We identify factors in emergency department (ED) patients presenting with apparent life-threatening events that distinguish those safe for discharge from those warranting hospitalization.

Data were prospectively collected on all subjects presenting to 4 EDs with apparent life-threatening events. Patients were observed for subsequent events or interventions, defined a priori, which would have mandated hospital admission (eg, hypoxia, apnea, bradycardia that is not self-resolving, or serious bacterial infection). For patients discharged from the ED, telephone follow-up was arranged. Classification and regression tree analysis was performed to delineate admission predictors.

A total of 832 subjects were enrolled. The overall median age was 31.5 days (interquartile range 10 to 90 days); 427 (51.3%) were male patients, and 513 (61.7%) arrived by emergency medical services. One hundred ninety-one (23.0%) infants had a significant intervention warranting hospitalization. One hundred thirty-seven patients (16.5%) met predetermined criteria that would obviously mandate hospital admission (eg, persistent hypoxia requiring oxygen) by the end of their ED stay. In addition to these patients for whom it was obvious that admission would be necessary in the ED, classification and regression tree analysis (receiver operating curve=0.90) yielded 2 factors predictive of hospitalization: having a significant medical history and having greater than 1 apparent life-threatening event in 24 hours. The sensitivity was 89.0% (95% confidence interval 83.5% to 92.9%); specificity was 61.9% (95% confidence interval 58.0% to 65.7%).

We found 3 variables (obvious need for admission, significant medical history, >1 apparent life-threatening event in 24 hours) that identified most but not all infants with apparent life-threatening events necessitating admission. These variables require external validation and reliability assessment before clinical implementation.