The Real Reason Your Patient Engagement Program Isn't Working
Years ago I built a mental health app for people with bipolar disorder. We had a beautiful design, evidence-based content, a smart product team, the whole thing. Engagement rate: 15%.
Then we called them. With an AI. 85%.
That gap, the one between an app and a phone call, is the entire reason HANA exists. And honestly, it's also the reason most of the readmission programs I see today aren't working.
A new stepped-wedge trial published in AJMC in March 2026 studied 5,490 discharges across a system using a popular post-discharge digital engagement program. Patients got automated messages through a web or mobile app. Symptom check-ins. Education. Nurse messaging.
Result? No reduction in readmissions. And for the high-risk group, readmissions actually went up.
Read that again. The patients who needed help the most got worse.
I'm not surprised. And neither, I suspect, are the nurses who designed the program.
Why don't digital engagement apps reduce readmissions on their own?
Because the people who most need follow-up are the people least likely to open an app. That's the whole story.
The trial population was older, multimorbid, often dealing with cognitive load, low digital health literacy, or shaky connectivity. An app assumes the patient is a participant. A 70-year-old woman three days post-discharge on six new meds is not a participant. She's surviving.
We've been pretending engagement is a UX problem. It's not. It's a channel problem.
What does an 85% engagement rate actually look like?
It looks like a phone ringing on day 2 after discharge. The patient picks up. A voice she recognises (because she's talked to it before) says "Hi Mary, it's HANA from Dr. Patel's clinic, just checking on how you're doing since the surgery."
Mary talks. The AI listens. It asks the questions the nurse would ask. Pain level. Medication compliance. Any swelling. Any shortness of breath. If anything sounds clinically off, the call escalates to a human within minutes.
That's how HANA delivers 85% weekly patient engagement across more than a million patient interactions, with zero critical adverse events to date. It's not a UX miracle. It's just meeting people where they actually are. On the phone.
Why does voice outperform every other channel?
Voice is the only channel that doesn't ask the patient to learn anything new. No app to download. No password to remember. No portal to navigate. The phone rings, you answer.
According to MGMA's 2025 Patient Experience Survey, structured post-visit follow-up programs cut 30-day readmissions by up to 35% for high-risk patients. But the structure only works if patients actually engage. SMS gets 22% on a good day. Apps get 12-15%.
Voice gets 80% plus. Across every age band, every income level, every language.
It's the most boring channel in the world. That's the point.
What should clinic owners stop investing in?
Stop buying digital engagement layers that assume the patient will come to them. You'll get great engagement metrics among the patients who don't need much follow-up, and almost zero engagement from the ones you actually need to reach.
I see this constantly with clinic owners I talk to. They've spent months piloting a patient portal that 18% of patients open. They renew the contract because that 18% group rates them well. Meanwhile readmissions for the bottom quartile stay flat or rise.
Look, the math is brutal. If your readmission cost is $15,000 per case and your follow-up program reaches the wrong 80%, you're paying for a marketing tool, not a clinical one. The unit economics only work when the channel reaches the patients who matter.
How do you build a follow-up program high-risk patients can actually use?
Three things. First, default to voice for the first 72 hours after discharge. SMS and app are fine as secondary channels, but they're not the spine. Second, every call has to escalate to a real human the moment anything sounds clinically off. AI listens. Humans intervene. That handoff has to be sub-five minutes.
Third (and this is the one most platforms get wrong), the system has to remember. If Mary said her ankle was swollen on day 2, the call on day 5 starts with "How's the ankle, Mary?" Not "Hi Mary, can you tell me your name and date of birth." Continuity is what builds trust. It's also what catches deterioration.
HANA's clinical case studies show this pattern across cardiology, OB/GYN, primary care, and post-surgical recovery. Same engagement curve in every specialty.
Key takeaways
The AJMC trial is one of the cleanest demonstrations I've seen that digital engagement, by itself, isn't enough. It's not that the technology was bad. The platform did what it was designed to do. It just couldn't reach the patients who needed it most.
If you run a clinic or a value-based care program, the question isn't "do we have a digital engagement layer." It's "how do we reach the 80% who never open the app." For most populations the answer is the phone, automated intelligently, with a human standing by.
The patients didn't fail your program. The channel did.
FAQ
Does voice AI replace human nurses? No. It absorbs the routine work, scheduling calls, symptom check-ins, medication reminders, so nurses can focus on the patients who actually need human judgment. In every HANA deployment, nurses become more clinically impactful, not less.
Is voice AI HIPAA compliant? Yes, when designed for it. HANA is HIPAA aligned, self-hosted, and has no dependency on third-party LLM providers. PHI never leaves the customer's environment. Read more about how we built it.
What's the ROI on a voice AI follow-up program? Our clinic deployments average 31:1 ROI through reduced no-shows, recovered revenue, and prevented readmissions. The math gets even better in value-based care contracts where preventing a single readmission saves $15,000 plus.
If you're running a clinic or a health plan and your follow-up program is stuck below 30% engagement, I'm happy to walk through what's working in real deployments. No deck. Just a conversation.