2019
DOI: 10.1002/sim.8089
|View full text |Cite
|
Sign up to set email alerts
|

Accounting for a decaying correlation structure in cluster randomized trials with continuous recruitment

Abstract: mate the required sample size under most trial configurations likely to occur in practice. Planning of CRTs requires consideration of the most appropriate within-cluster correlation structure to obtain a suitable sample size. KEYWORDS clinical trial design, cluster randomized trial, crossover design, sample size, stepped wedge design, within-cluster correlation 1918

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

0
68
0

Year Published

2019
2019
2024
2024

Publication Types

Select...
6

Relationship

3
3

Authors

Journals

citations
Cited by 42 publications
(68 citation statements)
references
References 19 publications
0
68
0
Order By: Relevance
“…For each of these correlation structures, the vector of cluster‐period means forms a sufficient statistic for the treatment and period effects, hence collapsing to cluster‐period means results in no loss of information, and Model can be written as trueY¯skt=1mtruei=1mYskti=βt+()θ+UskXst+CPskt+ηsk+ϵskt,1emηskNfalse(0,ση2false/mfalse),1emϵsktNfalse(0,σϵ2false/mfalse), and it is this model that we consider in the remainder of the work. If trueY¯sk=()trueY¯sk1,,trueY¯skTT, then varfalse(trueY¯skfalse)=()ITXsV()ITXsT+ση2mJT+σϵmIT=VCP+σPU-2pt()Xs1T+bold1XsT-2pt+σU2XsXsT+ση…”
Section: Model For Continuous Outcomes and Information Content Of Cellsmentioning
confidence: 99%
See 2 more Smart Citations
“…For each of these correlation structures, the vector of cluster‐period means forms a sufficient statistic for the treatment and period effects, hence collapsing to cluster‐period means results in no loss of information, and Model can be written as trueY¯skt=1mtruei=1mYskti=βt+()θ+UskXst+CPskt+ηsk+ϵskt,1emηskNfalse(0,ση2false/mfalse),1emϵsktNfalse(0,σϵ2false/mfalse), and it is this model that we consider in the remainder of the work. If trueY¯sk=()trueY¯sk1,,trueY¯skTT, then varfalse(trueY¯skfalse)=()ITXsV()ITXsT+ση2mJT+σϵmIT=VCP+σPU-2pt()Xs1T+bold1XsT-2pt+σU2XsXsT+ση…”
Section: Model For Continuous Outcomes and Information Content Of Cellsmentioning
confidence: 99%
“…For each of these correlation structures, the vector of cluster-period means forms a sufficient statistic for the treatment and period effects, 4 hence collapsing to cluster-period means results in no loss of information, and Model (1) can be written asȲ…”
Section: Model For Outcomes Incorporating Treatment Effect Heterogeneitymentioning
confidence: 99%
See 1 more Smart Citation
“…Recent papers have discussed time as a continuous phenomenon in longitudinal cluster randomized trials, where participants have outcomes that are measured in continuous time, rather than at a set of discrete time points common to all participants. Grantham et al discussed within‐cluster correlation structures in the context of continuous time, and Hooper and Copas discussed the need to clarify sampling schemes and the terminology used to refer to specific sampling schemes. If participants can have their observations recorded at any time, rather than at a set of discrete times common to all participants, then the correlation structures we have assumed for cluster‐level random effects are not likely to be satisfactory: these correlation structures imply that all participants within a period are exchangeable, and that any pair of participants measured in a period are more highly correlated than any pair of participants measured in distinct periods.…”
Section: Discussionmentioning
confidence: 99%
“…We consider models with Y kti that are “within‐period exchangeable”: that is, reordering the Y kti within periods leads to the same distribution for the vector of Y kti . The theorem in the Supplementary Appendix of Grantham et al then states that cluster‐period means Ykt=1mi=1mYkti are a sufficient statistic for θ. Collapsing to cluster‐period means, Ykt=1mi=1mYkti, gives: trueYkt=βt+θXkt+Ck+CPkt+ηk+ϵkt,0emηkNfalse(0,ση2false/mfalse),1emϵktNfalse(0,σϵ2false/mfalse),1emCkNfalse(0,σC2false),1emCPktNfalse(0,σCP2false). …”
Section: Sample Size Formulas For Open Cohort Longitudinal Cluster Ramentioning
confidence: 99%