Key takeaways
- Switching systems is complex and risky[1]; most of the risk sits in data mapping, cutover and training, not the new software.
- Inventory every type of data first, including appointments, balances, alerts, documents and templates, and decide what moves.
- Run at least one trial migration and reconcile counts and samples before the real cutover.
- Treat the migration as new processing of special category data: complete a DPIA and be ready to report a breach within 72 hours[2][3][4].
Why migrations go wrong
When the Competition and Markets Authority looked at GP clinical systems, it heard that switching supplier is considered a complex and risky process that can involve a large financial cost[1]. The same is true for dental and private clinic systems. The new product is rarely the problem. Data that does not map cleanly, appointments that vanish in the gap between systems, templates nobody rebuilt and staff who were trained too late cause most of the pain.
A migration is also a data protection event. You are copying special category health data from one system to another, often with a supplier or contractor involved. Planning it properly protects patients and gives you evidence that you acted responsibly if something goes wrong.
Start with an inventory of everything you hold
| Data type | Examples | Risk if lost or wrong | How to check |
|---|---|---|---|
| Demographics | Names, dates of birth, addresses, NHS numbers, contact preferences | Wrong patient, failed reminders | Counts and random samples |
| Clinical record | Problems, medications, allergies, notes, results, codes | Clinical harm | Clinician review of samples |
| Alerts and flags | Safeguarding, violent patient, reasonable adjustments, communication needs | Safety and dignity failures | Full list compared one by one |
| Documents | Letters, scanned records, images, consent forms | Missing history, legal exposure | Counts by type, open samples |
| Appointments | Future bookings, recalls, waiting lists | Missed care, angry patients | Count and compare day by day |
| Financial | Balances, deposits, insurer details, payment plans (private and dental) | Lost income, complaints | Totals reconciled to the penny |
| Configuration | Templates, protocols, searches, message wording | Workarounds and errors | Rebuild list with owners |
Governance before any data moves
Data concerning health is special category data under UK GDPR, and processing it is prohibited unless a specific condition applies[2]. A migration changes how and where that data is processed, so check your lawful basis, update your records of processing and confirm the roles: you remain the controller, and the suppliers doing the extraction and loading act as processors under contract.
Where processing is likely to result in a high risk, particularly when using new technologies, UK GDPR requires an impact assessment before the processing starts, and processing special category data on a large scale is one of the cases named[4]. A DPIA for the migration should cover how data is extracted, transferred, stored during transition, validated and deleted from the old system. It should also follow the Caldicott Principles that health and care organisations in England use to keep confidential information protected[5].
- Signed data processing terms with both the outgoing and incoming suppliers.
- An agreed, secure transfer method, never email or removable media without encryption.
- A named person accountable for the migration, and a clinical lead for clinical data checks.
- Agreed rules for what happens to data held by the old supplier after exit.
A migration plan in eight steps
Agree scope
Decide what migrates, what is archived read-only and what is not needed. Write down the reasons.
Map the data
With both suppliers, map each field and code from the old system to the new. Flag anything with no direct equivalent.
Clean first
Fix duplicates, missing NHS numbers, old addresses and unused templates before extraction. Clean data migrates better.
Trial migration
Load a full copy into a test environment. Never test on live patients.
Reconcile
Compare counts, totals and samples between old and new. Log every difference and resolve or accept it.
Rebuild configuration
Recreate templates, searches, rotas and message wording in the new system, and test them.
Train and rehearse
Train staff on the tasks they do daily, using the trial data, then rehearse cutover day.
Cut over and check
Freeze the old system, run the final load, repeat key checks and go live with extra support on hand.
Checks that catch real problems
- Patient counts by status (active, deceased, left) match, with any difference explained.
- Every alert and flag in the old system appears on the right patient in the new one.
- Allergies and current medications match for a clinician-reviewed sample of at least a few dozen records, including complex patients.
- Future appointments match day by day, clinician by clinician.
- Documents open, are attached to the right patient and keep their dates.
- Financial balances and deposits reconcile exactly for dental and private practices.
- Codes used for recalls and contract reporting produce the same search results in both systems.
Pick sample patients deliberately: children, people with safeguarding flags, long-term conditions, recent deaths, name changes and duplicate records are where errors hide.
Plan around the appointment horizon
Appointments are the data patients notice first. In NHS general practice, 45% of appointments in the year to August 2026 were booked for the same day and 81% were booked to take place within two weeks[6]. That means the diary for the fortnight around cutover holds most of the bookings that matter, and errors there will show up immediately.
Source: British Medical Association, 2026[6]
Show the data
| Item | Value |
|---|---|
| Same day | 45% |
| Within two weeks | 81% |
Reduce risk by limiting forward booking for a short period before cutover, exporting a printed or offline list of the next two weeks' appointments, and checking the migrated diary against it on the first morning. Dental and private practices with longer booking horizons should check further ahead.
Cutover and running both systems
Choose a cutover at a quiet time, such as a weekend, and avoid the start of a month when reports and claims are due. Keep the old system available read-only for an agreed period so staff can look up anything that did not migrate. Decide in advance how long that period lasts and who can access it, because a second live copy of patient data is a risk in its own right.
- Freeze changes in the old system at a set time and tell all staff.
- Have both suppliers available on cutover day and the first working day.
- Run a short huddle at the end of each of the first five working days to log problems.
- Keep a simple issue log with owner, priority and fix date.
Training and telling patients
Train as close to go-live as practical, so skills are fresh, but early enough to fix gaps. Short role-based sessions work better than long general demonstrations: reception books, moves and cancels appointments; clinicians write notes, prescribe and send letters; admin staff process documents and run searches. Give each role a one-page quick guide for the first week.
Patients need to know only what affects them. Tell them in advance if online booking, the patient app, text reminders or payment links will change, and whether they need to log in again or download anything. Put a short notice on your website and phone message for the cutover weekend, and expect a few extra calls in the first week while patients adjust.
If something goes wrong
Treat lost, corrupted or misdirected data as a potential personal data breach. The controller must notify the regulator without undue delay and, where feasible, within 72 hours of becoming aware of it, unless the breach is unlikely to result in a risk to people[3]. Serious UK GDPR infringements can lead to fines of up to £17.5 million or 4% of worldwide turnover[7]. Record what happened, what you did and why, even if you decide the breach is not reportable.
72 hours
deadline, where feasible, for notifying the regulator of a reportable personal data breach[3]
Example: a dental practice switch
Example: a fictional four-surgery dental practice moved to a new cloud system. The trial migration showed that patient balances matched, but recall dates for children had been loaded against the wrong field. The supplier fixed the mapping before the real cutover. On go-live morning, reception checked the next fortnight's diary against a printed list and found three appointments missing for a hygienist who had been set up under a different name. Because the old system was still available read-only, the bookings were restored before the patients arrived.
Closing down the old system
Do not switch off the old system the day after go-live. Agree a date once the read-only period ends, confirm in writing that all required data has migrated, export any archive you need to keep under your retention policy, and obtain written confirmation from the outgoing supplier that remaining copies have been securely deleted. Update your records of processing and the DPIA to show the change is complete.
Frequently asked questions
How long does a practice software migration take?
Plan for several months from decision to go-live for a clinical system, including a trial migration, configuration and training. The data work usually takes longer than expected.
Do we need a DPIA for a migration?
It is strongly advisable. UK GDPR requires one where processing is likely to be high risk, including large-scale processing of special category data[4].
Should we migrate everything?
Not always. Some old data can be archived read-only and retained under your retention policy. Decide and document what moves, what is archived and why.
What is the single most useful check?
A clinician reviewing a deliberately chosen sample of complex records in both systems side by side, alongside matching counts of patients, alerts and future appointments.
References
Competition and Markets Authority (2023). Anticipated acquisition by UnitedHealth Group Incorporated of EMIS Group Plc: summary of provisional findings (opens in a new tab) ↑
legislation.gov.uk (The National Archives) (2026). UK GDPR Article 9: Processing of special categories of personal data (opens in a new tab) ↑
legislation.gov.uk (The National Archives) (2026). UK GDPR Article 33: Notification of a personal data breach (opens in a new tab) ↑
legislation.gov.uk (The National Archives) (2026). UK GDPR Article 35: Data protection impact assessment (opens in a new tab) ↑
GOV.UK (National Data Guardian) (2026). The Caldicott Principles (opens in a new tab) ↑
British Medical Association (2026). Pressures in general practice data analysis (opens in a new tab) ↑
legislation.gov.uk (The National Archives) (2026). UK GDPR Article 83: General conditions for imposing administrative fines (opens in a new tab) ↑