Try out IECV with different models on different clinics

[rubenpeters91]

To get an idea about the heterogeneity of the data with respect to the clinics we have to evaluate the following:

1. Apply Internal/External CV using linear models (logistic regression) and current model where splits are made on clinic level
2. Train separate models per clinic and check performance
3. Use tree-based models with a clinic feature and test performance