Department of Biostatistics Seminar/Workshop Series

Efficient Use of EHR for Biomedical Translational Research

Tianxi Cai, ScD

Professor of Biostatistics, Harvard T.H. Chan School of Public Health

Professor of Biomedical Informatics, Harvard Medical School

Most researchers agree the randomized controlled trial (RCT) is the optimal study design in determining efficacy of a given intervention, and this is largely due to the realization that random allocation allows for the general comparability across study arms. The expected level of imbalance for any and all potential baseline confounders is zero, but any single clinical trial will exhibit some level imbalance in these baseline variables. Further, under certain scenarios, seemingly trivial levels of baseline variable imbalance may still impact statistical inference and study result interpretation; these issues may be mitigated through careful study planning with regard to these factors. Ideally, according to prominent literature and guidelines, these influential variables should be considered (1) in the design phase of any given trial with the use of covariate involvement in the treatment allocation scheme, and (2) in the analysis phase since adjustment may increase precision and decrease bias in intervention effect estimates. This talk will briefly review reasoning behind these guidelines, illustrate current practice in published randomized controlled trials through the results of a systematic review, and present a complex case study of a cluster-randomized controlled trial. The complexity of the clustered design coupled with logistical implementation constraints have the potential to magnify confounding issues in such studies. Thus, while careful consideration of baseline variables in design and analyses is important in any RCT, it is even more vital in the cluster-randomized setting. Anecdotal evidence suggests that the apparent gaps between theory and practice may be due to lack of education, poor planning or budget considerations, and/or practical constraints. Ultimately, successful design and analyses of any study will involve a compromise between theory/guidelines and practicality.
Topic revision: r1 - 12 Apr 2017, AshleeBartley

This site is powered by FoswikiCopyright © 2013-2022 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding Vanderbilt Biostatistics Wiki? Send feedback