Selecting Test Cases from the Electronic Health Record for Software Testing of Knowledge-Based Clinical Decision...

Area of Science:

• Medical Informatics
• Software Engineering
• Health Systems Research

Background:

• Software testing for knowledge-based clinical decision support systems (CDS) is critical due to high-stakes clinical applications.
• Existing testing methods are often labor-intensive, expensive, and may not ensure adequate coverage.
• Efficient test case selection is essential to balance testing thoroughness with resource constraints.

Purpose of the Study:

• To develop and demonstrate a generalizable approach for selecting test cases for knowledge-based CDS.
• To improve testing coverage while minimizing the testing burden for systems like ATHENA-CDS.
• To validate the proposed method using the ATHENA-CDS diabetes knowledgebase.

Main Methods:

• Developed a rule-based filtering approach to define system logic paths.
• Utilized a proportion heuristic to allocate test cases to identified system paths.
• Applied the method to the ATHENA-CDS diabetes knowledgebase using electronic health record data.

Main Results:

• Identified 1,086 electronic health record cases with glycemic control above target goals.
• Created 48 filters and 50 unique system paths to model system logic.
• Allocated 200 test cases across the defined paths, demonstrating comprehensive selection.

Conclusions:

• The demonstrated approach provides a generalizable and efficient method for selecting test cases for knowledge-based CDS.
• This strategy effectively mimics system logic, ensuring adequate test coverage.
• The method offers a practical solution to the challenges of testing complex clinical decision support systems.