Table 3.

Performance of risk models for complications

Prediction	# Cases (prevalence)	Model	AUROC (95% CI)	AUPRC (95% CI)	F1 Value (95% CI)
Any complication	4,351 (37.92%)	XGBoost	0.755 (0.744, 0.767)	0.651 (0.632, 0.669)	0.621 (0.602, 0.639)
		Logistic regression	0.747 (0.735, 0.760)	0.646 (0.628, 0.665)	0.629 (0.615, 0.644)
Heart failure	116 (1.01%)	XGBoost	0.835 (0.773, 0.887)	0.101 (0.055, 0.181)	0.141 (0.097, 0.190)
		Logistic regression	0.878 (0.834, 0.915)	0.101 (0.058, 0.184)	0.087 (0.071, 0.104)
Delirium	303 (2.64%)	XGBoost	0.827 (0.793, 0.857)	0.139 (0.099, 0.187)	0.189 (0.153, 0.225)
		Logistic regression	0.873 (0.851, 0.896)	0.181 (0.134, 0.233)	0.169 (0.150, 0.187)
Arrhythmia	341 (2.97%)	XGBoost	0.794 (0.764, 0.822)	0.122 (0.092, 0.165)	0.148 (0.129, 0.169)
		Logistic regression	0.831 (0.800, 0.859)	0.156 (0.121, 0.204)	0.155 (0.138, 0.171)
Kidney failure	505 (4.40%)	XGBoost	0.869 (0.846, 0.891)	0.336 (0.282, 0.390)	0.326 (0.293, 0.359)
		Logistic regression	0.883 (0.863, 0.901)	0.308 (0.258, 0.363)	0.285 (0.262, 0.309)