We are trying to define a cohort of patients with ischemic stroke in a data source that includes EHR structured data with ICD10CM source codes as well as concepts extracted from physician narratives. To do this, the concept set expression includes concept id 443454 (cerebral infarction), since it is important to capture the ICD10CM I63* codes that map to 443454 (cerebral infarction). However, I am concerned that by including 443454 (cerebral infarction) in the concept set expression, that we could potentially include patients with a cerebral infarction that was hemorrhagic (rather than ischemic) when extracting concepts from the physician narratives.
Initially I was thinking that maybe what was needed was to get more specific mappings for some of these I63* codes. For example, I63.81 (other cerebral infarction due to occlusion or stenosis of small artery) maps to 443454 (cerebral infarction) and 321887 (disorder of artery) which are not very specific.
But often it looks like there may not be an equivalent or more specific mapping that would be completely correct. Therefore, are there ever situations where it is acceptable to make mapping exceptions (for example, mapping some of these I63* codes to a concept like āischemic strokeā) to solve for this, perhaps in a way that is similar to adding custom concept_ids? Or is there a recommended approach for how to go about solving for this?