Monitoring machine learning (ML)-based risk prediction algorithms in the presence of confounding medical interventions