A correct match is only the first step
Even when two records belong to the same person, the same visit may appear in both sources. Treating those rows as separate events can double-count care. Conversely, a gap between sources may be a period of unobserved care rather than good health.
The linkage also needs a clear rule for changes in identifiers over time.
Describe what is observable
- Define the population eligible to appear in each source.
- Identify periods with reliable coverage.
- Resolve duplicated events using documented rules.
- Assess uncertain and missed matches.
These steps help distinguish “no event recorded during coverage” from “no usable coverage.”
Linkage error may be uneven
People whose names, addresses, or other identifying fields change may be harder to link. If linkage success differs across groups, the final sample can become selectively incomplete.
Report the linkage process and relevant success rates, not only the number of merged rows. Where feasible, examine whether conclusions change under reasonable alternative matching or observation rules.
The value of linkage is the additional information it provides. Its credibility comes from preserving uncertainty about the connections, rather than presenting every joined record as an equally certain and continuously observed history.
