Senior Scientific Investigator Kaiser Permanente Washington Health Research Institute University of Washington, Seattle, Washington Seattle, United States
Background: Mother-infant linkages are critical for studying infant outcomes related to medication or vaccine exposure in pregnancy using electronic health data. Many studies have linked mothers with infants using insurance claims data, but less is known about conducting linkage within electronic medical records (EMRs), an important source of real-world data.
Objectives: To explore and compare data sources for mother-infant linkage in one US healthcare system with access to both claims and clinical EMR data.
Methods: The setting was Kaiser Permanente Washington (KPWA), a US integrated healthcare system providing health insurance and care delivery in Washington state. We used CPT and ICD procedure codes to identify live births from 2001-2022 among people enrolled in KPWA. Infants were patients with a KPWA medical record number (MRN) born from 2001-2022. To create links, we explored a range of data sources and variables. On the insurance/claims side, we considered insurance subscriber numbers and name-address matching using last name, birth/delivery date, and standardized home address. We also identified Care Management databases used by the insurance system to track hospital admissions, including deliveries, which include a direct link between maternal and infant MRNs. In addition, we explored Epic EMR data such as parental proxy access to the online health portal and obstetric history information from outside healthcare systems received via Epic’s CareEverywhere. We calculated the percent of deliveries linked to an infant overall and by data source and calendar year.
Results: We identified 117,668 deliveries from 2001-2022; each year, between 93-98% could be linked to an infant. Care Management data identified a linked infant for 85% of deliveries; this proportion varied from 63-93% across study years due to transitions between Care Management systems. Name-address matching linked 68% of deliveries to an infant. Insurance subscriber number identified matched infants for 32% of deliveries from 2010 on (21% overall). Parental proxy data were available from 2005 on and linked 38% of deliveries to infants in this period (33% overall). External obstetric history data do not provide direct infant links but contain exact delivery date and certain infant characteristics, facilitating name-address matching. These data were available from 2017 on and covered 75% of deliveries by 2022.
Conclusions: Many administrative and clinical data sources can support mother-infant linkage, but linkage rates vary greatly between data sources and over time. Novel data sources including Care Management (claims/insurance) data and parental online proxy and external obstetric history data (EMR) may be useful to supplement traditional approaches to improve linkage yield and accuracy.
