Analysing cluster randomised controlled trials using GLMM, GEE1, GEE2, and QIF: results from four case studies

Table 3 A summary of the results obtained from fitting the different statistical models to the PoNDER trial data (N = 2659)

^aModel adjusted for EPDS score at 6 weeks, living alone (no or yes), previous history of major life events (no or yes), and any previous history of postnatal depression (no or yes). Note that SE Standard error, CI Confidence interval, ICC Intracluster correlation coefficient. GLMM Generalized linear mixed model, mGLM Marginal generalized linear model, GEE Generalized estimating equations, QIF Quadratic inference function
1. EPDS score at 6 months postnatally. The EPDS is scored on a 0 to 30 scale with higher scores indicating a greater risk of PND
2. Dichotomised EPDS score at 6 months postnatally of < 12 or \(\ge\) 12
3. The intervention effect for the continuous outcome is the difference in the mean 6-month EPDS scores between the intervention and control groups; with a negative mean difference favouring lower scores (better outcomes) in the intervention group. The intervention effect for the binary outcome is the odds ratio for an EPDS score of 12 or more in the intervention group compared to the control group with an odd ratio < 1 favouring better outcomes (lower odds of PND) in the intervention group

ISSN: 1471-2288