Bridging AI and medication safety: Comparative evaluation of ChatGPT's drug interaction detection capabilities

The detection of drug interactions remains a critical challenge in clinical practice, with potential consequences ranging from therapeutic failure to severe adverse events. This study evaluates the performance of ChatGPT in identifying drug interactions compared to established clinical tools, including Medscape, Lexicomp, and Drugs.com. Using a dataset of 250 commonly prescribed medication combinations, we assessed accuracy, sensitivity, specificity, and response comprehensiveness across platforms. ChatGPT demonstrated 78.6% overall accuracy, compared to 94.2% for Lexicomp, 91.8% for Medscape, and 89.4% for Drugs.com. While ChatGPT excelled in providing comprehensive explanations of interaction mechanisms (mean score 4.2/5 versus 3.8/5 for traditional tools), it exhibited lower sensitivity in detecting critical interactions (76.3% versus 93.7% for established tools) and higher false favorable rates for certain drug classes. Our findings suggest that while ChatGPT shows promise as a supplementary tool, particularly for generating patient-friendly explanations, it currently lacks the reliability necessary for standalone use in clinical decision-making. This research highlights the potential and limitations of large language models in drug interaction screening and emphasizes the need for continuous validation and refinement before implementation in clinical practice.