PURPOSE: Electronic health records (EHR) offer clinical details not available in claims data, which are typically used for studies of medications during pregnancy. Historically, medication prescriptions rather than fills have been documented in EHR. However, linked EHR-pharmacy databases incorporate medication fill information and could be used for perinatal medication safety studies.
METHODS: This study compared fills of common medications during pregnancy using linked EHR-pharmacy data and claims data from a healthcare system in Minnesota, USA among patients with a delivery (2017-2021) and insurance coverage. We assessed prevalence differences and agreement using percent agreement and Cohen's κ for 27 medications. The percent of individuals with a fill according to one data source within ±1 day of their first medication fill in the other data source was assessed.
RESULTS: Among the 12 440 pregnancies included, there was a slightly higher prevalence of all medications of interest in linked EHR-pharmacy versus claims data. The prevalence difference was 1.5% for the medication with the largest difference. Percent agreement was > 97% for all medications; κ   ≥  0.870. The percentage of individuals with a fill according to linked EHR-pharmacy data within ±1 day of their first medication fill in claims data was higher than the reverse (claims data fill within ±1 day of linked EHR-pharmacy data fill) for all medications.
CONCLUSIONS: Although agreement between sources was high, linked EHR-pharmacy data identified slightly more individuals with medication fills during pregnancy than claims data. Linked EHR-pharmacy data may provide comprehensive medication fill data for studying medications during pregnancy.
PLAIN LANGUAGE SUMMARY: Many studies of medication safety during pregnancy rely on health insurance claims data. These data may not have important details about patient health. EHRs provide more complete clinical information, and they have been linked with pharmacy data to add prescription fill records. We assessed how well medication fill records agreed between linked EHR-pharmacy data and claims data for 27 commonly used medications in pregnancy. Using data from a Minnesota healthcare system, we analyzed records from 12 440 pregnancies occurring from 2017 to 2021 with health insurance benefits, prenatal care, and postnatal care. We compared the frequency of medication fills in each data source and assessed agreement using statistical methods. The datasets usually agreed on whether a patient filled a particular medication during pregnancy. However, linked EHR-pharmacy data identified slightly more individuals with medication fills during pregnancy than claims data. This suggests that linked EHR-pharmacy data may be useful for perinatal medication safety research.