ML model predicts UTI antibiotic susceptibility using PCR results plus age, gender, location.

Provides actionable probabilities within <8 hours, before full susceptibility returns.

Built from 32,300 paired PCR–phenotype samples using light gradient boosting.

Contains 20 antibiotic-specific algorithms, aggregated into one clinical report.

Only reports predictions with >60% probability to avoid low-confidence guidance.

Strong early performance for nitrofurantoin, fosfomycin, and key beta-lactams.

Pilot launch planned for 2026, aiming to reduce empiric therapy and treatment failures.