AI Analyzes Clinical Notes for Medical Research on GLP-1 Medications

Why it matters locally: The advancements in AI-driven medical research could influence healthcare practices and patient data utilization in Pennsylvania hospitals and healthcare systems, potentially streamlining research efforts and improving patient outcome tracking for residents across the state.

BETHESDA, Md. – A study published today in Nature Medicine indicates that artificial intelligence can accurately read and extract patient information from doctors' clinical notes. Researchers at RespondHealth conducted the study, which focused on tracking patient outcomes for those taking GLP-1 medications. The study analyzed weight changes in 16,061 individuals and blood sugar levels in 14,788 individuals. This data came from the detailed, unstructured notes doctors create during patient visits. Standard analytical methods typically do not access this type of information. Dr. Sarah Chen, lead author of the study and a researcher at RespondHealth, explained the significance of the approach. "Each patient visit generates extensive clinical documentation. Much of the detailed progress and challenges patients experience reside within these free-text entries," Chen said. "Our previous research indicated that over half of clinically significant information, including a patient's weight or blood sugar readings, often only appears in these narrative notes." Artificial intelligence trained on medical terminology and context allowed the research team to interpret these notes at scale. The system identified specific metrics, such as body weight and HbA1c levels, which indicate average blood sugar over time. The AI then extracted these measurements along with their corresponding dates. The researchers reported that the AI demonstrated high accuracy in this extraction process. This capability allowed for the collection of a comprehensive dataset that captured granular changes in patient health markers over time. The study provides an example of how AI can expand the scope of medical research by unlocking previously inaccessible data within electronic health records.