Until recently, most patients would bring their health concerns to a clinician. Now, more patients are showing up for appointments after first consulting an AI chatbot about their symptoms, fact-checking what their doctors say against AI, or using AI in lieu of a clinician and not seeking care at all.
For example, a recent survey released by the insurer Aflac in August found three-quarters of Gen Z adults and nearly two-thirds of millennials turn to AI for health support before seeking professional care.
“A wellness paradox is emerging as Americans — particularly younger generations — feel more confident and in control of their health than ever before, yet they have never been less likely to see a doctor,” according to an Aflac press release issued with the survey.
Their survey of 2,000 adults found that 18% of Gen Z and 14% of millennials only schedule doctor’s appointments after exhausting AI or online tools, compared to 3% of Gen X and 2% of baby boomers. They also are more likely to delay or skip routine checkups and screenings (65% for Gen Z and 61% for millennials), largely because they feel healthy. Yet, these groups are more likely to spend at least $100 per month on items like vitamins and organic foods, mental health therapy, wellness retreats and more, the survey said.
Journalists covering AI and/or patient safety should note and write about this trend to inform their audience about potential risks.
The dangers of relying on AI
It’s not only these groups. Patients today are overall more likely to arrive at appointments already armed with ideas, opinions or potential treatment paths shaped by online searches or AI, according to a blog post on Quirk’s Media. Generative AI is particularly shaping how cancer patients prepare for visits, form questions and process information throughout their care journey, the post said, citing a study conducted by the market research firm Lieberman. They might use AI to better understand a diagnosis, interpret test results, explore treatment options and prepare for clinical conversations.
The majority of American adults view their test or imaging results online before talking to a doctor, the Wall Street Journal reported, with a growing share uploading results straight into AI chatbots. This has left some medical professionals concerned about misinterpretations, AI hallucinations and privacy issues, the article said.
The danger is that as people engage with AI, they can start to view the responses as credible, these sources note.
A major concern that has received extensive media coverage is mental health. Many patients use AI and chatbots as a substitute for care with a licensed mental health professional, which can delay care with a real clinician, said Megan Coder, vice president of the Society for Digital Mental Health, in an interview with HIMSS TV.
Large language models designed for general use may create their own version of a mental health professional that sounds very convincing but is laced with false or dangerous information, she said.
“We’re trying to refute what the patients have convinced themselves that they have based on whatever they’ve searched and reviewed,” Jason Goldman, M.D., a Florida internist, told the Wall Street Journal in a recent story. “You can get everything from ‘You have a cold’ to ‘You’re dying of cancer.’”
Goldman told the Journal he has seen patients using conversations with chatbots to justify stopping life-saving cholesterol medications or question recommended vaccines. Some patients have told Goldman they will compare his recommendations to AI.
Why turn to AI first?
Some 34% of U.S. adults reported asking AI chatbots about their health for a variety of reasons, according to a recent Pew Research Center survey. Those included:
- Speed
- Cost
- Finding a possible cause of their symptoms
- Seeking treatment information
- Seeking more information about a doctor’s diagnosis
- Getting a better understanding of lab results
- Finding information about a sensitive topic
- Deciding whether to go to a doctor
Nearly half (47%) of those who asked AI for health advice said the information they received was extremely or very helpful.
Other reasons include long wait times to see a doctor and trying to avoid copays, according to a recent Washington Post article that spoke with six patients who used AI for health advice or answers. Medical experts who reviewed each of these cases for the Post gave it mixed reviews, including concern over telling a patient with stomach pains and vomiting to take over-the-counter medications. They warned that chatbots may give poor advice if users say they aren’t too worried about symptoms.
Embracing this trend, some medical centers have begun offering their own AI-powered symptom checkers, as I wrote about for AHCJ in August. The potential benefit to these programs is that they are managed by medical centers who can direct patients to appropriate clinicians as needed.
Legal ramifications
When a chatbot gets health symptoms wrong and a patient gets hurt, who bears the responsibility? Meghan O’Connor, a health law partner at Quarles & Brady, told MedCity News the answer is still emerging.
“This is one of the most significant open questions in health care law right now,” she said in the article, “and the honest answer is that liability allocation is going to be highly fact-dependent and will likely take years of litigation to clarify.”
AI companies do make disclaimers stating their outputs aren’t medical advice, O’Connor told MedCity News, but that might not hold up in court if a chatbot was marketed in a way that encouraged patients to treat its answers as diagnostic guidance.
Courts are unlikely to blame patients for trusting an AI tool. But the story for providers might be different.
If a patient tells a provider they have been relying on inaccurate AI advice, a clinician’s failure to correct it could become a malpractice issue, O’Connor told MedCity. Her practical recommendation for providers is to correct misinformation, document the conversation and treat AI-sourced information the same as any patient-reported information that conflicts with their clinical judgment: address it directly.
Health systems should assume AI-related litigation is coming and prepare accordingly, medical malpractice attorney Rick Groffsky of Sommers Schwartz told Healthcare IT News. Physicians will remain at the center of malpractice analysis because courts still expect clinicians to exercise independent medical judgment, the article said.
Future suits may cite not just independent clinicians but health systems that chose to purchase or make AI tools available to clinicians, Groffsky said. He also recommended that health systems keep thorough documentation of how they evaluated AI products, including the training data used, who is responsible for oversight, and how they monitor performance after implementation.
Future lawsuits could arise from both an overreliance and underreliance on AI, he noted.
Story angles
There are several ways to continue following this story either on a local or national level, such as:
- Speak with physicians and other medical experts to see how frequently their patients are asking about/mentioning health information received from AI. How are they counseling these patients? What cautions do they have for other patients?
- Interview patients of different ages/ethnicities/regional areas to see how they are using AI for medical information, and what their reasons are. Did they have any notable health outcomes made better or worse by what AI suggested? Did their use of AI impact their relationship with their regular clinicians, and if so, how?
- Interview hospital and health system informatics officers about any AI tools they are developing or using: How are they vetting them? How are they coaching physicians on how to use or document any information returned by the tools?
- Look into different AI platforms used by people for medical information. What kinds of disclaimers do they state? What do they do with medical information uploaded by consumers?
- Follow legal suits. For example, a Florida man is suing ChatGPT because he claims it told him not to seek medical help after repeatedly asking about symptoms in the build-up to a near-fatal pulmonary embolism, BBC News reported.
Resources
- From diagnoses to treatments, why Americans use AI chatbots for health – Pew Research Center
- Doctors don’t want patients to read test results with AI. They’re doing it anyway – Wall Street Journal
- AI misdiagnoses raise new liability questions for health systems – Healthcare IT News
- Patients are bringing AI into the exam room – Quirk’s Media
- Behavioral health needs may go untreated due to AI – HIMSS TV
- What happens when patients trust AI over their doctor? – MedCity News
- Misleading AI-generated doctors pose ‘huge danger to public safety’ – The Guardian
- 1 in 3 Americans use chatbots for health advice. These 6 patients explain why – Washington Post
- AI-powered symptom checkers direct worried patients to appropriate care – AHCJ blog post
- AI chatbots and mental health: How to report responsibly on a new risk – AHCJ webinar
- Understanding ChatGPT Health: How large language models are being used to triage care – AHCJ webinar
