Design patterns for the development of electronic health record-driven phenotype extraction algorithms.

Area of Science:

• Biomedical Informatics
• Software Engineering
• Ontology Development

Background:

• Design patterns offer generalized solutions for recurring problems in software development and ontologies.
• Existing biomedical literature lacks generalized design patterns for phenotype algorithm development.

Purpose of the Study:

• To generalize common approaches in phenotype algorithms into reusable design patterns.
• To provide a catalog of design patterns for the informatics community.

Main Methods:

• Reviewed 24 electronic Medical Records and Genomics (eMERGE) phenotypes from the Phenotype KnowledgeBase (PheKB).
• Identified and generalized recurrent elements within these algorithms into candidate patterns.
• Validated candidate patterns through author consensus and attribute annotation.

Main Results:

• Identified 21 distinct phenotyping patterns from the reviewed algorithms.
• These patterns are available as an online data supplement for community use.

Conclusions:

• Repeatable patterns exist within phenotype algorithms.
• Codifying these patterns can educate algorithm developers, improving algorithm development time, portability, and accuracy.