Biostatistics & epidimiology Flashcards
(8 cards)
Which Type of Bias it is?
“In my case–control study of smoking and MI, 20 % of the smokers moved out of town and never showed for follow-up, while almost all nonsmokers stayed.”
Attrition (loss-to-follow-up) bias – unequal dropout distorts the exposure-outcome link.
Hook: “They attrit and the results don’t fit.”
which bias is it?
“I screened hospital in-patients for alcohol use and compared them with healthy controls from the community.”
Berkson (admission-rate) bias – hospitalized patients are not representative of the source population.
Hook: “You picked from the berk (ward) not the world.”
Which bias is it?
“People who choose to answer my mailed diet survey are more weight-conscious than those who toss it.”
Non-response (volunteer) bias – responders differ systematically from non-responders.
Which bias is it?
“Factory workers seem to have lower overall mortality than the general population in my cohort.”
Answer
Healthy-worker effect – employment selects fitter subjects, making exposures look safer.
Which bias is it?
“I screen a population, find cancer earlier, and 5-year survival looks better even though time of death is unchanged.”
Lead-time bias – earlier detection falsely inflates survival duration.
Hook: “The clock starts earlier, not the life ends later.”
Which type of bias is it?
“A stress test is labeled ‘positive’ whenever the cardiologist knows the patient is a long-time smoker.”
Diagnostic-suspicion (verification) bias – prior exposure info prompts more or different testing.
Hook: “Suspicion changes the decision.”
“My study shows coffee causes lung cancer, but once I stratify by smoking the association disappears.”
Confounding bias – a hidden variable (smoking) relates to both exposure (coffee) and outcome.
Hook: “Coffee’s innocent; smoking was the con.
Which type of bias is it?
In smokers, β-carotene supplementation appears harmful, but in nonsmokers it has no effect.”
Effect modification (interaction), not confounding. The stratum defines a real change in effect.
Rule to remember: If stratifying by the third variable makes the association different in size or direction, that’s effect modification. If it eliminates the association, that variable was a confounder.
