Duplicate Providers?

[ericaVoss]

Duplicate Providers?

CDM or THEMIS convention?

THEMIS

Table or Field level?

TABLE

Is this a general convention?

PROVIDER table

Summary of issues

• There is conversation online on what to do when multiple values (i.e., PROVIDER_NAME, NPI, DEA, SPECIALTY_SOURCE_VALUE, etc.) are given for a second PROVIDER. Asserting the THEMIS rule that each provider only get one record is important.

Summary of answer

This is the THEMIS rule: Multiple Provider Specialities #43

In the table, each provider should appear only once; duplication is not allowed. If a provider specializes in more than one area, record their primary or most frequently practiced specialty.

Related links

• Multiple Provider Specialities #43
  • This states we can only have one record
• CDM GitHub.io PROVIDER
• Provider table
  • Seems to be a fair amount of discussion about what happens when you have multiple records for one provider with same NPI

Other comments/notes

• N/A

[jiawei-qian]

@ericaVoss, I also noticed the https://github.com/OHDSI/Themis/issues/43 issue about “Multiple Provider Specialties”. It's important to remove duplication.

I just saw one OHDSI Forum post which has Christian's reply on how to choose the appropriate specialty. If a provider has multiple specialties in different areas (pediatrics and adult), try to take out some rare outliers and find the common denominator. (for reference) https://forums.ohdsi.org/t/question-about-provider-table/1795/7

(From my understanding, Christian means to find the common ancestor for these specialties in Provider domain in ATHENA hierarchy. Because sometimes, ETL developers don't have much medical background and source data may not show us which is the primary/most frequent specialty. For example, there are only 2 source records with different specialties for the same provider. It may be hard to choose which is the primary or most frequent specialty. As a mechanism, we can also use this rule if it is hard to choose a specialty from multiple specialties especially they are in different areas.)

Hope this helps!

[clairblacketer]

Thanks @ericaVoss and @jiawei-qian. Given the wording of this convention I am going to put it in the CDM documentation since it should apply to every CDM