Patient Matching Module Strategy Enhancements

Abstract

The Patient Matching Module currently provides a means of selecting patient characteristics and properties for creating a matching strategy. We have a set of statistics-gathering methods we can run on a set of data to give us what are called "field metrics". A subject matter expert can look at these metrics and tell us which fields are best for matching. With the latest work in this field, we can now predict what an expert would suggest using a random forest of decision trees.

Project Champions

Shaun Grannis

Objectives

• Adding the ability to run "field metrics" (the statistical algorithms) on the OpenMRS dataset, if this feature does not already exist.
• Adding the random forest decision tree and mechanisms for using it to determine a good strategy based on the derived field metrics.
• UI for using these features.

Extra Credit

Resources

• Project Repository:  https://github.com/GarimaAhuja/openmrs-module-patientmatching/tree/StrategyEnhancements
• Code to generate decision trees and encode them as XML trees:  https://github.com/GarimaAhuja/GettingDecisionTrees/
• Demo Video: http://www.youtube.com/watch?v=syiv3C3NT1c