Missing data and convenient assumptions.

T his issue contains the first of several planned statistical primers on methodological problems commonly encountered by outcomes researchers. Although several resources exist for the interested outcomes researcher, they are often scattered throughout different literatures and, in particular, are not translated to the " outcomes " setting. Much of the empirical basis of outcomes research involves observational data—a setting in which the researcher has no control over what interventions are given and why they are given—yet desires to understand the causes of the interventions. Many clinicians are trained to understand the results of clinical trials in which randomization is used to quantify the efficacy of medical treatments. However, outcomes research not only involves assessing the results of randomized trials to make clinical decisions but also assessing findings from observational studies of medical interventions and policies, financial arrangements, organizational systems, and combinations thereof. A solid understanding of the statistical methods used in outcomes research to aid in drawing conclusions is therefore required. The objectives of the Statistical Primer Series are several-fold. A key goal is to introduce statistical techniques of particular relevance to observational data analyses. Second, the series aims to provide an enhanced understanding of analytical methods specifically within the context of outcomes research. In this setting, representative examples of common problems will be used to illustrate methods. Third, because of the natural complexity of the empirical data that forms the basis of outcomes research, the series will be used to introduce new statistical techniques. Thus, an additional goal of the series is to expose outcomes researchers to modern statistical techniques faster, with the hope of increasing their diffusion into the clinical literature. Several methodological areas have been selected for the forthcoming articles in the Statistical Primer Series. Because of the prominent role causality plays in outcomes research, Dr Sue Marcus and colleagues (United States) describe founda-tional principles of causal inference. The key task in casual inference is to establish causation—to determine what would happen to a specific patient under different treatment options. Accomplishing this task requires a clear understanding of what constitutes a causal risk factor and after identifying such a risk factor, an approach to establishing with reasonable assurance its effect on patient outcomes in the observational setting. Outcomes researchers face particularly challenging tasks, not only because of the lack of randomization but also because of the added complexity attributed to the nature of the interventions. For example, …