Why Use Stratified Cross Validation in Epidemiology?

Epidemiological data often involves binary or categorical outcomes with significant class imbalances. For instance, in a study of a rare disease, the number of cases (positive class) may be much smaller than the number of controls (negative class). Using standard cross validation could lead to some folds having very few or no positive cases, resulting in biased estimates of model performance. Stratified cross validation helps to mitigate this issue by maintaining the same class distribution across all folds.