Using ChatGPT to Determine Eligibility for Clinical Trials

Recent advancements suggest that artificial intelligence, specifically ChatGPT, can significantly streamline the process of screening patients for clinical trial eligibility. A study published in the journal Machine Learning: Health demonstrates how AI models like GPT-4 can quickly analyze electronic health records (EHRs), including unstructured data such as physicians' notes, to identify suitable candidates for trials. This approach has been tested at UT Southwestern Medical Center, where ChatGPT assessed data from 74 patients to determine their eligibility for a head and neck cancer trial. Traditional screening methods are time-consuming, averaging around 40 minutes per patient, and often overlook eligible candidates due to limitations in manual review capacity. By leveraging AI, researchers were able to reduce screening times to between 1.4 and 12.4 minutes per patient, at a minimal cost. Three prompting methods were tested: Structured Output, Chain of Thought, and Self-Discover, with GPT-4 showing the highest accuracy despite slightly longer processing times and higher costs compared to GPT-3.5. Experts believe that AI-assisted screening can support human reviewers, accelerates patient enrollment, and reduces trial delays—ultimately expediting the development of new therapies. This innovative use of machine learning exemplifies the potential of AI to transform clinical research, making trials more efficient and accessible. The study also highlights ongoing advancements, such as AI systems capable of real-time adjustments in radiation therapy, pointing to a future where AI supports various facets of patient care and treatment planning.

Source: https://medicalxpress.com/news/2025-07-eligible-clinical-trial-chatgpt.html