Bias Flashcards
(16 cards)
Confounding
If stratified ORs/RRs are equal but differ from crude
Effect modification
If stratified ORs/RRs are not equal to one another
Information bias
A systematic distortion of findings that results from inaccuracies in the measurement of participant characteristics
Recall bias
Participants do not remember past exposures or events accurately
Examples: cases may be more likely to accurately recall an exposure than controls
Reporting bias
AKA social desirability bias; participants are reluctant to report exposures because of social norms
Observer bias
Observers may have preconceived expectations of what they should find in an examination; the person collecting the data may influence the accuracy of the data
Example: in abstracting electronic records, differential recording of exposures in the records of cases vs controls
Hawthorne effect
Behavior of participants changed because they are aware of being observed
Differential exposure misclassification
When the accuracy of the exposure differs in those with and without disease
Differential outcome misclassification
the accuracy of the outcome differs with respect to exposure status
Non differential misclassification
The misclassification is not different across the groups in your study
Any misclassification of the exposure occurs to the same degree across both cases and controls OR misclassification of the outcome occurs the same in both those exposed and unexposed
Differential misclassification
The misclassification is different across the groups in your study.
The misclassification of the exposure differs by disease status, or is different for cases and controls
The disease is misclassified differently and those who were exposed and those who are unexposed
Selection bias
A systematic distortion of findings that results from the process used to recruit study participants
Non-response bias
People who respond to a study often differ systematically from people who do not respond (people who choose to participate may have different characteristics than those who do not participate)
Non response can occur at recruitment or follow up
Berksonian boas
A form of selection bias that applies to hospital based epidemiologic studies.
People in the hospital are likely to suffer from multiple diseases or engage in unhealthy behaviors (eg, smoking, drinking, food related issues, etc). As a result, hospitalized study participants are not typical of the community population.
May result in larger odds ratios than actually exist
Neyman’s Bias
If persons with more severe disease die quickly, those available for the study are not as sick ; cases in study may not represent all those with the disease; may lower ORs and mask strength of causal association
Healthy worker effect
Employed workers are healthier than other segments of the population and have more favorable mortality experience. Ill and disabled people are less likely to be employed. As a result, workers may not be typical of the larger population. Estimates of association may be lower than actually exist
