The paper Cascades of Care After Incidental Findings in a US National Survey of Physicians is an important read for anyone who believes artificial intelligence is going to lower healthcare costs.

A “Cascade-of-Care” is a medical workup that expands exponentially after an incidental finding is discovered after a lab test or diagnostics test. This expansion—which occasionally unmasks a significant disease—instead can result in harm to the patient, the physician, or both.

Most physicians reported that cascades had caused their patients harm (86.7%), including psychological harm (68.4%), treatment burden (65.4%), financial burden (57.5%), dissatisfaction with care (27.6%), physical harm (15.6%), disrupted social relationships (8.7%), and death (0.2%). They reported that they personally experienced wasted time and effort (69.1%), frustration (52.5%), and anxiety (45.4%).

Roughly speaking, physicians will initiate a cascade-of-care (also known as as a mega-workup) for abnormalities they don’t think are significant for two reasons:

First, because there is a slim chance they are wrong. Every good physician I have worked with is never 100% sure of a diagnosis. They always leave some room for doubt.

Second, if they miss something they may get sued. This factor shouldn’t be mInimized. Look, even if a physician has the imaginary near-perfect judgment of Marcus Welby or Dr. Quinn, Medicine Woman, and they are right 99.9% of the time, that also means they are wrong 0.1% of the time, and if they are seeing 5000 patients a year, this means they are wrong about 5 patients a year. Multiply that by 20 years, and more likely than not, there will be a bad outcome during a physician’s career which will lead to a potentially devastating lawsuit. I think this is a major driver in cascades, and in fact, this article points out that about 1/2 of the physicians felt that one possible solution to care cascades would be malpractice reform.

Now, will AI-enabled EHRs help this?

It’s doubtful, and I contend it will make things much, much worse.

Look, the essence of the cascade problem is that with new technologies (more thorough lab tests, more fine-grained imaging, new diagnostic approaches) we are introducing more data and more complexity, and this avalanche of results requires the physician to practice ever more defensive medicine.

For example, when a 60-year-old man comes in with bronchitis and when he mentions the cough hurts his chest, and your friendly AI-enabled chart formally mentions and documents “also consider heart disease, angina” then I can guarantee that once that word “angina” is out there, on the chart, the physician will probably initiate— against his or her better judgment—a cardiac evaluation.

So, unless there is paradigm-shifting malpractice reform (which, by the way, will never happen) along with AI-enabled EHR technology, then we can expect either a dramatic increase in healthcare costs, driven by cascades, an accelerating increase in physician exodus, driven by stress, or both. Probably both.

Bottom line: These cascades of care exist, and more technology—especially EHR technology—is going to make things worse.