No Field for Lactation: Inferring Lactation Status From US Commercial Claims to Describe Postpartum Medication Dispensing

Purpose Lactation status is not recorded as a discrete data element in U.S. administrative claims databases, limiting research on lactation in health databases. We developed a claims-based phenotype to identify periods of probable lactation and used it to describe medications dispensed to lactating women enrolled in a national health plan. Methods This retrospective cohort study used data from the 2012-2014 Merative MarketScan Commercial Claims Database. Deliveries were identified by diagnosis-related group classification system. Lactation windows ran from the delivery date to the date of the last surrogate lactation indicator, such as breast pump and supply codes, the encounter code for supervision of a lactating mother, and mastitis. New prescriptions filled within a window were characterized by therapeutic group, prescriber type, lactation risk category, and availability of published lactation data. Results Of 566,284 women with a single delivery, 84,679 were determined to be lactating using the developed algorithm based on maternal phenotype. The median length of assignable lactation window was 17 days (IQR 5-47). Windows most often closed on a breast pump claim (53.2%) or a postpartum visit code (28.2%). Among 158,471 new prescriptions filled during lactation, central nervous system agents (54.5%), anti-infectives (17.5%), and hormonal products accounted for > 80%. Combined oral contraceptives represented 8.2% of hormonal prescriptions during lactation against 29.0% among women of unknown status. Obstetricians wrote approximately one third of prescriptions filled during lactation windows. Conclusions Lactation can be identified in administrative claims using codes already present in the data, making national description of dispensing during lactation possible.