Machine Learning Algorithms to predict MRSA

Antibiotic resistance is a major threat to health. Methicillin-resistant Staphylococcus aureus (MRSA) causes skin, respiratory, and invasive infections in both children and adults, yet raw clinical datasets with minimum inhibitory concentrations (MICs) remain underused. We will analyze MRSA using the ATLAS database, which contains roughly 6.5 million MIC determinations for 3,919 agent–pathogen combinations, compiled from about 633,000 patients across 70 countries between 2004 and 2017, together with clinical metadata. Building on our pediatric project, we broaden the scope internationally to characterize how MRSA susceptibility evolves over time, differences across countries and registries, and variation by infection type (e.g., bacteremia versus skin/soft-tissue).

Using oxacillin/cefoxitin MICs (which phenotypically define MRSA) and key antibiotics—vancomycin, linezolid, daptomycin, clindamycin, trimethoprim–sulfamethoxazole, and others—we will quantify trends, identify possible laboratory artifacts, and apply clustering to detect subpopulations with rising MICs as early-warning signals. Where age data are available, we will compare pediatric and adult profiles and contextualize Romania relative to other countries.

Impact: The results will support better empirical therapy choices, targeted antimicrobial-stewardship policies, and public-health planning.