Conclusion
In this study we demonstrated that fever response 1-hr after
administration of acetaminophen can be used to predict bacteremia among
pediatric cancer patients presenting with FN. We modeled this using both
binary logistic regression and CART machine learning analysis with
AUC-ROC of 0.70 and 0.71 respectively. While unlikely to provide
sufficient predictive value to impact clinical decision making in
isolation, our results highlight a promising, objective, and novel means
of prompt FN risk stratification that has the potential to be used
across a multitude of clinical settings. Future prospective,
large cohort, multicenter studies are needed to ultimately assess the
utility of fever responsiveness in the risk stratification in FN.
Conflicts of interest: None