Appointment no-show prediction

• Each week, every booked appointment gets a likelihood of being missed, with the main reasons in plain English, such as two recent missed appointments, a letter returned undelivered, or a long journey with no transport booked.
• Patients most likely to miss get more help, not less: an extra reminder through the NHS App, an easy way to rebook, travel and transport information, a check that their contact details are right, and for some, a phone call from the booking team.
• Patients who can't come are helped to cancel early, so the slot goes to someone else on the waiting list instead of sitting empty.
• Clinic leads can add carefully limited extra bookings to sessions where several missed appointments are expected, within rules they set for each clinic. Booking staff decide, and the model never double-books onto a named patient's slot.
• Missed-appointment rates and the help offered are tracked by deprivation, ethnicity, age and disability, so the trust can show that the gap is narrowing and no group is worse off.
• Nobody is discharged, deprioritised or moved down a list by the model. A clinician reviews every missed appointment before any decision, and a child who was not brought is always considered for safeguarding.

• ICOUK GDPR special category data and the Data Protection Act 2018
  →

  Appointment histories show which services someone uses, so they are health data. That means a DPIA before the pilot, a clear lawful basis and condition for processing, processing in UK regions or the trust's own data centre, a defined retention period, and a contract that stops providers training their models on the trust's data. Patients are told, through the privacy notice and the reminder itself, how their information is used to offer help.

• ICOAutomated decisions (UK GDPR, as amended by the Data (Use and Access) Act 2025)
  →

  Since 5 February 2026, the strict limit on solely automated significant decisions applies mainly where special category data is involved, and here it is. The design keeps every significant decision with a person: the model only triggers extra reminders and support, and discharge, priority and overbooking decisions are made by staff.

• National Data GuardianThe Caldicott principles
  →

  Use only the data the model needs, give access only to those who need it, and inform patients' expectations about how their data is used. The Caldicott Guardian signs off the data used and the purposes, and is consulted on any change to either.

• NHS EnglandThe national data opt-out
  →

  The opt-out does not apply when a provider uses its own patients' data to support their individual care, which covers offering a patient help to attend. It can apply if data is used for wider planning or research, or shared under section 251 support. Agree the position with information governance in discovery, and apply the opt-out wherever it is needed.

• NHS England and EHRCHealth inequalities duties, Core20PLUS5 and the Equality Act 2010
  →

  The planning framework asks ICBs to show how they will reduce health inequalities, and trusts are bound by the public sector equality duty. NHS England's Core20PLUS5 approach focuses on the most deprived 20% of the population and other groups with poorer access. The model's results are monitored for each of these groups, and the trust's equality impact assessment is updated before go-live.

• NHS EnglandDCB0129 and DCB0160 clinical safety, and the DSPT
  →

  Missed appointments can cause harm, so the clinical safety officer decides how the standards apply. If the trust builds the model, it is the manufacturer under DCB0129 as well as the deploying organisation under DCB0160. The standards are under review, so watch for revised versions. The new system is covered in the trust's Data Security and Protection Toolkit submission.

On a phone, the table shows the platform you pick. Sovereign / on-prem means open, portable components you can run in your own data centre or with a UK or European provider; IONOS is shown as one example, and others exist. What sovereign cloud means →

• A clear rule, written into the access policy: a likelihood score can only bring extra help. It is never used to discharge, deprioritise or offer worse slots.
• Every missed appointment is reviewed by a clinician before any discharge. A child who was not brought is always considered under the trust's safeguarding policy.
• Overbooking rules owned by each clinic lead, with caps, excluded clinic types (for example urgent suspected cancer and paediatric clinics), and a monthly review of overruns.
• A quarterly fairness review, chaired by the health inequalities lead, of missed-appointment rates, help offered and outcomes by deprivation, ethnicity, age and disability.
• A named owner for the model, the templates and the call scripts, with change control like any other system.
• Booking staff trained before go-live in what the score means, what it doesn't, and how to talk to patients about getting to their appointment.

1. 3–4 weeks

   Discovery

   Measure missed-appointment rates by specialty and patient group, check data quality and contact details, agree the national data opt-out position and the supplier route, and pick the pilot specialties.

2. Gate · must be true to passBaseline by patient group agreed, Caldicott Guardian and IG support, DPIA and equality impact assessment started, clinical safety officer named.

3. 10–12 weeks

   Pilot

   Score appointments in shadow mode first, then switch on extra help for half the high-likelihood patients and compare. No overbooking yet.

4. Gate · must be true to passMissed appointments down, no group worse off, calibration within limits for every group, and clinical safety and IG sign-off.

5. 4–6 weeks

   Production

   Extra help goes live in the pilot specialties, then overbooking is switched on for clinics whose leads opt in.

6. Gate · must be true to passMonitoring, fairness reporting, support model, rollback plan and booking team training in place.

7. Ongoing

   Scale

   Add specialties in waves, then reuse the pattern for diagnostics and community appointments.

8. Gate · must be true to passEach new specialty passes the same pilot gate, including the fairness check.

R ResponsibleA AccountableC ConsultedI Informed· not involved

Where I'd fit

1. 1

   Goes wrong

   The model learns that deprivation predicts missed appointments, and the score gets used to give poorer patients worse slots or fewer chances.

   How the design avoids it

   A written rule that the score only brings extra help, enforced in the workflow, plus fairness monitoring by group from the first week of the pilot.

2. 2

   Goes wrong

   Patients are blamed for missing appointments they were never told about.

   How the design avoids it

   Contact details checked against the Personal Demographics Service, delivery receipts from NHS Notify, and 'never received' treated as a system failure, not a patient one.

3. 3

   Goes wrong

   Overbooking turns into overrun clinics and frustrated staff.

   How the design avoids it

   Clinic leads set caps per session, high-risk clinic types are excluded, and overruns are reviewed every month.

4. 4

   Goes wrong

   A missed paediatric appointment is treated like any other, and a safeguarding signal is lost.

   How the design avoids it

   'Was not brought' is recorded separately, never triggers automatic discharge, and always goes to a clinician.

5. 5

   Goes wrong

   The model quietly goes stale after a change, such as more virtual clinics or patient-initiated follow-up.

   How the design avoids it

   Monthly drift checks, retraining with change control, and a named owner who is told when pathways change.

1. 2026/27 to 2028/29
   The planning framework asks providers to move all direct-to-patient messaging to NHS Notify, with NHS App notifications as the preferred channel. Migration starts in 2026/27 and should be complete by the end of 2028/29.

   Send every reminder through NHS Notify from the start, and design templates for the app first, with text and letter as fallbacks.

2. 2026/27
   An outpatient reset: fewer routine follow-ups, more Advice and Guidance, and digital patient-initiated follow-up to be in place by the end of 2028/29. Every trust must improve 18-week performance by at least 7% or reach 65%.

   Expect the mix of appointments to change. Patients on patient-initiated follow-up have no fixed booking, so retrain the model as pathways change and don't compare this year's rate with last year's without adjusting.

3. 2026–2027
   The ICO's code of practice on AI and automated decision-making, required by regulations in force since May 2026.

   Keep the decision log, the reasons shown to staff and the records of human review in a form you could show the ICO.

4. Late 2026 and 2027
   Revised DCB0129 and DCB0160 clinical safety standards are expected after the review that closed on 11 September 2026.

   Keep the hazard log and safety case structured and current, so moving to the new versions is an update, not a rewrite.

5. By December 2028
   The Health Bill, introduced in May 2026 and now in the Lords, would give the legal basis for a Single Patient Record and move NHS England's functions into DHSC. The government's target is to launch the Single Patient Record in the NHS App by December 2028.

   Build on national services such as NHS Notify and the Personal Demographics Service, not one-off integrations, so the design keeps working as national systems change.