Table 3.

Performance shown with discrimination and calibration evaluation methods in subgroup analysis

	Cohort	Number of patients	Incidence of outcomes	AUROC	AUPRC (relative to incidence)	Calibration in the large	Calibration slope	Weak calibration intercept	Hosmer–Lemeshow intercept
30-day mortality	All surgeries	57,521	4.48%	0.880	0.353 (7.8)	1.01	0.96	0.11	0.00
	Elective	51,695	3.1%	0.886	0.275 (8.9)	1.00	1.12	0.31	0.00
	Emergency	5691	17.1%	0.877	0.608 (3.6)	0.98	1.05	0.14	0.00
	Curative	45,135	3.0%	0.896	0.329 (11)	1.00	1.10	0.23	0.01
	Palliative	2251	15.0%	0.865	0.588 (3.9)	0.95	1.06	0.16	0.00
90-day mortality	All surgeries	57,521	6.65%	0.878	0.431 (6.4)	0.99	1.02	0.07	0.00
	Elective	51,695	4.7%	0.873	0.341 (7.1)	0.99	1.13	0.30	0.01
	Emergency	5691	23.6%	0.864	0.671 (2.8)	1.01	1.13	0.12	0.01
	Curative	45,135	4.5%	0.879	0.367 (8.3)	1.01	1.11	0.24	0.00
	Palliative	2251	25.1%	0.829	0.652 (2.6)	1.03	0.99	0.05	0.01
1-year mortality	All surgeries	57,521	12.9%	0.861	0.559 (4.3)	0.99	1.04	0.06	0.00
	Elective	51,695	10.1%	0.854	0.469 (4.6)	1.00	1.10	0.16	0.01
	Emergency	5691	37.8%	0.852	0.781 (2.1)	1.00	1.13	0.05	0.02
	Curative	45,135	9.0%	0.850	0.429 (4.8)	0.99	1.07	0.11	0.00
	Palliative	2251	46.6%	0.812	0.788 (1.7)	0.97	1.03	0.08	0.01

AUROC area under the receiver operating characteristic, AUPRC area under precision-recall curve