A Houston Methodist physician wants patients using ChatGPT Health, as long as they show him the chat

At a glance: Patients are using ChatGPT Health to decode medical records and investigate symptoms, and most welcome AI in their care as long as a clinician reviews the results. Jordan Dale, MD, chief medical information officer and chief health AI officer at Houston Methodist Hospital, explains why AI answers miss the mark when patients omit context and what to weigh before sharing records with tools outside HIPAA.

Earlier this year, OpenAI launched ChatGPT Health, making way for conversations about what it means for patient care and personal health data. And research shows that patients are here for this evolution in healthcare technology. In fact, a University of Michigan study published in JAMA Network Open showed that a majority of patients welcome the use of artificial intelligence in medicine, so long as a real person (i.e. their clinician) is still reviewing the information behind the screen. The study showed that even if these tools have proven to perform well, visible safeguards and demonstrated effectiveness are core components that will get patients and providers to more fully embrace their use, and there’s still a long road ahead to get there.   

Even before ChatGPT Health’s release, according to OpenAI, more than 300 million people were already turning to ChatGPT to ask questions about their health. What makes ChatGPT Health different is the ability for users to also upload their personal medical records and sync wellness apps in an effort to get answers about symptoms they may be experiencing, or just general health advice.  

According to Jordan Dale, MD, a hospitalist who also serves as chief medical information officer and chief health AI officer at Houston Methodist Hospital, it makes sense that patients would want to utilize tools like this to get to the bottom of symptoms they’re experiencing. After all, we’re all busy and it can take weeks or even months to get in with a primary care physician as a new patient, especially amid an ongoing primary doctor shortage. Additionally, AI tools can help patients to break down complex medical jargon that accompanies their medical records, such as a CT scan or EKG, into patient-friendly lay terms if they haven’t yet heard back from the provider to summarize the findings. 

But the missing piece of the puzzle in achieving an accurate diagnosis from an AI tool is painting a full picture of their situation, an aspect that, for many patients, is easier said than done, he says. That’s why, rather than simply being presented with data and information that a patient was given by an AI tool, he prefers to have a patient walk him through the interaction to see what questions and information they fed the tools to reach the diagnosis or recommendation it came to. 

“I think it’s a positive thing if we’re getting patients to engage in their healthcare, but at the same time, there is a rhythm of how clinicians go through looking at a clinical case, and we are very prescriptive about how we ask questions about your symptoms,” Dr. Dale says. “I’ve had instances where a patient mentions they got a different diagnosis from an AI tool, so I’ve looked at the questions they’ve asked and noted they forgot to provide their full list of medications, or I’ve asked them if their pain goes away when they perform a certain maneuver. Oftentimes, AI is only getting a snippet of your situation because that was easy to type in, and you have to really provide the full context to these tools to get the right answer.” 

With “AI” literally being a part of his job title, it goes without saying that Dr. Dale supports patients using AI tools to efficiently get accurate answers about their health. But he emphasizes the importance of walking your actual provider through your online interactions, especially if there’s a conflict between the answers it gave you and the diagnosis your clinician handed you. One of the biggest hurdles for patients and clinicians alike is the fact that most free AI tools don’t meet the privacy standards of healthcare providers and institutions, which can pose risks when it comes to protecting their personal information. Consumer health apps such as ChatGPT Health are also not covered by the Health Insurance Portability and Accountability Act (HIPAA), which is why a provider will usually be reluctant to simply take a patient’s health records and interact with AI with them in the office, Dr. Dale says.  

“Very few consumers read the privacy policies that outline whether your interactions are training the model or whether your data is being stored, which is where patients may want to think twice about sharing things like their medical records, even if it could prove useful for receiving personalized recommendations,” Dr. Dale says. “I think we’ve become very desensitized to that thanks to the social media era and now the availability of these tools, but it’s a true risk, and an individual has to weigh that risk for what information they’re providing.” 

AI’s role in addressing inequities in care is still unsettled, as women and minority patients have historically had their concerns routinely dismissed, often leading to poorer outcomes from late-stage or terminal diagnoses. Patients in rural areas might question whether their symptoms warrant a long drive to see a physician, but if a situation is nagging or urgent enough that it would be risky to delay care, Dr. Dale notes that patients shouldn’t rely on these tools alone.   

“Ideally, we’ll eventually see these tools attached to a health system or healthcare provider so that it can direct you to a virtual visit with a provider so they can actually treat you by prescribing medications or ordering X-rays to validate a diagnosis,” he says. “The true gap is still the benefit of the services they need, and that’s where there are still opportunities to see how we can partner with them to get patients the care that they need.” 

Emilia Benton is a Denver-based health and wellness journalist whose work has appeared in publications such as Runner’s World, SELF, and Women’s Health. She is also a 15-time marathoner, USATF Level 1-certified running coach, and an avid traveler.