Selection bias in observational studies

Abstract

Bias comes in many flavors. Observational studies are especially prone to its many forms, mainly due to the investigator's lack of control over the study. Selection bias may result when a study fails to select a representative sample from the population of interest, limiting the applicability of the study's results.1,2 In this issue of Neurology® , Marrie et al.3 report an analysis of questionnaires completed by participants of the North American Research Committee on Multiple Sclerosis (NARCOMS) Registry,4 a self-report database for patients with multiple sclerosis (MS). They obtained demographic and clinical information from a questionnaire participants completed at the time of enrollment in the registry, and comorbidity information from a follow-up questionnaire in 2006. The study suggested that comorbidity lengthens the delay between onset and diagnosis of MS, and that comorbidity increases the severity of disability at diagnosis. However, after accounting for the delay in diagnosis, the association between comorbidity …