Blue Cross says hospital AI tools added $942 million in healthcare costs

The Blue Cross Blue Shield Association, known as BCBSA, says hospitals’ use of artificial intelligence tools when filing insurance claims added $942 million in unexpected healthcare spending over a two-year period. The figure comes from an internal analysis by the insurer trade group, reported first by the New York Times.

According to BCBSA, the analysis found a sharp increase in patients being documented as having complex medical conditions. Hospitals rely on medical coding, the system of billing labels that describe a patient’s diagnosis and treatment, to bill insurers for care they provide. The group says it found no matching rise in the actual care patients received, calling it “a clear disconnect between coding and treatment.”

Disputes between hospitals and insurers over what treatments should cost are nothing new. But the New York Times cited the BCBSA analysis as the latest evidence that AI tools, used by both hospitals to code claims and insurers to review them, are making those disputes more expensive rather than less, as each side adopts software meant to gain an edge over the other.

Luke Chalker, BCBSA’s senior vice president, rejected the idea that the standoff between hospitals and insurers is an even fight. “It’s not a war,” he said. “It’s a completely one-sided blood bath,” with insurers describing themselves as the side losing out financially.

AI startups see a mixed outcome

Dr. Shiv Rao, founder of the AI startup Abridge, which builds tools used in clinical documentation, acknowledged the risk that automated coding and automated claims review could spiral into open conflict between competing systems. He described a scenario of “bots fighting bots, agents fighting agents” as “a horrible dystopic future nobody wants to live in.” Rao said the same technology could instead ease tensions between hospitals and insurers and cut costs, though he did not lay out how that outcome would come about.

The source material does not detail BCBSA’s methodology for calculating the $942 million figure, nor does it include a response from hospital groups or from the AI vendors whose coding tools were examined. It is also unclear which specific two-year period the analysis covers, or whether the rise in complex-condition coding was concentrated among particular hospital systems or particular AI products.

What to watch next is whether hospital groups or other insurers publish their own competing figures in response, and whether BCBSA makes the underlying analysis public so the $942 million estimate can be checked independently.