What is the real test for Cadence's AI wager?

An AI agent can read a blood-pressure trend. That part is not in dispute. What is in dispute is whether automating the human response preserves the part that works, or only the part that bills. Clinical automation carries a recognized hazard in the medical literature: the tendency to defer to the machine and miss the signal it was meant to catch, which is precisely the judgment the human layer was there to supply. Remote monitoring helps the right patients when a clinician acts on it. It bills i

Industry

Vol. 1·Sunday, June 28, 2026

Cadence Raised $1.2 Billion to Automate Chronic Care. The Evidence Points at the Humans.

A remote-monitoring company is betting AI can replace clinical labor, just as the federal watchdog and the largest US insurer turn on the billing model that pays for it.

Noah Ogbi10 min read

Tips, corrections, or questions? support@omniscient.media

TopicsAgents AI Policy Bio & Science

What does Cadence actually sell?

Cadence's product is remote physiologic monitoring for older adults with chronic illness. Patients receive connected devices, such as blood-pressure cuffs, that stream readings to Cadence's clinical team, which adjusts medications and steps in when vitals drift, all delivered under the partner health system's own brand.^[3] It covers hypertension, heart failure, and diabetes, and works with more than twenty health systems; the new round added Duke Health and Texas Health Resources.^[2] The company says "supervised AI agents" already monitor patient vitals daily, and reports that its annual recurring revenue tripled in 2025, figures it has not broken out in dollars.^[2]

The reason to automate is economic. Standard remote-monitoring reimbursement pays a provider on the order of $100 per patient per month, and Cadence has also joined a new Medicare experiment that pays more but only if it hits outcome targets, a model analysts noted would force participants to lean on AI to cut expensive human labor.^[4] Automating the clinicians is not a side project. It is how the unit economics are meant to work.

Where does the benefit actually come from?

Here is the first problem. The evidence that remote monitoring helps patients points repeatedly at the humans, not the hardware. A large individual patient data meta-analysis of blood-pressure self-monitoring trials found that monitoring alone produced little change - a mere 1.0 mmHg reduction in systolic blood pressure, statistically indistinguishable from zero; the meaningful reductions, up to 6.1 mmHg, came only when monitoring was paired with active support such as counseling and medication adjustment.^[5] Two of the largest randomized trials of remote monitoring for heart-failure patients, Tele-HF and BEAT-HF, found no significant reduction in 180-day hospital readmissions or deaths.^[6]^[7] The pattern across the literature is consistent: monitoring works when a clinician acts on the reading for a patient who is genuinely uncontrolled. The data is a trigger; the human response is the intervention.

If that is right, then automating the human layer is not automating the overhead. It is automating the active ingredient. Cadence would dispute the framing, and it has more evidence than most of its peers: it points to its own peer-reviewed studies reporting higher rates of blood-pressure control and fewer hospital admissions among its patients.^[2] But those studies are observational, comparing enrolled patients with eligible ones who did not enroll, a design that rewards selecting motivated patients as much as treating them. The honest reading is not that monitoring never helps. It is that what helps is the part Cadence wants to make cheaper.

Why are regulators and insurers backing away from the billing model?

The second problem is who pays, and why they are starting to stop. Remote monitoring is reimbursed through a set of monthly billing codes: enroll a patient, collect readings on enough days, bill a recurring fee, and bill again the next month for as long as the patient stays on the service.^[4] Revenue scales with the number of patients enrolled and the months they are retained, not with whether anyone gets better. That structure is the whole of the critique.

The federal watchdog has noticed. In September 2024, the Department of Health and Human Services' Office of Inspector General reported that remote-monitoring use in Medicare had grown dramatically and that about 43 percent of enrollees billed for it did not receive all three components of the service, and it pressed Medicare to add safeguards.^[8] A 2025 follow-up found Medicare payments for remote monitoring exceeded $500 million and developed billing measures to flag practices warranting scrutiny.^[12] As Ateev Mehrotra, a Harvard health-policy researcher, put it, "the current billing system encourages a clinician to do it for as many patients as possible," including patients who are already doing well, for whom "adding this on top of that is not going to be beneficial."^[9]

Then the largest private insurer moved. UnitedHealthcare announced plans to stop paying for remote monitoring of hypertension, diabetes, chronic obstructive pulmonary disease, depression, and a range of other conditions, keeping it only for heart failure and hypertensive disorders of pregnancy, on the stated ground that for the rest it is "not reasonable and necessary due to insufficient evidence of efficacy." The policy, originally scheduled for January 2026, was subsequently delayed following industry pushback, though UnitedHealthcare stated it still intends to implement the restrictions in 2026.^[10] A 2025 study in Health Affairs had already found that adopting remote monitoring raised a practice's Medicare revenue by roughly a fifth, a revenue effect independent of any change in outcomes.^[11] This is the backdrop against which STAT reported that Cadence's monthly-billing model "has come under scrutiny from the federal health department's watchdog and from insurers, including UnitedHealthcare," with critics arguing the framework is "ripe for abuse and may support low-quality care."^[1]