week 6 Flashcards
Name 2 type of errors and then their subsections that could occur in the epidemiological studies?
- Random error.
2. systematic error. (confounding) and bias- selection and information bias.
What is random error?
It is a chance error. May result in under or over estimation of the true value.
Mainly related to sample size. Smaller the sample size, greater the random error.
What is systematic error?
Estimate of association between
exposure (E) and outcome (O)
systematically different from
truth
• Error not related to sample size
Describe what is Bias?
Usually a result of study design or
conduct.
Cannot be corrected at the analysis stage.
Different study types are prone to different biases.
What is selection bias?
bias related to who is included in study
sample
What is information bias?
bias related to how the study variables
(exposure / outcome) are measured
What is selection bias? (2)
Occurs when there is a difference in the
characteristics of those selected and those not
selected for the study…
AND this selection is associated with exposure or
outcome of interest
What is selection bias in case control studies?
• occurs when the exposure of interest is
associated with the likelihood of controls (or
cases) being included in the study
Problem
• Results in biased estimate of exposure odds in
population from which sample is drawn
What is selection bias in cohort studies?
• Less of a problem than in case-control studies
• Main selection bias: loss to follow-up (attrition)
• Other selection biases: choice of unexposed
group
• E.g. healthy worker effects in occupational
studies
What is information bias ?
misclassification of exposure status, outcome status or both
What are the two types of information bias?
- Non-differential (random) misclassification:
• Of exposure independent of outcome or
• Of outcome independent of exposure - Differential (non-random) misclassification:
• Of exposure dependent on outcome or
• Of outcome dependent on exposure
What are the two main types of information bias?
- Recall (response) bias:
• Occurs when a respondent’s recall of exposures
varies according to whether they are a case or a
control - Observer (interviewer) bias:
• Occurs when awareness of case / control status
influences the assessment or recording of
exposure status
What is recall bias?
Cases or controls selectively likely to remember
or forget, or disclose or not disclose the
exposure
• This may be due to, e.g.:
• “effort after meaning”
• Social desirability bias
How to minimise recall bias?
Select controls from a group who similarly are
being treated for another disorder (reduces
“effort after meaning”
• Context of interview could be similar for cases
and controls
• Use more comprehensive measures of exposure
• Multiple informants
• Blinding of cases and controls to their group
What is observer bias?
Non-comparable information is obtained from
cases and controls, e.g. leading questions
• For cases & controls (in case-control studies)
• For exposed & unexposed (in cohort
studies)
• Limiting observer bias:
• Blinding observer to hypothesis/caseness
(often difficult)