Bots Fighting Bots, Patient in the Middle
Updated
As AI continues to invade every facet of daily life, the profit-driven health care industry in America has reached a strange place. Somewhere in the hidden machinery, choices are being made about everything that might be wrong with a patient and what those ailments should cost, and more and more, absolutely no doctor is involved in either half of those decisions. Instead, those decisions are being handled mechanically by two sets of algorithms. One set of rules works for the hospital and wants the patient’s bill to be higher. And the other works for the insurer and wants it lower. The patient, as usual, works for neither and is merely a pawn in a pricing game.
How do we know about the hospital software manipulation? Because Blue Cross Blue Shield did its own analysis. Released on September 24, the insurer’s report noted that, over a two-year span, coding tools used by hospitals added roughly $942 million to insurance plan costs by doing things like scanning through lab results and doctors’ notes, spotting secondary conditions like a slightly low sodium reading or anemia. The software then used that data to reclassify ordinary hospital stays as “medically complex.” As expected, a complex visit pays more: an average of $11,900 in additional charges across more than 55,000 cases.
Shockingly, roughly 70% of that $942 million, which amounts to more than $650 million in this case, came from diagnoses that changed the billing and nothing else. Not one additional thing happened to the patient; they were just unknowingly charged more. As noted by Blue Cross, if a patient was tagged with anemia, somewhere you would expect to see more transfusions. But there were none. It is clear that the invoice total increased, but did the condition live on, and was it treated? And the bigger question apparently ignored by both sides: was there a holistic and natural way to address this condition that was not shared?
But Blue Cross is not an innocent bystander in this scenario, and they are not a victim, either. As an insurer, their recent analysis, pinning rising health care costs on hospitals, is exactly the kind of report that any insurer wants on the record. It easily turns the company paying the claims into the injured party, and it gives Blue Cross a reason to deny more of those claims and charge more for the coverage. Even so, the findings can still be right. And the firm announcing them, Blue Cross, also makes money if they are believed.
But there is a hole in the evidence because Blue Cross used billing records for their analysis, not actual patient charts. Their report can successfully show that the codes changed, but it cannot show that the patients also did not change. Hospitals point at that hole, insisting that, despite apparently no change in the patient’s treatment plan, their software is catching real health conditions that busy staff missed. They argue that the expanded record is simply a more accurate picture of patients who were already complicated. OK. A quick glance at our unhealthy nation suggests that some of that may be true.
Besides pointing at the hole left by the insurers’ report, the hospitals’ other excuse is the one that makes this an interesting game. They argue that they need these AI billing tools in order to keep pace with the tools used by insurers. In other words, while hospital software hunts for every billable condition, insurer software works in the opposite direction, rapidly scrutinizing, downcoding, and denying claims. From 2022 to 2023, the American Hospital Association found that commercial claim denial rose by about 20 percent in a single year, which was partly driven by the software. Meanwhile, with both sides setting the patient aside, billing paperwork flies back and forth between the two faster than any person can read it. Summing up the game, Shiv Rao, founder of the medical-scribe company Abridge, shared, “bots fighting bots, agents fighting agents, a horrible dystopic future nobody wants to live in.”
And that is the real disastrous story here, hiding under the billing codes. As AI takes over, medicine in this country is quickly and quietly becoming a standoff between two server rooms. One profits by making every patient’s illness appear worse, and the other profits by making every patient claim appear cheaper. And, without ever laying an eye on them, they are both trading documents at machine speed over those patients. Essentially, health care has become an arms race. Not the nuclear kind, the AI kind, generating a great deal of motion, a great deal of expense, and no peace at either end.
Indeed, when hospital software pulls in an extra $11,900 out of a routine stay, that money has to come from somewhere. This isn’t rocket science—it comes back on the patient’s bill. Meaning a higher premium, a larger deductible, and often a balance due after the insurer’s competing software marks the care “not medically necessary.” To win the race, the hospital installed a bot army to fatten up the claim. On the flip side, the insurer has a bot army ready to go cut it. The real loser here is the patient when the invoice arrives in their mailbox.
Perhaps the strangest part of this entire debacle is that it has entirely stopped being about actual patient health care. None of the code has been written to make anyone well. The program flagging low sodium or anemia has no interest in taking the next step with the patient to fully evaluate and correct whatever issue exists. And the program denying the claim has no interest in the patient behind it. Again, both are fine-tuned to the single thing they are designed for—money. And the patient, no matter how sick or well, has been stuffed into a document that two conglomerates fight over. Indeed, a system in which one robotic algorithm invents complexity and another refuses to pay for it has neither gotten smarter nor made patients healthier. Instead, it has gotten faster, more expensive, and has completely forgotten about the patient.