6B-06 - Comparison of linked EHR-pharmacy dispensing data and health care claims during pregnancy

Background: Studies of medication utilization and safety during pregnancy often use fill information from administrative claims data. However, electronic health records (EHR) offer clinical details not available in claims. Automated medication information in EHR has often been limited to prescription orders. Recently, linked pharmacy-EHR databases have incorporated fill information. Comparison of perinatal medication fills from linked EHR-pharmacy data and claims data is needed to understand the strengths and limitations of these data for studies of pregnancy and postpartum medication safety.

Objectives: To assess agreement on medication fills during pregnancy from linked EHR-pharmacy data and claims data for the most common medications.

Methods: This study used EHR and claims data from a healthcare system in Minnesota, USA. Individuals with a livebirth or stillbirth between 2017 and 2021, continuous insurance from six months before pregnancy through delivery, and ≥1 prenatal and postnatal visit were included. Medication fills during pregnancy for 27 of the most common used prescription medications in pregnancy were assessed. Fills were obtained from claims data and from Surescripts® and DrFirst®, the linked EHR-pharmacy databases used by the health system. We assessed prevalence differences between the two sources for each medication. Agreement on medication fills during pregnancy between claims data and linked EHR-pharmacy data was evaluated using percent agreement and Cohen’s kappa. The percent of first medication fills in one data source that had a fill within +/- 1 day in the other data source was also assessed.

Results: Among 12,440 eligible individuals, most were non-Hispanic White (63.0%), and 35.2% had Medicaid insurance. Prevalence during pregnancy was higher according to linked pharmacy-EHR data than claims data for all medications, with differences ranging from 1.5% for ondansetron to 0.05% for nifedipine. Percent agreement between the two data sources was >97% and kappa was ≥0.87 for all medications. The percent of first medication fills identified from claims data during pregnancy that had a fill within 1 day according to linked EHR-pharmacy data ranged from 85.9% for aspirin to 97.7% for amoxicillin-potassium clavulanate. The percent of first medication fills identified from linked EHR-pharmacy data that had a fill within a day according to claims data ranged from 71.9% for aspirin to 90.3% for amoxicillin-potassium clavulanate.

Conclusions: Although agreement between data sources was high, linked pharmacy-EHR data identified more individuals with medication fills during pregnancy than claims data for each medication assessed. Results suggest that linked EHR-pharmacy databases may be reliable for studying medications during pregnancy.