Study Design & ProbabilityApril 18, 20264 min read

Comparison table: Bias types (selection, recall, lead-time, length-time)

Quick-hit shareable content for Bias types (selection, recall, lead-time, length-time). Include visual/mnemonic device + one-liner explanation. System: Biostatistics.

Bias questions on USMLE love to test whether you can spot how a study gets distorted—not just the right answer choice. The fastest way to rack up points is to recognize the signature pattern of each bias and match it to the vignette. Below is a quick-hit, shareable comparison of four high-yield bias types: selection, recall, lead-time, and length-time.


The “4 Biases You’ll Actually See” Cheat Sheet

Comparison Table (with mnemonics + one-liners)

Bias typeWhat it is (one-liner)Classic USMLE clueMnemonic / visualHigh-yield consequence
Selection biasWho gets into the study differs from the target population in a way related to exposure/outcome.Low follow-up, volunteer responders, comparing groups recruited differently, healthy worker effect.“SELECT = SICK/HEALTHY people picked differently” → imagine a bouncer choosing who enters the club.Distorts association between exposure and outcome (can exaggerate or hide true effect).
Recall biasCases remember exposures differently than controls (usually over-reporting).Case-control study, past exposure, “patients with disease report more exposure than healthy controls.”“RECALL = RE-Call your memories” → the case is “replaying” exposures on repeat.Creates differential misclassification → typically biases away from the null (but direction can vary).
Lead-time biasEarlier detection makes survival time seem longer without changing time of death.Screening test “improves 5-year survival” but mortality unchanged.“LEAD-time = you got a HEAD START” → stopwatch starts earlier.Inflates survival statistics; does not prove screening reduces mortality.
Length-time biasScreening preferentially finds slow-growing, indolent disease (long preclinical phase).Screen-detected cases have “better prognosis” because aggressive disease shows up between screens (interval cancers).“LENGTH = LONG-lasting tumors are found” → fishing net catches slow fish, fast ones slip by.Screening appears to find “gentler” disease → overestimates benefit of screening.

Quick Pattern Recognition (What the question stem is really asking)

Selection bias: “Your sample isn’t representative”

Look for:

  • Differential loss to follow-up (attrition)
    • Example: If sicker patients drop out of the treatment arm more than placebo, outcomes get skewed.
  • Volunteer bias: People who volunteer are systematically different (health-conscious, higher SES).
  • Healthy worker effect: Workers look healthier than the general population because severely ill people are less likely to be employed.

USMLE nugget: Selection bias can happen in any study design (cohort, RCT, cross-sectional), especially when enrollment or retention differs by exposure/outcome.


Recall bias: “Cases remember differently”

Look for:

  • Case-control design (most classic)
  • Outcomes already known, so subjects are digging through memory for exposures.

Board-style trigger: A mother of a child with a congenital condition is more likely to remember medication use during pregnancy than a mother of an unaffected child.


Lead-time bias: “Survival looks longer because the clock started earlier”

A screening test can:

  • Increase measured survival time (time from diagnosis → death)
  • Without changing actual lifespan (time from disease onset → death)

High-yield distinction:

  • Survival improves ≠ screening is helpful
  • Mortality decreases = stronger evidence screening helps

Length-time bias: “Screening catches slow disease”

Screening is more likely to detect:

  • Slow-growing diseases with a long asymptomatic phase (they “sit around” waiting to be found) Less likely to detect:
  • Aggressive diseases that become symptomatic quickly between screening intervals (“interval cancers”)

USMLE tie-in: This is why screen-detected cancers can appear to have better outcomes even if screening does not meaningfully reduce mortality.


Mini Mnemonic Set (Ultra-shareable)

  • Selection: “Who got selected?” (sampling problem)
  • Recall: “Who remembered what?” (memory problem)
  • Lead-time: “Clock started earlier.” (timing illusion)
  • Length-time: “Screening finds slow disease.” (biology/tempo illusion)

Rapid-Fire Practice (1-liners)

  1. A study recruits participants by posting gym flyersSelection bias (healthier volunteers).
  2. Patients with lung cancer report more asbestos exposure than controlsRecall bias.
  3. PSA screening increases 5-year survival but prostate cancer mortality unchangedLead-time bias.
  4. Mammography finds more low-grade tumors; aggressive ones appear between screeningsLength-time bias.

Test-Day Tip: Choose the bias by the “unit of weirdness”

  • Weirdness in who entered/stayedSelection
  • Weirdness in memory of past exposureRecall
  • Weirdness in survival time after earlier diagnosisLead-time
  • Weirdness in screen-detected cases being less aggressiveLength-time