Background Reducing prevalence and inequalities in child mental health problems (CMHP) will improve life chances and prevent the perpetuation of social inequalities. Research seeking to identify options for reducing health inequalities often aims to identify amenable risk factors on the pathway between socio-economic circumstance (SECs) and health. Such risk factors may create health inequalities via two mechanisms: differential exposure (risk factor is more prevalent in disadvantaged groups) and differential susceptibility (health impacts of the risk factor are worse in disadvantaged groups). Most research focusses on differential exposure, using mediation analysis. However, most mediation methods produce inaccurate estimates if differential susceptibility (i.e. an interaction) is present. Furthermore, differential exposure and differential susceptibility may reveal alternative or complementary policy actions. We used a novel effect decomposition method to examine the contribution of maternal psychological distress (MPD) to inequalities in CMHP in the nationally-representative UK Millennium Cohort Study (born 2000–2002, n∼18,000).

Methods SECs (exposure) were represented by maternal education (GCSE grades A*-C, yes/no, age 9 months), MPD (risk factor) by the Kessler scale (continuous, 3y), and CMHP (outcome) by the Strengths and Difficulties Questionnaire (borderline-abnormal, yes/no, 5y). The total effect (TE) of SECs on CMHP was estimated using risk ratios (RR) and decomposed, using Stata’s ‘Med4way’, into: the direct effect and effects via MPD due to: mediation (differential exposure), interaction (differential susceptibility), and mediated interaction (when SECs affect MPD and its impact on CMHP). Confidence intervals were estimated using non-parametric bootstrapping (1000 replications). We adjusted for baseline confounders (ethnicity, maternal age at first live birth) in a complete case sample (n=9,777). Sensitivity analyses examined bias from unmeasured intermediate confounding and attrition.

Results 10% children had borderline-abnormal CMH. The RR for the TE of SECs on CMHP was 1.79(1.58–2.04). Two thirds (66% [0.50–0.81]) of this was direct, i.e. not acting through MPD. Four percent (1.2–6.8%) was mediated (differential exposure), 28% (12.1–42.2%) was due to interaction (differential susceptibility), and 2% (0.6–4.3%) was from mediated interaction. Bias from unmeasured intermediate confounding and attrition appeared minimal.

Conclusion The direct effect of SECs on CMHP was large. Maternal psychological distress is also potentially important in the development of inequalities, predominantly due to differential susceptibility (and not differential exposure). Thus analyses which only consider mediating pathways may underestimate its role. Policies to improve MPD have the potential to reduce inequalities in CMHP. Future research to understand the factors that buffer advantaged mothers from the consequences of MPD may inform policy content and delivery.