Selection Bias

What leads to selection bias?

Using procedures to select study subjects that lead to a different result than one would've gotten from the entire population actually targeted.

Can selection bias occur in prospective cohort and experimental studies?

Yes, from differential loss to follow-up because this impacts which subjects are selected for analysis.

Spurious

fake

Internal Validity

was the study done without bias, confounding, or random error

Bias

A systematic error that results in an incorrect (invalid) estimate of the measure of association between exposure and outcome

List the studies bias can occur in:

All study types: experimental, cohort, and case-control

Three things to assess to determine validity of a study

Bias, Confounding, Random Error

Where in the study does bias occur?

In the design and conduct of the study

Who primarily introduces bias?

Investigator or study participants. It is not intentional, rather, something overlooked

Bias toward the null

Mask an association; OR or RR will move toward 1 when should be <1 or >1

Can bias be fixed in the analysis phase?

No, it can be evaluated for effect on outcome, but not fixed.

Bias away from the null

Spurious association; Assumes an association where one doesn't exists. OR or RR will be <1

Selection Bias

Study subjects don't represent target population.

What studies does selection bias most likely occur in?

case-control or retrospective cohort because exposure and outcome already occurred

External Validity

is the study generalizable to all populations, not just the study population