My friend recently sent me an article featured in the LA Times about a New York City-based healthcare startup that has received approval to launch an AI agent, in a very narrow scope, to replace the physician’s role to prescribe medications for humans. She said, “I’m scared and still want to talk to my human docs!”
Of course, I had to read the article to find out what she was talking about. I thought she must have misread it. Surely this couldn’t be happening in the real world! AI replacing physicians?
Well, no, not exactly… at least not quite yet. It turns out the startup Nolla Health has received authorization in the state of Utah to run a pilot program for adults with acne to receive fairly low-risk topical prescription medications from an AI agent without having to see a physician. The pilot requires a doctor to review and approve the first 100 prescriptions before they are sent to the pharmacy. The next 400 will be sent to the pharmacy and retrospectively reviewed by an MD on a weekly basis, and in the final phase a physician will review at least 10% of all prescriptions monthly.
Sounds fairly reasonable, right?
My initial reaction, I assume like many, was, “Whoa, wait a minute, haven’t these people seen Terminator 2?” This was followed very quickly by an image flashing in my head, one I’ve seen in many futuristic space movies, where a terminally ill or mortally wounded protagonist is “saved” by being placed in a health pod that scans, identifies, and treats (sometimes surgically) the cause, allowing for the beloved hero’s full recovery. The confounding part? Both outcomes stem from the same idea.
The question seems to be this: Is this a good development, another small step for man and giant leap for mankind towards a healthier future? Or are we standing at a precipice of something we don’t yet understand the implications of?
The role of AI and its unprecedented prevalence in our lives is a hot topic causing massive controversy across the globe. While I am not in any way an AI expert, and in fact was a late adopter of the technology, I am a healthcare professional who can see both sides of this conundrum. As with all things, there are two sides to every coin. Good or bad, right or wrong, charged or neutral, every viewpoint is based on the lens that the person looks through. With AI becoming so heavily integrated in our lives in such a relatively short period of time, without us being able to fully realize the long-term implications of its involvement, and when you add the potential of it becoming more deeply involved in our medical decisions, it can feel kind of like sitting on a powder keg in the dark and lighting a match to see.
Strong leadership is based on accountability as much as the ability to see all sides, weigh the viewpoints and options, and make a determination on how to proceed based on the values and desired outcomes. From my healthcare and leadership perspectives, the ability of an AI agent to prescribe medications is an intriguing concept. Although this article was about AI prescribing acne medications that carry a low risk, it raises the potential for a path towards more involved automated processes for diagnosing and prescribing down the road.
As some might say, “It’s a slippery slope.” While it could improve access to care twofold, both for those who can’t get in to see a doctor (where I currently live there is up to a 9-month wait list to see a neurologist, and when I lived in NJ my primary care physician was booked for 6 months straight) or for those who can’t afford it, while at the same time freeing up physicians for more “serious” cases requiring human analytical thinking, it also raises serious concerns about creating precedent and a gateway for AI to become more prevalent within clinical decision-making in the healthcare setting. And while it is not subject to the burnout or mental fatigue of healthcare professionals that can lead to harmful cross-prescribing of medications, it also lacks the intuition and “gut feelings” innate to humans that could save a life. I’ve heard more than a few stories of physicians who have uncovered serious pathologies early on by listening to that little voice inside that “something just wasn’t right,” even though the patient’s symptoms indicated a straightforward diagnosis. One of my patients’ throat cancer was diagnosed in Stage One when his PCP followed a hunch about a sore throat he’d mentioned in passing during his annual physical. There were no other symptoms and it was during the winter months when sore throats run rampant. He later stated he wouldn’t have sought medical attention for it. Would AI be able to do that? Based on its programming, perhaps, or perhaps not.
The implications are vast. Although this particular startup is “low risk,” what does it lead to next and how would we ensure the safety of our health with AI as a prescribing entity? Would it overprescribe medications “just to be safe,” potentially causing new strains of super bacteria or another version of the opioid dependency crisis? Might it overutilize diagnostic testing for the same reason? Or on the flip side, underutilize to avoid the scrutiny of overutilization? And how would it be held accountable for malpractice in the case of a misdiagnosis or an adverse event?
These pertinent questions present complex topics needing to be unpacked, topics that span and intertwine legislation, ethics, and the very fabric of capitalism. There is no doubt that AI has its place within our daily lives and healthcare. It provides massive quantities of information at our fingertips that can assist us in making more informed decisions, while freeing up our brains for more high-level mental processing of that information. In the healthcare setting, it is already providing clinicians with faster documentation solutions allowing for less time spent on administrative tasks and more time for patient care.
As we are in the nascent stages of realizing its full potential, the real question is whether we are ready to invite AI into the role of a prescribing physician. It may seem a far distance from the current pilot, but in actuality we are standing at the threshold of this brave new world.
It’s easy to jump to one side or the other of the argument. The more challenging, and required, task is to rise up to a 100,000-foot view, disengage from any emotional ties to the topic, and decide as a society where we want to go and how involved we want AI in our healthcare. The pros and cons must be listed and weighed. Debates must be had regarding the risks and benefits. Legislation must catch up to effectively regulate this new frontier. It should not be a short-term implementation, and its ramifications should not be underestimated.
Whether we see it as good or bad, the introduction of AI as a prescribing agent is upon us and may be the next evolution in our society. If so, we must be vigilant in providing the proper vision, regulations, and guidelines to maximize the chances of success while minimizing the potential for harm. We must remember that AI is built on human programming, and if we are to invite it into a more intimate role within our society and healthcare it should be with a solid understanding and well-defined parameters of what we want it to do for us, how we want it to contribute, and when to rein it in or allow it to do more. Regardless of whether AI becomes a more autonomous partner in healthcare, we are ultimately accountable for the decisions we entrust to it and must take that responsibility head-on.
As for me, I’m with my friend: I still want to talk to my human docs.
To read the full article, Utah allows AI startup to prescribe acne medication without a doctor’s direct oversight, click here.
