A Comparison of Pediatric Hospitalization Data in Cosmos and the Kids’ Inpatient Database

OBJECTIVE: To compare characteristics of pediatric hospitalizations in Epic Cosmos, an electronic health record-based database that aggregates voluntarily submitted data from across the United States, with nationally representative pediatric inpatient data in the Kids' Inpatient Database (KID). METHODS: We performed a retrospective cross-sectional study comparing hospitalizations for children (<18 years) in 2022 from KID and Cosmos. We compared proportions of key variables between these databases for demographics, primary diagnoses, and rates of medical complexity. RESULTS: We identified 1 878 764 hospitalizations in Cosmos and 5 140 395 hospitalizations in KID. Age and sex distribution aligned closely. Admissions from Cosmos had more missing race and ethnicity (31.4% vs 9.5%) and payor data (10.9% vs 0.0%) compared to KID. Cosmos hospitalizations more frequently exceeded 3 days in length (37.2% vs 32.2%) and had a higher proportion of admissions with medical complexity (14.3% vs 13.8%). Cosmos and KID demonstrated comparable distributions of primary diagnoses in each database, though Cosmos included a smaller proportion of liveborn admissions compared to KID (64.6% vs 68.6%). Cosmos had markedly fewer patients from the Western region than KID (14.1% vs 22.8%) but more from the South (43.9% vs 39.7%) and the Midwest (24.3% vs 21.0%). CONCLUSION: Cosmos showed core demographic similarities to KID, with some differences that may reflect its voluntary data submission model. We note differences to varying degrees in race and ethnicity, payor mix, regional representation, length of stay, medical complexity, and primary diagnoses. Future work should explore avenues to optimize the use of these databases in pediatric research.