Table 2.

Description of feature sets used in machine learning experiments

Name of Feature Set	Type of Feature	Description
Note_count	Integer	The number of clinical notes by a medical student for each admission
Age	Integer	Age of patient
Words (baseline)	Binary	Bag of Words features; “1” if word present; “0” if word absent
CUI	Binary	Concept code features; “1” if CUI present; “0” if CUI absent
CUI_NEG	Binary	Concept code with negation: If a CUI is negated in the note, the value of CUI is “0” and the value of CUI_NEG is “1”
CUI_SEC	Binary	Dyad of concept code (CUI) and section: “1” if CUI present in the section; “0” if CUI absent in the section
CUI_count	Integer	The count of each CUI in the notes for one admission
CUI_count_tfidf	Numeric	The TFIDF value of each CUI in the notes for one admission
STY	Binary	Semantic type features (as defined in the UMLS): “1” if semantic type present; “0” if semantic type absent