Lecture 5: Bias & Misclassification Flashcards
what 3 aspects of a study, must researchers evaluate for internal validity before they can declare a true association?
- confounders and effect modifiers
- bias and misclassification
- statistical significance
definition of bias
systematic error in study design or conduct, resulting in false results
how do we control for bias after a study has been conducted?
we can’t
must be done before study is implemented
what are the 3 elements of bias impact?
- source or type
- magnitude
- direction
the 2 main categories of bias
selection-related
measurement-related
selection-related bias
any aspect in the way the researcher selects study subjects differently
measurement-related bias
any aspect in the way the researcher collects data differently
types of selection bias
healthy-worker bias
self-selection bias
other names for responder bias
self-selection bias
participant bias
what type of bias is easily seen in prospective cohort studies?
healthy-worker bias
measurement bias can be ……?
subject related
observer related
screening related
subject related bias
type of measurement bias recall bias hawthorne effect contamination bias compliance bias lost to follow up bias
groups being studied have different withdrawal rates
lost to follow up bias
groups being interventionally studied have different adherence with study protocols
compliance/adherence bias
members of the control group, receive treatment or are exposed to the intervention being studied
contamination bias
take drugs they shouldn’t
individuals alter their behavior because they know they are apart of a study under observation
hawthorne effect
or
observer effect
a differential level of accuracy in provided information between study groups
recall bias or reporting bias
observer-related
type of measurement bias
interviewer bias
surveillance bias
a systematic difference in interpreting done by the researcher
interviewer bias
not always consciously done
different evaluation or diagnosis between study groups due to an observers expectations
diagnosis/surveillance/expectation bias
screening-related
type of measurement bias
lead-time bias
an apparent benefit from a healthcare screening due to the early detection of disease despite an unchanged clinical outcome
lead-time bias
source of measurement bias, especially observer-related bias
misclassification bias
error in classifying either disease or exposure status, or both
misclassification
the 2 types of misclassification
differential
non-differential
misclassification of exposure which is unrelated to the other exposure
non-differential = error in both groups equally
not different
non-differential
in non-differential bias, the measure of association is always….?
type of misclassification
OR is always moved towards 1.0
misclassification of exposure which is related to the other exposure
differential misclassification
in differential bias, the measure of association is…..?
can be inflated or depreciated
ways to control for bias
blinding/masking
multiple data sources
randomly allocate observers for data collection
minimize loss to follow up bias
random allocation and randomization are ways to control for….?
random = bias control randomization = confounding control
ways to control for confounding variables
restriction randomization matching stratification multivariate analysis
