Optimizing antimicrobial use remains a key challenge in hospitals, particularly amid the need for efficiency and high bed turnover. This study aimed to evaluate the clinical and economic impact of implementing an artificial intelligence (AI) algorithm for automated recommendation of antimicrobial route switch from parenteral to oral administration. A retrospective analysis was conducted over a 90-day period following implementation of Munai Health’s AI-driven stewardship solution. The platform integrates laboratory data, vital signs, prescriptions, and clinical notes into a unified digital interface, generating automated alerts for early antimicrobial route switch. Clinical pharmacy, infectious disease, and infection control teams were trained prior to deployment and began evaluating alerts from day one. Patients were divided into two groups: accepted alerts (with prescription modification) and pending alerts (without modification). Mean length of stay and economic impact were compared between groups. During the study period, 109 valid alerts were generated, of which 72 (66. 1%) were accepted. Mean hospital stay was significantly shorter in the accepted-alert group (7. 32 days) compared to the pending-alert group (9. 41 days), a mean difference of 2. 09 days (p = 0. 0379). Based on an average daily hospital cost of R1, 088. 86, direct savings were estimated at R2, 271. 32 per patient and R163, 535. 00 for the analyzed period. The projected annual savings totaled R663, 225. 28. In addition to cost reduction, benefits included decreased venous exposure, reduced healthcare-associated infection risk, and increased bed availability. Implementation of an AI algorithm to support clinical decision-making in antimicrobial route switch demonstrated positive clinical and economic impact. The tool proved effective in promoting rational antimicrobial use, enhancing hospital efficiency, and improving patient safety, representing a promising strategy for healthcare institutions seeking to combine technological innovation with sustainability.
Falci et al. (2026) studied this question.