Analysis and reporting of stepped wedge randomised controlled trials: synthesis and critical appraisal of published studies, 2010 to 2014

Calum Davey; James Hargreaves; Jennifer A Thompson; Andrew J Copas; Emma Beard; James J Lewis; Katherine L Fielding

doi:10.1186/s13063-015-0838-3

Table 2 Primary outcome analysis methods

From: Analysis and reporting of stepped wedge randomised controlled trials: synthesis and critical appraisal of published studies, 2010 to 2014

Author, year	Model fitted	Effect estimated	Vertical and/or horizontal analysis	Method to account for clustering	Repeated measures on individuals? Method used	Adjustment for confounding?	Adjustment for secular trends	Risk of lag	Accounting for lags	Risk of fidelity loss
Bacchieri, 2010 [29]	Poisson regression	Rate ratio	Both	Robust standard errors	Time-to-event	No	Fixed effect (no interaction)	None	None	Minor
Bashour, 2013 [22]	Linear regression	Mean difference	Both	Random effects	No	Yes	Fixed effect	None	None	Major
Bashour, 2013 [22]	Linear regression	Mean difference	Both	Random effects	No	Yes	(no interaction)	None	None	Major
Durovni, 2013 [12]	Cox regression	Rate ratio	Within-step only, plus frailty	Random effects	Time-to-event	Yes	Conditional	Major	None	Minor
Fuller, 2012 [23]	Logistic regression	Odds ratio	Both	Random effects	Yes, not explicitly accounted for	Yes	Fixed effects	Major	None (per-protocol analysis using observed lags)	Major
Fuller, 2012 [23]	Logistic regression	Odds ratio	Both	Random effects	Yes, not explicitly accounted for	Yes	(no interaction)	Major	None (per-protocol analysis using observed lags)	Major
Gruber 2013 [21]	Linear regression, Log-Poisson regression	Risk difference, risk ratio	Both	Robust standard errors, GEE	Yes, not explicitly accounted for	Yes, in secondary analyses	Fixed effect	Major	None	Low
Gruber 2013 [21]	Linear regression, Log-Poisson regression	Risk difference, risk ratio	Both	Robust standard errors, GEE	Yes, not explicitly accounted for	Yes, in secondary analyses	(no interaction)	Major	None	Low
Horner, 2012 [27]	Logistic regression	Odds ratio	Both	Random effects	Yes, not explicitly accounted for	Yes	Fixed effect	Minor	None	Minor
Horner, 2012 [27]	Logistic regression	Odds ratio	Both	Random effects	Yes, not explicitly accounted for	Yes	(no interaction)	Minor	None	Minor
Mhurchu, 2013 [24]	Logistic regression	Odds ratio	Both	Random effects	Yes, pupil-level summaries and random effect	Yes	Fixed effect	None	None	Minor
Mhurchu, 2013 [24]	Logistic regression	Odds ratio	Both	Random effects	Yes, pupil-level summaries and random effect	Yes	(no interaction)	None	None	Minor
Roy, 2013 [25]	Logistic regression	Odds ratio	Both	Random effects	No	Yes	None	None	None	Minor
Schultz, 2014 [28]	Logistic regression	Odds ratio	Both	None: GEE not possible, and RE did not converge	No	No	Fixed effect	Minor	None	None
Schultz, 2014 [28]	Logistic regression	Odds ratio	Both	None: GEE not possible, and RE did not converge	No	No	(no interaction)	Minor	None	None
Stern, 2014 [26]	Linear regression	Mean difference	Both	Random effects	Yes, multiple level random effects	Yes	Fixed effect with interaction	Major	None	Minor

Back to article page