Table 3.

Comparison of Model Statistical Performance and Calibration

Variable	Logistic model using STS score	Logistic model using UHC score
Results for total study population, n	2,171	2,171
Observed number of deaths, n (%)	58 (2.7)	58 (2.7)
Wald chi-squared p value for covariate	< 0.0001	< 0.0001
Statistical performance measures
Maximum adjusted R²	0.129	0.273
C statistic	0.812	0.885
Calibration by deciles of predicted risk
Observed/expected deaths in decile 1	0/3.1	0/2.8
Observed/expected deaths in decile 2	1/3.2	0/2.8
Observed/expected deaths in decile 3	1/3.3	1/2.8
Observed/expected deaths in decile 4	1/3.4	0/2.9
Observed/expected deaths in decile 5	2/3.6	1/2.9
Observed/expected deaths in decile 6	2/3.8	1/2.9
Observed/expected deaths in decile 7	5/4.2	4/3.0
Observed/expected deaths in decile 8	11/4.9	3/3.2
Observed/expected deaths in decile 9	10/6.3	6/3.8
Observed/expected deaths in decile 10	25/22.2	42/30.9
Hosmer-Lemeshow test p value (df = 8)	0.010	0.033
Results for patients without any complication, n	1,528	1,528
Observed number of deaths, n (%)	7 (0.5)	7 (0.5)
Wald chi-squared p value	0.006	0.745
Maximum adjusted R²	0.053	0.001
C statistic	0.794	0.494
Hosmer-Lemeshow test p value (df = 8)	0.446	0.240
Results for patients with any complication, n	643	643
Observed number of deaths, n (%)	51 (7.9)	51 (7.9)
Wald chi-square p value	<0.0001	< 0.0001
Maximum adjusted R²	0.108	0.283
C statistic	0.739	0.876
Hosmer-Lemeshow test p value (df = 8)	0.086	0.005

STS, Society of Thoracic Surgeons; UHC, University HealthSystem Consortium.