Instability in clinical risk stratification models using deep learning