Module 7 - 8 Flashcards
Bias
Any systematic error in the design, conduct, or analysis of a study that results in a mistaken estimate of an exposure’s effect on the risk of a health outcome.
Ex: bias can result from how people are selected to be in the study, how the disease or exposure status is classified in the study, or if there is confounding that is not accounted for in the study design or analysis
Selection Bias
A distortion in a measure of association that occurs due to how participants are selected to be in the study.
Ex: if the way in which cases and controls, or exposed and unexposed individuals, were selected is such that an apparent association is observed - even if, in reality, exposure and disease are no associated - the apparent association is due to selection bias.
Information Bias
Bias that arises from how information is obtained on the exposure or health outcome.
Ex. Information bias can occur when the means for obtaining information about the subjects in the study are flawed so that some of the information gathered regarding exposures and/or disease outcomes is incorrect
Confounding
A distortion of an association between an exposure and a health outcome by a third variable.
Ex. When we observe an association, we ask whether is ti causal or whether it is a result of confounding by a third factor that is both a risk factor for the disease and is associated with the putative exposure in question
Effect Modification (Effect Measure Modification)
When a measure of association such as a relative risk or odds ratio, changes over values of another variable
Ex. When effect measure modification is present, presenting pooled results can be misleading, and instead results should be presented separately for each level of the effect modifier
Selection Bias
Arises from including individuals in the study or dataset who were not supposed to be included, and excluding individuals from the study or dataset that should have been included.
This can also arise from nonresponse of potential study participants.
Participant losses during follow up - also called “emigrative selection bias”
Nonrepsonders
It is important to keep nonresponse in a study to a minimum - people who refuse to participate differ from those who do.
Emigrative Selection Bias
Participant losses during a study that bias the observed associations
Difference between selection bias and selecting subjects
Selection bias impacts the internal validity of the study and the legitimacy of the inference regarding the association of exposure and the outcome. It is a systematic error in selecting subjects as part of the exposed/unexposed or case/control
Selecting subjects impacts the generalizability or external validity of the study.
Exclusion Bias
Results when investigators apply different eligibility (inclusion) criteria to the cases and to the controls, with regard to which clinical conditions in the past would permit eligibility.
Compensation bias
When bias in selecting cases and control is of the same magnitude, compensating bias is achieved.
differential misclassification
A type of information misclassification bias in which the proportion of misclassification differs in different study groups. For example, misclassification of exposure may happen such that unexposed cases are misclassified as being exposed more often than the other way around.
Nondifferential Misclassification
A type of information misclassification that is a problem inherent to the data collection method. It typically results in a CIR or OR that is diluted and shifted toward 1.
Types and sources of information bias
Abstracting records
interviewing
surrogate interviews
Surveillance bias
Recal bias
Reporting bias
What must be true of a confounder
- It must be associated with the exposure of interest
- It must be associated with the outcome of interest
- It cannot be a result of, or caused by, the exposure of interest (temporality)
What are the approaches to handling confounding
Individual matching
Group matching
In the analysis of data:
Stratification
Adjustment
Information Bias
Collecting incorrect values on participant’s:
Exposure status
Disease status
Covariate status (e.g. confounder
Measurement errors are associated with
Continuous variables like age
Categorical variables are associated with
Misclassification (e.g., sex)
Information Bias cont
Cannot be eliminated with sample size
Impacts OR and RR
Incorrect recall
Inaccurate diagnostic equipment
Leading questions
Biased interviewers
Non-compliance
Inference (etiological irrelevance)
Categorical Information Bias
Misclassification incorrectly specifying someone’s disease or exposure status
Non-differential Misclassification
sensitivity and specificity of exposure classification is the same among cases and controls
Dilutes the true effect - moves closer to 1
Differential Misclassification
Sensitivity and Specificity of exposure classification is different among cases and controls
Can go towards or away from null
Recall Bias
They remember exposure status differently
Differential misclassification
