Inequalities in self rated health in the 1958 birth cohort: lifetime social circumstances or social mobility?BMJ 1996; 313 doi: http://dx.doi.org/10.1136/bmj.313.7055.449 (Published 24 August 1996) Cite this as: BMJ 1996;313:449
- Chris Power, senior lecturer in epidemiologya,
- Sharon Matthews, research assistanta,
- Orly Manor, lecturer in statisticsb
- a Department of Epidemiology and Biostatistics, Institute of Child Health, London WC1N 1EH
- b School of Public Health and Community Medicine, Hebrew University, Jerusalem, Israel
- Correspondence to: Dr Power.
- Accepted 15 July 1996
Objective: To investigate explanations for social inequalities in health with respect to health related social mobility and cumulative socioeconomic circumstances over the first three decades of life.
Design: Longitudinal follow up.
Setting: Great Britain.
Subjects: Data from the 1958 birth cohort study (all children born in England, Wales, and Scotland during 3-9 March 1958) were used, from the original birth survey and from sweeps at 16, 23, and 33 years.
Main outcome measures: Subjects' own ratings of their health; social differences in self rated health at age 33.
Results: Social mobility varied by health status, with those reporting poor health at age 23 having higher odds of downward mobility than of staying in same social class. Men with poor health were also less likely to be upwardly mobile. Prevalence of poor health at age 33 increased with decreasing social class: from 8.5% in classes I and II to 17.7% in classes IV and V among men, and from 9.4% to 18.8% among women. These social differences remained significant after adjustment for effects of social mobility. Health inequalities attenuated when adjusted for social class at birth, at age 16, or at 23 or for self rated health at age 23. When adjusted for all these variables simultaneously, social differences in self rated health at age 33 were substantially reduced and no longer significant.
Conclusions: Lifetime socioeconomic circumstances accounted for inequalities in self reported health at age 33, while social mobility did not have a major effect on health inequalities.
We examined the effects of cumulative exposure and social mobility on health at age 33 in a national birth cohort
Cumulative lifetime exposures, as represented by social class at birth and at ages 16 and 23 and prior ill health, seemed to have a major role in creating health inequalities
Health and social mobility were related (young adults with poor health tending to move downward), but social mobility was a minor explanation for inequalities in health at age 33
Future research on health inequalities needs to take account of factors operating throughout life
Commenting on systematic variations in mortality and morbidity across social groups in the United Kingdom, a recent Department of Health report concludes: “It is likely that cumulative differential exposure to health damaging or health promoting physical and social environments is the main explanation for observed variations in health and life expectancy, with health related social mobility, health damaging or health promoting behaviours, use of health services, and genetic or biological factors also contributing.”1 The report therefore recognises the role of health related social mobility (whereby unhealthy people drift down the social scale and healthy people drift up) but places this as secondary to cumulative differential exposures.
Investigating cumulative differential exposure is not without its challenge. One particular problem is the lack of appropriate health measures to indicate poor health at ages when mortality is uncommon. None the less, it is increasingly recognised that self rated health provides a useful measure of health status because it is associated with fitness2 and morbidity3 and predicts mortality.4 5 6 7 Thus, many studies of social differences in health focus on this measure.8 9
We investigated explanations for social inequalities in self rated health among 33 year olds in the 1958 birth cohort. We focused on the relative importance of health related social mobility and of cumulative socioeconomic circumstances. Because the sample had been studied longitudinally, there was information on social class at different life stages. We used these data, firstly, to indicate if and when social mobility had occurred and, secondly, to provide a proxy for the cumulative socioeconomic circumstances experienced by different social groups.
The 1958 birth cohort includes all children born in England, Wales, and Scotland during 3-9 March 1958. The study originated in the perinatal mortality study, the aim of which was to determine the social and obstetric factors associated with stillbirth and death in early infancy. Information was collected on 98% of births, totalling 17 414. Five subsequent follow up studies have been conducted at ages 7, 11, 16, 23, and 33, with 11 407 subjects included in the most recent sweep.10 As expected, sample attrition has been associated with underrepresentation of those with the most disadvantaged backgrounds. Such biases tend to be small: 19.4% of men responding at age 33 had been born into classes IV and V compared with 21.1% in the original sample; for women the figures are 20.5% and 21.4% respectively. Thus, those remaining in the study are considered to be generally representative of the original sample.3 10
Cohort members gave an overall assessment of their health during a personal interview at ages 23 and 33: they rated their health as excellent, good, fair, or poor. Small numbers rated their health as poor (112 and 198 at ages 23 and 33 respectively), and consequently this group was combined with those rating their health as fair. The combined group with fair or poor health rating is referred to here as having poor health.
Social position is indicated by social class based on occupation at different ages. Two measures of social class represent social origins—father's occupation at the time of the respondent's birth (using the registrar general's 1950 classification) and father's occupation when the subject was aged 16. To reduce the effects of sample attrition due to data missing at age 16, father's social class at age 11 was used if data at 16 years were not available. Two social class measures define the social destinations of the cohort members—own social class at ages 23 and 33 (using the registrar general's 1980 and 1990 classifications respectively). To avoid small numbers in the extreme groups, social class at each age was collapsed from the original six categories to four—that is, I and II, III non-manual, III manual, and IV and V. Both men and women were allocated to a social class on the basis of their own occupation.
At ages 23 and 33, most subjects were classified on the basis of a current occupation. For men, 85% of 23 year olds and 91% of 33 year olds had an occupation at the time of the interview; the figures for women were 69% and 72% respectively. The remainder was classified according to their most recent occupation. The percentage varied by social class. For example, at age 23 social class was based on current employment for 93% of those in classes I and II and for 69% of those in classes IV and V. The main results were unaffected by such differences.
Analyses were undertaken to assess whether there was a strong relation between self rated health at ages 23 and 33, whether social mobility was related to health at age 23, and whether social mobility explained the health gradient at age 33. Firstly, the association between self rated health at the two ages was assessed using gamma test.11 Secondly, the relation between health and social mobility was expressed as an odds ratio for subjects with poor health at age 23.12 Odds were estimated for moving upward or downward, relative to remaining in the same social class at age 23. Thirdly, to determine the effect of health related social mobility on the health gradient, we compared the gradient for the full sample with that for a subsample of subjects who were healthy at baseline (in this case age 23). In the subsample social gradients in health could not be attributed to health related social mobility because previous poor health had been excluded. The magnitude of the health gradient was summarised using the slope as an index of inequality, based on logistic regression and represented as an odds ratio (Power et al, unpublished data). The odds ratio for the slope provides an estimate of the proportional increase from the top and bottom of the social class scale.
Further analyses were made to test the influence of lifetime socioeconomic circumstances and social mobility on differences in self reported health at age 33. Thus, several logistic regression models were constructed to establish the effects of social class at birth and at ages 16 and 23. A potential problem in these logistic models is collinearity between the coefficients of the social class variables (because these variables could be strongly correlated), which may impose difficulties in separating the effects of each variable.13 14 We assessed the size of this problem and its impact on the estimated coefficients and found that the results were not affected by multicollinearity. The role of intergenerational mobility was examined by including an interaction term between father's social class at 16 and own class at 33, while that of intragenerational mobility was examined by including an interaction between social class at 23 and at 33. Estimates derived from these models are expressed as odds ratios of poor health in each class, with classes I and II as the reference category. The Wald statistic was used to test the relation between self rated health and class at age 33.12 The related P value indicates the significance of this relation before and after adjustment for other variables. All analyses were confirmed with the slope index.
ASSOCIATION BETWEEN SELF RATED HEALTH OVER TIME
At age 23, 9% of subjects (8% of men and 10% of women) rated their health as poor, and this increased to 12.9% at age 33 (12.5% of men and 13.2% of women). Table 1 shows strong associations between self rated health at the two ages, in both men and women. Overall, 57% of subjects had the same health rating, 27% reported a deterioration, and 15% reported an improved rating.
HEALTH RELATED SOCIAL MOBILITY
Table 2 shows the effect of health status at age 23 on subsequent social mobility. Odds of downward mobility by age 33 (relative to staying in the same social class) were consistently higher for men and women who reported poor health at age 23, particularly for those in classes I and II at age 23. In addition men who reported poor health showed consistently low odds ratios for upward mobility; for women, however, odds ratios for upward mobility did not differ significantly from unity. However, the numbers of subjects who were socially mobile, particularly downwards, and had poor health were small in comparison with the total sample.
HEALTH RELATED SOCIAL MOBILITY AND SOCIAL GRADIENTS
Table 3 shows that the prevalence of subjects in the full sample who reported poor health increased with decreasing social class at age 33: from 8.5% in classes I and II to 17.7% in classes IV and V among men, and from 9.4% to 18.8% among women. A gradient was also apparent for the subsample, which excluded subjects who reported poor health at age 23. The prevalence of reported poor health at age 33 was reduced in all classes in the subsample because of the strong association between self rated health at both ages (table 1), but the size of the gradient was similar (for men the slope index (odds ratio) was 3.21 for the full sample and 3.14 for the subsample; for women it was 3.16 and 3.14 respectively). Thus, the social gradient in reported poor health was not due to health related social mobility, because the size of the gradient was unaffected by excluding those with prior poor health.
We conducted further analyses to examine the impact of social mobility on health differences at age 33 and to assess the role of lifetime socioeconomic circumstances. Table 4 shows the odds of reported poor health for social classes III non-manual to IV and V relative to social classes I and II. These odds ratios were adjusted for prior self rated health (at age 23) and for social class at birth and at ages 16 and 23: each of these variables had some effect on the odds ratios (table 4). In general, the differences between social classes showed greatest reduction after adjustment for the most recent measure of social class. For example, for men in social classes IV and V, the odds of reported poor health fell from 2.52 to 1.85 after adjustment for class at age 23. Social differences in health at age 33 remained significant after adjustment for a single variable. However, when adjustment was made for all variables simultaneously the social differences in health were no longer significant. This would suggest that inequalities in self rated health at age 33 could be explained by the combined effect of socioeconomic circumstances at different life stages.
The effects of social mobility on the social gradients in poor health were analysed with interaction terms for intergenerational mobility (between father's social class at age 16 and own class at age 33) and intragenerational mobility (between own class at age 23 and class at age 33). If social mobility was a major explanation for differences in health, adjustment for these interactions should reduce the odds ratios substantially. Table 4 shows that this was not the case and significant differences in health remained after adjustments. For example, the odds ratio for men in classes IV and V was 2.49 after adjustment for the interaction term for intergenerational mobility, which is actually higher than the odds ratio of 2.23 when adjustment was made for only social class at age 16.
Our study suggests that health inequalities are due to differential lifetime socioeconomic circumstances and not primarily to health related social mobility. Although health status influences subsequent social mobility, people who are socially mobile because of poor health, especially those moving downward, are numerically too small to influence the overall levels of health in the social class they join.
LIMITATIONS OF SELF RATED HEALTH
These conclusions pertain to self rated health. As a subjective measure, this is likely to be affected by response set, whereby at a given level of objective ill health individuals report different levels of subjective ill health. If the social classes reported their health differently we would not be able to estimate reliably the size of health inequality. Although there is some evidence that, at a given level of objective ill health, lower income groups report worse health than higher income groups,15 others have found that the predictive effect of self rated health on mortality is similar in manual and non-manual groups.5 Self rated health is of interest because it is associated with other measures of current health and because it predicts later mortality.4 5 6 7 Even so, other health measures need to be examined, and further analyses of these data are planned.
CUMULATIVE DIFFERENTIAL EXPOSURE TO HEALTH RISKS
Although few studies have examined cumulative differential exposure to health risks, some evidence does exist. At a baseline survey in 1985-8, the Whitehall II study showed systematic variation in potential biological, behavioural, and psychosocial risk factors across grades of civil service employees, with adverse factors clustering in the lower grades.16 It is also relevant that multiple socioeconomic factors are associated with subsequent mortality in longitudinal analyses.17 18
Studies of the effects of lifetime socioeconomic circumstances are less common and only rarely reach back to childhood. An exception is Forsdahl's study, which used infant mortality as an index of deprivation in childhood.19 Forsdahl found that deaths from ischaemic heart disease in adults in the 1970s were correlated with the level of infant mortality around the time of birth, suggesting that early socioeconomic environment was relevant to adult health. Adverse effects of poor socioeconomic circumstances in childhood have been shown in more recent studies of morbidity and mortality in men, although effects on women seem to be inconsistent.20 21 Compared with socioeconomic circumstances in adulthood, childhood circumstances seem to have a weak effect on adult mortality. Nevertheless, social origins are likely to influence adult mortality through their effect on adult circumstances.22 23
Socioeconomic careers are, however, rarely considered in relation to subsequent adult health. An exception is Mare's analysis of socioeconomic status at several stages of people's life history and their subsequent mortality.24 This showed that socioeconomic origins have enduring effects on adult mortality through relations with later experiences, such as education, and subsequently with occupational and financial resources. The effect of cumulative socioeconomic circumstances is reflected in the finding that, in comparison with current occupation, a greater predictive effect on adult mortality is associated with longest held occupation.25 Similar cumulative effects seem to be operating during childhood. In a study of childhood growth, measures of persistent poverty were associated with childhood stunting (low height for age) and wasting (low weight for age), but single year measures of income were not.26 Investigations of lifetime socioeconomic circumstances also show that mortality risks associated with lower social class origins are to some extent ameliorated by higher socioeconomic position in later life.24 27 These studies underscore the need to take account of the accumulation of socioeconomic factors throughout life.
In interpreting our findings, it is important to recognise that social class is acting as a proxy for diverse factors in the social and physical environment. The use of class as a proxy in this context is justified because it is known to be linked to economic circumstances, reflecting earnings and conditions, career prospects, authority, and other aspects of work and market situations and, hence, access to other goods.28 29 There is also abundant evidence that social class differentiates groups according to factors relevant to health, such as smoking, breast feeding, and diet.30 However, the registrar general's social classification has several inadequacies as a measure of status, and several alternative classifications have been proposed.28 We will examine more specific factors in the social and physical environment, and thereby extend our research on social differences in self rated health at age 23.3
We have previously studied health related social mobility, focusing on intergenerational mobility (that is, from class of origin to that based on own social position at age 23).3 31 People who were upwardly mobile had better health, while downwardly mobile people had poorer health. Other studies have shown similar trends for intergenerational mobility (Rahkonen et al, unpublished data).32 Even so, mobility in the 1958 cohort did not seem to be a major factor in creating and maintaining social inequalities in health during early adulthood.3 33
Our present study extends our earlier analyses, in which the extent of intergenerational mobility may have been underestimated because many people had only just started their careers by age 23 and would therefore not have achieved their social destinations. None the less, our findings confirm the earlier results in showing a negligible effect of intergenerational social mobility on health inequalities. Our examination of health and intragenerational mobility (that is, between social class at ages 23 and 33) provides additional evidence for a downward drift among those with poorer health and, in men at least, an upward drift among those with better health. Similar trends have been reported elsewhere.34 Again, however, social mobility did not have a major impact on health inequalities.
Unlike some analyses that focused specifically on health related social mobility,35 36 the interactions we used denoted all social mobility. Thus, our analysis incorporated social mobility linked to health and that linked to other factors possibly reflecting health potential. When intragenerational social mobility was taken into account, health inequalities in women did not seem to attenuate and the slightly attenuated health differences in men were still significant. These results for social mobility need to be interpreted with those relating specifically to health, provided here in a comparison of the subsample who did not rate their health as poor at baseline and the full sample. This comparison suggests that health related social mobility did not exert a strong effect even among the men. This finding adds to evidence from other studies suggesting that health related social mobility has only a minor impact on health inequalities at several life stages.37
Funding CP is supported by the Canadian Institute for Advanced Research as a Weston Fellow. The study was supported by the Economic and Social Research Council (ESRC award No R000235189).
Conflict of interest None.