What is the rationale behind the inclusion of concept record counts in ATLAS? When we update the record counts using Achilles, a lot of record counts remain zero. For example the episode_object_concept_id. It seems that all of these concept id’s return a record count of zero, although they are available in the source database. Also, if not via Achilles, how can we retrieve record counts for the concepts that are not included by default?
Let’s say you’re trying to define a cohort and you have done some literature review and you’ve built a code-list that can identify the patient records for the condition of interest. You talk to your peer and review the code list, and there’s a disagreement about one code or another. Maybe it’s too specific, or maybe it’s a vague code (classic ICD9 x.8 codes). You fight and fight and fight over one of the codes…
If this code has 0 record count in your data, is it something you need to worry about? The record counts are used to understand ‘presence’ of the code in data. The exact number isn’t as important as understanding the magnitude (0, 100s, 1000s, or 1000000s).
The use of episode_object_concept_id came much later than the idea of record counts, where the focus of record counts was about what was being captured by a EHR system or if codes are being used at all in the data.