Identification of factors to predict low risk febrile neutropenia admissions in patients with acute myeloid leukemia (AML) following intensive chemotherapy.

Abstract

7031 Background: Febrile neutropenia (FN) is the most common reason for hospital readmission following chemotherapy for AML. Standard practice for FN in AML is hospitalization for rapid administration of IV antibiotics, thus FN is a major driver of healthcare resource utilization and cost. Whether this approach is justified in all situations is unknown. While FN risk models have been developed focusing on solid tumors, factors predicting uncomplicated courses of FN in AML, which may allow identification of patients suitable for outpatient management, have not been studied. We therefore examined whether baseline clinical characteristics could predict which AML patients with FN have a lower risk of progression to severe illness. Methods: We retrospectively reviewed medical records from 7/29/2014-12/24/2022 to identify adults with FN (fever >38.3C and white blood count <1000/µL). Included were patients with high-grade myeloid neoplasm (>10% blasts in blood/marrow) who received intensive chemotherapy (cytarabine dose >=100 2 /day) within 8 weeks. We collected baseline clinical and disease variables. Outcomes were: infections identified, hospital length of stay (LOS), intensive care unit (ICU) admission, and survival. A “low-risk [LR]” outcome was defined as a LOS <72hrs without ICU admission or inpatient death. Univariate and multivariable (MV) logistic regression models were used to assess covariate associations with outcomes. Results: We identified 350 FN admissions in 198 patients with a median age 62 (range: 21-87) yrs. 63% were newly diagnosed and 37% relapsed/refractory. 12% had prior transplant. Median duration of neutropenia prior to fever was 12 (1-287) days. 46% had transfusion <24 hours prior to fever. The median LOS was 5 (1-56) days. 9% of patients went to ICU and 6% died inpatient.Overall, 15% of admissions were classified as LR. 57% had an infection identified. The table shows factors in MV analysis that predicted an infection identified or high-risk (HR) admission. Conclusions: Only 15% of AML admissions for FN were LR. Factors that predict HR outcomes in standard FN models such the MASCC score—e.g. age, prior fungal infection —were not predictive of HR outcomes in AML. As a next step, we will compare our model to other published models used to predict poor outcomes after FN, which have not been developed or validated in AML. [Table: see text]