Although we cannot rule out a health selection

February 2001, Vol. 91, No. 2282 American Journal of Public Health

effect, we do not believe that it wholly accounts for the differences. Men in middle employ- ment grades who had low incomes in retire- ment were slightly more likely to have left the Civil Service for medical reasons (7% vs 4%) or because of redundancy (17% vs 15%), but these differences were not sufficient to account for a 2-fold increase in risk. While being men- tally fit might have increased the chances of rising a grade, the greater job control in a higher employment grade might have improved men- tal health.

The socioeconomic differentials found in this study probably underestimate those in the general population in that all of the men in the cohort had experienced relatively good em- ployment and pension provisions in the Civil Service. Moreover, the resurvey respondents had better self-perceived health than that re- ported in other studies. The mean scores for the mental health and physical performance scales were 82.1% and 77.3%, respectively, as com- pared with 79.7% and 64.4% found in popula- tion studies in 3 local districts in Britain36 and mean scores ranging from 68% to 73% and 54% to 72% in 6 localities in outer London.37

The findings in this article add to our pre- viously reported evidence2,9 of long-term so- cioeconomic effects on morbidity. Moreover, the further differentiation in outcomes by SES in retirement suggests that there is a con- tinuing accumulation of disadvantage in old age. Strong socioeconomic differentials were found among the survivors of a privileged and relatively healthy group.

Contributors E. Breeze helped design the documents, carried out the analyses, and drafted the paper. A. E. Fletcher, D. A. Leon, M. G. Marmot, R. J. Clarke, and M. J. Ship- ley all commented on drafts and approved the final version. R. J. Clarke was instrumental in initiating and designing the resurvey and coordinating the field- work; all of the authors participated in the steering committee for the study.

Acknowledgments The British Heart Foundation funded the resurvey, in- cluding support for Elizabeth Breeze and Martin J. Shipley. Michael J. Marmot was supported by a Med- ical Research Council research professorship. The sur- vey was approved by the ethics committees of the Lon- don School of Hygiene and Tropical Medicine, the University of Oxford, and University College London.

We would like to thank all of the participants who completed questionnaires. Assistance provided by staff from the health authorities and the Office for National Statistics was invaluable. We also grate- fully acknowledge the contributions of the team at the Clinical Trials Unit in Oxford (Rory Collins, Dr Linda Youngman, Pamela Bell, Paul Sherliker, and Smita Shah).

References 1. Goldblatt P. Mortality and alternative social clas-

sifications. In: Goldblatt O, ed. 1971–1981 Lon- gitudinal Study. Mortality and Social Organi- sation. London, England: Her Majesty’s Stationery Office; 1990:174–176.

2. Breeze E, Sloggett A, Fletcher A. Socioeco- nomic and demographic predictors of mortality and institutional residence among middle-aged and older people: results from the Longitudinal Study. J Epidemiol Community Health. 1999; 53:765–774.

3. Marmot MG, Shipley MJ. Do socioeconomic differences in mortality persist after retirement? 25 year follow-up of civil servants from the first Whitehall Study. BMJ. 1996;313:1177–1180.

4. Hattersley L. Expectation of life by social class. In: Drever F, Whitehead M, eds. Health In- equalities: Decennial Supplement. London, Eng- land: Her Majesty’s Stationery Office; 1997: 73–82. DS series 15.

5. Smith J, Harding S. Mortality of women and men using alternative social classifications. In: Drever F, Whitehead M, eds. Health Inequali- ties: Decennial Supplement. London, England: Her Majesty’s Stationery Office; 1997:168–185. DS series 15.

6. Kallan J. Effects of sociodemographic variables on adult mortality in the United States: com- parisons by sex, age and cause of death. Soc Biol. 1997;44:136–147.

7. Arber S, Ginn J. Gender and inequalities in health in later life. Soc Sci Med. 1993;36:33–46.

8. Prescott-Clarke P, Primatesta P. Health Survey for England 1995: Volume 1. Findings. London, England: Her Majesty’s Stationery Office; 1997.

9. Breeze E, Sloggett A, Fletcher A. Socioeco- nomic status and transitions in status in old age in relation to limiting long-term illness mea- sured at the 1991 census. Results from the UK Longitudinal Survey. Eur J Public Health. 1999; 9:265–270.

10. Marmot MG, Shipley MJ, Rose G. Inequalities in death—specific explanations of a general pat- tern? Lancet. 1984;1:1003–1006.

11. Marmot MG, Smith GD, Stansfeld S, et al. Health inequalities among British civil servants: the Whitehall II Study. Lancet. 1991;337: 1387–1393.

12. Hemingway H, Nicholson A, Stafford M, Roberts R, Marmot M. The impact of socio- economic status on health functioning as as- sessed by the SF-36 questionnaire: the White- hall II Study. Am J Public Health. 1997;87: 1484–1490.

13. Reid DD, Hamilton PJS, Keen H, Brett GZ, Jar- rett RJ, Rose G. Cardiorespiratory disease and diabetes among middle-aged male civil servants. Lancet. 1974:469–473.

14. Clarke R, Breeze E, Sherliker P, et al. Design, objectives, and lessons from a pilot 25 year fol- low up resurvey of survivors in the Whitehall Study of London civil servants. J Epidemiol Community Health. 1998;52:364–369.

15. Ware JE, Sherbourne CD. The MOS 36-Item Short-Form Health Survey (SF-36), I: concep- tual framework and item selection. Med Care. 1992;30:473–483.

16. Ware JE, Snow KK, Kosinski M, Gandek B. SF- 36 Health Survey: Manual and Interpretation Guide. Boston, Mass: Health Institute, New Eng- land Medical Center; 1993:18–19.

17. Morris R, Carstairs V. Which deprivation? A comparison of selected deprivation indices. J Public Health Med. 1991;13:318–326.

18. Davey Smith G, Shipley MJ, Rose G. Magni- tude and causes of socioeconomic differentials in mortality: further evidence from the Whitehall Study. J Epidemiol Community Health. 1990; 44:265–270.

19. Bulpitt CJ, Shipley MJ, Demirovic J, Ebi-Krys- ton KL, Markowe HLJ, Rose G. Predicting death from coronary heart disease using a question- naire. Int J Epidemiol. 1990;19:899–904.

20. Pinsky JL, Branch LG, Jette AM, et al. Fram- ingham Disability Study: relationship of dis- ability to cardiovascular risk among persons free of diagnosed cardiovascular disease. Am J Epi- demiol. 1985;122:644–656.

21. Jette AM, Pinsky JL, Branch LG, Wolf PA, Fein- leib M. The Framingham Disability Study: phys- ical disability among community-dwelling sur- vivors of stroke. J Clin Epidemiol. 1988;8: 719–726.

22. Pinsky JL, Jette AM, Branch LG, Kannel WB, Feinleib M. The Framingham Disability Study: relationship of various coronary heart disease manifestations to disability in older persons liv- ing in the community. Am J Public Health. 1990; 80:1363–1367.

23. Stata Statistical Software Release 5.0. College Station, Tex: Stata Corp; 1997.

24. Mossey JM, Shapiro E. Self-rated health: a pre- dictor of mortality among the elderly. Am J Pub- lic Health. 1982;72:800–808.

25. Schoenfeld DE, Malmrose LC, Blazer DG, Gold DT, Seeman TE. Self-rated health and mortal- ity in the high-functioning elderly—a closer look at healthy individuals: MacArthur field study of successful aging. J Gerontol. 1994;49: M109–M115.

26. McCallum J, Shadbolt B, Wang D. Self-rated health and survival: a 7-year follow-up study of Australian elderly. Am J Public Health. 1994; 84:1100–1105.

27. Scott WK, Macera CA, Cornman CB, Sharpe PA. Functional health status as a predictor of mortality in men and women over 65. J Clin Epidemiol. 1997;50:291–296.

28. Grand A, Grosclaude P, Bocquet H, Pous J, Al- barede JL. Disability, psychosocial factors and mortality among the elderly in a rural French population. J Clin Epidemiol. 1990;43:773–782.

29. Brazier JE, Walters SJ, Nicholl JP, Kohler B. Using the SF-36 and Euroqol on an elderly pop- ulation. Qual Life Res. 1996;5:195–204.

30. Andresen EM, Bowley N, Rothenberg MPA, Panzer R, Katz P. Test-retest performance of a mailed version of the Medical Outcomes Study 36-Item Short-Form Health Survey among older adults. Med Care. 1996;34:1165–1170.

31. Lyons RA, Perry HM, Littlepage BNC. Evi- dence for the validity of the Short-Form 36 ques- tionnaire (SF-36) in an elderly population. Age Ageing. 1994;23:182–184.

32. Brunner E. Stress and the biology of inequality. BMJ. 1997;314:1472–1476.

33. Bosma H, Marmot MG, Hemingway H, Nichol- son AC, Brunner E, Stansfeld SA. Low job con- trol and risk of coronary heart disease in the Whitehall II (prospective cohort) Study. BMJ. 1997;314:558–565.

February 2001, Vol. 91, No. 2 American Journal of Public Health 283

34. Marmot MG, Bosma H, Hemingway H, Brunner E, Stansfeld S. Contribution of job control and other risk factors to social variations in coronary heart disease incidence. Lancet. 1997;350:235–239.

35. Stansfeld SA, North FM, White I, Marmot MG.

Work characteristics and psychiatric disorder in civil servants in London. J Epidemiol Commu- nity Health. 1995;49:48–53.

36. Lyons RA, Crome P, Monaghan S, Killalea D, Daley JA. Health status and disability among

elderly people in three UK districts. Age Age- ing. 1997;26:203–209.

37. Ong BN, Jordan K. Health experiences of eld- erly people in an outer London area. Soc Res Online [serial online]. March 31, 1997; doc 2.