Lecture 28 - Confounding II Flashcards
Controlling confounding in the study analyses
- Stratification
- Multivariable analysis
What is stratification
It is a method to control for confounding by diving the potential confounder into strata and calculating the MOA for each stratum
Basic recipe for Stratification
- Calculate the crude MOA between exposure and outcome
- Divide potential confounder into strata
- For each strata, calculate MOA
- Compare the stratum-specific MOA
What if Crude MOA = stratum -specific MAO
There is no major confounding. You can use the crude measure of association
What if crude MOA ≠ stratum specific MOA
Confounding is present. Use adjusted MOA
The pros of Stratification
- Easy for small number of potential confounders with limited strata
- Can evaluate impact of confounding
- Can identify effect modification
Cons of stratification
- Can leave residual confounding
- Not feasible when dealing many potential confounders or many strata
What is multivariable analysis
A statistical method to estimate the MOA while controlling for potential confounder, often recognised by the term regression
- can work in situations where stratification won’t
Misc point potential issue
- Residual confounding
- Can only control what you’ve measured
Standardisation
- similar issues as stratification with multiple potential confounders/number of strata
- Multivariable analysis is often more efficient in analytic studies
Assessing confounding - how much change indicated confounding
General guideline is if controlling for confounder changes the MOA by 10% or more
Effect modification vs Confounding
Effect modification - The association between exposure and outcome differs across strata of the effect modifier. It is an important finding.
Confounding - A third factor distorting the association. A nuisance
Identifying effect modification
If stratum-specific MOA are different, effect modification is present. Report stratum-specifc MOA instead if Crude MOA
