Table 1.

Description of alternative traditional and enhanced risk prediction models.

Model	Model type	Feature description	Knowledge-driven feature count	Data-driven feature type
Traditional model 1	LR	Age, gender, disease risk scores	4	–
Traditional model 2	LR	Age, gender, disease risk scores, costs	14	–
Enhanced model	LR	Full knowledge- driven feature set	939	–
Non-sequential machine learning models	FNN, GBM	Full knowledge- driven feature set, data driven features	939	Single patient-level feature vectors
Sequential deep learning models	CNN, LSTM	Full knowledge- driven feature set, data driven features	939	Temporal patient-level feature vectors

CNN, Convolutional Neural Network, GBM, Gradient Boosting Model; LR, Logistic Regression; LSTM, Long Short-Term Memory.