Objective: To examine the associations between childhood socioeconomic and family circumstances, health and behavioural and cognitive development, and health and mental well-being outcomes in adulthood; exploring whether associations are different for cohorts born in 1958 and 1970, or for men and women.
Design: Pooled analysis of two prospective, population-based, British birth cohort studies.
Participants: 11 327 men and women born in 1958 and 11 177 men and women born in 1970 who responded in the adult follow-up investigations at ages 33 and 30 respectively.
Main outcome measures: Self-rated general health, Rutter malaise scale indicating mental well-being, and presence of a long-standing illness limiting daily activities; assessed at ages 33 and 30 for the 1958 and 1970 birth cohorts respectively.
Results: A diversity of family background (socioeconomic deprivation, housing tenure, family disruption and parental interest), health and development (cognition and behaviour) measures each provided powerful independent indications for general health and mental well-being. Indications for limiting long-standing illness in adulthood were focused most strongly upon health difficulties in childhood. Few interactions between either birth cohort or gender and childhood measures were observed, and excepting these interactions consistency in associations between the childhood measures and the outcomes by gender and cohort was observable.
Conclusions: This study emphasises the importance of cognitive and behavioural development in childhood, as well as deprivation, family background and childhood health in indicating future adult health and mental well-being, emphasising time-persistent effects and important indications for men and women.
Statistics from Altmetric.com
A substantial literature has detailed associations between childhood experiences and adult health and mental well-being. Clear indications for adult health have been provided by each of childhood social circumstances,1–10 health,11–16 birth weight,15 17 18 cognitive ability and educational attainment,19–21 and temperament and behaviour,22–26 though we are unaware of a study which considers all of these factors simultaneously. The primary goal of this study is to integrate measures of childhood family disadvantage—socioeconomic deprivation, family disruption, housing tenure and parental interest in education—and measures of child development—birth weight, health, cognition and behaviour—such that the relative indications of each net of the others may be explored. Our main hypothesis is that these domains all matter for the outcomes considered. Reports of adult health including self-perceived health, mental well-being and presence of limiting long-standing illness are considered as outcomes.
Explorations of patterns of health within the British population and elsewhere according to gender have provided evidence of increased mortality for males compared with increased experience of morbidity for females,27–29 particularly a substantial excess of psychological distress.30–32 Along with physiological differences, examinations of equity- and gender-related experiences suggest social explanations for gendered health inequalities.32–34 We explore whether evidence is provided for gender differences in the relationship between indicators of childhood disadvantage and development for adult health outcomes. In comparing the birth cohorts from 1958 and 1970 we investigate whether differences in the relationship between indicators of childhood disadvantage and development and adult health outcomes for these two cohorts are evidential, given the changes in health policy and provision and in social, demographic and economic conditions in Britain over the life course of these two birth cohorts.
Data from two prospective population-based British birth cohort studies, the 1958 National Child Development Study35 (NCDS) and the 1970 British Cohort Study36 (BCS), are combined. The analysis pools 11 327 participants from the NCDS for whom adult follow-up data were available at age 33 and 11 177 participants from the BCS for whom adult follow-up data were available at age 30. Self-reported general health, categorised as fair or poor versus excellent or good;37 poor mental well-being, indicated by a positive response to seven or more symptoms from the Rutter Malaise Inventory;38 39 and presence of a long-standing illness limiting daily activities40 were derived as indicators of adult health from interview data at these follow-up investigations.
Childhood investigations were carried out at birth and at ages 7, 11 and 16 in the NCDS, and at birth and at ages 5, 10 and 16 in the BCS. Detailed descriptions of the derivation of the childhood measures from these data are provided as supplementary information.
Summarising across each of the childhood follow-up investigations, measures were indicated for: socioeconomic status; housing tenure; disruption from a traditional nuclear family; behavioural problems indicated by parental report of traits of aggression, anxiety and restlessness from the Rutter Behaviour Scale;41 academic test scores within the lowest quartile on age-standardised assessments of mathematics and English ability; and health difficulties indicated by assessment of a disabling condition by a medical officer or health visitor, or parental report of absence from school for health reasons for 1 month or more in the previous year. Birth weight was categorised indicating low birth weights of up to 2.5 kg and between 2.5 and 3 kg. Parental interest in schooling was reported by the child’s teacher at age 11 for the NCDS, and at age 10 for the BCS. For each childhood measure, participants for whom there was no information at any of the relevant childhood follow-up investigations were included by explicitly coding a missing category.
Using logistic regression analysis, odds ratios, 95% confidence intervals and significance p values were estimated for association between each outcome and each childhood measure individually and in models including all childhood measures, each adjusted for cohort and gender. All analyses were carried out using a pooled dataset of the NCDS and BCS cohorts, with effects estimated separately by cohort or gender when there was evidence of interaction between the effect estimate for any childhood measure and either gender or cohort. Interactions were tested via the Wald test, with a p value of <0.01 considered to indicate statistical significance given the large sample size available and the multiple testing involved.42 All analyses were carried out using Stata.43
The distributions of adult health outcomes for men and women within each of the NCDS and BCS cohorts are presented in table 1, with some notable differences in the prevalence of each outcome according to cohort and gender. As has been previously documented for these birth cohorts44 there was a substantially higher proportion of women reporting high malaise scores than men, and a clear increase for BCS cohort members compared with the NCDS cohort.
There was substantial overlap in the reported outcomes: among those reporting fair/poor general health 35.7% reported poor mental well-being, malaise score ⩾7, and 27.3% long-standing illness limiting daily activities; among those with poor mental health 37.9% reported fair/poor general health and 20.9% long-standing limiting illness; and among those reporting long-standing limiting illness 52.3% reported fair/poor general health and 38.1% poor mental health (p<0.001, χ2 test for association between each of the outcomes).
The distributions of the childhood measures for men and women within each of the cohorts are presented in table 2. Members of the 1970 cohort were much less likely to have spent their childhood living in social housing, less likely to have experienced low parental interest in their education and somewhat less likely to have experienced health difficulties during childhood, but much more likely to have experienced family disruption. There were few strong gender differences in childhood experiences, though boys were more prone to behaviour problems.
Childhood socioeconomic deprivation, family housing tenure other than consistent home ownership, family disruption, lack of parental interest, behaviour problems, low academic test scores, low birth weight and health difficulties were each clearly associated with reporting fair or poor, as opposed to good or excellent, general health (table 3). The associations which were estimated by analysing each childhood measure individually, adjusting for cohort and gender, were somewhat attenuated in the full model considering all childhood measures; however, clear associations continued to be maintained. Association between socioeconomic deprivation and general health was apparent at the level of some deprivation for women; however, association was not apparent for men at this level (p = 0.002 and p = 0.001 for interaction in the individual and full models respectively). A slight indication of a weaker association between academic test scores and general health for the BCS compared with the NCDS cohort at the level of consistently low scores was observed (p = 0.01 and p = 0.02 for interaction in the individual and full models respectively). For all other childhood measures comparable associations for men and women and for each of the cohorts were estimated and no interactions were evident with p value <0.01.
Substantial differences in the prevalence of poor mental well-being, indicated by the malaise score, were apparent according to gender and cohort (table 1). The odds ratios estimated for cohort and gender indicated that adjusting for the childhood measures marginally increased the observable differences, rather than cohort and gender differences being explained by the childhood measures (table 4). Childhood socioeconomic deprivation, family housing tenure other than consistent home ownership, family disruption, lack of parental interest, behaviour problems, low academic test scores and health difficulties were each clearly associated with poor mental well-being in adulthood when estimated by analysing each childhood measure individually, adjusting for cohort and gender, and in the full model considering all childhood measures, although they were to some extent attenuated. The slight association between low birth weight and adult mental well-being observed in the individual analysis was almost entirely attenuated on consideration of other childhood measures, and statistical significance was not maintained. There was an indication of a weaker association between academic test scores and malaise score for the BCS than for the NCDS cohort at the level of consistently low scores (p<0.001 and p = 0.002 for interaction in the individual and full models respectively). For all other childhood measures the odds ratio estimates were similar for men and women, and for each of the cohorts, and no interactions with p value <0.01 were evident.
Limiting long-standing illness
Association was evident between each of the childhood measures and the presence of a long-standing illness which limited daily activities when each measure was considered individually, adjusting for cohort and gender (table 5). However, the full model including all childhood measures was dominated by powerful indications of health difficulties in childhood for the presence of a limiting long-standing illness in adulthood, particularly when health difficulties in childhood had been indicated repeatedly. Among the other childhood measures, only academic test scores showed a clearly significant association with limiting long-standing illness, with slight indications of association for birth weight, family disruption and behaviour problems in childhood. For all the childhood measures the odds ratio estimates were similar for men and women and by cohort, and no interactions with p<0.01 were evident.
This study emphasises the importance which may placed on cognitive and behavioural development in childhood, as well as deprivation and childhood health in indicating future adult health and mental well-being. A diversity of family background, health and development measures each provided powerful independent indications for general health and mental well-being, whereas indications for limiting long-standing illness in adulthood were focused most strongly upon health difficulties in childhood. Interaction between gender and childhood socioeconomic deprivation described an increasingly detrimental indication of general health for women compared with men. Interaction between birth cohort and consistently low academic test scores showed a less detrimental indication for poor mental well-being, and to some extent general health, for the later born 1970 birth cohort compared with the 1958 birth cohort. Little evidence of interaction between any other of the childhood measures and gender or birth cohort was provided, and consistency of estimates by gender or cohort emphasised the importance for men and women and that for the time persistence of the indications of these childhood measures.
Strengths and limitations
This study is unusual in the range of childhood measures considered. It incorporates key measures of childhood background—including socioeconomic deprivation, family disruption, housing tenure and parental interest in the child’s education—and powerful measures of developmental problems for the child—including low birth weight, health difficulties, low cognitive performance and behavioural difficulties. The strength of associations may be overstated when any of these factors are considered in isolation. However, the results of this study show clear evidence of persisting indications of childhood disadvantage and development for self-reported general health and poor mental well-being or general health, even after control all the other measures, and some persisting association between low birth weight and general health. When a more restricted measure of health—limiting long-standing illness—is considered, childhood health indicators dominate in the full model, yet there are some indications that educational test scores, birth weight, family disruption and behaviour problems remain significant predictors.
The advantages of prospective studies for obtaining reliable measures of many of the indicators used here are evident, utilising proximal measures of childhood behaviour or cognitive performance and recording childhood experiences without the potential for recall bias in the light of later circumstances. In addition, repeated measurements were captured on virtually all of these indicators at three different points during childhood, thus enabling discovery of gradients according to the severity and persistence of adversity. Although exact congruency between the detail and timing of some measures was not attainable between the two birth cohorts, comparable measures were carefully developed and the extent of evidence for changes between the cohorts in the associations of childhood indicators with adult health outcomes, or for gender differentials, systematically explored.
Limitations of the cohort study data included sample attrition45 and the prevalence of missing data.46 47 In derivation of the childhood summary measures we sought to include the full availability of data over each of the three childhood follow-up investigations. Many of the characteristics considered were measured in only one or two of the childhood follow-up investigations even where some information was missing, and thus at least some measure on the indicator in question was available. Missing information was explicitly coded, and participants were still included when information was missing in all relevant childhood waves thus maximising the inclusion of participants.
The indicators of childhood deprivation and development focused on indications of disadvantage since it is these groups who experience the worst adult health outcomes, whereas several other studies also distinguish particularly advantaged childhood circumstances which increase the gradients observed. Distinguishing groups from the professional or managerial classes, from more expensive owner-occupied housing or with high cognitive scores would serve to increase the contrasts observed and provide an alternative focus in interpreting the gradients of association. Moreover, in order to keep the analysis manageable elements that would often be separated (eg, externalising and internalising behaviours, social class and poverty, and medical diagnosis with health-related school absence) were combined. Alternative specifications have been explored for these combinations and for birth weight and the main results are not sensitive to these groupings. We have also considered the possibility of using weight for gestational age, but much higher levels of missing information raise concerns about selectivity.48 Further work might explore these refinements.
The outcomes considered each rely on self-reported health status, and although a large literature exists that demonstrates the relevance of such measures for indicating ill-health, for example,15 40 this is acknowledged as a weakness of the study. However, there are major advantages to exploring the contribution from a large prospective population study in which detailed clinical measures can be prohibitively expensive; and the study benefited from having the potential to contrast the three outcomes.
Comparison with other studies
Strong gradients of association between childhood socioeconomic conditions and adult health have been consistently observed in a number of British, and other, populations at various stages within their life course, with outcomes considered including all-cause mortality, general health measures and specific causes of mortality and morbidity.1–10 This study continues to provide clear evidence for association between childhood socioeconomic deprivation and adult general health and mental well-being, even considered within a broad context of child well-being including other aspects of family background, health and development. In contrast, childhood deprivation was less indicative for limiting long-standing illness than were the other indicators of health and development.
Poor cognitive development19–21 and behavioural difficulties22–26 in childhood have each been associated with an increased prevalence of poor physical and mental health outcomes in adulthood. Through using repeated developmental evaluations throughout childhood, and in demonstrating the continued importance of these developmental indicators within analyses integrating deprivation and health, this study provides a strong re-emphasis of prior findings, particularly for general health and mental well-being.
Consistently observable continuities between childhood and adult health11–16 were confirmed in this study, with the indications of health difficulties in childhood for the presence of a limiting long-standing illness in adulthood particularly being emphasised. Associations between low birth weight and each of the health outcomes were demonstrable, as has been previously indicated for adult health15 and mental well-being.18 These effects, however, were strongly attenuated on adjustment for other childhood measures, particularly for the mental health outcome. A stronger association between birth weight and health has been described for later adulthood than for early adulthood,15 suggesting that the developmental consequences of low birth weight for general health may have been less apparent at the early stage of adulthood considered in this study than they may become later.
Interpretation, unanswered questions and future research
This study emphasises the diversity of childhood experiences which are associated with adult health. Evidence is provided that there is not just a health gradient associated with socioeconomic status, but several gradients across disadvantage and development measures. As an outcome of social exclusion consequent to childhood adversity, this study places health within a literature which examines other social and economic outcomes such as family demography, welfare position, educational attainment, employment history and criminal behaviour.46 47 49 50 Continued observation of the NCDS and BCS cohorts as they progress through adulthood provides opportunities for extension of this research. It would be valuable to further trace some of the pathways through late adolescent and early adult experiences to the health outcomes, through experiences such as unemployment, socioeconomic status and own housing circumstances which have been shown to be powerful correlates of mental well-being and many other adult outcomes in the 1958 cohort.50 As few explanations were provided for gendered patterning of health inequalities, particularly mental well-being, these remain priorities for further research.
What is already known on this subject
Although there is evidence that childhood socioeconomic status, health, cognition and behaviour are each associated with adult health, it has not been clear whether these associations persist independently, net of the indications of the others; and few studies have systematically explored whether there are gender differences in these relationships, or changes between birth cohorts.
What this study adds
This study provides evidence that a diversity of family background (socioeconomic deprivation, housing tenure, family disruption and parental interest), health and development (cognition and behaviour) measures gives powerful independent childhood indications for general health and mental well-being in early adulthood, whereas limiting long-standing illness in adulthood is most strongly indicated by health difficulties in childhood. Excepting stronger indications of academic test scores for general health and mental well-being for the earlier 1958 birth cohort, and of socioeconomic deprivation for the general health of women, differential associations of the childhood measures with adult health are not apparent by gender, or between the 1958 and 1970 birth cohorts.
We thank Kathleen Kiernan, Wendy Sigle-Rushton, Carmen Huerta and Darcy Hango for their contributions to this work. The study carried out secondary data analysis using data received in an anonymous form with permissions from the distributor. Study data were provided by the Centre for Longitudinal Studies, Institute of Education, London, UK, and were made available through the Data Archive, University of Essex, Essex, UK.
Funding: Financial support for this work was provided to JH and FKM by the ESRC through the Gender Equality Network.
Competing interests: None.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.