- Oral presentation
- Open Access
Assessing sensitivity to assumptions in mixed effects analyses of stepped-wedge trials
© Davey and Thompson 2015
- Published: 16 November 2015
- Public Health
- Sensitivity Analysis
- Weighted Average
- Simulated Data
- Mixed Effect
Stepped-wedge trials are used to evaluate the impact of interventions. Researchers often use mixed effects regression to estimate effects. This method includes within-cluster - non-randomised - comparisons that requires assumptions about the secular trends.
We simulated data from stepped-wedge trials with different characteristics. We analysed these data using a within-step only approach and mixed effects regression, and evaluated their performance. The within-step only approach preserves randomisation by combining estimates of effect from within steps using a weighted average; we used non-parametric bootstrapping to generate inferential statistics. We introduced violations of the mixed effects model assumptions and investigated the effects on the two methods.
When the assumptions were met, the mixed effects method was more sensitive and specific than the within-step approach. Bias was introduced to the mixed effects results by interaction of the secular trend with the clusters, and with the intervention. The within-step approach remained unbiased even in extreme violations of these assumptions. Comparing the mixed effects estimate of effect with the within-step estimate helped identify violations of the assumptions.
We confirmed that mixed effects methods are more powerful than a within-step method when assumptions are met. Moderate to severe violations of assumptions led to bias, supporting the need for clear reporting standards and sensitivity analysis for stepped-wedge trials. Estimating the within-step effect can be useful for identifying bias.
Within-step analyses that preserve the randomisation should be used as a diagnostic to assess the validity of common mixed effects methods for analysing stepped-wedge trials.
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/4.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.