Causal Inference in Perinatal Epidemiology: Big Data’s False Promises

The age of large administrative datasets has arrived with both promise and peril, particularly in perinatal epidemiology. Electronic health records capturing millions of pregnancies and birth outcomes are now widely available, providing unprecedented statistical power and external validity to test a range of hypotheses, including some related to the long-term health effects of prenatal exposure to adverse metabolic conditions.1 Yet, statistical power is a paradoxical advantage.2 These vast datasets can just as easily amplify biased conclusions as they can advance our understanding of maternal and child health.