 Berkson's bias, also known as Berkson's paradox or Berkson's fallacy, is a type of selection bias. It can occur in hospital-based studies when the study population consists of individuals who are in the hospital for reasons unrelated to the exposure or outcome being studded. If the study investigating the relationship between diabetes and tuberculosis includes hospitalized patients with TB, it might appear that diabetes is negatively associated with TB because diabetic patients are more likely to be hospitalized. To mitigate Berkson's bias, researchers can consider the following strategies. Include out patients, include a control group, use population-based data and conduct sensitivity analyses.