Predicting need for hospitalization in acute pediatric asthma.

OBJECTIVES

To develop and validate predictive models to determine the need for hospitalization in children treated for acute asthma in the emergency department (ED).

METHODS

Prospective cohort study of children aged 2 years and older treated at 2 pediatric EDs for acute asthma. The primary outcome was successful ED discharge, defined as actual discharge from the ED and no readmission for asthma within 7 days, versus need for extended care. Among those defined as requiring extended care, a secondary outcome of inpatient care (>24 hours) or short-stay care (<24 hours) was defined. Logistic regression and recursive partitioning were used to create predictive models based on historical and clinical data from the ED visit. Models were developed with data from 1 ED and validated in the other.

RESULTS

There were 852 subjects in the derivation group and 369 in the validation group. A model including clinical score (Pediatric Asthma Severity Score) and number of albuterol treatments in the ED distinguished successful discharge from need for extended care with an area under the receiver-operator characteristic curve of 0.89 (95% confidence interval [CI], 0.87-0.92) in the derivation group and 0.92 (95% CI, 0.89-0.95) in the validation group. Using a score of 5 or more as a cutoff, the likelihood ratio positive was 5.2 (95% CI, 4.2-6.5), and the likelihood ratio negative was 0.22 (95% CI, 0.17-0.28). Among those predicted to need extended care, a classification tree using number of treatments in the ED, clinical score at end of ED treatment, and initial pulse oximetry correctly classified 63% (95% CI, 56-70) of the derivation group as short stay or inpatient, and 62% (95% CI, 55-68) of the validation group.

CONCLUSIONS

Successful discharge from the ED for children with acute asthma can be predicted accurately using a simple clinical model, potentially improving disposition decisions. However, predicting correct placement of patients requiring extended care is problematic.