Modeling and executing electronic health records driven phenotyping algorithms using the NQF Quality Data Model and...

Area of Science:

• Health Informatics
• Computational Biology
• Clinical Research Informatics

Background:

• Electronic health records (EHRs) are increasingly used, necessitating formal methods for EHR-driven phenotyping algorithms.
• The Quality Data Model (QDM) offers a standardized way to represent clinical data and algorithmic criteria from EHRs.
• Current QDM representations require manual interpretation and implementation for execution on EHRs.

Purpose of the Study:

• To investigate the use of the JBoss® Drools rules engine for automatically translating QDM criteria into executable rules.
• To develop a tool for converting QDM-defined phenotyping algorithms into Drools rules scripts for direct execution on EHR data.

Main Methods:

• Utilized the Apache Foundation's Unstructured Information Management Architecture (UIMA) platform to build a translator tool.
• Developed a framework to convert QDM criteria into executable Drools rules.
• Demonstrated the execution of these rules on real patient data from Mayo Clinic.

Main Results:

• Successfully developed and demonstrated a translator for QDM criteria to Drools rules.
• Showcased the execution of these rules on real patient data to identify cases of Coronary Artery Disease and Diabetes.
• Established a novel framework for executing QDM-modeled phenotyping criteria using a business rules management system.

Conclusions:

• The developed framework enables automatic translation and execution of QDM-based phenotyping algorithms on EHR data.
• This approach addresses the need for direct execution of phenotype definitions, reducing manual interpretation.
• This represents a significant advancement in automating EHR-driven clinical research and phenotyping.