145 - Identifying Pediatric Hospitalizations for MIS-C through Administrative Claims Data

Poster Number: 145
Publication Number: 145.416

Katherine Auger, Cincinnati Children's Hospital Medical Center, Cincinnati, OH, United States; Matthew Hall, Children's Hospital Association, Lenexa, Kansas, KS, United States; Staci D. Arnold, Emory University, Atlanta, GA, United States; Samina Bhumbra, IU School of Medicine - - Indianapolis, IN, Indianapolis, IN, United States; Mersine A. Bryan, University of Washington School of Medicine, Seattle, WA, United States; David M. Hartley, Cincinnati Children's Hospital, Cincinnati, OH, United States; Rebecca Ivancie, Stanford School of Medicine, Stanford, CA, United States; Harita Katragadda, University of Texas At Southwestern, Dallas, TX, United States; Katherine Kazmier, University of Washington/Seattle Children’s Hospital, Seattle, WA, United States; Seethal A. Jacob, Riley Hospital for Children at Indiana University Health, Indianapolis, IN, United States; Karen Jerardi, Cincinnati Children's Hospital Medical Center, Cincinnati, OH, United States; Matthew J. Molloy, Cincinnati Children's Hospital Medical Center, Cincinnati, OH, United States; Kavita Parikh, Children's National Health System, Washington DC, DC, United States; Troy Richardson, Analytics, Lenexa, KS, United States; Amanda Schondelmeyer, Cincinnati Children's Hospital Medical Center, Cincinnati, OH, United States; Samir S. Shah, Journal of Hospital Medicine, Cincinnati, OH, United States; Patrick Brady, Cincinnati Children's Hospital Medical Center, Cincinnati, OH, United States

Background:
Individual children’s hospitals care for a small number of patients with multisystem inflammatory syndrome in children (MIS-C), which makes it challenging to conduct generalizable research.

Objective: We sought to develop and validate diagnostic algorithms in administrative data to identify MIS-C hospitalizations, both before and after the introduction of an MIS-C diagnostic billing code.

Design/Methods:
We developed multiple algorithms using diagnosis and medication administration data in the Pediatric Health Information System (PHIS), excluding hospitalizations for children less than 1 month. We reviewed medical records at 7 geographically diverse hospitals to refine the algorithms. Our multidisciplinary team developed 8 algorithms and applied them to 2020 PHIS data. The Table details each algorithm (e.g., administration of IVIG plus anticoagulation). We compared potential cases of MIS-C identified by each algorithm to participating hospitals’ MIS-C list (used for public health reporting and considered the gold standard). We also compared hospitalizations for January-August 2021 once the MIS-C ICD-10 code (M35.81) became available. Finally, we determined the diagnostic accuracy of each algorithm.

Results:
Across the 7 different sites there were a total 263 hospitalizations for MIS-C in 2020 and 305 hospitalizations for MIS-C from January to August of 2021 on the hospitals’ MIS-C list. Sensitivity of the various algorithms from 2020 ranged from 43-82% (Table). False positives ranged from 12-59%. Overall agreement ranged from 41-88%. Algorithm H had the highest sensitivity (82%), with a reasonably low false positive rate (22%), and moderately high agreement (78%). Algorithm H combined two algorithms featuring a diagnosis code for COVID-19 and the other specified systemic involvement of connective tissue endorsed by the American Academy of Pediatrics in fall 2020 (detailed in Table).  Alternative algorithms had higher agreement but lower sensitivity. For hospitalizations in 2021, the sensitivity of the MIS-C diagnosis code was 99%, and the overall agreement of the code rate was 72% (Table).
Conclusion(s): We have identified high sensitivity algorithms to use for epidemiologic research and high agreement algorithms for treatment comparative effectiveness research. Accurate algorithms to identify MIS-C hospitalizations throughout the timespan of the pandemic will facilitate research important to understanding this novel entity.
Table: Algorithm Performance Across Hospitals