Objectives: Bronchiolitis is a leading cause of infant hospitalization in the United States; the mean length of stay (LOS) is 3.3 days. We sought to identify the initial clinical characteristics of bronchiolitis associated with admission and with longer LOS in a large multicenter clinical trial.
Methods: This study was a secondary analysis of a randomized trial conducted in 20 emergency departments in the Pediatric Emergency Care Applied Research Network. We examined age, sex, days of illness, Respiratory Distress Assessment Instrument score, vital signs, and oxygen saturation by pulse oximetry (SpO(2)) at presentation in 598 infants aged 2 to 12 months with moderate to severe bronchiolitis. We used classification and regression tree and logistic regression analyses to identify associations with admission and longer LOS (defined as LOS > 1 night).
Results: Of the 598 infants, 240 (40%) were hospitalized; two thirds underwent longer LOS. The best predictor of hospitalization was initial SpO(2) value of less than 94%, followed by Respiratory Distress Assessment Instrument score of greater than 11 and respiratory rate of greater than 60. For this model, the sensitivity was 56% (95% confidence interval, 50%-62%) and the specificity was 74% (95% confidence interval, 70%-79%). Among admitted patients, the only decision point for prediction of longer LOS was initial SpO(2) value of 97% or less.
Conclusions: A model using objective findings had limited accuracy for predicting hospitalization after emergency department evaluation for bronchiolitis. In these infants with moderate to severe bronchiolitis, however, initial SpO(2) was the best predictor of hospital admission and of longer LOS. Efforts to better define and manage hypoxemia in bronchiolitis may be helpful.