Selection Bias
Berkson Bias and how to avoid it
Question Author: Tomislav Meštrović, MD, PhD
Question #3: A cardiology specialist working in the university hospital center notices that a large number of his patients who have suffered from myocardial infarction also have some sort of malignancy in their medical history. In order to determine whether a relationship exists between myocardial infarction and malignant processes, he decides to pursue a case-control study. His selected cases consisted of a sample of patients hospitalized in an internal ward with a history of a myocardial infarction, while control cases were patients on the same ward without a history of a myocardial infarction. Their charts …show more content…
This may result in hospitalization rates that are approximately the same for the exposed and unexposed, as well as for case and control groups. The cardiologist would then found out that malignancy occurs just as often among individuals with other major diseases, as among those with myocardial infarction.
Wrong answer A: Diagnoses that are either positively or negatively linked to the risk factor under study should be excluded, as they may also give rise to bias. The use of two control groups may be an alternative approach for solving the problem of Berkson’s bias.
Wrong answer B: Increasing the number of cases and controls will not solve the fundamental problem of false association due to sampling of both cases and controls among hospitalized individuals from the same ward. The probabilistic calculations rely on the assumption that exposed (diseased) clinical conditions have an independent and additive impact on hospitalization rates.
Wrong answer C: Patients with diseases in the control group must have a probability of hospital admission akin to cases, otherwise it results in overrepresentation of exposed cases, Berkson's bias, and potentially other types of sampling