FOR THE NEXT GENERATION OF PHYSICIANS
← All USMLE guides
BIOSTATISTICS, EPIDEMIOLOGY AND ETHICS

Bias, confounding and validity in research.

Selection bias, recall bias, lead-time bias, length-time bias, confounding and ways to reduce them.

Bias questions describe a flaw in a study and ask for its name. Learn the common biases by definition.

Biases

  • Selection bias: nonrandom sampling or loss to follow-up (Berkson, attrition, healthy worker).
  • Recall bias: cases remember exposures differently than controls.
  • Measurement bias: the instrument is wrong; Hawthorne effect changes behavior when observed.
  • Lead-time bias: screening finds disease earlier and appears to lengthen survival.
  • Length-time bias: screening catches slow-growing disease preferentially.
  • Observer or interviewer bias: expectations change measurement.
  • Pygmalion effect: researcher expectations affect outcomes.

Confounding

A third variable associated with exposure and outcome distorts the relationship. Control with randomization, restriction, matching, stratification or multivariate analysis. Effect modification is a real difference by subgroup.

Validity

Internal validity is whether results are true for the study sample; external validity (generalizability) is whether they apply to other populations.

Keep going in biostatistics, epidemiology and ethics

Related on USMLE WEB

Frequently asked questions

What is lead-time bias?

Earlier diagnosis seems to improve survival without changing the outcome.

How does randomization help?

It balances known and unknown confounders.