Insurance AI Is Going To Court

United Health is facing a potential class action lawsuit alleging their usage of an artificial intelligence program that denies claims on faulty grounds has resulted in multiple deaths, including those of the plaintiffs whose estates are leading the suit.

‍

The core of this case involves nH Predict, which is an AI platform that UnitedHealth developed through a subsidiary and has been using since the fall of 2019, which, like many if not most artificial intelligences, is somewhat of a black box - meaning that its internal operations are proprietary and therefore the exact process of decision-making that the AI platform is conducting in any given instance is not publicly available information if it is traceable at all.

‍

The primary complaint alleged by the plaintiffs is that the nH Predict system estimates an impractically conservative amount of post-acute care that any given patient will require, including physical therapy or nursing home care following a traumatic illness and/or injury.

‍

While the post-acute care estimates produced by the AI are not determinative as to whether a claim is processed or denied, former employees have confirmed that case managers are expected to keep their recommended coverage within a prescribed window of the nH Predict range in making recommendations to staff physician medical reviewers who then make the ultimate coverage decision, overriding the prescribing physician when applicable.

‍

It’s also worth noting that according to one outside investigation into the matter, case manager recommendations are becoming even more tightly linked to nH Predict estimates. In 2022, case managers were expected to make recommendations on the length of covered nursing home stays within 3% of nH Predict estimates, whereas a 1% plus or minus window is expected of case managers as of 2023, for example. 

‍

The complaint also alleges that there are major flaws in the nH Predict system that are leading to erroneous claims denials. Plaintiffs argue that the system doesn’t take into account various factors that should be included among decision-making criteria, such as patient co-morbidities and extraneous factors like whether or not a patient caught a virus or infection while receiving care at a medical facility. 

‍

Further, when complaint denials have been challenged- whether through internal appeals or via administrative courts - around 90% have been reversed, which plaintiffs believe shows a clear pattern and that UnitedHealth is therefore knowingly benefiting from the improper denial of claims that should be covered by continuing to use the nH Predict estimates.

‍

Interestingly, even when claims denials are reversed, patients are often awarded only a relatively small amount of additional car. When that award expires, patients must refile their claims, which are often then denied on the same grounds as the original claim that was reversed, forcing patients to restart the cycle with the best case scenario outcome being an award of additional care that will quickly prove insufficient, as well. 

‍

Given these dynamics the sought-after class action lawsuit may be one of the more effective means to better align the incentives of the various stakeholders involved in a way that results in fewer claims that should be covered being denied, whether that be through better software or decreased reliance on it. 

‍

In any case, the implications involving how AI will be used in the insurance space going forward are significant, and we’ll be keeping an eye on developments. 

‍

You can read more about this topic and the pending lawsuit here.