Despite the growing popularity of propensity score (PS) methods in epidemiology, relatively little has been written in the epidemiologic literature about the problem of variable selection for PS models. The authors present the results of two simulation studies designed to help epidemiologists gain insight into the variable selection problem in a PS analysis. The simulation studies illustrate how the choice of variables that are included in a PS model can affect the bias, variance, and mean squared error of an estimated exposure effect. The results suggest that variables that are unrelated to the exposure but related to the outcome should always be included in a PS model. The inclusion of these variables will decrease the variance of an estimated exposure effect without increasing bias. In contrast, including variables that are related to the exposure but not to the outcome will increase the variance of the estimated exposure effect without decreasing bias. In very small studies, the inclusion of variables that are strongly related to the exposure but only weakly related to the outcome can be detrimental to an estimate in a mean squared error sense. The addition of these variables removes only a small amount of bias but can increase the variance of the estimated exposure effect. These simulation studies and other analytical results suggest that standard model-building tools designed to create good predictive models of the exposure will not always lead to optimal PS models, particularly in small studies.

If confirmed, these results suggest that conventional antipsychotic medications are at least as likely as atypical agents to increase the risk of death among elderly persons and that conventional drugs should not be used to replace atypical agents discontinued in response to the FDA warning.

Doubly robust estimation combines a form of outcome regression with a model for the exposure (i.e., the propensity score) to estimate the causal effect of an exposure on an outcome. When used individually to estimate a causal effect, both outcome regression and propensity score methods are unbiased only if the statistical model is correctly specified. The doubly robust estimator combines these 2 approaches such that only 1 of the 2 models need be correctly specified to obtain an unbiased effect estimator. In this introduction to doubly robust estimators, the authors present a conceptual overview of doubly robust estimation, a simple worked example, results from a simulation study examining performance of estimated and bootstrapped standard errors, and a discussion of the potential advantages and limitations of this method. The supplementary material for this paper, which is posted on the Journal's Web site (http://aje.oupjournals.org/), includes a demonstration of the doubly robust property (Web Appendix 1) and a description of a SAS macro (SAS Institute, Inc., Cary, North Carolina) for doubly robust estimation, available for download at http://www.unc.edu/~mfunk/dr/.

No abstract

Summary Instrumental variable (IV) methods have been proposed as a potential approach to the common problem of uncontrolled confounding in comparative studies of medical interventions, but IV methods are unfamiliar to many researchers. The goal of this article is to provide a non-technical, practical introduction to IV methods for comparative safety and effectiveness research. We outline the principles and basic assumptions necessary for valid IV estimation, discuss how to interpret the results of an IV study, provide a review of instruments that have been used in comparative effectiveness research, and suggest some minimal reporting standards for an IV analysis. Finally, we offer our perspective of the role of IV estimation vis-à-vis more traditional approaches based on statistical modeling of the exposure or outcome. We anticipate that IV methods will be often underpowered for drug safety studies of very rare outcomes, but may be potentially useful in studies of intended effects where uncontrolled confounding may be substantial.

Patients who adhere to preventive therapies may be more likely to engage in a broad spectrum of behaviors consistent with a healthy lifestyle. Because many of these behaviors cannot be measured easily, observational studies of outcomes associated with the long-term use of preventive therapies are subject to the so-called "healthy user bias." To better understand this effect, the authors examined the association between adherence to statin therapy and the use of preventive health services in a Pennsylvania cohort of 20,783 new users of statins between 1996 and 2004. After adjustment for age, gender, and various comorbid conditions, patients who filled two or more prescriptions for a statin during a 1-year ascertainment period were more likely than patients who filled only one prescription to receive prostate-specific antigen tests (hazard ratio (HR)=1.57, 95% confidence interval (CI): 1.17, 2.19), fecal occult blood tests (HR=1.31, 95% CI: 1.12, 1.53), screening mammograms (HR=1.22, 95% CI: 1.09, 1.38), influenza vaccinations (HR=1.21, 95% CI: 1.12, 1.31), and pneumococcal vaccinations (HR=1.46, 95% CI: 1.17, 1.83) during follow-up. These results suggest that patients who adhere to chronic therapies are more likely to seek out preventive health services, such as screening tests and vaccinations. Further work is needed to identify study design and analysis methods that can be used to minimize the healthy user bias in studies of preventive therapies.

Epidemiologic studies are increasingly used to investigate the safety and effectiveness of medical products and interventions. Appropriate adjustment for confounding in such studies is challenging because exposure is determined by a complex interaction of patient, physician, and healthcare system factors. The challenges of confounding control are particularly acute in studies using healthcare utilization databases where information on many potential confounding factors is lacking and the meaning of variables is often unclear. We discuss advantages and disadvantages of different approaches to confounder control in healthcare databases. In settings where considerable uncertainty surrounds the data or the causal mechanisms underlying the treatment assignment and outcome process, we suggest that researchers report a panel of results under various specifications of statistical models. Such reporting allows the reader to assess the sensitivity of the results to model assumptions that are often not supported by strong subject-matter knowledge.

Effect estimates from EPS models by simple LR were generally robust. NN models generally provided the least numerically biased estimates. C was not associated with the magnitude of bias but was with the increased SE.