15 terms

Selection Bias

STUDY
PLAY
What is selection bias?
Error introduced when the study population does not represent the target population.

Results from procedures used to choose study subjects. This gives results that differ from the result that would occur among individuals who were eligible but left out of the study.
When does selection bias occur?
Selection bias occurs at the stage of recruitment of participants and/or during the process of retaining them in the study.
Correcting selection bias
Selection bias is difficult to correct in the analysis, although one can do sensitivity analysis.
Selection bias is most likely in:
Case-control studies

Retrospective cohort studies

In both of these studies, both outcomes and exposures have occurred by the time study subjects are selected.
Selection bias may also occur in:
Experimental and prospective cohort studies due to differential loss to follow-up
Population Hierarchy
Target Population (to which the results may be generalized.)

->

Source population

->

Eligible population (intended sample;possible to get all)

->

Actual study population (Study sample successfully enrolled)
Unbiased sampling
Exposed and diseased group have equal probability of being included in the study.
Biased sampling
Exposed and diseased group have a lower probability of being included in the study: this leads to imbalance and bias.
Selection bias in case-control studies
Selection bias will occur if selection of cases and controls is based on differing criteria related to exposure status.
Selection bias in retrospective cohort studies
Selection bias occurs if the exposed and unexposed individuals are selected in a manner related to developing the outcome of interest.
How does selection bias occur in practice?
Inappropriate selection of controls in case-control studies (i.e., control selection bias)

Refusal, nonresponse, or agreement to participate in a study that is related to exposure and disease (i.e., self-selection bias)

Loss to follow-up related to exposure and disease

Selection of the general population as a comparison group in an occupational cohort study (i.e., health worker effect)

*These problems can bias study results towards either towards or away from the null.*
Healthy Worker Effect
"Healthy worker effect" is a special type of selection bias that occurs in cohort studies of occupational exposures when the general population is used as the comparison group.
Specifics of the Healthy Worker Effect
The general population consists of both healthy people and unhealthy people.

Note: Those who are not healthy are less likely to be employed, while the employed work force tends to have fewer sick people.

People with severe illnesses would be most likely to be excluded from employment, but not from the general population.

Thus, comparison of death rates between an employed group and the general population will be biased - typically showing that being in a dangerous occupation is protective.
Key facts about selection bias
Error that arises from systematic difference in selecting or following study subjects.

More common in case-control studies and retrospective studies because exposure and disease events have already occurred. Possible in other study designs.

Types include control selection bias in case-control studies; healthy-worker effect in occupational cohort studies; self-selection bias; differential loss-to-follow-up, and differential surveillance, diagnosis or referral of subjects

Little can be done to fix bias once it has occurred

Effect is to bias associations either towards or a way from the null
Methods for avoiding selection bias
Use similar criteria for selecting cases and controls

Ensure high participating rates

Use a variety of methods to maximize trace of study subjects

Take diagnostic and referral practices into accounting while designing studies