How Many Patient Calls Is Your Clinic Missing? The Number Nobody Owns
There's a number sitting in most clinics that nobody owns. Not the no show rate, not days in AR, not the star rating. The missed call rate.
Nobody owns it because it isn't anybody's fault. The front desk is three people deep with a walk in, the phone rings eleven times, and the woman on the other end, four days out from a knee replacement, decides she'll try again tomorrow.
She doesn't try again tomorrow.
I spent years building things patients were supposed to come to. Apps, portals, dashboards, all beautifully designed, all sitting there waiting to be opened. One of them, a mental health app for people with bipolar disorder, got 15% weekly engagement and taught me the most expensive lesson of my career: the surface where care happens is not a design choice, it's the whole outcome. We threw it out and called patients instead. Engagement went to 85%. Same content, same people, different channel. That's the short version of why HANA exists.
So let's talk about the channel everyone already has and nobody is measuring.
How many patient calls does a typical clinic actually miss?
More than leadership thinks, and the data now exists. In a survey of 302 medical practice leaders cited in Forbes this summer, 59% of practices reported handling more than 300 inbound calls per business day, and more than one in three said they miss 11% or more of calls at peak times.
Take the midpoint of that. Three hundred calls, 11% missed, thirty three patients a day who wanted something from you and didn't get it. That's 165 a week. Roughly 8,000 a year in a single mid size specialty group.
Most of those are refills and scheduling. Fine. But some fraction is not.
Why is a missed call a clinical event, not an admin one?
Because the patient who most needs to reach you is frequently the one least willing to try twice. Post op patients underreport. Older patients don't want to be a bother. Anyone who's ever felt like a nuisance in a waiting room will sit on a symptom for another day rather than call back.
That's the mechanism behind the readmission, and it's invisible in your data because a call that never connected leaves no record. You can't chart silence.
Vanderbilt's discharge care centre is the cleanest demonstration I've seen of what closing that gap is worth. Structured outreach across more than 80,000 discharges dropped 30 day readmissions from 10.6% to 9.9%, about 197 fewer readmissions a year at roughly $15,000 a pop. They didn't cure anything. They just made sure somebody made contact.
Why does the phone still beat the patient portal in 2026?
Because it asks nothing of the patient. No download, no password reset, no working out which of the four apps their health system made them install is the one with the discharge instructions in it.
I learned this crossing Australia with a circus, of all places. (feels like another lifetime, honestly) The acts pulling the biggest crowds weren't the technically hardest. Aerial silk is objectively more difficult than swinging fire on chains. Fire on chains got three times the audience in every town we played, because a bloke walking past a car park at nine at night could understand it in one second flat.
Healthcare keeps building aerial silk. Patients keep turning up for fire chains. A ringing phone is a fire chain.
Should clinics automate inbound or outbound first?
Outbound. Almost always outbound, and almost everyone starts with inbound because that's where the pain is loudest.
And that's backwards. Inbound is the hardest surface in the building: unpredictable, emotionally loaded, legally exposed, and the first conversation a distressed patient has with you. Outbound is bounded. Post op day two check ins, care management enrolment, no show recovery. Fixed clinical question set, known patient, known context, no ambiguity about what the call is for.
Start there and your staff get two weeks of transcripts to argue with before anything touches a patient in crisis. Then move to inbound with a team that already trusts the system. The deployments that scaled fastest all ran that order. The ones that didn't took twice as long to get anywhere useful.
What does a clinic actually get back in the first quarter?
Three things worth measuring, in this order. Completion rate, meaning the share of calls where the patient answered the clinical questions rather than just picking up. Escalation precision, meaning of the flags raised, how many were genuinely clinical. And staff hours returned, which is the number your CFO cares about and the one you should report last, because it's a consequence of the first two rather than a goal.
We've run more than a million patient interactions with zero critical adverse events across five countries and three languages. The clinical use cases with the fastest payback are the ones where silence is the dangerous signal: post discharge, post op, chronic care management enrolment. All the routing and escalation logic is documented openly in our docs and runs self hosted inside the provider's own infrastructure, open source, no OpenAI dependency, because a black box making clinical judgement calls on your patients is not a thing you should accept. When completion holds, the economics land around 31 to 1.
Key Takeaways
Every clinic has a missed call rate and almost none of them measure it, which is remarkable given that a third of practices are missing more than one call in ten at peak. Those calls aren't evenly distributed in clinical risk. The patient least likely to call back twice is frequently the one you most needed to hear from, and a call that never connected leaves no trace in the chart. The phone stays the highest completion channel in healthcare because it demands nothing of the patient. Automate outbound before inbound, because it's the cheaper place to make mistakes. And measure completion before you measure savings, because savings follow contact, not the other way around.
FAQ
Can AI calls handle clinical questions safely? Within a bounded scope, yes. A well designed outbound agent runs a fixed clinical question set and escalates to a human the moment a response falls outside expected parameters. It doesn't diagnose and it shouldn't try. The safety comes from the escalation logic, not the model.
Will older patients engage with an AI phone call? Often better than younger ones. There's no digital literacy barrier on a telephone, no app, no login. The 78 year old who'll never open a portal will absolutely answer her phone and tell you her incision feels warm.
How long does it take to go live? A bounded outbound programme like post op follow up runs live in two to four weeks once EHR access is sorted. Inbound takes longer, mostly because the policy decisions are harder than the engineering.
If you want to know your own missed call number, pull it. Then, if it's ugly, book a slot on my calendar and I'll walk you through what we'd do with it, including the deployments where it didn't go smoothly.