Analysis of Rolling Group Therapy Data Using Conditionally Autoregressive Priors

• Activity Number: 293
• Type: Topic Contributed
• Date/Time: Tuesday, August 3, 2010 : 8:30 AM to 10:20 AM
• Sponsor: Health Policy Statistics Section
• Author(s): Susan Paddock*+ and Sarah Hunter and Katherine Watkins and Daniel F. McCaffrey
• Companies: RAND Corporation and RAND Corporation and RAND Corporation and RAND Corporation
• Keywords: conditionally autoregressive prior ; Bayesian methods ; hierarchical modeling ; substance abuse treatment

Abstract

Group therapy for alcohol and other drug use disorders is often delivered under a rolling admissions policy, where new clients are continuously enrolled into the group as space permits. Despite the ubiquity of rolling admissions in practice, little guidance on the analysis of such data is available. We discuss the limitations of previously proposed approaches and improve upon them by fully modeling the interrelatedness of client outcomes using a hierarchical Bayesian model that assumes a conditionally autoregressive prior for session-level random effects. We demonstrate improved performance using our method with respect to coverage probabilities for parameter estimates and the enhanced ability to learn about the complex dynamics of rolling therapy groups. Our approach broadly applies to any group therapy setting where groups have changing client composition.