The Chart Read Itself and Found a Raise
Even now, in hospitals across the country, a piece of software is reading your chart more carefully than your doctor did, and it has found something: you are sicker than anyone actually treated you for. The Blue Cross Blue Shield Association — 31 insurers, 100 million members, presumably some accountants with very tired eyes — released an analysis on September 24 showing that AI-assisted coding tools have pushed the share of inpatient stays billed as "medically complex" from 37 percent in early 2023 to 40 percent by the end of 2025. In a development that will surprise no one who has been paying attention, the patients did not get sicker. The paperwork did.
The Smoking Gun Wears a Hospital Gown
Sources within the Health Insurance Community point to one detail that does the arguing for them: at hospitals leaning hardest on this software, the rate of "acute posthemorrhagic anemia" diagnoses — a real condition, treated with a real transfusion — hit 13.7 percent of cases, versus 9.9 percent everywhere else. And yet the actual transfusion rate at those same AI-heavy hospitals was lower: 16.9 percent, against 19.3 percent elsewhere. Somehow the hospitals found more of the disease and administered less of the cure. One presumes this is what a significant jump in diagnostic accuracy looks like from the billing department's chair.
BCBSA senior VP Luke Chalker offered the sort of quote that lands with the candor of a man reading a confession into a tape recorder: "the reimbursement mechanisms that exist allow this." Not encourage. Not reward. Allow — as though the software simply noticed a door was unlocked and, being a piece of software, walked through it 55,000 times, at an average of just under $11,900 a walk.
Meanwhile, on the Other Side of the Ledger
And yet — as if this were not enough — the hospitals are not shy about where the money for The Algorithm goes. McLaren Health Care's CFO, Dave Mazurkiewicz, confirmed the hospital's coding tool, SmarterDx, adds roughly $1 million a month to its books. Somewhere in an insurance company's basement, one presumes, a rival algorithm is being fed the same charts to try to deny exactly what the first algorithm found. Two machines, one patient, and a fee schedule stretched between them like a net.
This Has Happened Before, and Congress Remembers
Medicare tried this particular rodeo in 2008, introducing severity-based billing tiers meant to reward hospitals for treating genuinely sicker patients — and by 2017 had clawed back $11 billion after discovering hospitals had, shockingly, learned to code for severity rather than deliver it. The AI version of this trick just runs faster and reads more charts before lunch.
Peterson Health Technology Institute's Caroline Pearson and former federal health IT official David Brailer both note that hospitals may, in fairness, simply be correcting decades of *under*-coding — that AI is finally counting what always went uncounted. Which is the institutional equivalent of insisting the raise was overdue the exact week the software started asking for one.
Sources: FierceHealthcare — Hospitals' use of AI coding tools cost BCBSA plans $942M more for similar care: analysis · Business Model Analyst — Blue Cross Says Hospital AI Cost It $942 Million. Blame the Staircase
