AIs (LLMs) are now being used in hospitals…
The American Hospital Association is pushing back on a Blue Cross Blue Shield Association report that linked hospitals’ growing use of artificial intelligence-enabled documentation tools to higher healthcare spending, arguing the analysis does not prove AI is driving improper billing.
The Blue Cross Blue Shield Association estimated that coding-related changes resulted in $942 million in additional spending for participating Blue plans between 2023 and 2025 as hospitals increasingly classified patients as having more complex conditions.
But the AHA said Monday that the analysis lacks critical context and does not establish that hospitals incorrectly coded patients or that AI tools were responsible for the changes.
“The central question is not whether hospitals are reporting more diagnoses,” the AHA said. “The question is whether the patient’s medical record supports those diagnoses.”
Full article, HERE from Fox News.
Anybody want to bet the LLMs are using the IF/THEN matrix to ‘add’ diagnoses to patients?
For those who may not be familiar with it, IF/THEN is a use case where you get the IF (say a specific diagnosis) and the THEN says if the patient has ‘this’, it must also have ‘that’, e.g. the second diagnosis…
This is basic programming, and I can easily see this being done by the LLMs since they are being used to ‘ensure’ the hospitals are making a profit…
Other than the Shrine Hospitals and St. Jude, I cannot think of a hospital these days that isn’t a profit center…
sigh…
AI has also reduced the sepsis death rate, the leading cause of death in hospital care, in some east coast hospitals by over 50%, as well as freeing up nurses for patient care instead of monitoring. So it’s not all bad news.
Why would it reduce the sepsis death rate?
It is a tool that makes some specific good things easier, and some specific bad things easier.
AI could certainly do a better job of accurate coding and billing than we currently see..if it is programmed and instructed to be accurate. The people who currently do the billing are NOT the best and brightest. They are average employees who don’t make enough to care about doing an excellent job. Add in that many coders/billers are pressured by administrators to pad bills whenever the opportunity presents itself and the system is rife with cheating. As with most endeavors AI will perform as well or as poorly as those who implement it desire.
The complexity of the current diagnostic and treatment code system is such that if a practice is seeing anything more than a tiny subset of the codes–rather than the large subset that a hospital, let alone a really big hospital system sees–it’s impossible to accurately code your services with a typical office administrative staff. So you code on the side of caution, because of the hazards of overcoding, so you may well be underreimbursed. Probably are, if cognitive services are a big part of what you do.
That’s part of what’s killing old school private practice, and turning doctors essentially into employees of large groups. The big groups can wrangle the billing better.
What LLMs are doing is making it possible for a smaller number of admin workers to wrangle the codes better, and, even with strict compliance, for the doctors to be better paid for what they do. For now.
Is the VA a for profit hospital system?
Just simple malice or incompetence (but I repeat myself) in coding can cause a months long nightmare fighting the bill. AI will be the modern equivelant of what we were taught in the MS DOS era, GIGO.
I believe that the Catholic Sisters of Providence hospitals are also nonprofit.
If the use of LLMs isn’t tied up in Medicare, Medicaid, and other government billing fraud, I’ll be surprised. Very, VERY surprised.
All- Thanks for the comments! JFM, thanks I was not aware of that.
Peter B- Maybe… That assumes that all doctors NOT in the major systems are undercoding. I don’t believe that…
As far as the VA, I have no idea, but my ‘guess’ is that it is ‘semi’ non-profit.
George Steinbrenner looked down from heaven sad said to Cashmen, and Boone, “WHAT HAVE YOU DONE TO MY TEAM!”
Curious that the AHA talks about billing for diagnosis by their AI tools. I always thought hospitals and doctors got paid for treatment. Sending a bill for what the AI found in records after a patient leaves the hospital, without any treatment for an AI “diagnosis” sure sounds like fraud to me.
My BiL had a couple medical clinics, each with about 25 employees. A handful of doctors, same of nurses, but the largest group was the paper pushers. These were the people that tried to match the diagnoses to the government and insurance databases, so they could get paid for their work.
No idea how AI would fit in to this business, as he had to sell them and retire early. To retain his hospital privileges, he was forced to get a Covid shot, and this ruined his health. Never sick a day in his life, now he is constantly sick, so not feasible to work with the public or his employees.