Is drug insurance status an effect modifier in epidemiologic database studies? The case of maternal asthma and major congenital malformations.

Abstract

Our previous work on the association between maternal asthma and congenital malformations was based on cohorts formed by women with public drug insurance, i.e., over-represented by women with lower socioeconomic status, questioning the generalizability of our findings. This study aimed to evaluate whether or not drug insurance status, as a proxy of socioeconomic status, is an effect modifier for the association between maternal asthma and major congenital malformations. A cohort of 36,587 pregnancies from asthmatic women and 198,935 pregnancies from nonasthmatic women selected independently of their drug insurance status was reconstructed with Québec administrative databases (1998-2009). Asthmatic women were identified using a validated case definition of asthma. Cases of major congenital malformations were identified using diagnostic codes recorded in the hospitalization database. Drug insurance status at the beginning of pregnancy was classified into three groups: publicly insured with social welfare, publicly insured without social welfare, and privately insured. Adjusted odds ratios were estimated with generalized estimation equations, including an interaction term between maternal asthma and drug insurance status. The prevalence of congenital malformations was 6.8% among asthmatic women and 5.8% among nonasthmatics. The impact of asthma on the prevalence of congenital malformations was significantly greater in women publicly insured with social welfare (odds ratio = 1.42; 95% confidence interval, 1.25-1.61) than in the other two groups ([odds ratio = 1.10; 1.00-1.21] in the publicly insured without social welfare and [odds ratio = 1.13; 1.07-1.20] in the privately insured group). The increased risk of major congenital malformation associated with asthma was significantly higher among pregnant women publicly insured with social welfare than among those privately insured. As a result of this effect modification by drug insurance status, findings from Québec observational studies using databases mainly formed of patients publicly insured with social welfare may not be generalized to the entire population.