In California and Wisconsin (USA), OpenAI ChatGPT’s generative artificial intelligence helps doctors prepare answers to patients’ online questions. This is done to test whether AI can reduce the time medical staff spend on online requests, writes The Wall Street Journal.
UC San Diego Health and UW Health began testing in April. Stanford Health Care plans to join the project soon. In total, this tool is being piloted by about two dozen medical workers.
Marlene Millen, a primary care physician at UC San Diego Health, has been testing GPT in her mailbox for about a week. According to her, the first AI responses had to be heavily edited, so her team is working on that. Also, project participants add a unique way of communication. For example, if a patient mentions a recent trip, the draft response from the AI might include a line asking if everything went well.
“It adds a human touch that we would,” explains the doctor.
She also adds that the AI pilot project is exciting for some medical professionals. Millen notes that doctors are so exhausted that they are looking for any hope.
According to Wisconsin-based Epic, which developed the MyChart messaging tool for healthcare providers, logins reached 260 million in Q1 2023, up from 106 million in Q1 2020. Now, when clicking on a message from a patient, the AI shows the doctor a draft of a possible answer.
Seth Hain, Epic’s senior vice president of research and development, says that when preparing a draft response, the AI compares the information in the patient’s message with the abbreviated version of their electronic medical history. At the same time, medical data remains protected, in accordance with federal laws that require the protection of patient privacy.
Currently, the San Diego team prevents the AI from responding to any queries that require medical advice. Similarly, in Wisconsin, 10 UW Health physicians have allowed AI to answer a limited set of patient questions, including requests for prescriptions and requests for documentation or paperwork, according to Chero Goswami, UW Health’s chief information technology officer.
Administrators and doctors believe this tool could be transformative if it really works. However, if the drafts require too much fact-checking or editing, doctors may lose confidence in the new option.
We remind you that ChatGPT users can now turn off history of its chats, so OpenAI will not use data from chats to train artificial intelligence models. To do this, account holders must click a switch in their settings.