Health service research definition builder: An R Shiny application for exploring diagnosis codes associated with...

Area of Science:

• Health Services Research
• Data Science
• Medical Informatics

Background:

• Administrative health data studies often define patient cohorts using procedure and diagnosis codes.
• Cohort definition transparency can be limited for non-analysts, especially with restricted data access.

Purpose of the Study:

• To develop an interactive tool for transparently defining patient cohorts using administrative health data.
• To enable non-analyst users to interrogate claims data and diagnosis code groupings.

Main Methods:

• Developed a SAS and R Shiny interactive tool named Health Services Research (HSR) Definition Builder.
• Utilized a SAS program with a tree classifier to identify diagnosis codes associated with specific medical services.
• Employed a matched, random sample comparison to identify relevant diagnosis codes.

Main Results:

• The HSR Definition Builder application generates interactive tables and graphics displaying diagnosis code summaries.
• Users can dynamically adjust inclusion/exclusion criteria to estimate claim counts for selected services.
• Demonstrated application use on US Medicare claims data for knee arthroscopy, spinal fusion, and urinalysis in patients over 65.

Conclusions:

• The tool facilitates the development of preliminary and shareable cohort definitions for administrative health data research.
• Enhances cohort definition validation by examining the frequency of associated diagnosis codes.
• Reduces the risk of incorrect code inclusion or omission in final cohort definitions.