Insurers are using unregulated predictive algorithms to pinpoint the precise moment when they can plausibly cut off payment for the treatment of older patients
STAT : Tweets: @marshacollier and @maartenvsmeden Tweets: Marsha Collier / @marshacollier : How Medicare Advantage Plans Use #AI To Cut Off Care For Seniors Insurers are using unregulated predictive algorithms, under guise of scientific rigor, to pinpoint the precise moment when they can plausibly cut off payment for older patient's treatment https://www.statnews.com/... https://twitter.com/... Maarten van Smeden / @maartenvsmeden : “Behind the scenes, insurers are using unregulated predictive algorithms, under the guise of scientific rigor, to pinpoint the precise moment when they can plausibly cut off payment for an older patient's treatment.” Big and sad if true https://www.statnews.com/...
Context & Ripple Effects
The playbook STAT describes did not appear overnight: back in 2018, Arkansas automated health assessments were already cutting disabled patients off state-sponsored home care, and a 2020 investigation found service-targeting software infusing racial bias into who gets stepped-up care. What this report adds is the payer-side version — insurers running predictive models not to support clinicians but to time payment cutoffs under a veneer of scientific rigor.
The aftermath in the corpus shows the pressure point held: CMS issued a memo stating health insurers cannot use AI to determine or deny care on Medicare Advantage plans, while UnitedHealth's Optum is meanwhile building an AI-powered risk scoring system for Medicare patients that bypasses physician-submitted diagnosis codes — the same data pipeline this story exposes.
First-order effects
- Older patients on Medicare Advantage plans face treatment payments cut at the precise moment the insurer's model says continued coverage is no longer defensible, leaving families and treating physicians to fight the denial after the fact.
- Insurers gain a cost-control lever that looks actuarially neutral, because the algorithm supplies the 'scientific rigor' framing for what are effectively utilization decisions.
Second-order effects
- Rival Medicare Advantage insurers have every incentive to adopt similar denial-timing models, since each percentage point shaved from treatment payouts compounds against competitors still relying on manual review.
- Regulators are forced into reactive rulemaking — CMS's ban on AI-determined denials exists precisely because the industry moved first, putting the agency in the position of policing models it did not approve or audit.
Third-order effects
- If the pattern holds, the durable battleground becomes whether payment decisions require human accountability: expect a structural split between insurers whose algorithms set the default and regulators mandating clinician sign-off, with the burden of appeal falling on patients either way.
- The deeper shift is that health AI is migrating from clinical decision support toward financial gatekeeping — a trajectory running from Arkansas home-care cuts through hospital diagnostic tools to payer-side scoring systems like Optum's.
The trend: Predictive algorithms are moving upstream from advising doctors to controlling payment flows in American health insurance, with federal regulators legislating human oversight after each exposed deployment.