Table 3:

Predicted probability of 30-day mortality in the derivation and validation cohorts, by procedure category

Performance metric	CABG	AVR	CABG + AVR
Derivation cohort: FY2017 and FY2018
Predicted probability of 30-day mortality, %	1.80	1.95	4.15
C-statistic	0.8882	0.8500	0.8400
Optimism-corrected c-statistic	0.8659	0.7618	0.7644
Hosmer–Lemeshow χ² statistic p value	0.2	0.08	0.7
Brier score^*	0.0150	0.0136	0.0269
Calibration slope	0.8988	0.6932	0.6634
Validation cohort: FY2015–FY2016
Predicted probability of 30-day mortality, %	1.35	0.95	3.80
C-statistic	0.8801	0.7882	0.7591
Optimism-corrected c-statistic	Not applicable	Not applicable	Not applicable
Hosmer–Lemeshow χ² statistic p value	0.2	0.06	0.001
Brier score^*	0.0142	0.0145	0.0320
Calibration slope	0.9672	0.8426	0.9218
Validation cohort: FY2019
Predicted probability of 30-day mortality, %	2.65	1.10	1.55
C-statistic	0.8587	0.9008	0.8175
Optimism-corrected c-statistic	Not applicable	Not applicable	Not applicable
Hosmer–Lemeshow χ² statistic p value	0.7	0.8	0.01
Brier score^*	0.0136	0.0128	0.0309
Calibration slope	0.9491	0.9596	1.0061

Note: AVR = aortic valve replacement, CABG = coronary artery bypass graft, FY = fiscal year.
↵* The Brier score is the average squared prediction error, where lower values indicate better model performance.