KEY TAKEAWAYS

• A recent study claims that ChatGPT may answer patients’ medical questions with higher levels of quality and empathy than physicians.
• The authors propose that chatbots may serve as tools to draft responses which can be vetted by physicians, reducing the amount of time they must dedicate to virtual healthcare and improving the overall quality of medical care.

As highlighted in a recent article in JAMA Internal Medicine, physicians are increasingly facing burnout from the rising demands of virtual healthcare. Spurred on by the COVID-19 pandemic, patients are more frequently using messaging systems to ask medical questions. The study, by Dr John W Ayers et al, suggests that artificial intelligence (AI) assistants could help to tackle this workload.

Healthcare professionals prefer chatbot responses

The researchers randomly selected 195 interactions between patients and verified physicians from Reddit’s online forum r/AskDocs. They presented the patient questions to ChatGPT (in ‘fresh’ sessions) and asked healthcare professionals (HCPs), who were blinded to the origin of the responses, to evaluate those from both the clinicians and chatbot. HCPs:

• preferred ChatGPT responses 6% of the time,
• found ChatGPT responses to be of higher quality (78.5% ‘good’ or ‘very good’, compared with 22.1% of physicians),
• and more empathetic (45.1% ‘good’ or ‘very good’, compared with 4.6% of physicians).

Using chatbots could benefit patients and doctors

The authors suggest that reducing the workload of virtual healthcare could not only alleviate physician burnout but also free up time for physicians to spend on more complicated tasks. This could improve job satisfaction, as well as helping patients.

They also highlight that virtual messaging platforms are vital to ensure equitable healthcare access when patients are restricted from visiting clinics. Indeed, if overloaded services are unable to respond to patients in a timely manner, patients themselves may turn to AI to seek answers to their medical problems. To prevent the spread of misinformation, physicians must be able to vet AI-generated advice.

To prevent the spread of misinformation, physicians must be able to vet AI-generated advice.

We need further investigation into using chatbots in healthcare

The potential for chatbots to reform virtual medical care looks promising, with ChatGPT outperforming physicians in this study. Healthcare centres may in future adopt chatbot services to draft replies, which physicians review before sending to patients. However, the authors call for randomised controlled trials to assess the effects of incorporating AI into patient messaging services, in terms of both patient outcomes and potential benefits for physicians.

————————————————–