Reminders will not fix your no-show rate

Almost every clinic tries the same two things, in the same order. Send more reminders. Then charge for no-shows.

The peer-reviewed evidence says the first helps a little and the second does not work at all. Meanwhile the strongest predictor in the literature is something most practices never look at.

What the no-show rate actually is

The reference point worth knowing is a systematic review in Health Policy (Dantas et al., 2018), which pooled 105 studies. Its findings:

• Average no-show rate: 23%
• Range across studies: 4.0% (intravenous therapy) to 79.2% (physiotherapy)
• North America: 23.5%. Europe 19.3%, Asia 25.1%, Africa 43.0%
• By specialty, medians ranged from paediatrics at 17.0% up to physiotherapy at 57.3%

The spread matters more than the average. If a vendor quotes you an industry no-show rate without asking your specialty, they are quoting a number that does not describe you.

For a single-site view with unusually good longitudinal data, a study at a VA medical centre covering FY1997–2008 and 76,000+ veterans found a mean no-show rate of 18.8% across ten clinics, with gastroenterology highest at 25.7% and audiology lowest at 12.6%.

The finding almost nobody acts on

From the same systematic review, the two main determinants of non-attendance are:

1. Long lead time — the number of days between booking and the appointment
2. Prior no-show history

Read that again in operational terms. If you are booking six weeks out, you have a structural no-show problem, and no reminder cadence will solve it. The patient who booked in June has had six weeks for circumstances to change, for symptoms to resolve, for a different provider to have an opening next Tuesday.

The metric that describes this is third next available appointment — the standard access measure. Most practices track no-show rate obsessively and access almost never, which is backwards.

Reminders: real, modest, and not what you were told

Reminders do work. Just not the way the pitch deck says.

The Cochrane review of mobile phone messaging reminders (2013) found:

• Text versus no reminder: RR 1.14 (95% CI 1.03–1.26) — a real but modest improvement in attendance
• Text versus phone call: RR 0.99 (95% CI 0.95–1.02) — statistically equivalent
• Text plus postal versus postal alone: RR 1.10 (95% CI 1.02–1.19)
• Cost: text reminders cost 55–65% of phone reminders per attended appointment

A randomised controlled trial in an academic primary care clinic found much the same: 11.7% missed with text, 10.2% with phone — equivalent effectiveness, with text far cheaper (€0.07 per text against €0.08 per call plus €8,700 in administrative labour over six months).

So the honest argument for SMS reminders is a cost argument, not an efficacy one. Anyone selling you texting as a step-change in attendance is overselling it.

The 2023 meta-analysis in BMC Health Services Research is more sobering still. Across 61 studies, pooled SMS reminders gave an odds ratio of 1.21 with a confidence interval crossing 1 — and four SMS studies found no significant effect at all. The authors also note that 56 of the 61 studies tested reminders, which tells you how lopsided the research is. We have studied the intervention we already believed in.

For a sense of the real-world ceiling: at that VA centre, introducing a reminder-letter system moved the no-show rate from 18.17% to 16.96%. A centralised phone system moved it from 16.3% to 15.2%. Both are genuine improvements. Neither is transformation.

No-show fees do not work

This is the finding worth putting on a wall.

The same 2023 meta-analysis tested behavioural-economic interventions. A financial penalty produced no statistically significant reduction (p = 0.0895). Free rideshare transport had low uptake and negligible effect.

What did work: a $15 gift card, odds ratio 1.94 (95% CI 1.16–3.24), and reminders delivered in the patient's preferred language.

Clinics reach for penalties because they feel like accountability. The evidence says you will collect resentment, a billing task, and roughly the same attendance.

Online scheduling helps — but only if it is actually booking

This distinction is the most useful thing in this article for anyone about to buy software.

A 2025 study in Frontiers in Digital Health looked at two German ophthalmology settings:

• Private practice with direct patient booking (16,894 appointments): online-booked appointments had a median 1.8% no-show rate versus 5.9% for offline. It also cut unused appointments from 22.7% to 10.3%.
• University hospital using a request-and-triage model (81,173 appointments): online 14.3% versus offline 11.2%. Online was worse.

Same technology category. Opposite results. The difference is that one let patients actually book a slot, and the other made them submit a request that staff still had to process.

A "request an appointment" form is not self-scheduling. It is a lead form that generates work for your front desk and leaves the patient without a confirmed commitment.

Worth knowing that adoption is the real bottleneck: MGMA polling found 71% of practices have fewer than a quarter of their patients using digital self-scheduling, essentially flat year over year. The tool gets bought. The workflow never gets finished.

What is actually eating your front desk

MGMA polled practice leaders in March 2026 on their most time-consuming phone tasks. The ranking was:

1. Eligibility and prior authorisation — verification, coordination of benefits, submissions, status checks, denials
2. Scheduling — including exceptions that online booking does not cover
3. Intake and registration — specifically the rework loop: the callback to fix a missing policy number
4. Results, medication and portal questions
5. Prescription refills — described as one of the most disruptive interruption patterns

MGMA's own framing is the sharpest description of the problem we found: "Just answer the phone" becomes a daily tradeoff: serve the patient in front of you, or the patient on hold.

And the payer-side burden is measurable. The AMA's 2025 prior authorisation survey (1,000 practising US physicians, fielded December 2025) found:

• 40 prior authorisations per physician per week
• 13 hours weekly spent by physicians and staff on them
• 40% of physicians have staff working exclusively on PA
• Phone is still the most commonly used method for medical-service PAs
• Only 24% report their EHR offers electronic PA for prescriptions

Which points at an uncomfortable design truth: a beautiful booking widget in front of a fax-based prior-auth process moves the bottleneck rather than removing it.

The compliance part, stated plainly

Three things clinics routinely get wrong.

HIPAA does not forbid texting patients. HHS guidance is explicit: "appointment reminders are considered part of treatment of an individual and, therefore, can be made without an authorization." The constraints are the minimum necessary standard on content, a business associate agreement with your vendor, and TCPA — which is a different law governing automated calls and texts, with its own consent and frequency conditions.

Your AI vendor is a business associate, not a conduit. HHS explicitly lists a third-party AI chatbot performing services involving PHI — symptom assessment, medical reminders, appointment scheduling — as a business associate. The conduit exception is narrow: HHS states that "entities that access PHI on a regular or frequent basis to perform a service on behalf of a covered entity are not conduits." A telecom carrier is a conduit. A voice agent that listens, understands, books and logs is not. Any vendor claiming otherwise has not read the guidance.

Subcontractors are in scope too. A business associate must have its own BAA with its subcontractors. In practice that means your voice vendor's language-model provider, telephony provider and transcription provider are all part of the chain. Ask for the list. The answer tells you a great deal about how seriously the vendor takes this.

Two further items depending on where you operate: California's AB 3030 requires a disclaimer when generative AI produces patient communications containing clinical information — though it explicitly excludes administrative matters like scheduling and billing — and practices that are federally-assisted SUD programmes carry 42 CFR Part 2 obligations on top of HIPAA. None of this is legal advice; it is the shape of the questions to put to your counsel.

The numbers we refuse to quote

You will see "no-shows cost US healthcare $150 billion a year" everywhere, including in articles from organisations that ought to know better. It traces to a scheduling vendor with no published methodology. The same goes for "$200 per missed appointment," "$150,000 lost per physician per year," and essentially every clinic call-centre benchmark in circulation — 7% abandonment, 4.4-minute hold times, "85% of patients who don't get through never call back." Every one traces to a vendor blog with no method disclosed.

There is also no credible published benchmark for patient intake time in an outpatient or dental setting. We looked. The "25 minutes down to 7" figures are all vendor marketing.

So rather than quote a number at you, here is the arithmetic that is unattackable because it is yours:

Your no-show rate × slots per day × average collected revenue per visit × clinic days per year.

Pull the first number from your practice management system, not from an article. Then measure your third-next-available. If it is measured in weeks, start there — before anyone sells you a reminder system, and certainly before anyone sells you prediction.