Please explain the "then within" part of the cohort criteria.

[kaicode]

Sharing a question from someone in the community:

I was just reviewing the slides from the data analytics demonstrator & was wondering about the background to the cohorts model and where the “then within” parts come in?

[kaicode]

The data model is quite simple but it does include a date in the clinical encounter.
https://github.com/IHTSDO/health-data-analytics#data-model

The "then within" part of the cohort criteria requires the clinical encounter be within a number of days since the previous clinical encounter in the criteria. For example a cohort selection criteria says "Tumor necrosis factor alpha inhibitor product" then within "180" days "Bacterial pneumonia" then only patients who have a lung infection within six months of being prescribed the TNF drug will be included in the cohort.

The idea is that this temporal criteria can help filter out irrelevant cases. For example when assessing the effect of a drug prescription on the chance of disorder occurring you would not want to include patients who got a the disorder before the drug was prescribed or, potentially, those who got a disorder many years after the drug was prescribed.

There are plenty of ways the data model could be improved but we feel this is a start in the right direction and enough to demonstrate the potential for powerful data analysis using SNOMED CT.

Here is an example of adding time between clinical encounters in the demonstration data generation logic: https://github.com/IHTSDO/health-data-analytics/blob/1.1.0/src/main/java/org/snomed/heathanalytics/ingestion/exampledata/ExampleDataGenerator.java#L105

Kind regards,
Kai