Confounding Flashcards
What is confounding?
Give examples of potential confounders?
When the association between an exposure and outcome is distorted due to an independent factor (a confounder).
Independent = indep. from BOTH the outcome and exposure. AND not on the causal pathway.
Age, SE status, sex, ethnicity, smoking status, education, PA levels etc.
How can we control for confounding?
In study design phase AND in analysis phase.
How can we control in study design? (3 ways)
Randomization - only used in intervention studies. This leads to equal dist of known and unknown confounders.
Eg. parts are randomly assigned to control vs intervention.
Restriction: restrict known confounders from your study.
Eg. if being a men is a known confounder - use only women in the study pop. This does mean that the results can’t be extrapolated to populations of men.
Matching: in case -control studies. Pick a matching person in each case vs control group so condounders are dist. equally.
How can we control in analysis phase? (2 ways)
Stratification: This is calculating the measure of assoc differently for each group. Eg. to control for age.
NB. try to limit the number of stratum as each group lowers the sample size and increases chance.
Stat mdelling - regression analysis.
What is residual confouding?
Give an example.
When we don’t control enough for confounder during study design so we get residual confouding.
Example: study of throat cancer and alcohol consumption
Potential confounder: smoking status.
Try to control by comparing smokers vs non-smokers in the group by asking whether you smoke or not (Yes/No).
BUT
the number of cigarettes would have been a better measure. So we could have done more in the first instance to control for confounding and we get residual which would UNDERESTIMATE the effect of the confounder and bias the association
What effect does residual have on study finding?
Underestimates the effect of the confounder
Bias’s the assocation between exposure and outcome.
