[CH2] BIAS & CONFOUNDERS Flashcards
deviation of outcome measurement
errors
errors can be due to
sampling
biological variation
measurement process
errors can be eliminated by
larger samples
taking averages of at least 2 measurements
systematic induced error which results in under/over estimation of an association
bias
bias types
selection bias
information bias
selection bias types (5)
volunteer bias
healthy workers effect
neyman’s bias
Berkson bias
attrition bias
information bias types (7)
recall bias
reporting bias
interviewer bias
inter-observer bias
intra-observer bias
surveillance bias
detection bias
information bias can be controlled by (5)
1-blinding
2-using mail surveys
3-use multiple questions that ask the same important information
4-use multiple sources of information
5-using standard measurement instruments
selection bias can be controlled by (4)
1-random selection
2-use of incidence cases in case-control studies
3-avoid use of hospitalized cases from same department of intended disease
4-avoid use of end-stage cases except if they’re the target group
effect or association between an exposure & outcome is distorted by presence of 3rd factor related to both
confounder
is confounding a bias?
NO
characteristics of a confounder
1-must be a risk factor
2-associated w exposure
3-NOT intermediate step in causal path
4-NOT a result of exposure
BIAS: prevalence of diseased volunteers is higher than prevalence of diseased individuals in general population
Volunteer bias
BIAS: morbidity & mortality from a disease are lower among the sample of those present at work
healthy worker effect
BIAS: reflects both risk factors & survivorship
neyman bias (use of prevalent cases in case control studies)
BIAS: controls may have increased/decreased exposure to risk factors that leads to their diseases & not the one under study
berkson bias
BIAS: loss during follow up
attrition bias
BIAS: misinformation due to selective recall of previous activities & exposure
recall bias
BIAS: women who had complications during pregnancy recall pregnancy outcomes better than those w/ easier pregnancy experiences
recall bias
BIAS: exposures get under-reported because of attitudes/beliefs/perceptions
reporter bias
BIAS: drug abuse exposure is under reported due to attitudes & perceptions
reporter bias
BIAS: misinformation due to taking sides by interviewers
interviewer bias
BIAS: interviewer may ask extra questions, use of voice tone, body language, facial expressions to suggest preference to certain responses
interviewer bias
BIAS: variation b/w results obtained by 2+ different observers examining same material
inter-observer bias
BIAS: variation b/w results obtained by an observers examining same material more than once
intra-observer bias
BIAS: one group is followed up more closely by surveillance system
Surveillance bias
BIAS: systematic differences b/w groups due to differences in method of evaluation of outcome
detection bias
BIAS: exposed group is evaluated by lab investigations & X-rays while control group is evaluated w/ only an interview
detection bias
recall bias is common in
case control studies
% of loss that renders a study questionable and in need of sensitivity analysis
more than 20%
attrition bias is common in
cohort
clinical trials
attrition bias is common in
cohort
clinical trials
