New Study Demonstrates AI's Accuracy in Interpreting Unstructured Clinical Notes for Medical Research
A new study published in Nature Medicine, co-led by researchers at RespondHealth and Drexel University, demonstrates that Artificial Intelligence (AI) can accurately read and interpret unstructured clinical notes at scale. Historically, a significant portion of patient information, often found in doctors' written notes, has been inaccessible for medical research due to its unstructured nature. The study developed a system that converts this narrative text into computable and analyzable data, while maintaining traceability back to the original sentences. The AI system achieved 99.4% accuracy in a combined measure of correctness and completeness, outperforming the agreement rate between human physicians (94.7%). This breakthrough allows for the extraction of crucial details, such as depression scores, pain intensity, and even blood sugar readings, that are often only recorded in prose and not in coded medical fields.
