Lead Author(s): Jeff Martin, MD

Definition of Stratification

Forming separate strata or groups during study analysis or during study design.

Strata to Determine if Confounding Is Present

To determine if confounding is present we need to look at potential associations (diagram below) 0326_2x2crude.JPG

First, we look at the crude or unadjusted estimates, Second, to determine if confounding is present, we stratify for the potential confounding variable in question by You can determine an odds ratio (OR) separately in each stratum, depicted in this example where the potential confounder:

Adjusted Estimate from the Stratified Analyses

If there are multiple confounders, we need to create a stratum for each one.

Goal: Create a single unconfounded ( adjusted) estimate for the relationship in question Process: Summarize the unconfounded estimates from the two (or more) strata to form a single overall unconfounded adjusted estimate

Advantages of Stratification

The primary advantage is that this is a very straightforward and easy to comprehend approach. Many reviewers are phobic of fancy regression models and, hence, stratification, if you can do it is typically very easily understood.

Stratification is also a very easy way to evaluate for the presence of interaction.

Limitations of Stratification

Requires continuous variables to be discretized The first limitation is what do we do if we have continuous variables as our exposure or potential confounders, something like age, for example?

Deteriorates with multiple confounders

Solution to Limitations of Stratification

Mathematical modeling (multivariate regression) e.g. The solution to all of these limitations lies in the use of mathematical models also known as multivariate regression.