Research studies using measures of social stratification based on the three core dimensions of socioeconomic status, namely, level of education, occupational status and income, social health inequalities are being increasingly explored through the life-course epidemiological approach 1 2 3 4 5 . Adverse life events or circumstances can constitute pathways between social conditions and disparities in health outcomes (pathway model) or lead to social inequalities in economic resources, which in turn, are associated with poor health outcomes (i.e., accumulation model) 6 . Research studies in this field are contributing to the health inequality debate, thus providing data for policy makers to allocate human resources to disadvantaged families with children as a means of improving population health or to increase financial resources for families through public policies 7 8 .

In this research field, the family social environment that includes emotional and social support has been documented as one of the predictors of numerous health outcomes in childhood and adulthood 9 10 . The quality of family functioning, including emotional support, nonviolent interparental relationships and parents'emotional stability, has been shown to be associated with a significantly lower risk of psychosocial maladjustment in children and adolescents 11 12 . In the 35-year follow-up of the Harvard Mastery of Stress Study, Russek and Schwartz showed that the subjects who gave lower scores for "perceived caring" items while in college had coronary artery disease, hypertension, duodenal ulcer and alcoholism in midlife 13 . Yet, one of the critical situations where the family social environment and family cohesion are impaired is parental separation or divorce, which has increased over the last 40 years in industrialized countries. In the United States, the divorce rate was 3.6 per 1000 population (i.e., a total of 1.07 million divorces) in 2005, which is still one of the highest in the world 14 . In OECD (Organisation for Economic Co-operation and Development) countries, the crude divorce rate was 2.2 per 1000 population in 2001, and the number of divorces was 44.4 per 100 marriages in 2002. In France, divorces increased from 38.2 per 100 marriages in 1995-52.2 per 100 marriages in 2005 15 . These changes in family structure have been shown to have a significant impact on the well-being of children 16 17 18 and, later, on young adults in terms of psychosocial adjustment and social integration in adult life 19 20 21 22 23 24 25 . In a meta-analysis, Amato 26 27 found that children from divorced families experience higher levels of behavioural, psychological and cognitive development impairments than children from intact families. The short-term consequences include lower levels of scholastic performance 28 and a higher risk of psychosocial maladjustment, such as depression, conduct disorders, psychoactive substance use 29 30 31 and difficulties in relationships with family and peers. In a review of long-term studies, Wallerstein 32 reported that children from single parent families are more likely to divorce during their own adulthood than children raised in intact families. With regard to the dimensions of psychological well-being such as happiness, self-rated health and self-satisfaction in social life, and risky health behaviours and interpersonal relationships, children from disrupted families are more likely to be on the negative end of the rating scale 33 34 . More recently, the Hunt-study II, a Norwegian longitudinal study showed a higher risk of emotional disturbances and scholastic difficulties in adolescents from disrupted families 35 36 .

In the case of family breakup in childhood, several theoretical perspectives have been posited to explain the link between the family social environment and short and long-term health consequences 37 . The family stress model 38 assumes that parenting competence is impeded by parental psychological distress due to marital separation or economic difficulties. Children of separated parents with psychological distress are exposed to insecure marital attachment and marital conflict and come to feel that their own well-being is threatened. Accordingly, family-centred approaches aim to alleviate parental distress and to serve as a buffer in the relationship between the child and the custodial parent 39 . The investment theory argues that economic disadvantages negatively impact education and leisure spending that contribute to a child's well-being 40 . Indeed, parental separation leads to a possible dramatic decline in the standard of living in the custodial household. In social epidemiology, the life-course approach has been employed to demonstrate the association between material deprivation in childhood and further health inequalities in adulthood 41 42 43 On the other hand; Foster and Kalil showed that cognitive and behavioural outcomes did not vary by family structure among economically disadvantaged pre-schoolers, irrespective of race-ethnicity 44 . Therefore, the impact of social policies for single parent households in the last decade on the association between parental separation and psychosocial maladjustment in adolescence and later in adulthood remains to be demonstrated in the next years 45 .

In this context, investigating mediational pathways between parental separation or divorce in childhood and future self-perceived health could be a useful approach for promoting health in young adulthood. Although many studies have reported protective factors in pediatric populations, such as temperament, coping styles, family process factors, such as low-conflict and positive parenting, and supportive relationships with peers 24 35 only a few of them have tested these factors as potential mediators of the association between family breakup in childhood and health status in young adulthood. Specifically, the goal of our study was to analyze the association between parental divorce or separation before 18 years old and adult self-reported health status and to examine the mediating role of self-reported family functioning in childhood.

Of the 1006 young people aged 18-37 years old who completed the questionnaire, 13.6% reported a fair, poor, or very poor general health; 16.3%, a fair, poor, or very poor physical health; and 20.3%, a fair, poor, or very poor psychological or emotional health.

Respondents from disrupted families in childhood or reporting poorer family functioning during childhood ranked more at the negative end of the health scales (Table 1).

Table 2 shows changes in the odds ratios for the independent association between family variables during childhood (family breakup and family functioning variables) and self-rated health. Results support the association between family disruption in childhood and poor self-perceived general health status (OR = 1.93 [1.16-3.20]), poor self-rated physical health (OR = 1.98 [1.19-3.28]) and poor psychological well-being (OR = 1.79 [1.12-2.84]). These associations were adjusted for age, gender, marital status, and socioeconomic variables in adulthood. Each variable defining the quality of family functioning in childhood (maternal or paternal relationships and witnessing interparental violence) were also independently associated with perceived health in adulthood.

To find a potential mediation in our model between family breakup and self-perceived health, the associations were tested between exposure to parental breakup in childhood and family functioning variables during childhood adjusted for age, gender, marital status and socioeconomic status in adult life (Table 3). Having experienced parental separation or divorce was significantly associated with poor paternal relationships (OR = 3.36;CI = [2.44-4.63]) and/or poor maternal relationships (OR = 2.11;CI = [1.37-3.24]) and with having witnessed interparental violence (OR = 5.05;CI = [3.57-7.14]).

A path diagram representing the SEM results is shown in Figure 1. Family functioning during childhood and self-rated health were modelled as latent variables. An examination of the path coefficients shows that the occurrence of parental separation or divorce before 18 was significantly associated with the latent variable defining the poorer quality of family functioning (β = 0.499, p ≤ 0.001), which was also associated with poorer self-perceived health status (β = 0.426, p ≤ 0.001). The indirect standardized effect of parental separation or divorce before 18 on self perceived health in adulthood was 0.21 (0.499 × 0.426; p ≤ 0.001), whereas direct effect was -0.095 (p > 0.05). In a model without mediational variables--such as family functioning during childhood-(figure not shown), the direct standardized effect of parental separation or divorce before 18 years old on self perceived health in adulthood was estimated to be 0.12 (p ≤ 0.001). Therefore, the percentage of parental separation (or divorce before 18) effect mediated by family functioning variables in childhood was 63% (= 0.21/[0.21 + 0.12]). Overall model fit appears quite good: the root mean square error of approximation was within the acceptable range (RMSEA = 0.041 [0.031-0.05]) and the Bentler-Bonett non-normed fit index was 0.952.

Our first objective was to examine the association between the family social environment before age 18 years old and self-perceived health outcomes in young adulthood. This study was conducted among a representative sample of the Paris metropolitan area and is the first study of its kind in France. Of primary note, is the association found between the quality of the family social environment in childhood and self-rated health in adulthood. In addition, the analysis revealed that 63% of the effect of parental separation on self-rated health in young adulthood is due to the quality of the family functioning in childhood. In our study, the parental separation (or divorce before 18) is deleterious for self-rated health in adulthood only because of family violence and child-parents relationships.

These findings are consistent with other studies demonstrating the effect of earlier familial relationships on later health outcomes. Previous research showed that parental emotional support, interparental violence and parental distress were related to and mediated the association between family breakup in childhood and psychosocial maladjustment in adolescence 31 .

These findings support the life-course continuity proposition of exposure to relational disadvantages at an earlier age and negative health outcomes at different developmental stages during adolescence and young adulthood. Exposure to a risky family environment is associated with many different factors, which can modify the individual's life trajectory. For example, research studies have shown that many forms of childhood adversities due to parental separation are associated with an increased risk of depression in adolescence and early adulthood 51 52 53 54 and that early-onset depression is associated with a high recurrence risk. Depression or other childhood or adolescent psychopathologies can occur after parental separation and serve as a pathway between family breakup and adult mood disorders, 55 which can lower health ratings.

The second aim of this paper was to explore the mediating role of the family functioning variables in the association between family breakup in childhood and health status outcomes in adulthood. Our findings are consistent with the psychosocial perspective 33 and the mediational model supports related hypothetical mechanisms underlying the association between parental separation in childhood and health status in adulthood. In this connection, parent-child transactional processes are essential for ensuring cohesion, upon which the children's identity builds, 56 and continuity in the adaptation to different developmental tasks 57 58 . This first framework can be supplemented with a behavioural model through a social approach. The family is the first level of social support. It lays the foundation for the social network, which promotes growth of the capacity for experimentation, autonomy and acquiring independence during adolescence and young adulthood so that the child can respond actively to life's challenges 59 . Social relations, in particular, in families, influence individual behaviour by shaping norms around healthy attitudes and by using collective prevention resources 60 . As well, social networks provide social support, which influences physical and psychological health 61 . Parental separation triggers a period of family relationship reorganization and the renegotiation of the family's boundaries 62 . Those exposed to interparental violence often have threatened self-esteem, which is associated with depression and impaired social functioning 63 64 65 . Consequently, any impairment in social support due to overt conflict and hostility in families may have an impact on health status at older ages.

However, our study had several limitations. The main weakness of the analysis is the retrospective reporting of family variables (separation before age 18, family dysfunction and violence). It's difficult to determine the timing of each of these entangled variables which limits one's ability to interpret the path/mediation analyses done using structural equation models. The retrospective study design may have induced some recall bias. Current circumstances, social life or depressed mood can lead to a significant increase in reports of dysfunctional relationships with parents in childhood 66 . In the literature, the analysis of the validity of retrospective studies has shown that the prevalence of past events is underestimated 67 . But adverse childhood experiences can also be recalled to a greater degree by psychologically impaired individuals as a result of internal biographical coherence 68 . In particular, recalling domestic violence upon a single question may have been connoted by the subject's emotional state during the face-to-face interview, which, in the more psychologically fragile subjects, would have induced same-source bias.

Secondly, because of the cross-sectional nature of the data, we could not suggest causality between family breakup and family functioning. Family breakups are, to a large extent, detrimental to children because of the atmosphere in which the separation occurs and not merely because of the resulting quality of the interrelationships. Thirdly, we adjusted only for the current family structure and the usual socioeconomic variables but did not take into account all the other stressful events during childhood or in adult life (such as divorce or separation). Yet, marital instability is highly associated with both parental separation and negative health outcomes. Lastly, we did not introduce information about the atmosphere in which the separation occurred or the resulting quality of the interparental relationship, both of which could affect children.

This study sheds light on the mechanisms that mediate the relationship between family separation in childhood and health status in young adulthood. Our findings are consistent with the public health implications of family disruption in terms of interventions at three levels: familial (clinical support, family centred-approach), legal (mediation, joint custody), and micro-and macro-social. Additional research is thus needed to evaluate these interventions in the field of health promotion. Indeed, envisioning a healthy future for our children prompts us to pay close attention to the model through which they interact with their social environment. Doing this requires focusing both on the family constellation that we create for our children and the material conditions that we provide to them.

The authors declare that they have no competing interests.

CR participated in conceiving and designing the study, performed analyses with EC, interpreted the results, and drafted and revised the manuscript. EC participated in performing analyses. ER participated in conceiving the database and revising the manuscript. GK participated in revising the manuscript. IP participated in conceiving the database and in revising the manuscript. PC participated in conceiving the database, in interpreting the results and in revising the manuscript. All authors read and approved the final manuscript.