Intelligent prediction of RBC demand in trauma patients using decision tree methods

Table 2 Comparison between the LR, CRT and XGBoost models in predicting blood transfusion

Parameter type	Methods	AUC	Sensitivity	Specificity	Accuracy	Youden index	P-value
Non-invasive parameters	XGBoost	0.705	0.66	0.77	0.75	0.19	< 0.001
	LR	0.716	0.86	0.50	0.55	0.12
	CRT^*	0.692	0.89	0.42	0.48	0.16
All parameters	XGBoost	0.937	0.94	0.82	0.83	0.10	< 0.001
	LR^#	0.797	0.80	0.70	0.72	0.12
	CRT^#&	0.816	0.69	0.92	0.89	0.09

^*Non-invasive parameter prediction, there was a significant difference in the AUC between CRT and the XGBoost model (P < 0.05)
^#All parameter prediction, there was a significant difference in the AUC between LR and the XGBoost model (P < 0.05), and there was a significant difference in the AUC between CRT and the XGBoost model (P < 0.05)
^&All parameter prediction, there was a significant difference in the AUC between CRT and the LR model (P < 0.05)
AUC area under the curve, XGBoost eXtreme gradient boosting, LR logistic regression, CRT classification and regression tree

ISSN: 2054-9369