2857. A Statistical Model to Predict Invasive Fungal Disease in Pediatric Cancer and Hematopoietic Stem Cell Transplantation Patients

Background Invasive fungal diseases (IFDs) are devastating opportunistic infections that result in significant morbidity and death in pediatric cancer and hematopoietic stem cell transplantation (SCT) patients. Identification of risk factors for IFD will help clinical decisions relevant to the diagnosis and management of IFD in a timely manner. Despite this, data evaluating prediction risk tools for IFD in pediatric cancer are limited. Methods We conducted a retrospective review of pediatric oncology patients with a diagnosis of febrile neutropenia (FN) at UChicago Comer Children’s Hospital from July 2009 to December 2016. We analyzed 13 clinical, laboratory, and treatment-related risk factors for IFD including (age, gender, underlying diagnosis, SCT status, graft vs. host disease, chemotherapy in the last 2 weeks, temperature, height, fever duration, presence of hypotension, absolute neutrophil count, duration of neutropenia, absolute monocyte count, and the absolute lymphocyte count (ALC)). IFD was stratified as possible, probable, and proven according to the latest EORTC/MSG criteria (2008). Multivariable logistic regression risk prediction models were developed with separate analyses for all suspected IFD cases and only proven and probable cases. Results A total of 667 FN episodes (FNEs) were identified in 265 patients. IFD was diagnosed in 62 episodes (9.2%) of which 13 (1.9%) were proven, 27 (4%) probable, and 22 (3.3%) possible. Five variables obtained were significantly more common in IFD. Patients presenting with hypotension and fever >5 days were highly associated with IFD (P < 0.001). SCT receipts (P < 0.01), neutropenia longer than 10 days (P = 0.02), and ALC <300 mm3 at time of presentation (P = 0.03) were additional risk factors. The final model performs very well compared with other published models with a receiver operating characteristic–area under the curve (ROC-AUC) of 86.5 for all IFD cases and ROC-AUC of 84.5 for proven, probable IFD cases. Conclusion Our findings showed important clinical markers for the development of IFD in pediatric oncology patients. A predictive regression model including identified significant factors has been created. Risk stratification with prospective external validation using this model can be used to refine the clinical approach. Disclosures All Authors: No reported Disclosures.