Which biases are most associated with observational epidemiologic studies, with examples?

• A. Publication bias, Allocation bias, Confirmation bias
• B. Selection bias (nonrandom sampling), Information bias (misclassification), Confounding (unmeasured variables)
• C. Random error, Systematic error
• D. Measurement bias only

Answer: (B)

In observational epidemiology, you don’t control who is exposed, so biases come from the study design and data collection itself as well as from variables you didn’t measure. Selection bias happens when the group you study isn’t representative of the target population because of how participants are chosen or who agrees to participate. For example, in a case-control study of a risk factor, if cases and controls differ in their likelihood of recalling past exposures, the observed association can be distorted. Information bias arises from errors in measuring exposure or outcome, such as relying on people’s memory for past behaviors or using imperfect diagnostic criteria, which can misclassify whether someone was exposed or whether they had the disease. Confounding occurs when another variable is linked to both the exposure and the outcome and isn’t accounted for, such as age or socioeconomic status, giving a distorted view of the exposure’s effect. These three—selection bias, information bias, and confounding—are the biases most closely tied to observational studies, whereas other biases like publication bias or general error categories do not reflect the inherent design challenges of these studies.