Unobserved Confounding in Associations between Childhood Household Dysfunction and Health and Social Outcomes in Young Adulthood

Joonas Pitkänen , University of Helsinki
Lauren Bishop, University of Helsinki
Pekka Martikainen, University of Helsinki

Childhood household dysfunction is a well-known risk marker of several adverse outcomes. However, less is known about unobserved confounding emerging from shared familial factors in these associations. The study is based on register data on all individuals born in Finland in 1987–2000 and residing in Finland at age 15 (N=835,448). We considered parental hospital-presenting substance use and psychiatric disorders, parental imprisonment, parental death, social assistance receipt and union dissolution at ages 0–14 as childhood household dysfunction. The study participants were followed from age 15 onwards for hospital-presenting psychiatric disorders and substance use, psychotropic medication purchases, violent and property crime and not being in education, employment or training. Cox regression was used as the statistical method and cousin comparisons were conducted to account for unobserved confounding. All the studied indicators increased the risk of the studied outcomes in the population-level models, with hazard ratios ranging from 1.2 to 2.5. There was a clear dose-response relationship between the number of household dysfunction indicators and the hazard of the outcomes. Most of the associations attenuated in cousin-comparisons but remained statistically significant. There were clear differences in the attenuation percentages by outcomes and by specific indicators of household dysfunction. These findings suggest that unobserved confounding may create upward bias in the associations between household dysfunction and subsequent outcomes, but the amount of confounding depends on the outcome and the specific indicators under examination. Nevertheless, unobserved factors shared within families have a role in the associations between childhood household dysfunction and later outcomes.

See extended abstract

 Presented in Session P2. Health, Mortality, Ageing - Aperitivo