Effect of an Artificial Intelligence-Assisted Antibiotic Susceptibility Test on Reducing the Mortality Rate of and Antibiotics Use in Patients with Bacteremia: A Prospective Observational Cohort Study

Importance Although various computational approaches have been proposed for pathogen detection, no artificial intelligence (AI)-assisted antibiotic susceptibility test (AST) system has been robustly validated and truly deployed in a clinical laboratory in real-world settings. Consequently, the clinical impact of this type of medical AI system remains unknown.Objective To evaluate the clinical impact of an AI-assisted AST prediction system deployed in a clinical laboratory.DesignThe AI-assisted AST prediction system XBugHunter was extensively validated (internal validation, time-wise validation, and independent testing) using data collected between May 22, 2013, and June 30, 2019. The clinical impact was evaluated based on a prospective observation during the deployment phase, which was from February 1, 2020, to September 30, 2020.Setting Data were collected from two tertiary medical centers in Taiwan, and the AI system was deployed in a tertiary medical center.Participants A total of 90,064 consecutive patients with ESKAPE (Enterococcus faecium, Staphylococcus aureus, Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, and Enterobacter spp.) infection were included in the development and validation study. During the deployment, a prospective observational cohort of 155 consecutive patients with Staphylococcus aureus bacteremia (SAB) was analyzed, whereas the historical control cohort included 455 patients.Exposures AST prediction from XBugHunter.Main Outcomes and Measures The diagnostic performance of XBugHunter was evaluated during the validation study. The clinical impact was evaluated in terms of the reduction in inappropriate antibiotic prescriptions, AST turnaround time, and mortality rate of SAB during the deployment.Results In the deployment, the predictive sensitivity and specificity for S. aureus (oxacillin) by XBugHunter were 0.95 (95% CI, 0.82–0.98) and 0.97 (95% CI, 0.94–0.99), respectively. The turnaround time reduction in the reporting of the AST results of SAB was 34.06 h. Death within 28 days was the outcome in 21 of the 155 SAB patients (13.55%) in the XBugHunter intervention group, which was significantly lower than the 28-day mortality rate (17.58% [80/455]) in the historical control cohort in 2019 without XBugHunter (p = 0.042). Regarding the antibiotic prescriptions, 45.16% [70/155] of the SAB patients had a change in their empirical antibiotics based on the predicted AST results. Totally, 114.77 defined daily doses of inappropriate antibiotics were avoided in treating SAB by deploying XBugHunter.Conclusions and Relevance This work has demonstrated the clinical impact of XBugHunter through extensive validation protocols. Additionally, among the patients with SAB, XBugHunter can prevent inappropriate antibiotic use, and this adjustment in antibiotic treatment can yield a lower mortality rate.