Epidemiology · Reference
What is selection bias?
Selection bias is a systematic error that arises when the people included in a study, or retained in it, differ from the population they are meant to represent in ways that distort the association between exposure and outcome. It is a threat to validity, not a problem solved by larger samples.
Where selection bias comes from
Selection bias occurs when the procedure for choosing or keeping participants depends, however indirectly, on both the exposure and the outcome, so that the studied sample gives a distorted picture of their relationship. It can enter at recruitment — for example if volunteers differ systematically from non-volunteers — or during follow-up, if people leave the study in ways related to exposure and outcome. Crucially it is a systematic error: it shifts the estimate in a particular direction and persists no matter how large the sample, which distinguishes it from random sampling error.
Common forms
Several named patterns recur. Self-selection (volunteer) bias arises when those who choose to take part differ from those who do not. Loss to follow-up (attrition) bias occurs when dropout in a cohort is related to both exposure and outcome.
In case-control studies, control-selection bias arises when controls are not drawn from the same population that produced the cases. The healthy-worker effect is a classic example in occupational studies, where employed populations are healthier than the general population. Berkson’s bias can distort associations in hospital-based studies because hospitalisation depends on multiple conditions at once.
Why it matters and how it is addressed
Selection bias threatens the internal validity of a study: if it is present, the observed association may not reflect the true relationship even within the sample studied. Because it cannot be fixed after the fact by statistical adjustment in the way some confounding can, the main defences are at the design stage — defining the source population clearly, selecting controls from that population, maximising follow-up, and minimising differential participation. Where bias may remain, good practice is to reason explicitly about its likely direction and magnitude.
Selection bias versus confounding and information bias
It helps to separate the major threats to validity. Selection bias stems from who is in the study; information (measurement) bias stems from how variables are measured or reported once people are in; and confounding stems from a third factor associated with both exposure and outcome. All three can distort findings, but they arise at different points and call for different remedies. Reporting guidelines such as STROBE ask authors to describe how participants were selected precisely so readers can judge the risk of selection bias.
Key facts
At a glance
- Definition: Systematic error from how subjects are selected/retained
- Affects: Internal validity of the study
- Not fixed by: Larger samples (it is systematic, not random)
- Examples: Volunteer, attrition, control-selection, healthy-worker
- Best tackled at: The design stage
Common questions
FAQ
What is selection bias?+
Selection bias is a systematic error that occurs when the people included in or retained by a study differ from the population they should represent in ways that distort the exposure–outcome association. Because it is systematic rather than random, it cannot be removed simply by enlarging the sample.
How is selection bias different from confounding?+
Selection bias arises from who is included in or kept in the study, whereas confounding arises from a third variable associated with both exposure and outcome. Selection bias is mainly prevented at the design stage, while some confounding can be addressed analytically through adjustment.
What is the healthy-worker effect?+
The healthy-worker effect is a form of selection bias in occupational studies where employed people tend to be healthier than the general population, because those who are unwell are less likely to be in work. This can make workplace exposures look less harmful than they are.
The step most authors miss
Doing CRediT right? Don’t stop at the statement.
A CRediT statement credits you inside one paper. The recognition CRediT was built for happens when those roles are tied to you, persistently. Sign in with your ORCID — free — and claim your CRediT contributions on casrai.org, the home of the standard. They become a verified, portable part of your identity, not a line that disappears into one PDF.
Free: claim your contributions, then export a journal-ready CRediT statement, schema.org structured data, JATS XML, CSV or BibTeX — and preview your public profile. A membership publishes that profile publicly and verifies the journals you serve.
