Bias Flashcards
Status-Quo Bias
- Default option is ranked more highly
- Hesitancy to change due to aversion to loss
- Endowment effect (overvalue what you own)
Observed Association Explanations
- Observed association is true
- Due to chance
- Resulted due to bias
- Resulted due to confounding
Precision
- Lack of random error (chance)
- Study size
- Study efficiency
Validity
- Lack of systemic error
- Internal validity (bias, confounding)
- External validity (generalizability)
Internal Validity
- Validity of inferences drawn that relate to study population
- Ability of test to measure what it sets out to measure
- Whether the experimental treatment makes a difference in this specific experimental instance
External Validity
- Inferences about the extent to which a casual relationship holds over variations in persons/settings/treatments/etc
- Validity of inferences drawn as they relate to groups OUTSIDE the study population
- Extent to which the study findings can be generalized
Bias
- Any trend in the collection, analysis, interpretation, publication or review if data that can lead to conclusions are systemically different from the truth
- Systematic error: incorrect estimate of association (RR or OR is wrong)
Why worry about Bias/Confounding?
- Can cause incorrect results
- Over or underestimate association
- Change direction of association
- Quasi-experimental research is prone to bias and confounding
- Due to limited ability to remove the effect of other risk factors, ascertain exposure and outcome, representative sample
Validity
- Randomization doesn’t eliminate threats to validity
- Minimizes threats to internal validity
- External validity can still be jeopardized due to stringent criteria
Sources of Bias
- Caused by investigators or participants
- During design/conduct of study
- Occur in cross-sectional, case-control, cohort, or interventional studies
Major Types of Bias
- Selection Bias
- Information Bias
- Sampling or ascertainment bias
- Response/Participation Bias
- Lost to Follow-up Bias
- Publication Bias
Selection Bias
- Distortion in measure of association
- Results form manner in which subjects are selected into study population
- Results differ from if you had enrolled entire population
Information Bias
- Flaw in measuring exposure or outcome
- Collection of incorrect information from subjects
Sampling/Ascertainmetn Bias
- Healthcare access bias
- Length-bias sampling
- Exclusion
- Detection
Response/Participation Bias
- Differences in characteristics between participants in study and non-participants
- Non-participants having more severe symptoms or conditions not responding to surveys for example
Exclusion Bias
- Applying exclusions to only one group being studied (controls or cases)
- Can lead to over or under estimation
Types of Information Bias
- Observer/interviewer bias
- Surveillance/detection bias
- Recall bias
- Reporting bias
- Response fatigue Bias
Observer/Interviewer Bias
- Knowledge of hypothesis
- Disease or exposure status knowledge
- Gathering selective data
Surveillance/Detection Bias
-Disease ascertainment might be better in a more closely monitored population
Recall Bias
- Accuracy problems due to memory
- Differential recall - subject with disease are more likely to remember exposure in the past
Reporting Bias
-Selective suppression or revealing of information
Response Fatigue Bias
-Incorrect answers given at the end of a long questionnaire
Misclassification Bias
- Type of Information Bias
- Subjects are incorrectly catagorized with response to either exposure or outcome
- Placed into the wrong group
- Can introduce bias towards or away from the null
Non-differential Misclassification
-Proportion of subjects misclassified is about the same in each group
Differential Misclassification
- Information is more accurate in one study group
- Causes bias
Methods to Control Bias
- Select groups by similar mechanisms or sources
- Choose study population that is easy to follow
- Maximize follow-up
- Standardize data collection
- Interviewers/data collectors should be blinded
- Facilitate recall
