Hi, Builders,
Question : we’re working on an ETL with NLP information about diagoses, which has the following columns:
term,
location,
attribute,
sentiment
The term,location and attribute is pretty simple, it would have things like ‘cancer’ in term, ‘liver’ in location, and some attribute of the observation in attribute, so we can use those information to create a mapping to concepts.
The question I have involves the use of the ‘sentiment’ column. The sentiment column has values such as ‘have’, ‘have.not’, ‘suspected’, ‘diagnosed’, ‘complain’, and others.
We have decided to make the ‘positive sounding’ sentiments (have, confirmed, diagnosed, etc) as condition_occurrence records in the cdm. The specific value I’d like to ask the community is the sentiment of ‘complain’.
Do you think that we should use a note that describes the patient that ‘complain about term X’ as a condition_occurrence (ie: they definitely had the condition and deserve a diagnosis code of it) or is ‘complain’ sentiment not strong enough to consider it an actual diagnosis?
-Chris