Mike's Blog

A recent Stat article highlighted early results from the CMS’ WISeR (Wasteful and Inappropriate Service Reduction) Model, an AI-enabled prior authorization pilot now underway in six states. The program is being piloted in Arizona, New Jersey, Oklahoma, Ohio, Texas, and Washington, which includes ten CTSAs. The goals are to reduce fraud, waste, and unnecessary care, by targeting 13 medical services, several of which are in the category of orthopedic pain management. So far, surveys suggest that approval for these services requires two to four times longer than in the past. Senator Maria Cantwell (D-Washington) released a report providing details on the program and its impact, collected by the Washington State Hospital Association. Cantwell also directly addressed her concerns with Health and Human Services Secretary Robert F. Kennedy Jr. during a recent Senate Finance hearing.

While the academic community is focused on the application of AI into overall health care, especially clinical decision support, administrative applications are clearly in the mix. A report released last month by the Peterson Health Technology Institute paints a rather disturbing picture. Their focus was on AI and its role in reducing administrative costs. They estimate approximately $350B of administrative waste annually with $266B due to ‘administrative complexity’ and the rest to fraud and abuse. The goal of injecting AI into administrative activities in healthcare is efficiency, but these initial results look a lot like amplification of complexity, the exact opposite of the intention.

From their executive summary:

• “…rapid AI deployment by both providers and health plans to support prior authorization and medical billing transactions risks increasing levels of system activity without reducing costs” 
• “…AI automation could increase the volume of prior authorization back-and-forth, rather than making the process more efficient”
• “…AI-assisted coding tools could accelerate coding intensity and charge capture, which—even if accurate—would have an inflationary impact on healthcare costs”

None of this is unexpected, but patients and their health care providers will be stuck between health care systems and insurers while their respective CFOs battle each other in an ever increasingly complex AI arms race. What should also be expected is lots of collateral damage, as well as friendly fire. Buckle up!

On a more pleasant note, I attended the ACTS conference in Milwaukee. Overall, it was a great meeting. In addition to a fantastic program, the numerous hallway conversations were most welcome. Maybe I’m just old school, but 3D 1:1 interactions always seem much more productive and satisfying relative to Zoom or Teams.

One interesting observation was the coffee station. Usually, there is a colorful sampling of available ‘sweeteners’ (white for sucrose, pink for saccharin, blue for aspartame, yellow for sucralose, green for erythritol, etc.). After sucrose, only sucralose and erythritol were offered by the Baird Center. There have been reports of dysbiosis with sucralose usage in both Microorganisms and Inflammatory bowel diseases, but the emerging science on erythritol presented by the American Heart Association and Nutrients is a bit more worrisome. The association of erythritol blood levels with cardiovascular events is striking (2X for top 25% compared to bottom 25%). NIH has even commented on the risk. In addition, further work has provided a mechanism with erythritol enhancing platelet reactivity.

It’s easy to see how erythritol gained its status as a non-nutritive sweetener. It’s about 65% as sweet as sugar. Erythritol is a four-carbon sugar alcohol that is absorbed into the bloodstream and then excreted through the kidneys unmodified; there is no insulin response and little if any caloric value is derived. Sounds perfect; unfortunately, its circulation in the bloodstream provides ample opportunity for platelet interaction that is the likely basis for the increased cardiovascular risk.

Now, it’s simple to avoid erythritol in the diet (avoid the green sweetener), but the history with another dietary health hazard, trans fatty acids (TFAs), is instructive and raises some questions as to how to proceed with research investigating possible adverse effects of non-pharmaceutical interventions. There’s a nice review of the TFAs detailing early suggestions in The Lancet in the 1950’s of adverse effects, through the 70’s and 80’s with evaluations and reviews declaring TFAs safe, including the Institute of Medicine and the US Surgeon General, into the 90’s when a slew of studies correlated TFA consumption to increased risk of coronary heart disease presented in American Journal of Epidemiology, The Lancet, and The New England Journal of Medicine. By 2006, The New England Journal of Medicine published that, one in five myocardial infarctions and 50K death per year were attributed to TFAs which prompted the FDA to require food labelling for TFAs. The FDA ban on TFAs did not go into effect until 2018. Nearly half a century elapsed between early signals and decisive policy, including 25 years of lobbying efforts.

The whole fandango with TFA begs the question of how we accumulate robust evidence to support adverse effects such that we can effect change. We have comprehensive and rigorous processes for determining interventions that are clinically beneficial, but for things that we do or ingest which may be counterproductive for our health, we struggle. Clearly, no one is going to conduct a placebo controlled randomized trial of erythritol (“we’re giving you this to see if it is bad for you”), so how can we efficiently generate high quality data (without resorting to RCTs) to drive policies that could lead to eliminating detrimental ingredients in our food supply? Seems like a translational science challenge that CTSAs can address. Functionally, this comes down to a category I refer to as “how do we know what we know?” In the case of medical evidence, RCTs reign supreme. We need additional tools of comparable quality.

What I’m reading: Devolution by Max Brooks. A group of technologically sophisticated city dwellers establish a ‘back to nature’ community in the Pacific northwest that is tremendously tech dependent (fiber optic internet connections; drone delivery of groceries, etc.). A volcanic eruption (Mt Rainer) isolates the community which lacks basic survival skills; then the trouble begins. With current global conditions revealing the fragility of current technological supply chains (have helium concerns impacted MRI utilization yet?), the parallels are obvious. Optimizing a system for efficiency (typically with cost as the measure) does not guarantee performance of that system under stress.

The whole country rests on a system that sacrifices resilience for comfort.
- Max Brooks, Devolution