Statistical Analyses
Data were analyzed using a modified Poisson regression approach that allowed for estimation of risk ratios instead of odds ratios24. Models were adjusted for child’s sex, maternal education level (<high school, high school, >high school), mother’s age, average PM2.5 during pregnancy, and average PM2.5 during the first year postpartum. We then examined effect modification by child sex (male/female) for respiratory outcomes by including the relevant interaction terms. Models with p-values <0.1 for interaction terms were then stratified by child sex. In sensitivity analysis, we also adjusted for average PM2.5 during the first two years postpartum, breastfeeding, and for SES instead of maternal education. Analyses were performed in R Version 3.5.1 (R Development Core Team) and SPSS version 24 (Chicago, IL).