{{ But nearly every report has missed one detail: That figure was derived from urgent care encounters, not from chronic medication renewals, which are the focus of the Utah pilot. These are different clinical tasks. And 99.2 percent accuracy, applied at scale, still means errors affecting thousands of patients. }}
What’s the accuracy for a live doctor? I bet if I had an AI doc, I’d still have my leg.
WP was paywalled, but Google AI came up with this…
Key Aspects of the Utah AI Prescription Program:
Technology & Process: Patients use a web portal to request refills for ~190 commonly prescribed drugs. The AI verifies identity, location, and queries symptoms/side effects before approving.
Safety Measures: For each drug class, a human physician reviews the first 250 prescriptions before the system acts autonomously. Cases involving, for example, controlled substances are escalated to human doctors.
Regulatory Status: Authorized by the Utah Department of Commerce’s Office of Artificial Intelligence Policy (OAIP) under the AI Learning Laboratory Program, making it the first such program in the U.S..
Criticism: Some, including Public Citizen, have criticized the use of terms like “AI doctor” to describe the software, arguing that autonomous AI should not replace human medical judgment.
Cost: The service currently charges a $4 fee per renewal.
This pilot represents a significant shift towards using AI in medical decision-making, aiming to improve medication adherence and reduce administrative burdens.
“In Utah, an artificial intelligence [platform called Doctronic](https://www.doctronic.ai/) is legally renewing prescription medications for patients without physician involvement. For a $4 fee, it conducts the intake, evaluates the patient and sends the prescription to a pharmacy.
Operating under a state regulatory exemption, this is a live commercial program handling real prescriptions for real patients with plans to expand to a dozen more states this year.
Yet the entire scientific case for its safety rests on a single unvalidated study.”
“That study is a preprint posted to the online site medRxiv in July. It has not been peer-reviewed. All eight of its authors hold equity in Doctronic. The company’s own internal committee conducted the ethics review. Colleagues requesting the underlying data were denied it. Eight months later, the study still hasn’t been published in a peer-reviewed journal.
That is the evidentiary foundation for the first autonomous AI prescribing program in the United States. And it is cause for serious concern.”
Excerpt From
“Opinion | Don’t trust this $4 solution for getting a prescription”
Anytime I see something touting AI, I always fall back on the old GIGO acronym (Garbage In Garbage Out) because AI is still a human coded program with built it human bias. I can’t possibly see anything going wrong with this (note heavy sarcasm).
Or, maybe AI would have told the docs to take the wrong leg and then you would be without either.
Two Texas lawsuits claim TruDi’s AI misled surgeons near carotid arteries, causing blood clots and strokes. The suits allege the AI was rushed to market with only an 80% accuracy goal, despite surgeon warnings about the risks.
What happens if your high blood pressure increases? Or your diabetic blood sugar increases? Etc etc. Seeing a med professional to monitor these events is important. It can be made more efficient. Automating renewals may be a bridge too far.
Most recurring prescriptions get written for 90 day supply w 3 refills. So you see Dr at least once per year. I’d say that is a reasonable minimum. Maybe automated refill is ok for selected situations like birth control. But not a good idea for many patients.
How many people “coding” AI have a medical background. Not zero but close enough. Seeing/treating a patient isn’t cookbook medicine strictly following an algorithm. Despite what insurance/government regulations and guidelines insist. You can’t just upload a bunch of medical texts and expect to get the correct outcomes.
I could write a book about diagnoses I’ve made that went against what the patient was telling me verses what their body language and non-verbal communication was showing me. Don’t think AI is anywhere near that ability.
I wonder how many physicians are able to keep up with all the medical literature. AI should make it possible to keep track of new findings. And apply them when traditional treatments are ineffective.
AI has much potential for rapid progress when used responsibly.
The OP and replies in that context point out that AI generated prescriptions without any oversight are in all probability not responsible usage of the technology in the here and now.
See, I don’t follow how this is of benefit to anyone other than the doctronic start up in the proposed form The safety net that’s allegedly in place allows only for renewal of prescriptions deemed to be “low risk” and for patients with chronic conditions who’ve already had said medications prescribed by an actual physician.
Now I’m not even going to mention that these folk tend to be the ones who probably need some degree of monitoring….that much is obvious. These folk already have such a service in operation. Per my experience, if my prescription runs out of refills (rare for me because both my PCP and cardiologist always ask/check at time of follow up appointment) the pharmacy itself contacts the prescribing physician for the appropriate renewal. For free. Gratis. Without charge. Can’t imagine squandering $4 and the time to fill put any questions on the app for the very same thing.
I’m sure anyone else who requires ongoing prescriptions has the exact same service available to them.
So, although the blather and alleged safeguards around this supposed new service in Utah emahasize thatthere’sno doctoring being done, and because I cannot see how it’s of benefit in the form proposed, I had a little look-see at the doctronic site. Just because. In case there’s anyone who hasn’t yet either…..
Now that doesn’t look to me like an outfit that’s not trying to muscle into actually becoming what it calls itself in the blurb. In the process of hunting around for more insight I stumbled across this….
I don’t find this too surprising since, as I’ve mentioned before a time or six, I have practical experience of this very type of manipulation. Come to think of it, in the field of prescription finagling, by the family albatross…..
Edit:….because I’d read up on the details of the Utah experiment, I was under the impression that meds like opioids etc were not being included in this pilot scheme. This second article alleges otherwise.
So, where’s the bamboozle one wonders?
All this “information” is sloshingaround the internet ready to be included in a Google AI overview for anyone to glom onto and believe.
So, a follow up to my mention of misuse of online prescription services…..and the relative ease in obtaining prescription only meds without real oversight.
“Anti aging”/hormone replacement/juicing is one such area…..with multiple “clinics” offering prescriptions for testosterone etc. One such being….
I discovered the existence of these places….and their popularity among the “bro” community….in the months prior to my daughter’s divorce. In sifting through the business financials, among some of the staggering charges on the hospital credit card (opened by the Lump of Foul Deformity) I spotted a recurring monthly fee for the above site. Dating back close to 18 months and coinciding (on hindsight) with quite a shift in The Lump’s already nasty behavior along with a big increase in muscle growth. Followed shortly thereafter with the appearance of subscriptions to dating sites [[ yes….all on one of the business credit cards!!!]]
Although the site claims physician overview, it turned out to be of the signing off on an online questionnaire and mailed in blood sample type. I’d just been “educated“ on the value of generative AI in concocting convincing …. if fictitious …. narratives and I was already aware of the relative ease of dropping testosterone levels by a small but adequate amount to game a blood test.
Seems like the sort of nitpicking thing I do (worrit over minutiae) but, again with something of a pay off come the time for legal shenanigans. In the pre hearing deposition, his pettyfogging shyster was attempting a few questions with my daughter in the form of ad hominem attacks that she wasn’t quite able to diffuse (Family court isn’t about arriving at the truth or an equitable and fair outcome…..it’s designed to crush the other partner and some lawyers and their puppeteers/clients revel in the process) …..until daughter’s lawyer asked specifically about any additional medication to his admitted antidepressants and ketamine treatment. He “plead the fifth”….which, apparently, in family court invites an inference of self incrimination. Anyways, it certainly tempered his counsel’s enthusiasm for character assassination come the Real Thing.
You won’t get this sort of insight on AI overview. It also has something of a public health implication beyond one family’s misery too, as I’ve discovered that there are more folk than I realised walking around with undiagnosed/incorrectly diagnosed psychopathies, and the ability to self medicate with stuff that’s even more consequential than ethanol should be of concern to all of us.
I’m sure much depends on the specialty. And there is keeping up with academic literature and practical literature. Unfortunately we are inundated with the former and little of the later.
For example, my specialty of anesthesia, years ago read about the effects of a new anesthetic gas on rabbit myocardium. Ok, maybe 10 or so years down the road this might work its way to the outpatient surgery center setting. Or, see papers on “do this, this, and this” and the patient is discharged from recovery room 2.3 minutes faster. They then try to extrapolate those findings into “real money” saved. Little application to real practice.
The biggest bombshell in my specialty was actually a negative event. One character had done a ton and written a ton on pain management. The problem, they fudged the data just enough to get statistically significant results. So all that research had to be repeated essentially.
AI is still GIGO. Except trying to use it to sift through research will be difficult sifting through a bunch of garbage input to find something practical to use in the everyday clinical setting.
I like the way Google sorts what it presents based on how many clicks from users on each item presented. Perhaps AI can learn what info users find most interesting to become more on target.
In oncology, AI generated treatment plans have higher cure rates.
{{ In oncology, the use of artificial intelligence (AI) represents a transformative advancement in cancer diagnosis, prognosis, and treatment. The use of AI in biomarker discovery improves precision medicine by uncovering biomarker signatures that are essential for early detection and treatment of diseases within vast and diverse datasets. }}