Abstract
INTRODUCTION: The safety and effectiveness of COVID-19 vaccination during pregnancy are commonly assessed using administrative health records, which may misclassify vaccine exposures. OBJECTIVE: The aim of this study was to compare the capture of COVID-19 vaccines in commercial insurance claims records and electronic health records (EHR) before and during pregnancy in a cohort in the United States and demonstrate the implications of observed misclassification on vaccine safety and effectiveness estimates. METHODS: Analysis of de-identified data from the Optum Labs Data Warehouse (OLDW) included pregnancies ending after December 11, 2020, and beginning before January 1, 2023, among people aged 15-54 years who were continuously enrolled in health insurance and had EHR data available. We compared the capture of COVID-19 vaccines from insurance claims and EHR databases to the combined and deduplicated doses using clinical procedure and drug codes, assessing records over time and by gestational age, dose number, vaccine manufacturer, and socioeconomic variables. RESULTS: For 29,663 eligible pregnancies, we identified 25,124 COVID-19 vaccine doses (87% in claims and 35% in EHR). About 44% had received one or more doses of a COVID-19 vaccine before or during pregnancy in the combined data, compared with 41% in the insurance claims data and 16% in the EHR data. Records of vaccination relative to the combined data improved in the claims data and declined in the EHR data over time. Vaccine manufacturer information from records without procedure codes in EHR data was unreliable, and a greater proportion of combined doses were recorded in EHR when the cohort was restricted to people with more frequent EHR encounters. The magnitudes of exposure misclassification observed, particularly in EHR data, could substantially bias estimates of vaccine effectiveness and safety. CONCLUSIONS: Insurance claims data captured a larger proportion of COVID-19 vaccine doses than EHR data, and this proportion increased over time. Combining data from multiple administrative health records sources can improve COVID-19 vaccine measurement before and during pregnancy and may be important to reduce bias in studies of vaccine effects.