Intra-cluster correlation cut-off values for selecting the stepped-wedge cluster randomized trial
Open Access
Author:
Ma, Yining
Graduate Program:
Public Health Sciences
Degree:
Master of Science
Document Type:
Master Thesis
Date of Defense:
March 14, 2019
Committee Members:
Vernon Michael Chinchilli, Thesis Advisor/Co-Advisor
Keywords:
stepped-wedge cluster randomized trial intra-cluster correlation cluster randomized trial linear mixed-effects model simulation study
Abstract:
The stepped-wedge cluster randomized trial is an alternative to a parallel cluster randomized trial. Compared to a parallel cluster randomized trial, a stepped-wedge design refers to the randomization of one or more clusters transitioning from the control to the intervention at regularly-spaced time intervals. This paper explores the smallest value of the intra-cluster correlation for selecting a stepped-wedge cluster randomized trial instead of a parallel cluster randomized design. We generated data to attain 1,000 datasets for each simulation scenario. We designated nine time periods for the stepped-wedge stages. We adapted the same set of 1,000 simulated datasets for the parallel cluster randomized trials by randomizing four control and four intervention groups. We allowed the intra-cluster correlation to range from 0.01 to 0.20 to compare statistical power between the stepped-wedge cluster randomized design and the parallel cluster randomized design. Our results show that the stepped-wedge cluster randomized design can yield higher statistical power than the parallel cluster randomized design when the intra-cluster correlation coefficient exceeds 0.02.
