Reviewing and summarizing electronic health records can be time-consuming for medical experts. Mobilize Center co-lead Akshay Chaudhari and a group of researchers adapted eight large language models (LLMs) to summarize clinical reports for four different clinical tasks: radiology reports, patient questions, progress notes, and conversations between doctors and patients. Results were evaluated by 10 doctors to evaluate completeness, correctness, and conciseness. In most cases, summaries from the best-adapted LLMs were either equivalent or better than summaries generated by medical experts. One potential drawback of these LLMs is they do not account for how different types of doctors may prefer different styles or lengths of summaries. Nevertheless, this research provides evidence that LLMs could be a valuable tool for summarizing clinical information, allowing doctors to focus more on patient care.
Categories: News
