Identifying and Validating Pediatric Hospitalizations for MIS-C Through Administrative Data

• Published in: Pediatrics (Evanston) Vol. 151; no. 5
• Main Authors: Auger, Katherine A, Hall, Matt, Arnold, Staci D, Bhumbra, Samina, Bryan, Mersine A, Hartley, David, Ivancie, Rebecca, Katragadda, Harita, Kazmier, Katie, Jacob, Seethal A, Jerardi, Karen E, Molloy, Matthew J, Parikh, Kavita, Schondelmeyer, Amanda C, Shah, Samir S, Brady, Patrick W
• Format: Journal Article
• Language: English
• Published: United States American Academy of Pediatrics 01-05-2023
• Subjects: Algorithms; Child; Databases, Factual; Epidemiology; Hospitalization; Hospitals; Hospitals, Pediatric; Humans; Inflammation; International Classification of Diseases; Medical Records; Multisystem inflammatory syndrome in children; Patients; Pediatrics; Predictive Value of Tests; Public health
• ISSN: 0031-4005; 1098-4275
• DOI: 10.1542/peds.2022-059872

Individual children's hospitals care for a small number of patients with multisystem inflammatory syndrome in children (MIS-C). Administrative databases offer an opportunity to conduct generalizable research; however, identifying patients with MIS-C is challenging.
We developed and validated algorithms to identify MIS-C hospitalizations in administrative databases. We developed 10 approaches using diagnostic codes and medication billing data and applied them to the Pediatric Health Information System from January 2020 to August 2021. We reviewed medical records at 7 geographically diverse hospitals to compare potential cases of MIS-C identified by algorithms to each participating hospital's list of patients with MIS-C (used for public health reporting).
The sites had 245 hospitalizations for MIS-C in 2020 and 358 additional MIS-C hospitalizations through August 2021. One algorithm for the identification of cases in 2020 had a sensitivity of 82%, a low false positive rate of 22%, and a positive predictive value (PPV) of 78%. For hospitalizations in 2021, the sensitivity of the MIS-C diagnosis code was 98% with 84% PPV.
We developed high-sensitivity algorithms to use for epidemiologic research and high-PPV algorithms for comparative effectiveness research. Accurate algorithms to identify MIS-C hospitalizations can facilitate important research for understanding this novel entity as it evolves during new waves.