This document discusses different types of bias that can occur in studies, including systematic error that is not affected by study size (e.g., confounding, selection bias, information bias) and random error that decreases with increasing study size. It provides examples of confounding, selection bias in a case-control study of smoking and stroke risk, and information bias from misclassification or researcher expectations.
This document discusses different types of bias that can occur in studies, including systematic error that is not affected by study size (e.g., confounding, selection bias, information bias) and random error that decreases with increasing study size. It provides examples of confounding, selection bias in a case-control study of smoking and stroke risk, and information bias from misclassification or researcher expectations.
This document discusses different types of bias that can occur in studies, including systematic error that is not affected by study size (e.g., confounding, selection bias, information bias) and random error that decreases with increasing study size. It provides examples of confounding, selection bias in a case-control study of smoking and stroke risk, and information bias from misclassification or researcher expectations.
This document discusses different types of bias that can occur in studies, including systematic error that is not affected by study size (e.g., confounding, selection bias, information bias) and random error that decreases with increasing study size. It provides examples of confounding, selection bias in a case-control study of smoking and stroke risk, and information bias from misclassification or researcher expectations.
à Decreases with increasing study size • Bias towards the null (observed effect larger than true effect)
• Bias away from the null (observed
effect larger than true effect) Bias and DIRECTION • Confounding: Third factor that changes the association between exposure and outcome • Selection bias: Resulting form Types of Bias selection or retention of the participants/study population • Information bias: poor measurement, of exposure, outcome or confounders Confounder • Medication and risk of heart failure • Medication prescribed for hypertension Example • Reduce hypertension ~ reduce heart failure confounding Case-control study
Does smoking increase risk of stroke?
Selection Cases: patients admitted for stroke bias: example Controls: patients admitted for everything else Problem? Cases and controls from same source population in relation to exposure Exposure associated with hospitalisation (less healthy) RESULT? Case-control àUnderestimation of effect smoking on risk of stroke
= Selection bias Healthy workers bias
Attrition (loss of participants) Other
Volunteer bias selection biases Misclassification to exposure or disease status
Researcher knows assignment of
Information treatment group bias: example à High expectations ~ more likely to identifying better outcome Other Recall bias information biases Hawthorne effect (observer bias)