Comparing the Kids' Inpatient Database and National Inpatient Sample for Pediatric Research.

Pediatric researchers use Agency for Healthcare Research and Quality (AHRQ) Kids' Inpatient Database (KID) and National Inpatient Sample (NIS) to analyze the national resource use and outcomes of hospitalized children. Inherent KID-NIS sampling design differences may yield disparate findings. We compared discharge counts and length of stay (LOS) between KID and NIS for common and rare reasons for hospitalization. Retrospective analysis of differences in discharges counts and geometric mean LOS for children ages 0-20 years from KID and NIS in 2019, measured for normal newborns and 331 additional reasons for admission, distinguished by All-Payer Refined Diagnosis Related Groups (APR-DRG) and categorized in deciles by annual discharge volume. We followed AHRQ instructions for data clustering, stratification, and weighting to accommodate the KID and NIS designs, including random samples of 80% and 20% of pediatric discharges, respectively, per hospital. KID-NIS differences in national estimates for total annual discharge counts differed by only 0.5% for normal newborns and 3.7% for all other admission reasons in children. KID-NIS differences remained small aside from reasons for admission in the two lowest volume deciles: 9.5% (SD 7.9%) for admission volumes 200-520; 41.1% (SD 64.2%) for volumes <200. KID-NIS LOS differences for these two-lowest volume deciles were 7.9% (SD 7.1%) and 26.0% (SD 29.3%), respectively. Although KID-NIS differences in discharge counts and LOS were small for high-volume admissions, the differences increased with reasons for admission that had annual discharge volumes approximately 500 or less. For study populations with discharge counts <500, KID may be preferred, given its higher sampling of discharges per hospital.

Informations de copyright

Copyright © 2023 Academic Pediatric Association. Published by Elsevier Inc. All rights reserved.

Déclaration de conflit d'intérêts

Declaration of Competing Interest The authors have no conflicts of interest to disclose.