When a new version of an established score or scale is released, it should be modelled as a separate archetype, and not as a specialisation. By “new version” we mean a non backwards compatible revision that isn’t just a correction of the previous version. For example: NEWS → NEWS2, AVPU → ACVPU
When there’s expected to be a series of versions of a tool, consider modelling the archetype as a generic structure without the details that change between versions. For example: TNM