LAUSR

Countless articles and textbooks have been written on the topic of missing data. In 2010, the National Research Council published recommendations for the prevention and treatment of missing data in clinical trials, which were developed by a panel of experts assembled by the U.S. Food and Drug Administration (1). However, reviews on the topic provide evidence that learning is slow and the effect of these written resources on published articles is hardly noticeable (2, 3). Many factors contribute to this problem. One solution could involve creating new mechanisms to deliver the information directly to the communities of scientists at the front lines. In our continued effort to explain analytic methods to clinical researchers in new ways (4, 5), we joined forces with Sense About Science USA to communicate concepts of missing data through graphic design. Sense About Science USA enlisted the data visualization expertise of Accurat. The product, Missing Data (http://labs.annals.org/missingdata), is an early result of our collaboration. Missing Data combines illustrations of key concepts (provided in the left pane of the screen) with explanatory text (in the right pane). Material updates as users move down the page, and users can also jump to specific sections by clicking on the outline in the center of the page. White vertical lines denote observed data items, and red forward slashes represent missing data items. This interactive Web site transitions from instructive scenarios where researchers commonly encounter missing data to descriptions of statistical concepts, such as missing completely at random, missing at random, and missing not at random, in order to set a foundation for understanding proper analytic methods. Two case studies based on published Annals articles allow users to see how this knowledge is applied in practice (6, 7). The module explores how researchers might go wrong and explains why a simple fix will not work. Missing data are ubiquitous in clinical research, and extracting valid conclusions from incomplete data sets requires thought and knowledge. We hope this interactive guide will help clinical researchers to better understand their missing data so that they can minimize lost data when possible, handle it more optimally within analyses, carry out meaningful sensitivity analysis, and involve statisticians when needed. We also hope this module stimulates new approaches to communicating statistical concepts to researchers.