Table 3.
Model performance comparison.
| Model 1a | Model 2 | Model 3 | Model 4 | |
|---|---|---|---|---|
| Discrimination | ||||
| C statistics (95% CI) | 0.364 (0.196, 0.601) | 0.421 (0.242, 0.523) | 0.526 (0.341, 0.622) | 0.625 (0.59, 0.66) |
| Calibration | ||||
| H-L | X2 = 12.623, p = 0.5947 | X2 = 9.156, p = 0.609 | X2 = 7.263, p = 0.764 | X2 = 4.539, p = 0.806 |
| Reclassification | ||||
| IDI, 95%CI | Reference | 0.005 (−0.012, 0.051) | 0.006 (−0.009, 0.058) | 0.090 (0.049, 0.133) |
| NRI, 95%CI | Reference | 0.102 (−0.167, 0.521) | 0.201 (−0.148, 0.407) | 0.457 (0.323, 0.573) |
CI, confidence interval; H-L, Hosmer–Lemeshow statistic; NRI, net reclassification improvement; IDI, integrated discrimination improvement.
a
Model 1 included the following variables: age, mean arterial pressure, albumin, C-reactive protein, and phosphorus; Model 2 = Model 1 + pulse wave velocity; Model 3 = Model 1 + galectin-3; Model 4 = Model 1 + pulse wave velocity + galectin-3.