Evaluation of identifier field agreement in linked neonatal records.

OBJECTIVE: To better address barriers arising from missing and unreliable identifiers in neonatal medical records, we evaluated agreement and discordance among traditional and non-traditional linkage fields within a linked neonatal data set. STUDY DESIGN: The retrospective, descriptive analysis represents infants born from 2013 to 2015. We linked children's hospital neonatal physician billing records to newborn medical records originating from an academic delivery hospital and evaluated rates of agreement, discordance and missingness for a set of 12 identifier field pairs used in the linkage algorithm. RESULTS: We linked 7293 of 7404 physician billing records (98.5%), all of which were deemed valid upon manual review. Linked records contained a mean of 9.1 matching and 1.6 non-matching identifier pairs. Only 4.8% had complete agreement among all 12 identifier pairs. CONCLUSION: Our approach to selection of linkage variables and data formatting preparatory to linkage have generalizability, which may inform future neonatal and perinatal record linkage efforts.

Duke Scholars

Author Keith Allen Marsolo Population Health Sciences