Prognosis prediction models for post-stroke depression: a protocol for systematic review, meta-analysis, and critical appraisal

Table 2 Summary of measuring performance of prognostic prediction models

Items	Performance measures/statistics	Values for better performance	Visualization
Discrimination	AUC-ROC	Higher	ROC curve
	ACC	Higher
	BER	Lower
	D statistics	Lower
	MCC	Higher
	F1 score	Higher
	Log-rank	Lower, p > .05
	AUC-PRC	Higher	PRC curve
Calibration	Calibration curve, slope, and intercept	Slope closer to 1 and intercept closer to 0	Calibration plot
	Reliability-deviation	Lower
	Reliability–within-bin variation	Lower
	Reliability–within-bin covariance	Higher
	Resolution	Higher
	Predictive range	Higher
	Emax	Lower
	Eavg	Lower
	Hosmer–Lemeshow test	Higher
	Total O:E ratios	Higher
Classification	RMSE via neighborhood estimate	Lower	Reclassification scatter plot
	R² statistic	Higher
	TPR	Higher
	TNR	Higher
	NRI	Higher
	IDI	Higher
Overall performance	Brier score	Lower
	Brier skill score	Higher
	Prediction squared error	Lower
	Decision Curve Analysis	/	DCA plot

AUC-ROC Area under receiver operating characteristic curve, TPR Sensitivity, ACC Accuracy, BER Balanced error rate, MCC Matthews correlation coefficient, TNR Specificity, AUC-PRC Area under precision-recall curve, RMSE Root-mean-squared error, NRI Net reclassification improvement, IDI Integrated discrimination improvement

ISSN: 2046-4053