Hello,
I am looking for an enhancement to Usagi that enables the user to choose whether Usagi looks at two columns or one column when matching to concept ids. Currently, Usagi just looks at the source code to do the matching piece to concept ids. However, there are times when I need it to look at two columns. In this case, these are observation data that involve question/answer value pairs. Examples of these are here:
Most of these had to be mapped manually because Usagi only took the source_code into consideration where both the source_code and source term are needed to make a match. I am asking for the option (through a user prompt) to have ‘additional information’ column included in the matching algorithm in additional to the source code when needed. So the source term you see in the examples above would go in the additional information column for matching purposes. However, it would also need to be in the source term column so it gets exported with the current stcm file and can be loaded into source_to_concept_map table for linking. So the ETL join would be observation_source_value=source_code and value_as_string=source term.
I am not in favor of concatenating the source code and source term fields into one string eg ‘ALLERGEN_DESC CODEINE’ , ‘ALLERGEN_DESC PENICILLIN G’ , ‘SOCIAL ADL: SLEEP CONCERN NO’ in the source code for the following reasons:
- readability. I and users have found these long strings hard to read especially the longer the allergen name and family history measures like ‘Family History - Parkinson’s Decease’
- better ETL performance eliminating need to join on string columns that are too long
- ) more efficient querying without having to use like statements to find a particular measure. Currently, it is much cleaner and efficient when you can search for observation_source_value=’ 'ALLERGEN_DESC’ and value_as_string=’CODEINE’ because we have numerous history metrics we are putting in observation.
- Plus it will cause more manual work to get this format with me having to separate into source_code and source_name for it to read cleaner after the mapping is done, create that new file, and then create another source to concept map. Concatenating it would be like trying to parse note text in queries.