Knowledge Translation in Statistical Methods: Applied Myth-Busting in the Field of Missing Data

• Activity Number: 506
• Type: Invited
• Date/Time: Wednesday, August 12, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Biometrics Section
• Author(s): Melanie L. Bell* and Mallorie Fiero and Nicholas Horton and Chiu-Hsieh Hsu
• Companies: The University of Arizona and The University of Arizona and Amherst College and The University of Arizona
• Keywords: Differential dropout ; knowledge translation ; missing data ; RCT

Abstract

Despite a rich literature on missing data and dropout in the field of statistics, there are gaps in translation of this knowledge to applied statisticians and researchers. We carried out surveys of randomized controlled trials and cluster RCTs to assess approaches to missing data and sensitivity analysis. Missing data were common, and statistical approaches, largely assuming MCAR, for RCTs have not changed much, when compared to a similar review from 10 years ago. The knowledge translation gap may be responsible for the persistence of missing data myths. Myth 1: if attrition rates in longitudinal clinical trials are similar between study arms, there is no need for concern about bias. Myth 2: if attrition rates are dissimilar between study arms, then results will be biased. We show that these myths are false using a simulation study. Statisticians doing statistical research need to consider how to 1) answer questions of import to applied statisticians and researchers, and 2) translate their findings to improve applied research.