Development and validation of a risk score for predicting hospitalization in children with influenza virus infection.

OBJECTIVE

Influenza virus infections cause significant morbidity and often result in hospitalization in children. Many children with influenza seek care in emergency settings during seasonal influenza epidemics. We hypothesized that certain features could predict the need for hospitalization in children with influenza.

METHODS

Retrospective cohort study of all children 18 years or younger seen at a children's hospital with laboratory-confirmed influenza infection between July 2001 and June 2004. Medical records of children with confirmed influenza virus infection were reviewed. Predictors of admission were identified using logistic regression models. An influenza risk score system was created and validated based on 4 predictors.

RESULTS

We identified 1230 children with laboratory proven influenza virus infection, 541 were hospitalized. Multivariate logistic regression demonstrated that 4 predictors were independently strongly associated with hospitalization. In the clinical prediction rule for children with influenza who were hospitalized, history of a high-risk medical condition (odds ratio [OR], 4.06; 95% confidence interval [CI], 2.91-5.68) was worth 2 points. Respiratory distress on physical examination (OR, 2.33; 95% CI, 1.61-3.38) was worth 1 point. Radiographic evidence of focal pneumonia (OR, 7.82; 95% CI, 3.62-16.92) was worth 3 points and influenza B infection (OR, 3.99; 95% CI, 2.57-6.21) was worth 2 points. High-risk children with influenza with a total risk score of 3 to 8 had an 86% probability of hospitalization.

CONCLUSIONS

The presence of a high-risk medical condition, respiratory distress on physical examination, radiographic evidence of focal pneumonia, and influenza B infection were the 4 strongest predictors of hospitalization. The risk score assigned to a child with influenza may provide a disposition tool for predicting hospitalization in children in seasonal influenza epidemics.