Remote Monitoring Just Failed a Randomized Trial. The Monitoring Wasn't the Problem.

Years ago I sat in front of a Stripe dashboard and watched a DTC company I'd built do a million dollars in a single day. I should have felt something like triumph. What I actually felt, somewhere under the adrenaline, was a small cold voice saying: the product isn't good.

It wasn't. We were spending our way past it. Scale is a fantastic anaesthetic. It numbs the thing that's wrong until the thing that's wrong is structural.

I thought about that dashboard when I read the ACCOMPLISH trial results this summer, because health systems are about to do the same thing with remote monitoring, at enormous cost, and the evidence just showed up early.

What did the sepsis remote monitoring trial actually find?

Nothing. That's the finding.

A randomized clinical trial across 19 hospitals, 1,286 adults recovering at home after sepsis or serious lower respiratory infection, four different remote monitoring configurations tested against usual care. Published in JAMA Network Open in June 2026. Remote monitoring did not increase days spent alive at home. Readmission rates were statistically indistinguishable across every arm.

And for patients 65 and older, the monitoring arms did worse. Fewer days at home than usual care, with an inferiority probability of 99.6% in the standard-response group.

Read that again if you're about to sign an RPM contract. The CMS-reimbursed intervention made things worse for the exact population CMS reimburses it for.

Why would monitoring ever make outcomes worse?

Because collecting a signal and acting on a signal are different products, and only one of them is hard.

The trial sent patients questions twice weekly. Responses generated alerts. Alerts were categorised medium or high. A team responded. On paper that's a closed loop. In practice, only sixty percent of assigned patients ever enrolled, questions arrived on a schedule set by the protocol rather than by the patient's body, and a "response" could mean a callback landing days after the moment that mattered.

You can build every box on that diagram and still have nothing that touches a patient at the hour their ankles swell.

What's the difference between a signal and a loop?

A loop closes. That's the whole definition and almost nobody meets it.

Most remote monitoring programs are dashboards with a staffing plan stapled on. Data flows in, a nurse reviews a queue, someone eventually dials a number the patient doesn't answer because it shows up as an unknown caller. The loop opens and then just sits there, open, generating billable telemetry.

Compare it to what happened in the heart failure program Cardiac Solutions ran, where patients who engaged with more than half their outreach readmitted at 7.7% against 21.6% for those who didn't. Same clinical protocol on both sides of that gap. The difference was contact that actually landed.

Monitoring is not the intervention. Contact is.

Does timing matter more than intensity?

More than almost anything, and the trial design is instructive here precisely because it varied the wrong dial. It tested low-intensity versus high-intensity questionnaires. It tested standard versus enhanced response teams. It did not meaningfully test speed.

Meanwhile a Kaiser Permanente study of nearly 12,000 heart failure patients found contact inside 7 days of discharge was associated with roughly 19% lower odds of readmission, and contact after day 7 showed no significant benefit at all. Not less benefit. None.

So you can double your question set, hire a nurse practitioner-led response team, buy the wearables, and still miss, because the thing that mattered was reaching the patient on Tuesday instead of the following Monday. My daughter is ten and she's better at this than most health systems. When something's wrong she tells me that day. She doesn't wait for the weekly check-in.

What should a health system build instead?

Build for reach first and sophistication second. That ordering is uncomfortable for clinical leadership and it's still correct.

That means outreach triggered automatically at discharge, not assembled from a list on Monday. It means contact on a channel people actually answer, in the language they actually speak, which for a huge share of patients means a voice conversation rather than a portal notification. It means escalation that routes a red flag to a human clinician inside minutes. And it means every contact writing back to the record so the loop has a visible end.

HANA runs this across 5 countries and 3 languages, at 85% weekly engagement against a 15 to 20% baseline, over a million interactions, zero critical adverse events. It's open-source and self-hosted with no OpenAI dependency, which matters more to hospital governance committees than any feature we ship. The deployment results and the research are both public. So is the technical documentation, if your integration team wants to poke at it before anyone talks to sales.

Key Takeaways

A rigorous randomized trial found remote monitoring did not improve days at home after sepsis, and actively reduced them for patients over 65. That isn't an argument against technology, it's an argument against buying telemetry and calling it care. The programs that work close the loop: automatic triggering, fast contact on a channel patients answer, human escalation on red flags, writeback to the record. The programs that don't work generate beautiful dashboards while the seven-day window quietly closes. Scale hides problems. It hid mine for about eighteen months and a lot of money.

FAQ

Does this mean remote patient monitoring doesn't work?

It means monitoring alone doesn't. The trial tested question intensity and response team composition, not contact speed or reach. Programs that pair monitoring with fast, high-reach outreach show consistently better results than either component on its own.

Why did older patients do worse in the trial?

The authors don't claim a mechanism, and neither will I. One plausible reading is that alert-driven escalation pulled some patients into care settings they'd otherwise have avoided. It's a strong argument for tailoring monitoring to the patient rather than the billing code.

What's the highest-yield change a health system can make this quarter?

Shorten time-to-first-contact after discharge. The seven-day window is well evidenced and most systems miss it for structural reasons, not clinical ones. Fix reach before you buy anything else.

If you want to pressure-test your own post-discharge reach numbers against what we see in production, grab a slot on my calendar.