Oct 2 Flashcards
What is bias?
- any systematic error in the design, conduct, or analysis of a study that results in a mistaken estimate of an exposure’s effect on the risk of the outcome
- this bias is sufficient to explain the measure of association (creates an association where a true association doesn’t exist or it doesn’t find an association where a true association exists)
What is selection bias?
-are participants in the study similar in all important aspects except for the exposure/disease?
What is information bias?
- “observation bias” or “measurement bias”
- is information about the outcome or exposure obtained in the same manner in cases and controls?
What is confounding bias?
-when results of study are accounted for by a factor associated with both the exposure and the outcome but not directly in the causal pathway (eg. association between drinking and lung cancer- smoking is a confounder. Drinking doesn’t cause people to smoke then get lung cancer but people who smoke tend to drink)
What study designs can be affected by selection bias?
- case control studies
- cohort studies
How should controls be selected in case control studies to avoid selection bias?
- controls are selected to reflect the exposure pattern in the source population of cases
- bias is introduced when one selects: an inappropriate control group or an appropriate control group BUT participation in the study is related to exposure status (ex: something in env suspect is causing cancer- people believe that they were exposed to whatever it is in the environment may be more likely to participate in a study looking at whether or not they have cancer therefore reason for participating in study is related to exposure pattern)
Why are retrospective studies susceptible to selection bias?
- all cases of disease have already occurred so people know exposure and outcome status
- their answers may change based on knowing what the outcome is
- those with most interest in participation would both have been exposed and have the disease
- selection of exposed or unexposed is related to outcome of interest (bias occurs then)
What effect would be observed on the relative risk if health records of people with an exposure and disease were kept compared to others?
- selection bias
- could have kept these records over others if we had a suspicion that the exposure was leading to outcome
- this leads to OVERESTIMATE of relative risk
If you were running a company and your workers were thought to be exposed to a harmful exposure, which health records would you get rid of to hide the evidence?
- keep only small portion of records of people who were exposed and had the disease and larger portion of everyone else because
- this will give an UNDERESTIMATE of relative risk
What is volunteer bias?
- type of selection bias
- ads to participate in research studies that will pay, give bonus marks, etc.
- people wanting to participate are often students, lower SES, etc. who might not be representative of the entire population
What is the healthy volunteer bias?
- type of selection bias
- people who tend to volunteer to participate in studies tend to care about/are interested in their health
What is non response bias?
- type of selection bias
- people who we invite to participate and ignore our invitation are different from those who respond “yes” or “no”
- call in the evening asking to complete a survey; people who have been at work all day are probably less likely to want to participate whereas elderly people might say they have the time. You then end up with more elderly women in your study which wasn’t the intention
What is exclusion bias?
- type of selection bias
- biases that result from the post-randomization exclusion of patients from a trial and subsequent analyses (don’t provide full data, outliers, etc)
- left with selected subset of people who we initially invited to be in that trial and fit in that criteria
- reasons for exclusion are related to treatment and/or outcome and others may be applied differentially according to treatment allocation
How can exclusion bias occur?
- ineligibility fixed at randomization but then reassessed after randomization (maybe you didn’t have enough people in study, etc)
- might have early outcome
- violate protocol
- missing data
- some aspect of their participation is not representative of what we wanted them to do in the first place
What is the healthy worker effect?
- type of selection bias
- workers usually exhibit lower overall death rates than the general population because the severly ill and chronically disabled are ordinarily excluded from employment
- most studies indicate HWE will reduce the association between exposure and outcome by an average of 20-30% (underestimate of true association because we are not including sickest people)
