Prediction Model for Croup Admission Need

Abstract

OBJECTIVES This study aims to generate a predictive model stratifying the probability of requiring hospitalization and inpatient respiratory intervention for croup patients presenting to the emergency department (ED), and secondarily to compare the model’s performance with that of ED providers. METHODS Retrospective data was collected on croup patients presenting to the EDs of 2 pediatric and 1 community hospital from 2019 to 2020, including demographics, preexisting conditions, and history of croup. The ED length of stay, previous dexamethasone administration, time to ED dexamethasone, number of ED racemic epinephrine doses, viral testing, and ED revisits were also recorded. Westley croup scores were derived at ED presentation and final disposition. For admitted patients, any respiratory interventions were recorded. Admission need was defined as either admitted and required an inpatient intervention or not admitted with ED revisit. A prediction model for admission need was fit using L1-penalized logistic regression. RESULTS We included 2951 patients in the study, 68 (2.3%) of which needed admission. The model’s predictors were disposition Westley croup scores, number of ED racemic epinephrine doses, previous dexamethasone administration, and history of intubation. The model’s sensitivity was 66%, specificity was 91%, positive predictive value was 15%, and negative predictive value was 99%. ED providers’ performance had a sensitivity of 72%, a specificity of 94%, a positive predictive value of 23%, and a negative predictive value of 99%. CONCLUSIONS The croup admission need predictive model appears to support clinical decision making in the ED, with the potential to improve decision making when pediatric expertise is limited.