Implementation of a machine learning application in preoperative risk assessment for hip repair surgery

BMC Anesthesiology

Table 3 Predictive performance of machine learning algorithms on primary composite adverse outcomes^*

Algorithm	Accuracy	Sensitivity	Specificity	AUROC (95%CI)
Logistic Regression	0.699	0.710	0.699	0.794 (0.718–0.869)
Random Forest	0.690	0.677	0.690	0.776 (0.704–0.848)
SVM	0.716	0.710	0.716	0.768 (0.677–0.860)
KNN	0.706	0.516	0.711	0.644 (0.542–0.746)
lightGBM	0.703	0.710	0.703	0.786 (0.706–0.867)
MLP	0.691	0.677	0.692	0.777 (0.684–0.859)
XGBoost	0.638	0.645	0.638	0.734 (0.636–0.831)

^*Total 102 patients had primary composite adverse outcomes (Primary composite adverse outcomes included in-hospital mortality (and death in 48 h after discharge), sepsis, acute myocardial infarction, acute stroke, respiratory, liver and renal failure
Abbreviations: AUROC area under receiver operating characteristic curve, CI confidence interval, SVM support vector machine, KNN K nearest neighbor, light GBM light gradient boosting machine, MLP multi-layer perception, XGBoost extreme gradient boosting

ISSN: 1471-2253