There's a requirement to model scales - e.g. pain scales - in the clinical domain using DV_QUANTITY with a 'qualified real' property, but where units do not seem appropriate. So in this case, it would be modelled with a real and no units or blank unit.
However, the openEHR specification requires all DV_QUANTITY to have units and the units must not be blank, which will not meet the above requirement.
How does having DV_SCALE influence this task? Within Better there is a group of people claiming we absolutely need DV_NUMERIC or something like that, because using DV_QUANTITY makes no semantic sense and using unit '1' is cumbersome, but they have yet to produce a real use-case that needs it, apart from ratios (which are quantities and indeed have no unit because it cancels out), so I am curious about opinion of other SEC members and experts.
We have real needs to score a number that is not a ratio, without a unit. This is mainly for scores - and we have quite a lot of those. Some of these could be solved with a DV_SCALE, because they are a number with a range with a clear distinct meaning per range. Others are just a number, for a human to interpret the meaning of.
The problem of DV_SCALE for these kinds of things is that it does not have precision, so we cannot as easily say ‘this is a scale with a precision of 2, and every real number between 1 and 3 means …’., so DV_SCALE would currently mainly work well for predefined values, as ‘exactly 1.5 means ..’ and ‘exactly 2.3 means …'.
We currently use DV_QUANTITY with units “1” - and it works. Our tooling takes away the cumbersomeness by offering a dedicated option for a quantity without a unit, which then automatically becomes “1”. In these cases we use a separate DV_CODED_TEXT in case there is also a symbol attached.
With a very limited set of scores there’s also the possibility that there are several ways/variants to interpret a given score, so a DV_SCALE may not be of use there because of the fixed relation between a range and the symbol - but I guess that could be solved with specialisations.
There are definitely places where a simple real number is required, and that is why ' Qualified real' i.e “1” (which comes from UCUM) exists.
One simple option would be to drop the units mandation, such that an empty units is assumed to be equivalent to “1” i.e a simple unitless real number. Given that we are no longer tied to UCUM that might make sense.
and by definition, DV_SCALE does involved pre-defined and fixed numerics.
I like Ian’s suggestion to make the units optional.