Stratified sampling study designs are commonly implemented to improve the likelihood of observing interesting sub-populations that might be missed under standard, simple random sampling designs. However, for analyses, the sampling scheme must to be acknowledged appropriately to ensure unbiased and efficient estimation of population parameters and their standard errors. In many circumstances, it is common to assume stratum definitions are measured without error and thus the generalization from the sample to the population is relatively straightforward. But what if the variables used to define the sampling strata are measured inaccurately, thus resulting in misclassification of stratum membership and what effect might this have on the population summaries? In this talk, I will introduce the motivation for these questions which includes the creation of a maximum variation sampling scheme wherein strata are defined using both EHR and census data. I will then review design-based analytical approaches to survey analysis, introduce methods to account for stratum misclassification and summarize results from a simulation study designed to investigate the effects of misclassification when applying design-based methods.