Stanford Health Policy’s Michelle Mello and Sherri Rose assert in this JAMA Forum article that the use of artificial intelligence (AI) would seem like a logical tool to help with the drudgery of evaluating insurance coverage and claims. But results so far have been sobering, leading to class-action lawsuits and congressional committees demanding answers.

“Utilization review by health insurers is the type of problem that seems, on the surface, to cry out for solutions using artificial intelligence (AI)," the two health policy professors write. "The staggering volume and complexity, inefficiency, and decision-making that requires sophisticated evaluation (yet feels like administrative drudgery for insurance plan staff) make insurance reviews an attractive target for using AI. The market has responded robustly; however, the results illustrate that seemingly perfect opportunities for using AI can become clear examples of how algorithms can go awry when humans do not provide the expected bulwark against error.

“Medicare Advantage plans have become emblematic of such problems. Investigative journalists brought to light health plans’ use of algorithms to curtail postacute care with scant human oversight. These reports stoked the ire of congressional committees already provoked by other evidence of insurers’ wrongful denials of prior authorization requests."

Read the Full JAMA Forum Commentary