Coronary Artery Disease Phenotype Detection in an Academic Hospital System Setting

Area of Science:

• Cardiology
• Medical Informatics
• Genetics

Background:

• Coronary artery disease (CAD) poses a significant health burden in the U.S., particularly in West Virginia.
• Undiagnosed familial hypercholesterolemia (FH) is a key contributor to CAD.
• Identifying a CAD phenotype is crucial for detecting FH families.

Purpose of the Study:

• To develop and validate an algorithm for detecting CAD phenotypes using discrete data elements from electronic health records (EHRs).
• To assess the algorithm's accuracy and sensitivity across diverse patient cohorts.

Main Methods:

• Developed an algorithm (CAD Algorithm I) to search EHR data, including diagnosis codes (ICD-9/10) and procedure codes (CPT).
• Applied the algorithm to two distinct patient cohorts (n=500 each).
• Systematically reviewed EHRs to confirm CAD diagnoses, refined the algorithm (CAD Algorithm II), and re-evaluated its performance.

Main Results:

• CAD Algorithm I demonstrated 89.6% accuracy and 94.6% sensitivity.
• The revised CAD Algorithm II achieved 92.4% accuracy and 98.2% sensitivity in group 1, and 93% accuracy in group 2.
• Specific diagnostic and procedural codes proved more effective than generic terms for phenotype detection.

Conclusions:

• CAD Algorithm II accurately and sensitively detects CAD on a large scale across varied patient populations.
• This algorithm can be utilized for patient registry monitoring within EHR systems.
• It serves as a valuable tool for identifying individuals with potential FH.