OHDSI Home | Forums | Wiki | Github

Classification type drug groups in source data

We’re working to map data from global hip fracture registries and one of the key research points is the use of bone protective medication/osteoprotectives both before and after admission.

Some registries don’t have any further detail in the source other than bone protection meds given/not given (which only links to classification concept 35807481), and some provide slightly more detail but still in ‘classification’ form (e.g bisphosphonates or ‘calcium and/or vitamin d’)

I’m assuming there’s not much to be done for cases where the responses are more generic, but wanted to ask just in case anyone has any ideas!


Unfortunately, that is correct. Registries can be more narrow or richer, but they inherently capture only what they were designed at the beginning to capture. If new questions come up requiring other data - too bad. In that way they are similar to clinical trials - they can only answer one or two questions.

The classifications and whether they can be recorded as data - we have been struggling with that for a while. The reason we don’t allow it is that we have alternative classifications of the same thing, which defeats the idea of a unique concept. For example, you could use the concept of “opioids” from VA Class, ATC, SNOMED or ETC. They contain slightly different drugs. But which one is correct? The registry will not tell you. Therefore, classifications are not a precise reproduction of reality.

1 Like