What is bias?

Bias is any systematic error that reduces the validity of study results. There are several types of bias, including information bias (includes missing information and misclassification) and selection bias (occurs when the study population does not represent the target population). Information bias also includes “immortal time bias”, which occurs when follow-up of the exposed group includes an unexposed time interval during which the study endpoint can’t occur, and individuals must survive long enough to receive the study drug or intervention.

What is confounding?

Confounding may occur when a particular variable (eg, age) influences both the likelihood of a specific exposure (eg, treatment under study) as well as the outcome of interest (eg, occurrence of MI). When such variables are unequally distributed between treatment groups, confounding is present.

Randomization in RCT minimizes confounding by baseline characteristics, but confounding is an issue that must be addressed in RWE studies through study design and/or analytic methods.