Table 1.
Discrimination and calibration of the studied scores for predicting mortality in colorectal patients
| O | E | O:E | Discrimination | Calibration | ||||
|---|---|---|---|---|---|---|---|---|
| AUC (95%CI) | P* | H-L | df | P** | ||||
| POSSUM | 11 | 117 | 0.09 | 0.558 (0.412–0.703) | 0.575 | 9.362 | 8 | 0.313 |
| P-POSSUM | 11 | 30 | 0.37 | 0.557 (0.404–0.711) | 0.576 | 6.343 | 8 | 0.609 |
| CR-POSSUM | 11 | 98 | 0.11 | 0.607 (0.476–0.738) | 0.296 | 3.601 | 8 | 0.891 |
| E-POSSUM | 11 | 40 | 0.28 | 0.561 (0.406–0.717) | 0.551 | 11.863 | 8 | 0.157 |
| E-PASS | 11 | 41 | 0.27 | 0.458 (0.25–0.667) | 0.683 | 1.884 | 3 | 0.597 |
O observed 30-day mortality rate, E expected 30-day mortality rate, O:E observed to expected mortality ratio (value closest to 1 indicate more accurate model)
Discrimination is measured by area under receiver-operator characteristic curve (AUC) with 95% confidence intervals, *P of the test of difference between scores for the AUC as described by DeLong et al.
Calibration is measured by the Hosmer-Lemeshow test: smaller H-L Chi square values, and larger **P values represent better model calibration