Generative AI explains echo results to heart patients
Researchers have used ChatGPT, OpenAI’s massively popular artificial intelligence (AI) chatbot, to rewrite echocardiogram reports in plain language that patients can easily understand. The group shared its experience in JACC: Cardiovascular Imaging.[1]
“The immediate release of echocardiogram reports to patients, as mandated by the 21st Century Cures Act, may cause unnecessary worry or confusion until clinicians provide explanations,” wrote senior author Lior Jankelson, MD, PhD, an electrophysiologist and AI specialist with NYU Langone Health and NYU Grossman School of Medicine, and colleagues.
Jankelson et al. received special access to OpenAI’s ChatGPT technology to ensure they could run their experiments using real patient data while still following all necessary privacy regulations. They then used ChatGPT to generate 100 patient-friendly echo reports, asking a team of five board-certified echocardiographs to evaluate the reports for accuracy, relevance and understandability.
Overall, the human imagers said 84% of AI explanations were “all true” and the other 16% were “mostly correct.” In addition, while 76% contained “all of the important information,” another 15% contained most of it, 7% contained approximately half of it and just 2% contained less than half. None of the missing information was found to be “potentially dangerous.”
“Our study, the first to evaluate GPT4 in this way, shows that generative AI models can be effective in helping clinicians to explain echocardiogram results to patients,” Jankelson said in a statement. “Fast, accurate explanations may lessen patient worry and reduce the sometimes overwhelming volume of patient messages to clinicians.”
“If dependable enough, AI tools could help clinicians explain results at the moment they are released,” added first author Jacob Martin, MD, a cardiology fellow with NYU Langone. “Our plan moving forward is to measure the impact of explanations drafted by AI and refined by clinicians on patient anxiety, satisfaction, and clinician workload.”
The group also found, as one may expect, that people without clinical backgrounds found the patient-friendly reports preferable to the original reports in nearly every instance.
“Our next step will be to integrate these refined tools into clinical practice to enhance patient care and reduce clinician workload,” Martin said.
Read the team’s full analysis here.
Michael has more than 16 years of experience as a professional writer and editor. He has written at length about cardiology, radiology, artificial intelligence and other key healthcare topics.