Identifying enteral nutrition using administrative data in pediatric patients with eating disorders: A development and validation study

Background Enteral nutrition (EN) is a common nutrition intervention for a variety of conditions, including eating disorders; however, validated methods for capturing administrative data are lacking. We used administrative billing data to identify the use of EN among patients hospitalized with eating disorders and validated using clinical documentation at our institution. Methods Using billing data from the Pediatric Health Information System (PHIS), we identified patients aged 9–20 years hospitalized at our institution from 2018 to 2022 with a primary diagnosis of an eating disorder. EN was identified using International Classification of Diseases, Tenth Revision procedure codes and Clinical Transaction Classification codes for clinical services, supply, and/or other transacted services. We verified enteral feeding information from PHIS via medical record review to identify discrepancies. We calculated the performance of EN identified by PHIS data compared with medical record review including sensitivity, specificity, overall accuracy, and kappa. Results We included N = 483 hospitalizations for patients with eating disorders. Based on PHIS data, 16.2% received EN during admission vs 30.0% from medical record review. The overall accuracy = 85.3%, sensitivity = 52.4%, specificity = 99.4%, positive predictive value = 97.4%, negative predictive value = 83.0%, and κ = 0.60. Conclusion Our results from a single center suggest that EN may be underestimated using administrative data; however, specificity and positive predictive value were high with moderate agreement among patients with eating disorders. These findings have implications in health services research allowing for identification in PHIS and other data sources and quantifying the potential underestimation.