Types of Bias
What is selection bias?
A distortion in the measure of association due to procedures used to select participants that are not representative of the target population.
What is the potential source of bias: Participants in a diet study self-report how many calories they consume.
Information bias, specifically recall or reporting bias.
How can you reduce recall bias?
Use objective records when possible, or shorten the recall period.
What is a confounder?
A variable that is associated with both the exposure and the outcome, and is not in the causal pathway.
What is bias toward the null?
A bias that weakens the observed association, making it appear closer to no effect.
What is recall bias?
A systematic error due to differences in accuracy or completeness of recollections retrieved by study participants about past events or exposures.
Case-control bias scenario: Smokers are more likely to remember exposures than non-smokers. What bias is present?
Recall bias.
How can you mask interviewers?
Use double-blind procedures or standardize interview scripts and training.
What are the three criteria for a confounder?
1) Associated with exposure, 2) Associated with outcome, 3) Not an intermediate step in the causal pathway.
What is bias away from the null?
A bias that exaggerates the observed association, making it appear stronger than it truly is.
What is interviewer bias?
A bias that occurs when interviewers subconsciously or consciously influence the responses of participants.
Cohort study with exposure error: Workers are classified as exposed based on job title, but some don't actually handle chemicals.
Misclassification bias (non-differential if equal across groups).
How to minimize misclassification?
Use validated measurement tools, training, and consistent definitions.
How do positive and negative confounding affect the observed association between exposure and outcome?
Positive confounding exaggerates the true association; negative confounding underestimates it.
What is the impact of poor recall?
Can cause recall bias, often leading to misclassification and bias in either direction.
What is information bias?
Systematic error due to inaccurate measurement or classification of study variables.
Who was misclassified? A survey uses outdated diagnostic criteria to determine disease status.
The cases were misclassified due to outdated disease definitions (information bias).
Matching vs. Restriction? What are they used for?
Both are strategies to control confounding at the design stage. Matching pairs exposed and unexposed on confounders; restriction limits study population by confounder status.
What is confounding by indication?
When the reason a patient receives treatment (the indication) is also a risk factor for the outcome. Common in observational studies.
In what direction can misclassification bias the observed association?
It can bias the association toward or away from the null, depending on whether the misclassification is non-differential (usually toward the null) or differential (can be in either direction).
What is misclassification?
Systematic error due to inaccurate measurement or classification of study variables.
Confounding or bias? Older people in a study are more likely to take vitamins and also more likely to develop heart disease.
Confounding by age.
One way to reduce confounding?
Design stage: Randomization, restriction, and matching
Analysis stage: Stratification, multivariable analysis, and standardization
How can you control for confounding at the analysis phase?
Stratification, multivariable adjustment, or propensity score matching.
What is the direction of confounding?
It depends on the relationship between the confounder, exposure, and outcome; can bias toward or away from the null.