Predicting the risk of bacteremia in childen with fever and neutropenia.

Abstract

PURPOSE: We sought to identify factors assessable at the time of admission for fever and neutropenia that predict bacteremia in children with cancer. PATIENTS AND METHODS: One hundred fifteen consecutive episodes of fever and absolute neutrophil count (ANC) less than 500/microliter in 72 children with cancer were studied prospectively to determine the risk of bacteremia using data assessable at the time of presentation. After exploratory analysis identified admission temperature and absolute monocyte count (AMoC) as the strongest predictive factors, recursive partitioning was used to determine cutpoints for these variables that resulted in discrimination between episodes associated with a lower or higher risk of bacteremia. RESULTS: There were 24 episodes of bacteremia (21% of episodes). Episodes were grouped using the cutpoints for AMoC and temperature: 17% were classified as low risk for bacteremia (AMoC > or = 100/microliter), 65% as intermediate risk (AMoC < 100/microliter and temperature < 39.0 degrees C), and 18% as high risk (AMoC < 100/microliter and temperature > or = 39.0 degrees C). No episodes classified as low risk were associated with bacteremia; 19% of intermediate-risk and 48% of high-risk episodes were associated with bacteremia. The odds ratio of bacteremia for the high-risk versus the intermediate-risk group is 4.4 (95% confidence interval, 1.6 to 12.9). The risk classification was validated using data from 57 different episodes of fever and neutropenia treated in the same hospital. CONCLUSION: Three levels of risk for bacteremia are defined using the AMoC and temperature at the time of admission for fever and neutropenia. Trials now should be conducted to test whether these factors may be used to assign some children to less intensive or outpatient antibiotic therapy at the time of presentation with fever and neutropenia.