Year: 2018
Type: dissertation
Abstract: This thesis aims to extend the existing knowledge and enhance the methodological quality of future stepped-wedge cluster randomised trial (SW-CRTs). \n \nA systematic review of published SW-CRTs shows that pre-trial sample sizes calculations display a poor standard of reporting, with little adherence to published guidelines. The methodological rigor is often substandard, with inappropriate methods often used to determine sample size. \n \nIn SW-CRTs, it is assumed that the correlation between observations is independent of the timing of them. We test the validity of this assumption by outlining a method to estimate the within-period and inter-period correlation. A case study illustrates what these correlations may look like in practice. \n \nThe impact of varying cluster size in a SW-CRT is then demonstrated by comparing a design with unequal cluster size to a design with equal cluster size. A simulation study provides evidence that the SW-CRT is affected less, on average, than a P-CRT by varying cluster size. However, the potential power in a SW-CRT with unequal cluster sizes is extremely variable. A practical method for estimated power in a SW-CRT with varying cluster size is then illustrated through a Stata function. \n
Author: James Martín
Institution: University of Birmingham
Language: English
Subfield: Statistics and Probability
Field: Mathematics
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2800707914
