Background: One of the most common complications of chemotherapy treatment is febrile neutropenia (FN), affecting 50% of patients with a mortality rate as high as 21%. Oftentimes, fever is the first and only presenting symptom. These patients should receive antibiotics within 60 minutes to improve prognosis; however, the national mean time to antibiotic administration (TTA) ranges from 102 to 300 minutes. This is due to emergency department overcrowding, clinical care variability, and diagnostic delay.
Currently, there are no pediatric-specific risk models or predictors of TTA in place for FN. A unique protocol was instituted in a local children's hospital to achieve rapid TTA where patients are evaluated in the pediatric hematology oncology clinic, bypassing the emergency department. An order-set for FN is then activated for nurses to conduct patient stabilization, blood culture collection, and antibiotic administration.
Method: 18 variables in 328 pediatric oncology FN encounters over five years under this protocol were analyzed. A random forest model was then created to find specific key variables contributing to delayed TTA. Age, temperature, white blood count (WBC), absolute neutrophil count (ANC), time to blood culture, and time to line access were all key contributing variables. Turning points progression for each contributing variable in the delay categories of less than 30 min, 31-45 min, 46-60 min, and over 60 min were determined.
Results: This information was then integrated into a mobile app designed to flag patients at certain turning points to predict TTA, optimizing antibiotic administration and aftercare, with the use of a continuous live dashboard available for healthcare personnel. The application had an 89.73% accuracy with an 85.65% F1 score.
The novel triage system may be replicated in other institutions to reduce TTA and improve outcomes of febrile neutropenia in pediatric oncology patients.
