Ethnicity (Hispanic/Latino versus non-Hispanic/Latino) and race are also included in the data. Often, races that are prone to poor socioeconomic status have worse outcomes in healthcare. Let's leave the unimputed ethnicity and race variables (ETHUN and RACEUN) as is. We can remove the redundant RACER variable as well as the imputed versions of ethnicity and race (ETHIM and RACERETH):

X_train.drop(['ETHIM','RACER','RACERETH'], axis=1, inplace=True)
X_test.drop(['ETHIM','RACER','RACERETH'], axis=1, inplace=True)