Bias and Confounding Flashcards
if a study has a RR=4.3 and 95% CI (4.0-4.8) the association could be caused by:
random error
systematic error
true association between exposure and outcome
validity
absence of systematic error in a study result
what is a valid measure of association
will have same value as the true measure in the source population, except for error due to random variation
bias
extent to which a measure of association from a study differs from the true measure of association in the source population
T/F bias is for differences due to systematic and random errors
false: only systematic errors
T/F bias can make a study’s conclusion invalid
true
internal validity
study result is valid with respect to the population under study
- study population
- source population
external validity
study result is valid to a wider population beyond to study and/or source population
AKA generalizability
study population
subjects in the study
source population
population from which the subjects were drawn
other populations (=target population)
populations to which we may want to generalize our results
2 types of bias
non-differential
differential
non-differential bias
equally affects groups
differential bias
affects one group more than another
- diseases are biased, but not the non-diseased
2 general sources of bias
selection
information
selection bias
sample is different from the population
information bias
error in measurement
AKA misclassification bias
confounding
unknown factor distorts the relationship between the exposure and outcome
selection bias in cross-sectional descriptive (prevalence) studies
sample would have more (or less) disease than true prevalence in the source population
- can over/underestimate the amount of disease in the source population
selection bias in case-control studies
case (diseased) or control (non-diseased) samples have more (or less) exposure than the diseased or non-diseased groups in the source population
selection bias in cohort studies
exposed or non-exposed samples have a higher (or lower) disease incidence than the exposed and non-exposed groups in the target populations
self-selection bias
based on volunteers-may not be representative of the population as a whole
diagnostic bias
diagnosis of disease may be influenced by the vet’s knowledge of the exposure and their expectation of disease
how can you reduce diagnostic bias
have a clear, well-defines case definition
use as many objective parameters as possible
blinding of the exposure status of the animals
T/F if the error leads to misclassification they can lead to errors in the measure of association
true
information bias in cross-sectional (prevalence) studies
may result in prevalence estimate in the sample being different than the true prevalence in the target population
information bias in case control studies
error in measurement of the exposure in the diseased or non diseased may bias the association
how to reduce informational bias in case control and cross-sectional studies
evaluate accuracy of measuring tools and adjust estimates to reflect the error
information bias in cohort studies
error in measurement of the disease in the exposed or non-exposed
examples of informational bias
observer variation
deficiency of tools and technical errors
recall bias
reporting bias
confounding
distortion of the underlying relationship between an exposure and an outcome by a third factor
T/F third factor influences both the exposure and the outcome
true
T/F confounding is different than bias
false: confounding is a special type of bias
what 3 conditions must be met to be a confounder
associated with the exposure
associated with the outcome
not in the causal pathway between the exposure and the outcome
T/F before the study starts you can predict the confounder
true
how to reduce the confounder
match the study
restriction
randomization
confounding variable after study has been completed
stratify it
stratify
partition the results based on the confounding factor
ex: split sexes, type of practice etc