ARTIFICIAL INTELLIGENCE–GUIDED PRECISION ANTIBIOTIC STEWARDSHIP FOR COMBATING ANTIMICROBIAL RESISTANCE IN PAKISTANI TERTIARY CARE HOSPITALS

Abstract

Antimicrobial resistance (AMR) is a major challenge for tertiary care hospitals, where inappropriate antibiotic use can accelerate resistant infections. This study examined an artificial intelligence (AI)-guided precision antibiotic stewardship framework for AMR risk prediction in Pakistani tertiary care hospitals. A quantitative predictive design integrated clinical, microbiological, hospitalization, and previous antibiotic-use indicators. Multiple machine-learning algorithms were evaluated using standard performance metrics, while explainable AI was applied to identify important predictors. The findings showed that Gradient Boosting achieved the highest performance (accuracy = 89.3%; AUC = .95). Previous antibiotic exposure, microbiological resistance indicators, previous hospitalization, and clinical severity were the major predictors. Healthcare professionals also reported relatively high perceived usefulness and intention to use AI-assisted stewardship. The findings suggest that explainable AI can complement conventional stewardship by supporting individualized AMR-risk assessment and more informed antibiotic selection.