Background: Within the U.S. FDA’s Sentinel System, linkage of mother and infant data is critical for the assessment of medication safety during pregnancy. Recently, U.S. Medicaid/CHIP data in the new Transformed Medicaid Statistical Information System (T-MSIS) format were converted to the Sentinel Common Data Model and an initial mother-infant linkage was performed. Linkage rates using 2000-07 Medicaid data from an earlier data system were reported at 48%.
Objectives: To assess mother-infant linkage rates by U.S. Medicaid jurisdiction and to explore jurisdiction-specific reasons for differential results.
Methods: We used T-MSIS analytic files (TAF) from 2014 to 2018 for this linkage. A jurisdiction—state, territory, or the District of Columbia—was included for analysis provided it had acceptable data quality in at least one of the study years. We identified deliveries to mothers enrolled in Medicaid/CHIP at the time of birth, subject to a 180-day minimum prior enrollment period without evidence of another delivery; and we identified all infants enrolled in Medicaid/CHIP within 365 days of birth. Because no personal identifiers such as name or address were available, linkage was performed by matching jurisdiction-specific case numbers, as a proxy for family identifiers, and date of birth/delivery. Imperfect date matches were allowed if the birth date was proximate to the mother’s delivery date or within the mother’s hospital stay, in the case of inpatient deliveries. We calculated the linkage rate as the number of deliveries linked to at least one eligible infant divided by the number of eligible deliveries. We also explored potential data-based and jurisdiction-specific Medicaid program rules to explain differences in linkage rates.
Results: Among 53 different jurisdictions, 1.9 million (64.0%) of the over 2.9 million eligible delivery records were successfully linked to infant records. There was considerable variability in linkage rates: 32 jurisdictions had a linkage rate over 75%, 10 had a linkage rate between 50% and 75%, and only 11 had a linkage rate under 50%. Several of the jurisdictions with poor or zero linkage rates had case numbers that were unique or nearly unique among beneficiaries, indicating that these identifiers were not acting as family identifiers. Other jurisdictions had high rates of missing case numbers, depressing potential linkage rates. At least one jurisdiction had substantially many more link-eligible infants than deliveries in their data, which was related to how that state covers the healthcare of low-income women and their newborns.
Conclusions: In the U.S. Medicaid/CHIP TAF data, mother-infant linkage was successful for most jurisdictions. In other jurisdictions, alternative methods for linkage would need to be explored.
