Intended or unintended consequences of mapping

Even a dual authored, dual reviewed, clinically assured equivalence map may not provide a comprehensive and long term solution in every case.

The table below shows a well formed map.

Source ID	Source term	Relationship type	Target ID	Target term
A01	Asthma	equivalent	195967001	Asthma
C11	Common cold	equivalent	82272006	Common cold
D58	Dislocation of joint	equivalent	108367008	Dislocation of joint
E123	Ear infection	equivalent	129127001	Ear infection

The consequence here is that the new SNOMED CT content reflects only what was pre-existing, with no change in scope or range of clinical content, and with the features of the legacy source system maintained (and these may be 'convenience' terms, or hand-picked favourites, with none or some internal structure or they may be disjoint. Sometimes these source code sets are arbitrary or idiosyncratic.

Despite investing effort in producing a well formed map we achieve only what we already have, with merely a change in codes/identifiers. There is no improvement in clinical utility, standardisation, interoperation with other standard SNOMED CT-enabled systems, and analytics that might exploit the logical definitions of SNOMED CT are not so readily performed using subsumption or ECL queries.

In this scenario, the SNOMED CT map target concepts can be regarded as the minimum viable 'mapset' of concepts that will support migration or longitudinal data analyses. But to provide future utility, for integration, interoperability among and between systems and settings, and to support SNOMED CT-aware analytics, this mapset would need to be expanded to form a more comprehensive and sustainable ValueSet. Mapping is a great first step but it is not the whole journey.

Map products that contain broader or narrower relationships may influence how data is analysed and interpreted. Reports of patient data that are formulated using such maps will likely reflect at least some of those map features.

This may result in some slippage in understanding or direct comparability between reports formed directly using the old source code set and mapped data reports, and/or between mapped data reports and reports that are formulated using SNOMED CT encoded data directly. It may appear that data analyses are 'skewed' when a map product is in use. It is recommended that map performance is evaluated and compared with both retrospective and prospective analysis options to determine the impact of the map product.

Again, all of these consequences, intended or not, are choice-points that jurisdictions and users might consider when embarking upon and designing a roadmap for their mapping strategies.