Publications

Artificial intelligence is an obvious fit within the insurance industry. Insurance is rooted in assessing the probability of certain outcomes and AI is designed to leverage increasingly larger stores of data to improve decision making.

A recent putative class action lawsuit against UnitedHealth Group alleges the Company used AI to wrongly deny care to large numbers of its Medicare patients. The linked article and others ascribe the problem to AI. However, a closer look at the complaint shows that the issue isn’t really AI but UnitedHealth’s allegedly problematic use of statistics.

AI is featured so prominently in the complaint that it appears 67 times over the course of 47 pages. The complaint alleges that UnitedHealth used “artificial intelligence (AI) in place of real medical professionals to wrongfully deny elderly patients care owed to them under Medicare Advantage plans by overriding their treating physicians’ determinations as to medically necessary care based on an AI model that Defendants know has a 90% error rate.” Specifically, Plaintiffs allege that the insurance group denied or shortened rehab stays for patients. Plaintiffs accuse UnitedHealth of making “rigid and unrealistic” decisions that are not individualized and tailored to the patient. 

Although the complaint provides very little detail on the underlying AI technology involved, it describes an algorithm that “compares a patient’s diagnosis, age, living situation, and physical function to similar patients in a database of six million patients it compiled over the years of working with providers to predict patients’ medical needs, estimated length of stay, and target discharge date.” This description sounds very similar to basic predictive modeling; looking at historical data to determine the key variables that influence a health outcome and using that analysis to predict an outcome in a given situation. Although Plaintiffs ascribe this mathematical decision making to AI, their claims actually take issue with predictive modeling that has long been a part of the health insurance industry. Indeed, the algorithm at issue in this case has been used by UnitedHealth since at least 2019. Whether it was used in an appropriate manner by UnitedHealth is an entirely separate question. Health care providers, patients and even regulators have taken issue with the methods used by UnitedHealth and other insurers to make decisions about whether, and to what extent, they will cover health care services.   

Whatever comes of the UnitedHealth case, it highlights a key takeaway for clients: Media coverage and literature on AI leans heavily towards doom. As a result, plaintiffs will continue to cast their lawsuits as AI-based—whether they are or not—to capitalize on public anxiety about this nascent technology. This will eventually change as AI technology and its uses become more established and understood—by judges and the public alike—but in the meantime, defendants will need to actively work to neutralize the perceived threat. This means that defendants should be prepared to explain technologies simply in their filings, breaking down component parts and functions. Defendants should also be able to identify the human mechanisms of control. More generally, businesses using big data, predictive analytics, and AI would be wise to begin a concerted effort at educating their clients and the public more generally about these technologies. Such efforts will help reduce the time it takes to get to a place where the invocation of AI does not immediately cast a negative light upon a common business practice.