AI in Healthcare: Mixed Results Emerge Amidst Enthusiasm

Despite initial excitement, recent studies cast doubts on generative AI's ability to reduce physician burnout. Evaluations of AI systems for tasks like patient messaging, medical billing, and editing replies to patient messages have shown limitations and potential risks.

Amidst the growing adoption of artificial intelligence (AI) in various industries, healthcare has emerged as a fertile ground for innovation and potential transformation. AI-powered solutions have been touted for their ability to automate tasks, improve efficiency, and enhance patient care. However, recent research has dampened some of the initial enthusiasm surrounding generative AI's role in reducing physician burnout.

One study, published in The Lancet Digital Health, examined the use of AI for electronic patient messaging. Researchers prompted a large language model (LLM) to respond to simulated questions from cancer patients and compared its output to responses from board-certified radiation oncologists. The results revealed that the LLM drafts posed a risk of severe harm in nearly 11% of the survey responses, with one response even suggesting death.

Another study, published in the journal NEJM AI, evaluated four different LLMs for performance and error patterns when querying medical billing codes. The research found that all tested LLMs performed poorly on medical code querying, often generating codes that conveyed imprecise or fabricated information. The study concluded that LLMs are not appropriate for use on medical coding tasks without additional research.

A third study, published in JAMA Network, evaluated AI-drafted replies to patient messages and physicians' time spent editing them. The assumption was that generative AI drafts would lessen a physician's time spent on these tasks, but the results showed otherwise. Researchers found that AI-drafted replies were associated with significantly increased read time, no change in reply time, significantly increased reply length, and only some perceived benefits.

David Atashroo, M.D., chief medical officer of Qventus, an AI-powered surgical management solution, believes that AI has immense potential in healthcare, but realistic expectations must be set. He suggests that AI can take on lower-risk, highly automatable tasks traditionally performed by essential yet often overlooked administrative staff.

Safety and efficacy are paramount in AI applications in healthcare. Atashroo emphasizes the need for rigorous quality checks, regular assessments by human experts, and transparency in the development and implementation of AI technologies to ensure responsibility and reliability.

While AI remains a promising tool in healthcare, recent studies indicate a need for further research and development before it can effectively address physician burnout. AI-powered solutions must undergo rigorous testing and validation to ensure accuracy and safety before widespread adoption. Collaborations between AI developers, healthcare professionals, and regulators are crucial to harness the full potential of AI in transforming healthcare while mitigating potential risks.