Patient no-show prediction that scores every booked slot from your own attendance history. High-risk patients get a call or reschedule offer, low-risk patients get a text, and freed slots go to the waitlist before the day starts.
Patient no-show prediction is a machine learning model that estimates the probability each booked appointment will be missed, using factors such as lead time, past attendance, appointment type, reminder responses and travel distance. It outputs a daily risk list, a recommended action per appointment (text, call, reschedule offer or safe overbook) and a ranked waitlist to fill slots that open up.
Most clinics send the same reminder to everyone and double-book by gut feel. That spends staff time on patients who were always coming, and still leaves gaps that a waiting patient would have taken.
Of NHS hospital outpatient appointments were missed in England in 2024–25, out of 146.1 million appointments.[2]
Estimated yearly cost of missed appointments to the NHS in England at 2024 levels, according to NHS England.[1]
Of US medical groups saw higher no-show rates so far in 2026 compared with 2025, in an MGMA poll.[3]
Overbooking alone has uncertain evidence in a systematic review of prediction-led no-show interventions, so it needs guardrails.[4]
Trained on two to three years of your own appointments. Scores update as the visit approaches and as patients respond to reminders.
Cheap reminders for everyone, and staff time spent only where the model says it changes the outcome. Patients can confirm, cancel or rebook in one reply.
When a slot opens, the next suitable waitlisted patients get an offer automatically. Overbooking is suggested only where the model is confident and capped per session.
We extract 24–36 months of appointments from your scheduling system into your cloud account and agree the baseline no-show and filled-slot rates by specialty.
We train and back-test the model, then review error rates across age bands, postcodes or ZIP codes and payer types before any patient is contacted.
Daily risk lists go to the front desk. Targeted reminders and waitlist offers run; a comparable clinic or week stays on the current process as the control.
We compare no-show rate, filled-slot rate and staff calls made against control and baseline, and give you a written go/no-go.
| Criterion | Same reminder for all | Reminder platform with rules | Stratgik build + run |
|---|---|---|---|
| Who gets staff attention | Everyone or no one | Rule-based lists (e.g. new patients) | Patients the model scores as high risk, with reasons |
| Empty slots | Found on the day | Patient can cancel by text | Released early and offered to the waitlist |
| Overbooking | Gut feel | Fixed percentage | Per-session, confidence-based, capped and approved |
| Fairness checks | None | Rarely | Monthly review of actions and outcomes by patient group |
| Setup effort | None | Low; good for single-site practices | 21-day pilot on your attendance history |
The research is consistent on one point: sending more of the same reminder helps a little, while spending human effort on the right patients helps a lot.
Set your volumes and visit value. The share of no-shows recovered through prevention and backfill is an assumption; the pilot measures it against a control clinic.
Illustrative estimate using your inputs and stated assumptions, not a quote or guarantee. The pilot measures the real figure against your baseline.
21 days · one clinic or specialty
per month · production for one site
per month · multi-site or hospital outpatients
SMS, WhatsApp and voice messaging, model usage and cloud hosting billed at cost and reported monthly; taxes excluded. GBP and AED prices are indicative conversions from USD.
Patient no-show prediction is accurate enough to rank appointments by risk, not to say for certain who will miss. What matters is whether the top of the list contains most of the misses. In the pilot we back-test on your history and report how many no-shows fall in the riskiest 10% and 20% of bookings, by specialty, before anyone acts on the scores.
It can if built carelessly, which is why fairness checks are part of the design. We exclude race, ethnicity and other protected characteristics as inputs, test error rates across age, area and payer groups, and use high scores to offer help such as calls, transport or video visits rather than to penalise. Overbooking is capped and reviewed monthly by patient group.
Overbooking can recover capacity, but on its own the evidence is uncertain and it risks long waits when everyone turns up. We treat it as a last step: reminders and backfill first, then a small, capped overbook only in sessions where the model is confident several patients will miss. Staff approve each one, and we track wait times on overbooked sessions.
Stratgik's 21-day pilot for one clinic is a fixed $12,000. Production runs from $2,500 a month per site, and multi-site or hospital outpatient deployments start from $7,000 a month. Messaging and cloud usage are billed at cost. Most clinics compare that against the calculator's figure for recovered visit revenue.
It works with the scheduling system you already use through its API, HL7 feeds or scheduled report exports. Common connections include Epic Cadence, Oracle Health, athenahealth, eClinicalWorks and NextGen in the US, and SystmOne, EMIS and hospital PAS systems in the UK. Reminders go out through your existing messaging provider or one we set up in your account.
The production pilot takes 21 days for one clinic: four days to extract history and set a baseline, five days to train and check fairness, about nine days live with a control, and three days to measure. Rolling out to more sites usually takes one to two weeks each, since the model is recalibrated per site.
Book a 30-minute call. We will pick the clinic, the control and the filled-slot metric for a 21-day pilot.