Classification and Regression Tree and Computer Adaptive Testing in Cardiac Rehabilitation: Instrument Validation...

Area of Science:

• Health Informatics
• Medical Assessment
• Rehabilitation Medicine

Background:

• Patient questionnaires in cardiac rehabilitation (CR) can be lengthy, increasing patient burden.
• Shorter assessments are needed to maintain data quality and patient engagement.
• Computerized adaptive testing (CAT) and classification and regression trees (CART) show promise for reducing questionnaire length.

Purpose of the Study:

• To compare the effectiveness of CAT and CART in shortening needs assessment procedure (NAP) questionnaires for CR patients.
• To evaluate if reduced questionnaire length impacts data quality in domains like anxiety, depression, quality of life, and social support.
• To determine which method, CAT or CART, is better suited for optimizing CR assessments.

Main Methods:

• Utilized data from 2837 CR patients within a multicenter, web-based Cardiac Rehabilitation Decision Support System (CARDSS).
• Assessed CAT and CART methods based on their ability to shorten the NAP questionnaire.
• Evaluated performance using sensitivity and specificity metrics.

Main Results:

• CART achieved a 72% reduction in NAP questionnaire length, compared to 36% with CAT.
• CART demonstrated higher sensitivity (0.877) and specificity (0.777) than CAT (0.817 and 0.765, respectively).
• Regression trees within CART showed lower performance (sensitivity 0.743, specificity 0.40).

Conclusions:

• Both CAT and CART can significantly shorten CR needs assessment questionnaires.
• CART is the superior method, offering greater length reduction and higher accuracy.
• Integrating CART into CR platforms like MyCARDSS can provide real-time patient progress feedback for clinicians.