Bias Flashcards
What is bias?
A systematic error in an epidemiological study which either underestimates or overestimates the factor because of a deficiency in the design or excecution of the study
How can bias affect a study?
Biases can lead to incorrect conclusions about what the study or trial showed
Two main groups of bias
Information bias
Selection bias
Selection bias
Error in identifying the individuals or groups taking part in the study (wrong people)
Information bias
Error in the measurement of information on exposures or outcomes in the study participants
Sampling bias
Non-random selection of the sample, there is not an equal chance of selection from the population
Allocation bias
If randomisation to treatment or control in an RCT is not random
Responder bias
People willing to participate in studies to do with health are different from those who do not want to take part
e.g. Obesity study: non-responders more likely to be overweight.
Attrition bias
Participants who do not complete the follow-up investigations may be systematically different from those who do follow up
Publication bias
Studies with positive results are more likely to be published than those with negative results. This can mean systematic reviews are biased towards positive results
Healthy worker effect
In occupational cohort studies, the occupational cohort and reference population may not be comparable because occupational cohorts are on average more healthy than the general population
Recall bias
Systematic difference in recall about potential exposure by cases and controls
Reporting bias (aka interview bias, observer bias)
Can occur if the interviewer is aware of the status of the subject (e.g. case or control, drug or placebo) and may steer the participant to give the ‘correct’ answer
Data collection clerks who are aware of which exposures are under scrutiny may look harder for a record fo exposure in cases vs. controls
Social desirability response bias
Where the subjects give an answer that they feel is what the researcher would like to hear. Particularly re: embarrassing topics (sexual partners, cigarretes, alcohol)
What effect can information bias have on the results of a study?
Non-differential misclassification (random errors due to inaccurate measurement) favours the null hypothesis, and will attenuate the estimate of the effect
Systematic bias can either attenuate or exaggerate the estimate of effect.
