Missed Appointments: What AI Actually Changes About No-Shows

What AI fixes, and what it does not

AI does not bring in a patient who has decided not to come. It removes forgetting, opens cancellation at any hour, and refills the slot that was freed. Three mechanisms. Two of the three causes. The most complete Cochrane review on the subject measures attendance at 67.8% with no reminder, 78.6% with an SMS reminder and 80.3% with a phone call. A reminder therefore shifts roughly one absence in three. The other two thirds belong to how the schedule is built, not to the patient. An evaluation covering 25 general practices in East London puts it bluntly: to cut non-attendance, the appointment system has to change, not the patient. The rest of this article separates the three causes of a missed appointment. It states what acts on each one. And it names what no reminder will ever repair.

27 million appointments, and what the number hides

France’s most quoted figure comes from the national medical council, the Conseil national de l’Ordre des médecins. In its statement of 26 January 2023, the council reports that every week, 6% to 10% of patients fail to show up. That amounts to nearly two hours of consultation time lost per week per doctor. Extrapolated nationally, close to 27 million missed appointments a year. Around two thirds of those absences are said to concern a first appointment. That is an extrapolation, not a measurement. The strongest field survey published in France comes from URPS Île-de-France, the regional union of private-practice doctors for the Paris region: 2,240 respondents, 8 to 21 July 2022. 69% see one to two missed appointments a day, 22% see three to five. Lost consultation time runs from 20 minutes to an hour and a half a day. 80% of respondents call it a major problem in their daily practice. 57% see it rising. Internationally, the systematic review by Parsons and colleagues reports, across nineteen studies, rates from 3.3% to 48.1%, median 12.9%. There is no such thing as a no-show rate. There is an appointment system that produces one.

Three absences that have nothing in common

A missed appointment is not a behaviour. It is three different situations filed under one word. The systematic review by Parsons, Bryce and Atherton, published in 2021 in the British Journal of General Practice, pooled 26 studies on non-attendance in general practice. Twelve of them examine the reasons. Nine report the reasons patients gave. Work or family and childcare commitments come first, in seven studies. Then forgetting and transport difficulties, five studies each. Next, being too unwell that day and the weather, three studies each. Last, feeling better by the time of the appointment, two studies. The same review isolates a second family of reasons, this time attributable to the practice. The relationship with the clinician, five studies. The booking system itself, three studies. Miscommunication about what had been agreed, three studies. The Cochrane review, for its part, opens its abstract with an unambiguous sentence: patients’ forgetfulness is one of the main reasons for missed appointments. Three causes. Three levers with nothing in common.

• Forgetting: the patient means to come, memory failed
• Being unable to come: the patient knows the night before, the practice does not
• Not daring to cancel: the patient drops out silently
• Booking three clinicians at once: the appointment was never firm

Forgetting: the only cause a reminder erases

The Cochrane review on mobile phone messaging reminders pools eight randomised trials and 6,615 participants. Seven of those trials, 5,841 participants, compare SMS with no reminder: risk ratio 1.14, 95% confidence interval 1.03 to 1.26. Three others, 2,509 participants, compare SMS with a phone call: risk ratio 0.99, interval 0.95 to 1.02. The channel therefore matters less than the message landing. A randomised trial at Geneva University Hospitals on 6,450 patients confirms it: 11.7% missed appointments with SMS against 10.2% with a call, a non-significant gap. Over six months the SMS setup cost 230 euros, the call setup 8,910 euros. Almost all of that difference is reception staff time spent dialling. The wording of the message, on the other hand, weighs heavily. Across 161,587 members of the Israeli health fund Clalit, the control wording left 21.1% missed appointments. The best variant tested came down to 14.2%. Same appointment, same channel, close to seven points apart.

Cancelling must be as easy as booking

Here is the asymmetry almost nobody fixes. In the Île-de-France survey, 69% of missed appointments had been booked through an online platform, 25% through reception. Booking is open day and night, three clicks, no human contact needed. Cancelling still rests, very often, on calling a saturated switchboard during opening hours. A patient who realises at 9 pm on a Sunday that they cannot make it has no way to tell anyone. Monday morning, they are at work. By Tuesday the appointment is gone. The British review cited above explicitly files the booking system among the practice-side causes of non-attendance, alongside miscommunication about what had been agreed. A phone agent that answers at any hour removes that asymmetry. The patient cancels in thirty seconds, no queue, no need to justify anything. Nobody needs the reason for the cancellation. What matters is that the practice learns about it before the day of the appointment.

• Answer evenings, weekends and during consultations
• Cancel with no switchboard, no queue, no reason required
• Rebook inside the same call, before hanging up
• Write the cancellation into the calendar, not into voicemail

A freed slot earns nothing until it is refilled

A cancellation two days out earns nothing while the slot stays open. It has to be refilled, fast. The most instructive evaluation comes from Tower Hamlets, East London: 25 general practices, more than four million appointments analysed from April 2014 to March 2019. The mean non-attendance rate fell from 7% to 5.2% in two years using classic measures, SMS, phone reminders, publishing missed-appointment statistics. One single practice touched the structure of its schedule, cutting the forward booking horizon from 28 days to one working day. Its rate dropped from 7.8% to 3.9%. The authors conclude that forward booking time, counted in days, is the best predictor of a practice’s non-attendance rate. A waiting list acts on exactly that lever. It turns an appointment booked three weeks ahead into an appointment booked this morning for tomorrow. Calling the waiting patients, offering the slot, confirming, updating the calendar. The work is repetitive and time-critical. It rarely fits inside a reception team’s day.

What a reminder does not repair

Read the Cochrane figures the other way round. With an SMS reminder, attendance tops out at 78.6%. One patient in five still misses the appointment. A reminder touches neither the person who booked three clinicians at once, nor the one who gave up on care, nor the one with no transport that day. The French medical council notes that close to two thirds of absences are said to concern a first appointment. The Île-de-France survey points the same way: 80% of respondents name new patients as the most likely to miss. The patients on their own registered list are named the most reliable. A continuing relationship does more than a text message. The literature, for its part, has tested almost nothing else. A 2023 review counts 61 studies on non-attendance: 56 cover reminders, only two cover an incentive or a penalty. The punitive route has already failed in France. Article 52 of the 2025 social security financing act created a charge for missed appointments. The Constitutional Council struck it down on 28 February 2025, because parliament had defined neither the nature of the penalty, nor its amount, nor the conditions for applying it.

What a reminder is allowed to say

Telling a patient they have an appointment with a clinician means processing health data. So the message has no business carrying the reason for the consultation. Clinician, date, time. Nothing else. A text read on a locked screen, in an open-plan office or by a relative, must reveal nothing about the care. Hosting, meanwhile, follows a rule specific to French law. Article L. 1111-8 of the French public health code requires that any personal health data collected during prevention, diagnosis, care or medico-social follow-up be hosted by a provider holding the French HDS health-data hosting certification. That certification is issued by accredited bodies and overseen by the national digital health agency. A practice in Germany or Spain answers to the GDPR and to its own national law. Neither imposes an equivalent hosting certification. An agent that calls patients, sends reminders and refills slots stores names, phone numbers and consultation times. Those three items together are health data. The question to put to any supplier fits on one line: where is this data hosted, and under which certification?

Sources

1. Conseil national de l’Ordre des médecins, communiqué « Rendez-vous médicaux non honorés », 26 janvier 2023 https://www.conseil-national.medecin.fr/publications/communiques-presse/rendez-medicaux-honores
2. URPS Médecins libéraux Île-de-France, enquête « Rendez-vous non honorés » du 8 au 21 juillet 2022, 2 240 répondants, communiqué du 5 février 2024 https://www.urps-med-idf.org/wp-content/uploads/2024/02/20240205_urps_CP_etude_rendez-vous_non_honores.pdf
3. Gurol-Urganci et al., Cochrane Database of Systematic Reviews, CD007458.pub3, 2013 https://www.cochranelibrary.com/cdsr/doi/10.1002/14651858.CD007458.pub3/full
4. Junod Perron et al., BMC Health Services Research 13:125, Hôpitaux universitaires de Genève, 2013 https://pmc.ncbi.nlm.nih.gov/articles/PMC3623700/
5. Parsons, Bryce & Atherton, British Journal of General Practice 71(707), e406-e412, 2021 https://pmc.ncbi.nlm.nih.gov/articles/PMC8103926/
6. Margham, Williams, Steadman & Hull, British Journal of General Practice 71(702), e31-e38, 2021 https://pmc.ncbi.nlm.nih.gov/articles/PMC7716879/
7. Berliner Senderey et al., Clalit Health Services, PLOS ONE 15(6), e0234817, 2020 https://pmc.ncbi.nlm.nih.gov/articles/PMC7310733/
8. Werner et al., BMC Health Services Research 23:1136, 2023 https://pmc.ncbi.nlm.nih.gov/articles/PMC10594857/
9. Conseil constitutionnel, décision n° 2025-875 DC du 28 février 2025, article 52 LFSS 2025 https://www.conseil-constitutionnel.fr/decision/2025/2025875DC.htm
10. Agence du Numérique en Santé, certification HDS, article L. 1111-8 du code de la santé publique, 2016 https://esante.gouv.fr/produits-services/hds