Abstract
BACKGROUND: The World Health Organization has identified Stenotrophomonas maltophilia (SM) as a high-risk antibiotic-resistant pathogen. Notably, determining the effectiveness of current antibiotics against SM is challenging, leading to improper therapy and the spread of resistance. This study assessed how an artificial intelligence-clinical decision support system (AI-CDSS) utilizing mass spectrometry data to predict resistance enhances prescribing decisions and boosts survival. METHODS: This randomized controlled trial (ISRCTN16278872) involved 400 healthcare professionals, with 1,600 SM infections randomized in a 1:1 ratio to either standard practice (control, n = 800) or an AI-CDSS predicting resistance 1 day earlier (intervention, n = 800). Outcomes were assessed by healthcare professionals using structured surveys on days 3, 5, 7, and 14 after treatment initiation. Patient mortality was analyzed over a 14-day follow-up period. RESULTS: The AI-CDSS group demonstrated significantly higher confidence (p < 0.001) in antibiotic prescription, decision-making efficiency, and appropriate antibiotic selection across all time points. Mortality was lower in the AI-CDSS group (92/800, 11.5%) than in the control group (121/800, 15.1%) (p = 0.03). Effective antibiotic choices and reliance on the AI-CDSS during the critical early stages of treatment contributed to improved patient outcomes. CONCLUSIONS: Implementation of the AI-CDSS in a clinical trial setting enhances prescribing confidence, improves decision-making and antibiotic selection, reduces mortality, and demonstrates clinical potential. TRIAL REGISTRATION: ISRCTN, ISRCTN16278872. Registered 28 June 2024, https://www.isrctn.com/ISRCTN16278872 .