Table 3.

Mixed-effects meta-regression results for AUC.

Variable	Category	Estimate	SE	t-value	d.f.	p-value
Tool type	Logistic regression - reference	0.857	0.010
	Simple score	-0.029	0.012	-2.425	381	0.016
	Cox regression	-0.001	0.014	-0.033	381	0.979
	SVM	0.040	0.026	1.576	381	0.116
	Random forest	0.017	0.011	1.548	381	0.122
	Boosting	0.032	0.011	2.986	381	0.003
	Neural Network	0.039	0.020	1.956	381	0.051
	Deep Learning	0.024	0.014	1.680	381	0.094
Region	Europe (reference)
	USA	-0.009	0.010	-0.886	381	0.376
	China	0.069	0.009	7.798	381	< 0.001
	Other region	0.006	0.009	0.660	381	0.510
Rate of severe cases		-0.146	0.027	-5.411	381	< 0.001
The rate of severe cases squared		0.478	0.103	4.649	381	< 0.001
CRP used	No (reference)
	Yes	0.022	0.007	3.134	381	0.002

SVM: Support Vector Machine; Boosting: any tool that relies on boosting, e.g., XGBoost, or Gradient Boosting Machine; CRP: C-reactive protein; Rate of severe cases is centred at 25%.