OHDSI Home | Forums | Wiki | Github

Diagnosis completion rates in underlying EMR

Researchers
#1

Hi everyone - my site in australia ( austin health in melbourne) is undertaking our first ETL from CERNER Millenium to OMOP CDM.

I was wondering if OHDSI members have any information regarding comparative completion rates of diagnoses / problems fields in their EMR??

I am not sure if US institutions have higher rates of completion due to the high levels of audit for billing - this is not nearly such an issue in Australian public institutions although from a medical care perspective there is a lot of emphasis on accuracy and completion - nevertheless - we seem to struggle to convince our junior medical staff to use structured fields rather than free text when doing their documentation.

How big an issue is this for use of OMOP in research??

Thanks for responding

Graeme Hart

#2

Hello @GraemeHart and congratulations on joining the journey!

Completing the ETL is the largest and IMO the most important step of the journey. The data need to be accurately inserted into the CDM in a standardized format so they can be used by researchers. And having coded data is very important for research, especially Conditions. I don’t think I have seen an OHDSI study that didn’t have a Condition as part of the query.

Big is dependent on three variables. What percentage of your Condition data are uncoded? How many unique text strings does the data contain? And do you have an analyst on your team with a medical background? (MD, RN, etc.) Or can you borrow a medical coder or analyst to map these text strings to standard concept_ids?

t