How to reduce did-not-attend rates with AI

A did-not-attend, or DNA, is one of the most expensive quiet failures in a hospital. A slot was funded, a clinician was rostered, a room was booked, and nobody came. The appointment cannot be reclaimed, the patient is still on the list, and the cost lands twice: once for the wasted capacity, and again when that patient has to be rescheduled into a queue that is now a little longer for everyone behind them.

The instinct in most services is to send more reminders. More texts, an extra email, maybe an automated call the night before. It feels like action, and it moves the number a little. Then it stops moving. Understanding why it stops is the whole point of this piece.

A DNA is an operations problem, not a patient failing

It is tempting to frame missed appointments as patients being careless or entitled, and some services quietly do. That framing is both wrong and useless, because it points you at blame instead of at a fix. A patient who does not attend has almost always failed to complete a process that the service designed. If the process is hard to complete, more people fail it.

Treat the DNA rate as a metric your operation owns, in the same way you own theatre utilisation or clinic overbooking. Once it is an operations number rather than a patient character flaw, the questions change. Not "why won't people turn up," but "where in our booking journey do people fall out, and what would make it easier to stay in."

You do not have a compliance problem. You have a slot your process could not defend.

The real causes of a missed slot

When you actually ask non-attenders, the reasons cluster into a short list, and almost none of them are indifference:

• They forgot, or the date drifted. An appointment booked eight weeks ago against a letter is easy to lose. Life reorganised itself around it and nobody re-confirmed.
• They could no longer make it and had no easy way to say so. A phone line that rings out, an inbox nobody watches, a rebooking process that means starting again. The path of least resistance becomes silence.
• The appointment was no longer needed or wanted. Symptoms resolved, they were seen elsewhere, or they decided against the pathway. The service was never told.
• Access and anxiety. Transport, work, childcare, cost, or straightforward fear of what the appointment means. These are real and often unspoken.
• The contact details are stale. The reminder went to a number that changed a year ago and nobody ever heard it.

Notice that several of these are not really DNAs at all. A patient who no longer needs the slot, or who would happily give it up, is a slot you could have handed to someone else if only you had known in time. That is recoverable capacity hiding inside your DNA number.

Why one-way reminders plateau

A reminder is a broadcast. It tells the patient something and asks nothing back that the system can act on. That is why the first reminder helps and the tenth does not. It solves exactly one of the causes above, the honest forgetter, and does nothing for the other four.

The patient who cannot make it still has no easy way to tell you. The patient who no longer needs the slot has no reason to reply to a text. The patient on a dead phone number never sees it. You can add channels and change the timing, and you will squeeze out small gains, but you hit a floor because a one-way message cannot resolve a two-way problem. Confirmation, cancellation, and rebooking are conversations, and a reminder is a monologue.

What actually works

The services that meaningfully move their DNA rate all do variations of the same thing: they turn the reminder into a genuine two-way contact and make the easy actions frictionless.

• Active confirmation. Do not just tell the patient the date, ask them to confirm it. A patient who has actively said "yes, I'll be there" is far more likely to attend than one who was merely informed.
• An effortless out. Make cancelling and rebooking as easy as attending. Counterintuitively, giving people a simple way to cancel raises attendance, because the slot they give up gets filled and the appointments that remain are ones people actually intend to keep.
• Immediate rebooking. The moment a patient says a date does not work, offer another one there and then. A cancellation that ends in a new booking is not a loss, it is a save.
• Reach everyone, not the reachable few. The patients most likely to DNA are often the hardest to reach, so a channel that only works for the digitally engaged systematically misses the people who need contacting most.

The problem with all of this is obvious to anyone who has run a clinic. Phoning every patient, confirming, rebooking the ones who cannot make it, and chasing the ones who do not pick up, is enormous, repetitive work. It is exactly the work that gets dropped first when a team is short-staffed, which is precisely when DNAs climb. See how we approach this in reduce missed appointments.

Where AI fits

This is the gap AI is genuinely suited to close, because the work is structured, high volume, and conversational. An autonomous voice agent can call every patient on an upcoming list, in a warm and natural voice, confirm the ones who are coming, take a cancellation from the ones who are not, and offer a new date on the spot. It does this for the whole list, not the top slice a human team has time for, and it does it consistently on evenings and weekends when people actually pick up.

The important design choice is where the human stays in the loop. The agent handles confirmation, cancellation, and rebooking, and the moment anything clinical or distressing surfaces, it transfers to a person. It automates the coordination, not the care. Everything it captures, confirmed, cancelled, rebooked, or unreachable, is written straight back to the booking system so the clinic has a live picture rather than a pile of voicemails to work through. This is the heart of our approach to patient outreach.

A reminder tells a patient about a slot. A conversation defends it.

How to measure it

You cannot improve a number you are not honestly tracking, and DNA is one of the most commonly fudged metrics in a hospital because the definition slips. Nail it down first.

• Set a clean baseline. Decide exactly what counts as a DNA versus a cancellation, and measure it the same way every week. A cancellation with notice is a very different event from a no-show, because the first gives you a chance to refill the slot.
• Segment it. DNA rates vary enormously by specialty, appointment type, first versus follow-up, and how far ahead the booking was made. An overall number hides the clinics where the problem actually lives.
• Track the recovery, not just the failure. Count how many slots freed by early cancellation you managed to refill. That is the number that turns a DNA programme into recovered capacity.
• Watch the leading indicators. Confirmation rate and reach rate move before the DNA rate does. If you are confirming more patients and reaching more of the list, the attendance number will follow.

Run it as a proper before-and-after. Establish the baseline, change one thing, and measure against the same definition. Done honestly, the effect of moving from broadcast reminders to two-way contact is usually large enough that you do not need to torture the data to see it.

Missed appointments are not a story about unreliable patients. They are a story about a booking process that could not close the loop, and about a volume of contact work no team can staff by hand. Point automation at that, keep humans on the clinical decisions, and the empty slots start to fill themselves.

Luke
Founder, Neutropy

Want to see this on your own lists? Book a demo and we'll walk through it with your clinic's numbers.