We suspect that our clinical notes may have richer information than our coded data. So, if we want to assess potential disparities in timeliness of follow-up of documentation of pulmonary nodules, we envision the following approach:
- Create list of patients with clinical note confirming pulmonary nodules (and date of the note)
- Use that list of patients to design cohorts assessing timeliness of follow-up (e.g. characterize with/without timely follow-up, assess incidence rates, and do PLE for outcomes analyses)
Our challenge is that we have not ingested our clinical notes into OHDSI yet. Using Databricks, we can do complicated regex searches in a matter of minutes. However, until we characterize people based upon those notes, we won’t be sure whether we truly want to add that information as Observations. Plus, we refresh our OMOP data monthly, but have ad-hoc questions about our clinical notes daily, and want quick turn-around on such questions.
So, what we’re envisioning is using queries of notes to create a dummy cohort - and populate that patient list in our cohort table; and then use that dummy cohort # as part the Atlas logic.
One option that comes to mind is to have an Atlas widget that lets us pick a population from our cohort table and use them anywhere within a cohort definition (e.g. as entry events or inclusion/exclusion criteria). That way, we could use external processes to populate a list of patients and cohort start dates into our cohort table (which we could do real-time), and then enrich it with OMOP data (which we refresh less frequently).
I know this would not support network studies. However, that functionality could be quite helpful for operational analytics and rapid quality improvement efforts within our hospitals.
Are there others in the community who would see value from such an enhancement to Atlas?
Has anyone in the community found a good strategy for tackling this need?