Social class in childhood and general health in adulthood: questionnaire study of contribution of psychological attributesBMJ 1999; 318 doi: https://doi.org/10.1136/bmj.318.7175.18 (Published 02 January 1999) Cite this as: BMJ 1999;318:18
- Hans Bosma, senior researcher (, )
- H Dike van de Mheen, senior researcher,
- Johan P Mackenbach, professor.
- Erasmus University Rotterdam, Department of Public Health, PO Box 1738, 3000 DR Rotterdam, Netherlands
- Correspondence to: Dr Bosma
- Accepted 20 October 1998
Objective: To determine the contribution of psychological attributes (personality characteristics and coping styles) to the association between social class in childhood and adult health among men and women.
Design: Partly retrospective, partly cross sectional study conducted in the framework of the Dutch GLOBE study.
Subjects: Sample of general population from south east Netherlands consisting of 2174men and women aged 25-74years. Baseline self reported data from 1991provided information on childhood and adult social class, psychological attributes, and general health.
Main outcome measure: Self rated poor health.
Results: Independent of adult social class, low childhood social class was related to self rated poor health (odds ratio 1.67(95% confidence interval 1.02to 2.75) for subjects whose fathers were unskilled manual workers versus subjects whose fathers were higher grade professionals). Subjects whose fathers were manual workers generally had more unfavourable personality profiles and more negative coping styles. External locus of control, neuroticism, and the absence of active problem focused coping explained about half of the association between childhood social class and self rated poor health. The findings were independent of adult social class and height.
Conclusions: A higher prevalence of negative personality profiles and adverse coping styles in subjects who grew up in lower social classes explains part of the association between social class in childhood and adult health. This finding underlines the importance of psychological mechanisms in the examination of the negative effects of adverse socioeconomic conditions in childhood.
Regardless of adult social class, low social class in childhood is related to poor general health in adulthood
Adverse personality profiles and negative coping styles are more common in people who grew up in lower social classes
Psychological attributes, such as low perceived control, explain a substantial part of the direct association between childhood social class and adult health
Psychological mechanisms may explain adverse health outcomes in adults who have a low socioeconomic background
There is now increasing evidence that differences in adult health are partly caused by socioeconomic factors during early life and upbringing. 1 1The causal mechanisms relating adverse socioeconomic conditions in childhood and poor health in adulthood have not yet been examined extensively. Biological mechanisms, such as hypertension and hypercholesterolaemia, have been suggested,6–8 and there is some evidence for mechanisms related to lifestyle, such as smoking and physical activity. 3 9 12 With few exceptions the contribution of psychological attributes, such as personality factors and coping styles, has hardly been examined. 10–13 14 Psychological attributes are partially rooted in environmental conditions in childhood,(learning) experiences, and rearing styles.14–16 There is now also increasing evidence that psychological attributes influence health through behavioural mechanisms (for example, smoking) or direct physiological mechanisms, or both.17 Unhealthy personality factors and coping strategies may, therefore, be mechanisms through which adverse socioeconomic conditions in childhood contribute to poor health in adulthood. Baseline data from the Dutch GLOBE study (a Dutch study on health and living conditions of the population of Eindhoven and its surroundings) allowed us to examine whether childhood social class is related to psychological attributes and whether psychological attributes are mediating factors in the association between social class in childhood and adult health. To obtain the direct effects, we accounted for adult social class separately.18
Subjects and methods
Data were collected within the framework of the GLOBE study. A postal survey was conducted in 1991among 27 070 non-institutionalised inhabitants (aged 15-74years) of Eindhoven and a number of surrounding municipalities, all in the south eastern part of the Netherlands. Stratified by age and post code, the sample was randomly drawn from the municipal population registries. People aged 45-70years old and people from the highest and lowest social classes were overrepresented to allow specific analyses for middle aged subjects and to increase socioeconomic contrasts. The response rate was 70.1%, which resulted in a study population of 18973. The response rates were not substantially different for age, sex, marital status, level of urbanisation, or social class.19 A few months later a subsample of those who responded to the postal questionnaire was approached for a more extensive oral interview. This subsample consisted of 3529randomly chosen respondents to the postal questionnaire, of whom 2802agreed to be interviewed (79.4%). The response rates hardly differed for relevant demographic characteristics. More details on the study design can be found elsewhere.19 To allow for a more valid measurement of adult social class the analyses were restricted to men and women older than 24years (n=2462). The analyses were based on subjects who reported both their father's and their own adult occupational level (n=2174).
Social class, psychological attributes, and health outcome
Adult social class was measured by the respondent's current or last occupational level. Childhood social class was measured retrospectively by the father's occupational level when the respondent was 12years old. If the father was not in paid employment the father's last occupation in paid employment was used. The occupations of both the respondent and the father were classified according to the scheme of Erikson, Goldthorpe, and Portocarero.20 The psychological attributes were assessed by questionnaires on personality characteristics and coping styles. The personality characteristics were external locus of control (low perceived control), 21 22 neuroticism (emotional distress),23 parochialism (a narrow, local, and non-scientific attitude),24–26and orientation towards the future. 25 26 Seven typical styles of coping were distinguished: active problem focusing, avoidance behaviour, depressive reaction pattern, social support seeking, palliative reaction pattern, disclosure of emotions, and optimism.27 The items were summed for the separate scales and then divided into thirds.
A perception of “less than good” general health was used as a general measure of health (n=621; 29%). This was based on the subject's answer to the question: “How do you rate your health in general?” (very good; good; fair; sometimes good and sometimes poor; poor). Perceived general health has been shown to be strongly related to physical health and to survival. 28 29 A “less than good” perceived general health is hereafter referred to as poor general health.
Childhood social class and the psychological attributes were related by logistic regression analysis with adjustment for age and sex. For this analysis the psychological attributes were divided into two categories by combining the lowest two thirds. This analysis provided adjusted estimated percentages. Childhood social class and adult health were also related by logistic regression analysis controlled for age, sex, marital status (married, single, widowed/divorced), religious affiliation (none, Catholic, Protestant/other), and level of urbanisation (four levels). To examine the contribution of psychological attributes to the association of childhood social class and adult health, the psychological attributes (in thirds) were separately introduced into a model with childhood social class and confounders. As we were primarily interested in the direct effects of childhood social class on adult health, all analyses were separately adjusted for adult social class. Self reported height as a proxy for factors in early life was also separately controlled for. As the findings for men and women were similar, we combined data for both and controlled for sex.
Table 1shows that childhood social class was related to personality. External locus of control (42% v 15%), parochialism (35% v 12%), neuroticism (34% v 23%), and lack of future orientation (41% v 27%) were more prevalent in the lowest compared with the highest social classes in childhood. The prevalence decreased almost linearly from lowest to highest class but was somewhat less clear for neuroticism. Low childhood social class was also related to particular coping styles— that is, a lower prevalence of active problem focused coping (20% v 40%), a somewhat more frequent depressive reaction pattern (26% v 18%), avoidance (28% v 22%), less often seeking social support (21% v 36%), and less often expressing emotions (26% v 35%). Except for active problem focused coping, these associations were not linear. When we controlled for adult social class the associations were weaker, but most remained.
Table 2 shows that childhood social class was related to poor general health. Subjects with a low childhood social class were 2.1times as likely to rate their health as poor than subjects with a high childhood social class. Adjustment of this odds ratio for adult social class decreased the odds ratio to 1.67,which was still significant.
Two personality factors and one coping style contributed to the association between childhood social class and adult health: external locus of control, neuroticism, and the absence of active problem focused coping. The other personality and coping factors individually explained less than 10% of the gradient (data not shown). External locus of control had the strongest contribution to the socioeconomic gradient in poor general health. When external locus of control was taken into account, the odds ratio for subjects whose fathers were unskilled manual workers decreased by 46% in the model without adult social class controlled for (odds ratio 1.60in model 1) and by 33% in the model with adult social class controlled for (1.45in model 2). This was closely followed by neuroticism (32% and 34%, respectively). Active problem focused coping explained a smaller part of the increased risk — align=baseline border=0>namely, 18% and 8%, respectively. When the three psychological attributes were considered simultaneously, about half of the association of low childhood social class with poor general health could be explained by a higher prevalence of external locus of control and neuroticism and a lower prevalence of active coping styles in subjects with a low childhood social class.
Height hardly affected any of our findings (data not shown). Less than 7% of the association between childhood social class and adult health was explained by height. Further adjustment for external locus of control, neuroticism, and active coping explained an additional 50% of the association. This is similar to the results of analyses without adjustment for height. Furthermore, we found no evidence for interactions between childhood and adult social class or between childhood or adult social class and personality or coping styles (data not shown).
We have shown that particular personality factors and coping styles substantially contribute to the direct association between social class in childhood and adult health. Subjects whose fathers were unskilled manual workers generally had more unfavourable personality profiles and negative coping styles. External locus of control, neuroticism, and the absence of active problem focused coping explained half of the association between low social class in childhood and poor general health. The lack of influence of height may suggest a contributing role of psychological attributes and not just of fetal development and early growth.7 Given the finding that childhood social class may be particularly related to cardiovascular diseases,1 it is worth mentioning that we found similar results for reporting a severe heart condition or stroke (not shown). These findings, however, were not significant because of small numbers. Our findings suggest that psychological attributes are worth examining when the associations between social class in childhood and adult health are studied. 10 13 14
These findings indicate that personality is partially rooted in childhood social class. Rearing styles differ among social classes, resulting in long term effects on behaviour, emotion, and cognition. 15 16 Children from high class backgrounds may more easily experience and learn a sense of mastery and control because their parents have more resources (for example, money and knowledge).30–32This may underlie our findings with locus of control and active problem focused coping. Similarly, neuroticism may also reflect or be the consequence of a perceived lack of control over outcomes and events.33 Neurotic people may more easily internalise emotions instead of taking problem oriented approaches. Our findings emphasise the importance of control related psychological factors for the development of socioeconomic inequalities in health.34 Further research should examine whether perceived control is related to physical health through its impact on health behaviours or physiological mechanisms, or both. Our findings also suggest that intervention programmes should take account of relevant psychological pathways — align=baseline border=0>for example, through emphasising and increasing control beliefs in people from lower socioeconomic backgrounds — align=baseline border=0>as advocated by the empowerment approach.35 Increasing control beliefs in those without real control may be futile, however, because intervention on control beliefs per se may do little to the structural determinants which generate these beliefs. 10 35
A few methodological issues should be considered. Firstly, the design was cross sectional, whereas the ideal design would be to follow a cohort from birth into adulthood. This would allow a better examination of the causal pathways between childhood social class, psychological attributes, adult social class, and adult health. Our theoretical causal model was based on previous research that showed little effect of adult health on adult social class. 36 37 Another assumption was that personality affects adult health and not vice versa. Although there is clear evidence of such a causal relation,17 reverse causation cannot be completely excluded.38 Furthermore, we were primarily interested in the direct effects of childhood social class— align=baseline border=0>that is, independent of adult social class. If specific personality traits affect occupational achievements, however, the contribution of personality and coping to any association will be underestimated when adult social class is adjusted for. On the other hand, adult socioeconomic conditions are likely to affect components of adult personality and coping styles. This suggests the importance of controlling for adult social class. Given this dilemma we presented findings both adjusted and unadjusted for adult social class.
Secondly, all measures were self reported, which may have resulted in overestimated associations because of negative affectivity.39 This tendency to complain may have particularly affected the self reported general health outcome, but it is less likely to have affected findings with the more “objective” health outcome— align=baseline border=0>that is, self reported cardiovascular diseases (mentioned above; data not shown). The role of neuroticism is relevant here as it is sometimes used as a proxy measure for negative affectivity, and it contributed strongly to the association (both general health and cardiovascular health). Given the association with the more “objective” health outcome (cardiovascular diseases) we prefer to interpret neuroticism as reflecting worry about lack of control33 instead of negative affectivity. Worry about lack of control has recently been shown to be correlated with an increased risk of coronary heart disease.40 To exclude the possibility of reporting bias as much as possible, further research should use alternative, more objective information on the health outcomes.
Thirdly, self reported height was an imperfect measure of early life factors related to prenatal and postnatal growth, and related biological consequences, because it was self reported and because adult height reflects more than just early growth.41 Fourthly, the inclusion of other psychological mechanisms, such as feelings of parental caring,42childhood conscientiousness,43 and attachment via hostility,44 may have had an additional contribution to the association between childhood social class and adult health. Finally, the 288people for whom data on social class in childhood or adulthood were missing had lower educational levels and poor general health than those who responded to both questions (not shown). This partial non-response is likely to have resulted in underestimated associations between childhood social class and adult health.
Adverse socioeconomic conditions in childhood are related to poor general health in adulthood. Unhealthy psychological attributes (personality factors and coping styles) are more common in people who reported low childhood social class. Specific psychological attributes contribute substantially to the association between childhood social class and adult health. Perceived control may be the underlying psychological characteristic. When the influence of childhood socioeconomic conditions on adult health is examined the role of specific psychological attributes is worth further exploration, in addition to factors related to fetal development, early growth, and biological and behavioural mechanisms.
The study was conducted in close collaboration with the Public Health Services of the Dutch city of Eindhoven and the region of South East Brabant. We thank Michel Provoost and Ilse Oonk for carefully constructing the database and Carola Schrijvers and Mariel Droomers for their useful comments on previous drafts of the paper.
Contributors: HB was the main author and carried out the analyses, interpreted data, and responded to referees' comments. HDvdeM formulated the hypothesis, helped with writing and responding to referees' comments, carried out preliminary analyses, and was partly responsible for data collection. JPM was principal investigator, formulated the hypothesis, helped with writing and responding to referees' comments, was responsible for data collection and is guarantor for the study.
Funding: Dutch Ministry of Public Health, Welfare, and Sports and the Dutch Prevention Fund.
Competing interests: None declared.