Done or just delivered? A clinical records migration
Westbridge NHS Trust is migrating 2.3 million patient records to a new clinical system. The project initiation documentation states that the migration product must be 'accurate and user-friendly.' The supplier's test summary reports a 99.2% record match, but the quality register has not been updated for seven weeks, and acceptance records exist for only 8 of the 20 quality criteria. The senior user wants to approve the end-of-stage assessment because the regulator has fixed a go-live date in 12 weeks. The project manager knows that the data-quality tolerance was never formally agreed. Two clinician representatives say they have not seen the migrated data, and one quality review of the data mapping has been cancelled twice. The supplier argues that 99.2% is 'within normal industry tolerance' and that the remaining discrepancies are 'minor.'
1. What should the project manager do about the end-of-stage assessment?
Read the worked answer
The project manager cannot support an end-of-stage assessment that accepts products whose quality activities are incomplete. The quality register is seven weeks out of date, acceptance records cover only 8 of 20 criteria, the data-quality tolerance was never agreed, and a required quality review has been cancelled twice. Approving the stage would make the board's decision unreliable.
The project manager should:
- Bring the quality register up to date and record the status of every quality criterion, including the cancelled review.
- Re-plan the missing quality activities, especially clinician review of migrated data and the data-mapping quality review.
- Agree a measurable data-quality tolerance with the senior user before judging the 99.2% result.
- Assess whether the remaining work can be completed within stage tolerance. If it cannot, prepare an exception report for the board rather than forcing the stage closed.
- Record the regulator's fixed date as a constraint and show the board the consequences of accepting incomplete quality evidence.
If the board decides to proceed, it should do so with explicit acceptance of the residual risk, not by treating the stage as complete. The project manager should also update the quality management approach so that acceptance records and tolerances are agreed before testing, not after.
2. Evaluate the supplier's argument that 99.2% is within normal industry tolerance and that the remaining discrepancies are minor.
Read the worked answer
The argument cannot be accepted as it stands. 'Normal industry tolerance' is not a substitute for an agreed project tolerance. The project manager does not know whether 0.8% is acceptable because the data-quality tolerance was never formally agreed. On 2.3 million records, 0.8% is approximately 18,400 records, which is not self-evidently minor in a clinical system.
'Minor' also needs definition. A discrepancy in a patient's allergy field is not equivalent to a formatting difference in an address. The project manager should ask the senior user and clinicians to define data-quality criteria by dimension, such as completeness, accuracy, consistency and timeliness, and to set thresholds for each. The supplier's test summary should then be assessed against those criteria, not against a general industry claim.
If an agreed tolerance is met, the remaining discrepancies should be handled through a documented remediation plan with clinical risk assessment. If no tolerance is met or agreed, the product cannot be accepted, and the project manager should escalate the issue and consider an exception. The decision belongs to the senior user, informed by clinical judgement, not to the supplier.
3. How should the project manager turn 'user-friendly' into something that can be tested and accepted?
Read the worked answer
An acceptance criterion of 'user-friendly' is not testable. The project manager should derive measurable criteria from the product description and the users' real tasks. This means working with the senior user and clinician representatives to define what users must be able to do, under what conditions, and with what acceptable performance.
Useful criteria might include: a clinician can find a patient record within a specified time; a defined set of tasks can be completed without assistance; error rates stay below an agreed threshold; and users can complete key workflows after a defined training period. These criteria should be recorded in the product description and the quality register, with the quality method, reviewer and acceptance record named.
The project manager should also plan the quality activities that will produce the evidence, such as usability testing, observed clinical scenarios and user acceptance testing. The senior user should approve the criteria before testing begins. If the criteria are not met, the product is not accepted, even if the go-live date is fixed. Changing the date is a board decision; changing the definition of 'done' without evidence is not.