Observational Studies

University Of Pennsylvania, Philadelphia PA

Investigators

Abstract

An observational study is an empiric investigation of the effects of a treatment, policy, intervention, or exposure which was not randomly assigned to subjects, as it would be in a randomized experiment. Observational studies are common in most fields that study people, because harmful or unwanted treatments cannot be imposed on human subjects for experimental purposes. The central difficulty in an observational study is that, because treatments were not randomly assigned, the subjects receiving different treatments may not be comparable, so differing outcomes after treatment may not be effects caused by the treatment. If the treatment groups differ before treatment in ways that have been measured, there is an overt bias that often can be removed by adjustments, such as matching. Typically, there is concern that treatment groups differed before treatment in ways that were not measured, that is, concern about hidden biases that cannot be removed by adjustments. Several topics are to be investigated, two of which are described here. First, a sensitivity analysis asks how unobserved biases of various magnitudes might alter substantive conclusions. Studies vary markedly in sensitivity to unobserved biases: some are sensitive to small biases that may be difficult to avoid, while other studies are not sensitive to quite large biases, so their conclusions are on much firmer grounds. A method of sensitivity analysis will be developed for a large class of estimators, known as m-estimators, which include as special cases most common types of estimators. Second, there is interference between units when the treatment given to one unit affects other units. Interference is common when units are people who communicate, compete, or spread disease, or when units are different opportunities to apply a treatment to one or a few organisms, as is true of most studies that examine a few people using fMRI and PET as these people perform various cognitive tasks. Confidence intervals for treatment effects in the presence of interference will be developed. In public policy and public health, in medicine and clinical psychology, observational studies are the basis for many, perhaps most, policies and treatment decisions that affect people. The two goals of improved statistical methods for observational studies are better design of observational studies to produce more accurate evaluations of competing policies and better analyses to yield more realistic appraisals of the uncertainties that remain.