In April of this year, a groundbreaking study published in JAMA Internal Medicine sent shockwaves through the healthcare community. The study, led by researchers at the University of California, San Diego (UCSD), investigated the effectiveness of ChatGPT, an AI-powered chatbot, in responding to real patient questions. The results were astonishing: ChatGPT's responses were not only preferred over those of physicians but also rated significantly higher for both quality and empathy. Since then, this research has sparked intense discussions among healthcare professionals, policymakers, and researchers about the role of AI in healthcare and public health.

The Study: ChatGPT vs. Physicians

The study aimed to address a pressing issue in healthcare – the overburdened workload of chief medical officers (CMOs) due to an influx of patient messages. Dr. Ayers and his team wanted to determine if ChatGPT, in its early form (3.5), could effectively answer patient questions and alleviate this burden. To do this, they compared ChatGPT's ability to answer questions, both in terms of quality and empathy, with actual doctors.

To gather data, the researchers turned to a Reddit forum called AskDocs, where individuals submit medical questions, and licensed healthcare professionals provide anonymous and free responses. The study then compared these real patient questions and physician responses with those generated by ChatGPT. The results were astonishing: healthcare professionals preferred ChatGPT's responses four times more often, rated them as very good or good in quality four times more often, and judged them as empathetic ten times more frequently compared to responses from physicians.

The takeaway here is not that AI chatbots outperform physicians in terms of empathy, but rather that AI tools like ChatGPT have the potential to address the growing issue of patient inquiries, improve response quality, and potentially relieve some of the burden on healthcare providers.

Improving Patient Outcomes through Messaging

Dr. Ayers emphasized that the primary goal of this study was to improve patient outcomes. The study suggests that AI-powered messaging tools can be evaluated in randomized controlled trials to determine their impact on various clinical conditions. For example, these tools could help post-myocardial infarction (MI) heart failure patients by providing personalized messages based on their electronic health records (EHRs), helping them adhere to clinical feedback, and updating their care plans as needed.

One of the key takeaways from this research is that AI-assisted messaging has the potential to enhance patient-provider communication, transforming the way patients and healthcare providers interact. Instead of merely receiving and responding to messages, healthcare providers can send proactive, personalized messages based on patients' EHR data, improving patient engagement and adherence to clinical advice.

Disclosure and Doctor in the Loop

Regarding disclosure, Dr. Ayers explained that the current practice involves disclosing that the original message was generated by AI but enhanced by a doctor. This "doctor in the loop" approach ensures that AI-generated content is refined, edited, and improved before reaching the patient. This method not only ensures message quality but also streamlines the triage process, allowing healthcare providers to allocate more time to clinical tasks that require their expertise.

The Potential in Public Health

Dr. Ayers also highlighted the potential of AI in public health messaging. AI-powered tools can provide precision messaging by tailoring messages to individuals based on their health records, improving the speed and effectiveness of communication. This is particularly crucial in addressing issues like suicide, addiction, and other behavioral medicine concerns, which are significant drivers of premature mortality.

The Future of AI in Healthcare and Public Health

The White House Executive Order on AI has directed the Department of Health and Human Services (HHS) to develop AI-specific regulatory regimes by late April 2024. Dr. Ayers expressed the importance of focusing on clinical endpoints when evaluating AI in healthcare. He argued that regulators should prioritize patient outcomes over process indicators, such as code quality and safety. By setting the goal of improving patient outcomes, regulators can ensure that AI technologies contribute positively to healthcare.

In conclusion, the study led by Dr. John Ayers and his team at UCSD has shed light on the transformative potential of AI in healthcare and public health. While AI chatbots like ChatGPT have shown promise in improving response quality and patient engagement, their ultimate value will be determined by rigorous scientific studies with clinical endpoints. As AI continues to evolve, healthcare providers, policymakers, and researchers must prioritize patient outcomes to ensure that AI technologies enhance the quality of care and improve public health. The future of healthcare and public health communication is undoubtedly undergoing a paradigm shift, and AI is at the forefront of this revolution.

Source: Read the full article on JAMA