ch 10 Flashcards
SB = selection bias
What should we do before we make a judgment about causation and generalization?
Determine if associations are valid
internal valdity v generalizability
- Internal validity - do observed results accurately reflect the true association?
- Generalizability (external validity) - to whom can results be applied?
Requires internal validity
What are two key components of validity that are needed?
- If a study lacks internal validity, external validity is irrelevant
- We do not compromise internal validity to achieve external validity (generalizability)
2 types of error
- Random error - error is due to chance
- Systematic error - error due to recognizable sources
Three types of systematic error?
selection bias
information bias
Confounding
What is bias?
- systematic error in the design or conduct of a study that leads to an incorrect estimate of the association
How can bias occur? which studies? not inherent of? rare?
- Can be caused by the investigator or study participants during the design or conduct of the study
- experimental, cohort, case-control, cross-sectional, and ecological studies
- study population
- few have no bias or error
what are the two effects of bias?
- Creates the appearance of an association when there is none or masks one that exists
- Selection bias (SB) and information bias cannot be fixed; confounding can be fixed to a point.
What are some solutions to reduce bias? Effects what?
- Limit in study design.
- Limit in the study cohort (during data collection)
- Critically evaluate after the study has been completed and discuss its effects
= effects on :
Type and sources
Direction
Magnitude
Impact on study results -> Impact on study interpretation
What are the two types of bias?
selection bias (who is in the 2x2 table)
information bias (where in 2x2 table)
What is selection bias? why does this happen?
- occurs when the study population is not a random selection of the target population.
- Results from procedures used to select subjects for a study
At the time of recruitment
During the process of retaining subjects in a study
Which study designs does selection bias occur in? Why does it happen in each?
- case-control: selection/participation of cases and controls is related to exposure status
- cohort/experimental: selection/participation of exposed and unexposed subjects is related to disease status
When is SB more likely to occur? why?
Case-control and retrospective because D and E have already occurred by the time the subject is selected for a study
What are the multiple ways SB can happen?
- Control selection bias
- differential participation
- Loss to follow-up
- Refusal
- Nonresponse
- self-selection bias
- Healthy worker effect
- Different surveillance, diagnosis, or referral of subjects according to their E and D status
What is control selection bias? Happens when? prevented by? Which study uses this?
- a bias that occurs if controls are more or less likely to be selected if they are exposed (or unexposed)
- This happens when controls fail to represent the exposure distribution in the source population from which the cases arose because controls do not accurately represent the same population as the case.
- Prevented by using identical selection criteria for cases and controls
- case-control
What is Differential participation? prevented? When does different participation happen?
- ( case-control and cohort)
- Bias that can occur if willingness or ability to participate is related to both exposure and disease status
- Prevented by obtaining high participation rates for all groups
what is differntail loss t follwo up? Issue because? Prevented by?
- (cohort and experimental)
- The bias that occurs if study participants exit the study for reasons related to both exposure and disease
- reduces the power of the study and doesn’t know if they develop disease or not
- since the outcome cannot be known without good follow up, must maintain high participation rate (Maximmzin retention and trace study subjects)
What is self-selection bias? prevention method
- When participants voluntarily choose t participate in a study or program group rather than being randomly selected into one
- Conducting a community case-control study or limited sever cases
How can we reduce selection and information bias?
- Little (or nothing) can be done to fix selection bias once it has occurred
- Must be avoided through careful study design and conduct (see strategies for individual biases)
- can not be “controlled” in the analysis
What are the two directions of bias? What do the positive associations do?
- Toward null
Positive association is biased towards H0, and true association is underestimated. - Away from null
Positive association is biased away from null, and true association is overestimated.
What is information bias?
as a bias that occurs when the information collected for a study is inconsistent or flawed, potentially leading to inaccuracies in the study’s results
- Occurs after participants have entered your study
Which two cases do information bias get into?
- Case-control study - different techniques are used to collect information from cases and controls
- Cohort study - different procedures are used to collect information from exposed and unexposed groups
Which study does information bias occur more likely in?
case-control and retrospective because exposure and outcomes (retrospective) have already occurred when the subject is selected for the study
What are the types of information bias?
- recall bias
- interviewer bias
- measurement error or misclassification
what is recall bias?
- Bias occurs if people with the disease remember/report their exposure differently than people without the disease.
Which two studies is recall bias seen in?
- Case-control - cases are more or less likely to recall prior exposures than controls - The number of cases and controls do not change.
- Retrospective cohort - exposed participants are more or less likely to recall prior diseases than unexposed participants.
What is the solution for recall bias?
- Use sick controls
- standardized, close-ended questionnaires
- Use pre-existing data (medical records)
which direction is information bias go towards for null hypothesis?
Depends on which subjects have less accurate recall
What is interviewer bias?
- Bias occurs if there is a systematic difference in soliciting, recording, or interpreting information.
- Toward or away from null
Which study designs can interviewer bias be seen in?
- Case-control - the interviewer is influenced by the participants’ case or control status (the interviewer is aware of the disease status of the subject and questions differently about E)
- Cohort and experimental - participants influence interviewer treatment or exposure status (interviewer knows about E or Treatment status)
What are the solutions for interviewer bias?
- masking but is impossible because the information they learn while talking with subjects.
- Train interviewers
- standardized questionnaires
- good data
what is a misclassification error?
- Toward or away from null
- Bias that can occur if study participants are placed into the wrong exposure or disease category (error in classifying )
What are the sources for misclassification errors?
- self-report s
- Errors in medical records
- Errors in how data are captured
- Non-specific disease or exposure definitions
which study designs can misclassfication error be seen in?
Case-control and retrospective studies:
Relevant exposure may have occurred way before the study started and was hard for subjects to remember - poor recall
what are the two types of misclassification errors?
- non-differntial
- differential
What is non-differential bias?
- bias towards the null
- That the extent of the misclassification is the same for both groups
- Classification on one axis (E or D) is not related to the other
what happens when the expsoure or disease is misclassified?
- E: the exposure misclassification is the same for cases/controls or diseases/not diseased
- the extent of the disease misclassification is the same for exposed and unexposed persons
what is differential bias?
- Bias towards or away from null
- Classification of one axis (E or D) is related to the other axis (E or D)
what is the solution for nondifferential and differential bias?
Accurate definition of E and D
Accurate data sources
Multiple measurements
validation
