Table 5.
Comparison of the four models.
| Pseudo–R2 | Model CLIN | Model CLIN-F | Model CLIN-COG | Model CLIN-COG-F | ||
|---|---|---|---|---|---|---|
| Likelihood ratio tests | Model CLIN | 0.242 | χ 2(1) = 8.45, p = 0.004* | χ 2(3) = 38.81, p < 0.001* | χ 2(4) = 44.80, p < 0.001* | |
| Model CLIN-F | 0.256 | χ 2(2) = 30.37, p < 0.001* | χ 2(3) = 36.36, p < 0.001* | |||
| Model CLIN-COG | 0.304 | χ 2(1) = 5.99, p = 0.014* | ||||
| Model CLIN-COG-F | 0.313 | |||||
| AUC [95%-CI] | ||||||
| DeLong's tests for two ROC-curves in a correlated population | Model CLIN | 0.760 [0.721–0.800] | z = −1.41, p = 0.159 | z= −3.61, p< 0.001* | z= −3.50, p< 0.001* | |
| Model CLIN-F | 0.769 [0.731–0.808] | z= −2.67, p= 0.008* | z= −3.26, p= 0.001* | |||
| Model CLIN-COG | 0.798 [0.762–0.835] | z = −0.67, p = 0.504 | ||||
| Model CLIN-COG-F | 0.801 [0.765–0.838] | |||||
| Threshold | Sensitivity | Specificity | Positive predictive value PPV | Negative predictive value NPV | ||
| Sensitivity/Specificity analysis (Method: maximal youden-index) | Model CLIN | 0.237 | 0.691 | 0.703 | 0.417 | 0.881 |
| Model CLIN-F | 0.249 | 0.663 | 0.734 | 0.433 | 0.878 | |
| Model CLIN-COG | 0.299 | 0.635 | 0.819 | 0.518 | 0.875 | |
| Model CLIN-COG-F | 0.290 | 0.619 | 0.841 | 0.544 | 0.878 |
Comparison of all less complex models regarding goodness of fit of the underlying multivariate regression analysis and differences between the AUCs of these four models. Columns represent the more complex model; hence, if significance is reached, the model stated in the column has better goodness of fit or greater AUC than the model defined by the name of the row. Analysis of sensitivity and specificity under a maximal Youden-Index can be found below as well. Statistically significant tests results are highlighted in bold type, significant p-values are marked with *, and p-values at the significance level of .05 to 0.1 are marked with (*).