Comparing Statistical Models Used in Analysis of a Discrete Choice Experiment

• Activity Number: 675
• Type: Contributed
• Date/Time: Thursday, August 5, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Ji Cheng*+ and Lehana Thabane and Deborah Marshall and John K. Marshall
• Companies: St. Joseph's Healthcare Hamilton and McMaster University and University of Calgary and McMaster University
• Keywords: discrete choice experiment ; patient preference ; cancer screening ; model comparison

Abstract

Increasingly, discrete choice experiments (DCE) have been applied in health research to evaluate stakeholders' preferences among alternative programs or treatments. This study compares four commonly applied statistical methods including multinomial logit, nested logit, multinomial probit and mixed logit models using preference data on colorectal cancer screening (CRCS) collected from patients in Ontario. Patients preferences were obtained for CRCS among six alternatives (virtual colonoscopy, colonoscopy, barium enema, sigmoidoscopy, fecal DNA test and fecal occult blood) based on five key attributes (process, pain, preparation, sensitivity and specificity and cost), and screening up-take rate was estimated when the preferred choice and screening opt-out choice were both available. The results from above models are compared empirically on the consistency and robustness.