Why Did the Largest Remote Monitoring Trial Ever Run Fail to Cut Readmissions?

I built a mental health app once. For people with bipolar disorder. Mood tracking, sleep logging, medication reminders, the whole thing, and it was beautiful, and clinicians loved the dashboard, and we shipped it with real pride.

Fifteen percent engagement.

Fifteen. Out of a hundred people who needed it, eighty five never opened it twice. I sat with that number for a long time before I understood what it was telling me, which is that I'd built a very good thing for the six people who were already going to be fine.

So when I read the ACCOMPLISH results in June, I didn't feel smug. I felt seen.

What did the ACCOMPLISH trial actually find?

It found nothing. Which is the point. The ACCOMPLISH trial, published in JAMA Network Open, randomised 1,286 adults discharged after sepsis or a lower respiratory tract infection across nineteen hospitals, tested four different remote monitoring designs against UPMC's usual care, and found that none of them increased days alive at home at ninety days.

Readmission rates: 37.8% for usual care. Between 36.3% and 44.2% for the monitoring arms.

Nothing moved.

Why is that result more interesting than a positive one?

Because of the subgroup. In patients sixty five and older, remote monitoring was associated with fewer days at home. Not neutral. Worse. UPMC's own writeup calls it paradoxical and says it warrants immediate clinical attention, which is about as loud as a health system ever gets in a press release.

Look at the enrollment numbers and you can see why. Of 887 patients assigned to monitoring, 529 actually enrolled. Fifty nine percent. Then they answered 56% of the questionnaires they were sent.

So the intervention was, at best, half delivered.

So is remote monitoring a bad idea?

No. The delivery mechanism is the bad idea. CMS reimburses remote monitoring, so health systems bought remote monitoring, and what they bought was an app, a portal, a scale that talks to a phone. Every one of those things asks the patient to come to the technology.

Compare it to a heart failure outreach programme run by Cardiac Solutions, where automated messages went out daily and no app download was required. Overall 30-day readmissions landed at 15.5% against a 24% national baseline. But split the cohort by engagement and the real finding appears. Highly engaged patients: 7.7%. Low engagement: 21.6%.

Same programme. Same content. Nearly triple the readmission rate.

Engagement isn't a nice-to-have bolted onto the intervention. Engagement is the intervention.

What does getting this wrong actually cost?

A readmission runs $15,000 to $20,000. That's the number the Cardiac Solutions team put on it, and it's a number your CFO already knows by heart.

Now do the arithmetic on ACCOMPLISH. 887 patients assigned to monitoring. 358 of them never enrolled at all. Every one of those was budgeted for, staffed for, licensed for, and reported on, and not one of them received the intervention.

That's not a clinical failure. That's a procurement failure wearing a clinical costume. We publish what voice follow-up costs per patient precisely because that comparison should be easy for you to run yourself, and the clinics doing it see roughly 31:1 return, which sounds ridiculous until you remember the alternative is a $17,000 readmission.

What actually reaches the patients who don't engage?

A phone call. Boring, I know.

We learned it the hard way. After the 15% app, we started calling patients with a voice agent instead, and engagement went to 85%, and it stayed there across more than a million patient interactions with zero critical adverse events. The technology got simpler and the outcome got better, which is the opposite of what I expected and roughly what the Australian circus taught me years ago. The performers pulling the biggest crowds weren't doing aerial silk. They were spinning fire chains. Accessible beats impressive. Every single time.

An eighty year old recovering from sepsis will not learn your portal. She will answer her phone.

What should a health system actually do with this?

Stop measuring deployment and start measuring contact. Not "how many patients were enrolled in RPM," which is a procurement metric dressed up as a clinical one, but "how many patients did we genuinely talk to in the seventy two hours after discharge, and what did they tell us."

That's the whole reframe.

The virtual nursing data from nine hospitals published in npj Digital Medicine points the same direction. Discharges supported by a virtual nurse had a 3.7% thirty-day ED readmission rate versus 13.3% for traditional discharge. A human voice, delivered remotely, at the exact moment of transition. Not a sensor.

Voice scales in a way nursing headcount can't, which is the only reason any of this is a technology problem at all. The unglamorous version is in the deployments we've documented.

Key Takeaways

ACCOMPLISH is the strongest evidence yet that remote monitoring, as currently designed, doesn't reduce readmissions after serious infection, and may actively harm older patients. The failure isn't in the concept of post-discharge contact, which works. It's in the delivery layer. Programmes that require the patient to open an app, complete a questionnaire, or navigate a portal systematically lose the patients who need them most, then report the survivors as the outcome. If your post-discharge programme runs 59% enrollment and 56% response, you don't have a monitoring programme. You have a self-selection filter with a dashboard attached.

The fix is to meet patients where they already are, which is on the phone, answering calls, the way they have done their entire lives.

FAQ

Does remote patient monitoring reduce hospital readmissions? For sepsis and lower respiratory tract infection, the ACCOMPLISH trial found it doesn't. None of four RPM designs beat usual care, and older patients had fewer days at home. Evidence is stronger in narrower conditions like heart failure, and even there the effect tracks engagement rather than monitoring itself.

Why do older patients do worse with remote monitoring? The trial authors flagged it as needing further investigation. The likely mechanism is a combination of low enrollment, low questionnaire completion, and alert-driven readmissions in a population where a red flag triggers a reflex trip to hospital rather than a targeted intervention.

What's the difference between monitoring and engagement? Monitoring collects data from the patients who participate. Engagement is the share of your population you actually reach. A programme can have excellent monitoring and terrible engagement, which is exactly what most post-discharge programmes look like once you check the denominator. HANA runs at 85% weekly engagement because contact is initiated by us, not by the patient.

If you're running a post-discharge programme and you don't know your true contact rate, let's spend twenty minutes on it.