Predicting venous thromboembolism in hospitalized trauma patients: a combination of the Caprini score and data-driven machine learning model

Table 3 Prediction Metrics of Different Models based on EHR Data & Caprini Scores (Compared with the Caprini RAM)

Methods	No. of Selected features	Accuracy	Precision	TPR_c (Recall)	FPR_c	AUC
Caprini + Lasso + RF	5	0.653	0.362	0.757	0.290	0.799
Caprini + Ridge + RF	50	0.643	0.346	0.746	0.320	0.792
Caprini + ElasticNet + RF	6	0.644	0.357	0.757	0.290	0.799
Caprini + LR + RF	13	0.671	0.366	0.722	0.270	0.800
Caprini + MIE + RF	43	0.637	0.344	0.760	0.330	0.793
Caprini RAM	–	0.750	0.444	0.667	0.227	0.773

ISSN: 1471-227X