Constrained Randomization Within Group-Randomized Trials: Implications for Comparative Effectiveness Research

• Activity Number: 187
• Type: Contributed
• Date/Time: Monday, August 2, 2010 : 10:30 AM to 12:20 PM
• Sponsor: ENAR
• Author(s): Paul J. Nietert*+ and Ruth G. Jenkins and Lynne S. Nemeth and Andrea M. Wessell and Peter M. Miller and Steven M. Ornstein
• Companies: Medical University of South Carolina and Medical University of South Carolina and Medical University of South Carolina and Medical University of South Carolina and Medical University of South Carolina and Medical University of South Carolina
• Keywords: randomization ; comparative effectiveness research ; hierarchical models

Abstract

Group randomized trials of community interventions are likely to become standard in comparative effectiveness research. These often involve groups of people, such as patients who belong to various ambulatory care practices. In such situations, covariates that may confound the intervention's impact are often measurable prior to randomization. We propose a constrained randomization algorithm to achieve covariate balance, with consideration given for the number of randomization units, number of covariates, and correlation among outcomes. The algorithm involves a trade-off between stratified randomization, which may prevent interaction effects from influencing the trial outcome but which may be impractical with large numbers of covariates, and a novel approach, which achieves marginal covariate balance but with no guarantee for balance within strata. Several applications will be presented.