Table 2.

Results for K-CNN

Features	AUC	Precision	Recall	F1 Score	Specificity	Youden	Best ED
word	0.882	0.484	0.312	0.408	0.974	0.286	300
word + str	0.888	0.537	0.468	0.500	0.968	0.436	500
word + CUI	0.869	0.440	0.468	0.454	0.953	0.421	300
word + CUIsem	0.835	0.553	0.298	0.387	0.981	0.279	300
word + CUI + str	0.838	0.477	0.369	0.416	0.968	0.337	200

The best f1 score is highlighted in bold. Abbreviation: word: word embedding; str: Structured Data; CUI: Clinical Unified Indentifiers embedding; CUIsem: semantic selected CUI; ED: embedding dimension; AUC: area under the receiver operating characteristic curve. Precision, recall, f1 score, sensitivity, specificity and Youden’s J statistic are showing the value for positive class as distant recurrence.