Session 5 &6 Flashcards
Selection Bias
Non-exchangeability that arises from the processes through which individuals select into or are selected into the study or final analytic sample
Can be both an internal validity and external validity problem
Interval validity - If selection bias causes non-exchangeability in your sample, then the association does not equal a causal effect
External validity - If selection bias only causes a difference between the association in your study and the association in a different population, then both associations can be causal effects, just different causal effects.
Cohort Study
A study design in which you begin with people who are free of the outcome of interest and categorize or select your comparison groups on the basis of exposure. Then, you observe the cohort over time to see who develops the disease of interest
Population Cohort Study
A cohort study tied to a naturally occurring specified population in which you define both exposed and unexposed individuals.
Special Exposure Cohort
A cohort study in which the exposed individuals are identified through a particular location where the number of exposed individuals is enriched. A typical example is when exposure to a particular chemical is generally rare, but is common among workers in some occupational setting. The exposed individuals may be selected from that setting.
End of Study Case Ascertainment
A study in which we attempt to identify all the cases that occurred at any time during the risk period by means of a single assessment at the end of the risk period.
Ongoing (case registry) Ascertainment
A study in which the researcher continually monitors the treatment experience of the cohort during the risk period most typically by means of a case registry.
Periodic Case Ascertainment
A study in which the cohort is evaluated for the outcome at two or more time points during the risk period.
Imperfect Markers of a Confounder
The footprints that confounders typically leave behind in a study - they are associated with the exposure and associated with the disease in the source population. Slightly different markers are used across epidemiologic textbooks. Some textbooks say that confounders are associated with the disease in the unexposed and are not in the causal chain between the exposure and the outcome, or not a consequence of the exposure.
Study Artifacts
Sources of non-comparability that arise from the way a study was conducted (measurement error, selection bias) or analyzed (model misspecification)
Surrogate Confounder
A variable that is associated with a confounder but is not a de-confounder. Control for this type of variable will mitigate but not eliminate confounding caused by the true confounder which it is associated.
