Background and Objectives: As a clinical database, the newly released public use data set (PUD) from the Pediatric Emergency Care Applied Research Network (PECARN) Registry contains rich patient-level and operational granularity unavailable in large administrative databases. To determine whether the PECARN Registry PUD population is representative of the broader cohort of patients treated at US children’s hospitals, we sought to compare the PECARN Registry PUD to the widely used Pediatric Health Information System (PHIS) administrative database across demographic, system-based, and clinical ED visit characteristics. Methods: We analyzed 14,875,508 PHIS and 3,196,288 PECARN Registry PUD patient emergency department (ED) encounters from 2017 to 2021. We compared the databases for patient demographics, ED visit characteristics, and diagnosis groupings, as well as the number of medications administered, imaging rates, and diagnosis-specific quality indicators using standardized effect size differences. As 9 of the PECARN Registry PUD hospitals were also included in the PHIS database, we performed a secondary analysis to determine if the overlap inclusion of these 9 hospitals influenced effect sizes, comparing PHIS data from the 9 PECARN Registry PUD hospitals against PHIS data for the remaining non-PECARN Registry PUD hospitals. Results: Nearly all effect size difference comparisons between PHIS and the PECARN Registry PUD were classified as small (Cohen’s h <0.2). Only the proportion of non-Hispanic patients (68.0% vs. 77.2%; Cohen’s h : 0.208, 99% CI: 0.206, 0.210) and head injury visits with CT scans (23.5% vs. 15.6%; Cohen’s h : 0.200, 99% CI: 0.190, 0.211) had moderate effect size differences (Cohen’s h 0.2 to 0.5) that did not dissipate in the secondary analysis. Conclusions: Comparisons between the clinical PECARN Registry PUD and administrative PHIS databases demonstrated substantial agreement, suggesting that the PECARN Registry PUD is generalizable and representative of national children’s hospital populations.
Kappy et al. (Wed,) studied this question.