Lecture 12 - Bias & Misclassification Flashcards
Define bias.
Systematic (non-random) error in study design or conduct that lead to a distortion in the association between exposure and outcome
What can be done about be done about bias before a study? After?
Before:
-steps can be taken to avoid bias or minimize its impact
After:
-NOTHING!
What are the 3 elements of bias impact?
- Source/Type
- Magnitude/Strength
- Direction
What effect can bias have on the magnitude/strength of association?
Bias can entirely account for a weak association but is unlikely to account for a large association.
What effect can bias have on the direction of association?
Bias can result in over or underestimation of the measure of association either enhancing or minimizing the true relationship.
What are the two main categories of bias?
- selection
- measurement
What is selection related bias?
When the way that subjects are selected or acquired creates a systematic difference.
Common in comparative studies when the group are not similar enough or do not adequately represent the population.
What is measurement related bias?
When the way that information is collected or the subjects are measured/observed creates a systematic difference.
What are examples of selection bias?
- Healthy-worker bias
- Self selection/responder bias
What is healthy worker bias?
Type of selection bias in which a group of workers may not be representative because those who are unhealthy due to work may no longer be working.
Commonly seen in prospective cohort studies in which.
What is self-selection/responder bias?
Type of selection bias in which those who volunteer to participate in a study might not be representative of the rest of the population.
What are the types of measurement bias?
- subject-related
- observer-related
- screening-related
What is recall/reporting bias?
Diseased or exposed subjects may be more likely to recall detail OR exaggerate responses
What is the Hawthorne effect?
When individuals in a study alter their behavior because they know they are being observed.
What is contamination bias?
When individuals in the control group receive treatment outside of protocol similar to intervention being studied.
What is compliance/adherence bias?
When groups being studied have different compliance/adherence to study protocol or treatment.
What is lost to follow-up bias?
When groups have different withdrawal rates or lost to follow-up rates. Also when there are differences in those who stay and those who withdraw.
What is interviewer bias?
When the interviewers knowledge may affect how questions are asked or information is obtained differently between groups.
-also-
When interventions or treatment are administered differently due to differences in skill or compliance of study personel
What is diagnosis/surveillance bias?
When the observers preconceived expectations affect the evaluation, classification, diagnosis, or observation differently between groups.
What is lead-time bias?
When screening due to the study leads to earlier detection of disease than normal which can lead to the attribution of a greater than actual benefit of the treatment.
What is misclassification?
Error in classifying either disease or exposure status
What are the types of misclassification?
- non-differential
- differential
What is non-differential misclassification?
- misclassificaiton occurs equally in both groups
- attenutate measure of association; moves closer to 1.0
What is differential misclassification?
- misclassification occurs more frequently in one group than the other
- inflates or attenuated the measure of association; moves closer OR away from 1.0
What are methods for controlling bias?
- blinding/masking
- multiple sources to gather data
- randomly allocate observers/interviewers
- use technology as much as possible
- attempt to minimize loss to follow-up
