Optimal Allocation for Comparison of Proportions and Robust Designs in Cluster-Randomized Trials

• Activity Number: 138
• Type: Contributed
• Date/Time: Monday, August 10, 2015 : 8:30 AM to 10:20 AM
• Sponsor: WNAR
• Author(s): Sheng Wu* and Weng Kee Wong and Catherine Crespi
• Companies: UCLA and UCLA and UCLA
• Keywords: cluster randomized trials ; robust design ; efficient design ; binary outcome ; sampling ratio

Abstract

Cluster randomized trials (CRTs) are increasingly used for research in many fields including public health, education, social studies and ethnic disparity studies. Equal allocation designs are often used in CRTs but they may not be efficient when cost consideration is taken into account. In this paper, we consider two-arm CRTs with binary outcomes and develop various optimal allocation strategies for comparison of risk difference, relative risk and odds ratio. Optimal allocation is dependent on unknown success probability and intra correlation coefficient (ICC). When those parameters are misspecified, investigators may obtain incorrect allocation. In order to construct robust designs which are not sensitive to misspecification of those parameters, a maximin and a Bayesian method are provided. We show that robust designs usually have higher efficiency than equal allocation designs in CRTs.