Requiring continuous enrollment during follow-up? A call for robust research methods

Critical evidence for the evaluation and implementation of managed care strategies relies on researchers examining the health outcomes and costs associated with various pharmacological and nonpharmacological treatments and interventions. These studies usually leverage secondary data sources, such as administrative claims or electronic health records, and it is customary for researchers to design studies that explicitly exclude study participants that may be lost to follow-up. These exclusion criteria can be commonly referred to as continuous enrollment requirements and are ubiquitous in health outcomes research because such criteria are straightforward to implement and justify. However, requiring study participants to be continuously enrolled after the start of follow-up may lead to time-related biases, which can lead to lack of validity in study results. This article illustrates a type of time-related bias known as immortal time bias and explains how requirements for continuous enrollment during follow-up can contribute to this and other time-related biases. We examine the reasons individuals are usually lost to follow-up-death, insurance termination, or switching of insurance-and consider the various mechanisms by which these events can bias study estimates. We then delineate various considerations for evaluating the impact of attrition on various studies and conclude with a brief discussion of methods for avoiding time-related biases. We discuss recommendations for researchers to report useful attrition tables, reevaluate their study designs, and take steps to limit susceptibility to time-related biases. Additionally, resources for redesigning studies vulnerable to bias are presented along with some analytical strategies.