Prospective Validation of a Risk Prediction Model for Severe Sepsis in Children With Cancer and High-risk Febrile Neutropenia

Abstract

We previously created a risk prediction model for severe sepsis not clinically apparent during the first 24 hours of hospitalization in children with high-risk febrile neutropenia (HRFN), which identified 3 variables, age ≥ 12 years, serum C-reactive protein (CRP) ≥ 90 mg/L and interleukin-8 ≥ 300 pg/mL, evaluated at the time of admission and at 24 hours of hospitalization. The combination of these 3 variables identified a risk for severe sepsis ranging from 8% to 73% with a relative risk of 3.15 (95% confidence interval: 1.1-9.06). The aim of this study was to validate prospectively our risk prediction model for severe sepsis in a new cohort of children with cancer and HRFN. Predictors of severe sepsis identified in our previous model (age, CRP and interleukin-8) were evaluated at admission and at 24 hours of hospitalization in a new cohort of children with HRFN between April 2009 and July 2011. Diagnosis of severe sepsis, not clinically apparent during the first 24 hours of hospitalization, was made after discharge by a blind evaluator. A total of 447 HRFN episodes were studied, of which 76 (17%) had a diagnosis of severe sepsis. The combination of age ≥ 12 years, CRP ≥ 90 mg/L and interleukin-8 ≥ 300 pg/mL at admission and/or at 24 hours in the new cohort identified a risk for severe sepsis ranging from 7% to 46% with an RR of 6.7 (95% CI: 2.3-19.5). We validated a risk prediction model for severe sepsis applicable to children with HRFN episodes within the first 24 hours of admission. We propose to incorporate this model in the initial patient assessment to offer a more selective management for children at risk for severe sepsis.