Four pathways into a biased sample
Sampling bias occurs when enrollment criteria exclude a subgroup, skewing results away from the true population effect. Attrition bias happens when participants drop out non-randomly, leaving a healthier or more motivated subset. Healthy-worker bias arises in occupational studies because workers are pre-selected to be healthy enough to work. Volunteer bias distorts findings when volunteers differ systematically from non-volunteers (often healthier, more educated, more health-conscious).
Each type produces a sample that no longer represents the source population, making results appear more favorable (or more harmful) than reality.
Detecting bias in the enrolled cohort
Comparing the characteristics of enrolled versus excluded participants often reveals selection bias. If a heart disease trial enrolled only young men without comorbidities because sicker patients were deemed too risky, results will overestimate the drug's benefit in the general population of heart patients. Reporting who declined enrollment is a simple transparency measure that helps readers judge generalizability.