Skip to main content

Table 2 Characteristics of the determinants of (and) the statistical methods used for analysing the primary outcomes in cluster trials

From: Statistical analysis of publicly funded cluster randomised controlled trials: a review of the National Institute for Health Research Journals Library

Characteristics n %
Type of follow-up RCT (N = 86)
Closed cohort follow-up 76 88
Open cohort follow-up 4 5
Cross-sectional 4 5
Repeated cross-sectional 2 2
Data type of primary outcome (N = 100)
Continuous 63 63
Binary 28 28
Counts 5 5
Time to event 2 2
Percentage 2 2
Method of adjusting for clustering (N = 100)
Cluster-level analysis:   
 Standard generalized linear model 2 2
Individual-level analysis:   
 Generalized linear mixed model 80 80
 Robust standard errors 7 7
 Generalized estimating equations 6 6
Clustering not accounted for:   
 Statistical hypothesis test statistic—chi-square 1 1
  Standard generalized linear model 4 4
Specific statistical model (N = 100)
Linear regression 57 57
Logistic regression 25 25
Analysis of covariance 6 6
Relative sensitivity 1 1
Negative binomial regression 2 2
Analysis of proportions 1 1
Cox Proportional Hazards model 2 2
Poisson regression 4 4
Weibull regression model 1 1
Chi-square test 1 1
Random component of GLMM (N = 80)
Random intercept 76 95
Shared frailty 1 1
Random intercept and slope (repeated measures) 3 4
Correlation structure in GEE (N = 6)
Exchangeable correlation 5 83
Correlation structure not reported 1 17
  1. N = total number of trials; n = counts observed in each level of a category; RCT = randomised controlled trial; GLMM = generalized linear mixed model; GEE = generalized estimating equations. Not reported means that the information of interest was not considered and/or provided in the trial