Confounding Control

The study uses propensity score-based fine stratification weighting to address confounding. The primary analysis estimates relative risk and 95% confidence intervals using generalised linear regression. Adjusted covariates include maternal age, delivery year, child characteristics, birth weight, healthcare use, obstetric conditions, socioeconomic indicators, comorbidity, and other imaging exposures. Covariate balance is judged adequate when the absolute standardised difference is below 0.1. Robust variance estimators account for women contributing multiple pregnancies.