A cluster-randomised trial of a multifaceted quality improvement intervention in Brazilian intensive care units (Checklist-ICU trial): statistical analysis plan

作者：Damiani Lucas P^*; Cavalcanti Alexandre B; Moreira Frederico R; Machado Flavia; Bozza Fernando A; Salluh Jorge I F; Campagnucci Valquiria P; Normilio Silva Karina; Chiattone Viviane C; Angus Derek C; Berwanger Otavio; Chang Chung Chou H

来源：Critical Care and Resuscitation, 2015, 17(2): 113-121.

Background: The Checklist During Multidisciplinary Visits for Reduction of Mortality in Intensive Care Units (Checklist-ICU) trial is a pragmatic, two-arm, cluster-randomised trial involving 118 intensive care units in Brazil, with the primary objective of determining if a multifaceted quality-improvement intervention with a daily checklist, definition of daily care goals during multidisciplinary daily rounds and clinician prompts can reduce in hospital mortality.
Objective: To describe our trial statistical analysis plan (SAP).
Methods: This is an ongoing trial conducted in two phases. In the preparatory observational phase, we collect three sets of baseline data: ICU characteristics; patient characteristics, processes of care and outcomes; and completed safety attitudes questionnaires (SAQs). In the randomised phase, ICUs are assigned to the experimental or control arms and we collect patient data and repeat the SAQ.
Results: Our SAP includes the prespecified model for the primary and secondary outcome analyses, which account for the cluster-randomised design and availability of baseline data. We also detail the multiple mediation models that we will use to assess our secondary hypothesis (that the effect of the intervention on inhospital mortality is mediated not only through care processes targeted by the checklist, but also through changes in safety culture). We describe our approach to sensitivity and subgroup analyses and missing data.
Conclusion: We report our SAP before closing our study database and starting analysis. We anticipate that this should prevent analysis bias and enhance the utility of results.

• 出版日期2015-6