.jpg){kind=link}
- Published
- Updated
The same person may appear at several sites or in repeated data transfers within a PMCF registry. Unrecognised duplication can distort participant counts, device exposure and event rates. Deduplication should not be equated with deleting records.
Define the assessment
Define whether the record unit is a person, procedure, device or visit. Several devices in one person do not necessarily represent duplication. Design matching alongside privacy, permitted data use and error risk, with authorised review of uncertain matches.
Worked example
For example, an implant recipient attends another site, which creates a new record. Both records suggest the same person, but a device replacement also occurred. The team does not delete one based on a similar birth date. An authorised process links the person while preserving distinct device periods and source records from both sites.
Practical checks
- Distinguish person counts from device, procedure and visit counts in the data dictionary.
- Make possible, confirmed and rejected matching decisions traceable.
- Reconcile events, follow-up time and denominator calculations after linking records.
Decision and limitations
Acceptance depends on retaining traceability to original records and the matching decision. Do not present uncertain matching as error-free deduplication. Explain denominator changes, confirm comparable record units and avoid weakening privacy controls to solve data-quality problems.
Frequently asked questions
- Are similar identifying details sufficient to delete a record?
- No. Matching needs authorised reasoning; similarity can produce an incorrect merge.
- Are two devices in one person duplicates?
- Not necessarily. Model people, devices and exposure periods as separate data units.