Friday, June 10, 2016

#CM4H16 Patient Matching with @AdamCulby

Patient Matching on FHIR

Adam Culbertson, Innovator in Residence on Patient Matching, HIMSS

Patient matching: Comparing data from multiple sources using attributes to identify common records. ie. John Smith is the same John Smith.

Patient Matching in probabilistic matching creates a confidence score.

Data + Algorithms = Linked Data.

email is increasingly used.

Metrics for Algorithm Performance

Lack of transparency in how matching algorithms perform

Varied claims

Need beter transparency

Need reporting on Match rates

There are good metrics.

Precision / Recall

  • Are two things the same thing?

Possibilities are:
True Positive
True Negative

False Negative
False Positive

Blocking is the process of selecting matches on a particular field before a more detailed comparison is done on the full record.

FHIR is a great solution for Structure

FHIR Enabled Matching

FHIR can be used to create Synthetic data that can be tuned to specific use cases.

FHIR Search Parameters can be used to Create matching queries.

Potential to create a FHIR Extension.

Creating a Test Harness offers opportunities to build a tool to assess best matching tool for a particular use case.

via WordPress