GPT-Based Drug-to-Disease (DtD) Checker as a Tool for Optimizing Antibiotic Use and Reducing Resistance

Background: Inappropriate antibiotic prescribing is a major driver of antimicrobial resistance (AMR) and adverse drug effects. Traditional decision support systems often lack real-time adaptability and struggle to integrate complex patient-specific data. The advent of GPT-based Drug-to-Disease (DtD) checkers offers a novel AI-driven approach to assist clinicians in optimizing antibiotic therapy by ensuring accurate drug-disease matching, minimizing resistance risks, and preventing adverse effects.

Methods: We propose an AI-powered DtD checker based on a large language model (LLM) architecture, trained on clinical guidelines, antibiograms, pharmacokinetics, and patient-specific parameters (e.g., renal function, comorbidities, and prior antibiotic exposure). The model leverages natural language processing (NLP) and deep learning algorithms to cross-reference antibiotic choices with patient characteristics, flagging inappropriate prescriptions and suggesting evidence-based alternatives. Its performance was validated against real-world antibiotic prescribing data in hospitalized patients.

Results: The GPT-based DtD checker significantly reduced inappropriate antibiotic prescriptions by 35%, enhanced compliance with antimicrobial stewardship guidelines, and decreased the incidence of drug-related adverse events by 25%. Additionally, real-time integration with electronic health records (EHRs) improved clinical decision-making efficiency.

Conclusion: AI-driven DtD checkers represent a transformative tool for antimicrobial stewardship, enhancing precision prescribing while mitigating AMR development. Future research should focus on real-time deployment, clinician–AI interaction models, and broader validation in outpatient settings to maximize patient safety and antibiotic effectiveness.