To evaluate broad-spectrum intravenous antibiotic use before and after the implementation of a revised febrile neutropenia management algorithm in a population of adults with hematologic malignancies.

Quasi-experimental study.

Patients admitted between 2014 and 2018 to the Adult Malignant Hematology service of an acute-care hospital in the United States.

Aggregate data for adult malignant hematology service were obtained for population-level antibiotic use: days of therapy (DOT), C. difficile infections, bacterial bloodstream infections, intensive care unit (ICU) length of stay, and in-hospital mortality. All rates are reported per 1,000 patient days before the implementation of an febrile neutropenia management algorithm (July 2014–May 2016) and after the intervention (June 2016–December 2018). These data were compared using interrupted time series analysis.

In total, 2,014 patients comprised 6,788 encounters and 89,612 patient days during the study period. Broad-spectrum intravenous (IV) antibiotic use decreased by 5.7% with immediate reductions in meropenem and vancomycin use by 22 (P = .02) and 15 (P = .001) DOT per 1,000 patient days, respectively. Bacterial bloodstream infection rates significantly increased following algorithm implementation. No differences were observed in the use of other antibiotics or safety outcomes including C. difficile infection, ICU length of stay, and in-hospital mortality.

Reductions in vancomycin and meropenem were observed following the implementation of a more stringent febrile neutropenia management algorithm, without evidence of adverse outcomes. Successful implementation occurred through a collaborative effort and continues to be a core reinforcement strategy at our institution. Future studies evaluating patient-level data may identify further stewardship opportunities in this population.

A significant proportion of inpatient antimicrobial prescriptions are inappropriate. Post-prescription review with feedback has been shown to be an effective means of reducing inappropriate antimicrobial use. However, implementation is resource intensive. Our aim was to evaluate the performance of traditional statistical models and machine-learning models designed to predict which patients receiving broad-spectrum antibiotics require a stewardship intervention.

We performed a single-center retrospective cohort study of inpatients who received an antimicrobial tracked by the antimicrobial stewardship program. Data were extracted from the electronic medical record and were used to develop logistic regression and boosted-tree models to predict whether antibiotic therapy required stewardship intervention on any given day as compared to the criterion standard of note left by the antimicrobial stewardship team in the patient’s chart. We measured the performance of these models using area under the receiver operating characteristic curves (AUROC), and we evaluated it using a hold-out validation cohort.

Both the logistic regression and boosted-tree models demonstrated fair discriminatory power with AUROCs of 0.73 (95% confidence interval [CI], 0.69–0.77) and 0.75 (95% CI, 0.72–0.79), respectively (P = .07). Both models demonstrated good calibration. The number of patients that would need to be reviewed to identify 1 patient who required stewardship intervention was high for both models (41.7–45.5 for models tuned to a sensitivity of 85%).

Complex models can be developed to predict which patients require a stewardship intervention. However, further work is required to develop models with adequate discriminatory power to be applicable to real-world antimicrobial stewardship practice.