BIOL 437 Week Eight P.3 (Statistical Methods in Analytical Epidemiology) Flashcards
double cohort
- 2 distinct populations are invovled with different levels of an exposure of interest
- employed when expsoure is rare and a relatively small number of people are affected
selecting the study cohort
-from population
>chose those at risk of becoming a case
exclude from study cohort
- those who already have a disease outcome of interest
- those not at risk
- those with latent infections or recurring diseases
restriction
- used to improve validity of study
- selecting cohorts with limited exposure, narrow behaviours
- need to come from population where sampling can be effectively conducted
- limits generalization, but often improves feasibility and focus
bias in cohort studies
- Selection bias
- Confounding
>more a concern in double cohort studies - Misclassification
selection bias types
- Healthy worker effect
- Volunteer bias
- Loss to follow-up
healthy worker effect
- when workers represent exposed group and sample from general population represents the unexposed
- workers tend to be healthier on average than the general population
loss to follow-up
- lose contact, resulting in unavailable outcome data
- common problem
- general rule: validity of study requires a loss of follow-up not exceeding 20%
reasons for loss to follow-up
- refusal to participate
- unable to locate
- unable to be interviewed
- death
confounding
- more a problem in double-cohort studies
- can influence associates in both case-control and cohort studies
differential misclassification
- related to exposure
- if exposure classification influences differential accuracy in ascertaining outcome info
- overestimates association
non-differential misclassification
-not related to exposure
-through inaccuracies in classifying exposure status
>occur similarly between exposed and unexposed
-underestimates association
-error affects both groups and can make them more similar
avoid healthy worker bias
-selecting a comparison group made up of workers, only unexposed
minimize misclassification
- refine definition of exposed and unexposed
- avoid exposure classifications that result in differential outcome ascertainment
minimize loss to follow-up bias
- restrict study to those likely to remain
- collect personal identifying info
- make periodic contact and provide incentives
control confounding at study design level
-restriction to avoid bias due to confounding
-in double cohort:
>choose comparisions as a like as possible to the exposed population
control confounding at analysis level
-collect data on potential confounders at begining
>adjust through stratification and multiple regression techniques
effect modification
- when association between exposure and outcome is modified by the level of an extrinsic factor beyond random variation
- may occur in cohort or case-cohort
effect modifier
-extrinsic variable
effect modification can influence
-relationship between variables in cohort and case-control data
>can influence OR, RR, and rate ratios
confounding vs. effect modification
- treated differently
- control for confounding
- present results from effect modificaiton
- sometimes a variable can act as a confounder and an effect modifier
effect-modifying variable
- influences association between 2 other variables in an informative way
- can be informative and meaningful
if OR crude > OR1=OR2
-positive confounding
if OR crude < OR1=OR2
-negative confounding
