Table 3. Measures of model accuracy.
| Basic baseline prognostic model a | Basic baseline prognostic model plus baseline levels of ln NT-proBNP a (Model 1) | Basic baseline prognostic model plus 12 months levels of ln NT-proBNP a (Model 2) | Model 1 plus 12 months relative changes of ln NT-proBNP as continuous variable b (Model 3) | Model 1 plus 12 months relative change of ln NT-proBNP in categories b (Model 4) | |
|---|---|---|---|---|---|
| Model fit | |||||
| Likelihood Ratio (χ2) (df) | 78.2 (17) | 87.02 (18) | 98.9 (18) | 97.4 (19) | 96.4 (21) |
| Discrimination | |||||
| C-statistic [95% CI] | 0.69 [0.64, 0.74] | 0.71 [0.66, 0.76] | 0.72 [0.67, 0.77] | 0.72 [0.67, 0.77] | 0.71 [0.66, 0.76] |
| Reclassification | |||||
| Integrated Discrimination Improvement [95% CI] | 0.019 [0.006, 0.033] | 0.047 [0.027, 0.066] | 0.043 [0.025, 0.061] | 0.039 [0.022, 0.056] | |
| Continuous NRI | 0.487 | 0.597 | 0.568 | 0.535 | |
| NRI [95% CI] according to risk strata <10%, 10–20%, > 20% | 0.047 [−0.039, 0.133] | 0.098 [0.002, 0.194] | 0.086 [−0.009, 0.182] | 0.077 [−0.015, 0.170] | |
| Expected number of subjects to be reclassified | |||||
| Subjects with CVD-event | nup/ndown | 12/9 | 13/12 | 13/12 | 11/12 |
| Subjects without CVD.event | nup/ndown | 65/79 | 80/141 | 83/136 | 77/136 |
aadjusted for age, sex, BMI, blood pressure, LV function, current smoking, history of diabetes, total cholesterol, HDL-cholesterol, GFR std, use of statins
badjusted for baseline levels of ln NT-proBNP, age, sex, BMI, blood pressure, LV function, current smoking, history of diabetes, total cholesterol, HDL-cholesterol, GFR std, use of statins