In Palo Alto, California, during a recent medical checkup, I discovered that the doctor came equipped with an AI-powered tool. This was an opportunity to see the potential and challenges of this technology firsthand.
Dr. Christopher Sharp of Stanford Health Care said, opening an app on his smartphone: “Before we begin, I’d like to ask you a quick question. I’m using a technology that records our conversation and uses artificial intelligence to summarize it and prepare my clinical notes.”
During the examination, Sharp deliberately spoke aloud to record blood pressure and other findings so the AI could process them. He also uses AI to draft initial responses to patient messages, including treatment advice.
Artificial intelligence has already become part of the patient-doctor relationship. Over the past year, millions of people have received care from healthcare providers who use AI to handle routine tasks. The goal is to reduce physician burnout, speed up treatment, and potentially avoid errors.
AI in your clinic
While the use of AI sounds exciting, its rapid adoption in a sensitive field like medicine raises concerns. Errors caused by this type of intelligence, known as “hallucinations,” can be serious. For example, one study showed that 20% of answers provided by ChatGPT to medical questions were “inappropriate.” A physician might inadvertently pass along inaccurate recommendations.
There are also other risks. For instance, transcription systems have been shown to invent details that were never mentioned, increasing the potential for misunderstandings.
Current developments
Companies like Epic Systems use AI tools to write reports and summarize patient visits, and there are more than 100 new products in development. Startups like Glass Health are also innovating diagnostic and treatment recommendations, while K Health offers patient advice via its chatbots.
However, the biggest concern is that most of these tools do not require Food and Drug Administration approval because they do not provide definitive medical decisions. Doctors are required to review the results, but how thorough is that review?
The upside and challenges
During my experience with Dr. Sharp, there were clear benefits. Using the transcription technology (DAX Copilot), the doctor was able to maintain eye contact the entire time, rather than focusing on a computer screen as is usually the case. This technology records and summarizes conversations, saving time. However, some technical errors remain, raising questions about reliability.
Even when it comes to answering patient messages, there is still plenty of work to do. In one instance, the AI provided food allergy treatment recommendations that included using a steroid cream on the lips, which Sharp rejected, saying: “I wouldn’t recommend this treatment; lips are extremely sensitive.”
Are we ready for this technology?
Ultimately, how much you trust AI in medicine depends on your trust in your physician. While some argue that these tools may ease administrative burdens, the question remains: do these tools weaken doctors’ ability to make sound decisions?
If you are concerned about the use of AI, you can ask to review the notes or summaries yourself. In the end, technology can be a powerful tool if used cautiously and within well-considered limits.
Despite the drawbacks, researchers agree that artificial intelligence can improve the efficiency of the healthcare system if tested and proven beneficial clearly. Therefore, we must be cautious and not rely solely on these tools before they reach the required level of maturity.
