Table 2.
Performance of the AKDa prognostic mortality models for elderly patients.
| Cohort and models | AUROCb (95% CI) | Cutoff | Sensitivity (95% CI) | Specificity (95% CI) | PPVc (95% CI) | NPVd (95% CI) | |||||||
| Training cohort | |||||||||||||
|
|
LRMe | 0.698 (0.675-0.721) | 0.308 | 0.683 (0.661-0.704) | 0.619 (0.584-0.653) | 0.451 (0.421-0.481) | 0.810 (0.790-0.829) | ||||||
|
|
XGBoostf | 0.870 (0.853-0.886) | 0.312 | 0.772 (0.752-0.791) | 0.793 (0.763-0.821) | 0.594 (0.564-0.624) | 0.899 (0.883-0.913) | ||||||
|
|
LightGBMg | 0.861 (0.843-0.878) | 0.334 | 0.798 (0.779-0.816) | 0.754 (0.722-0.784) | 0.610 (0.579-0.641) | 0.885 (0.869-0.899) | ||||||
|
|
MLPh | 0.731 (0.709-0.753) | 0.332 | 0.737 (0.716-0.757) | 0.603 (0.568-0.637) | 0.491 (0.459-0.523) | 0.815 (0.796-0.833) | ||||||
|
|
RFi | 0.844 (0.826-0.862) | 0.351 | 0.887 (0.872-0.901) | 0.632 (0.597-0.666) | 0.702 (0.667-0.736) | 0.851 (0.835-0.867) | ||||||
|
|
KNNj | 0.717 (0.695-0.740) | 0.313 | 0.662 (0.640-0.683) | 0.648 (0.614-0.682) | 0.447 (0.418-0.476) | 0.817 (0.797-0.837) | ||||||
| Internal validation cohort | |||||||||||||
|
|
LRM | 0.720 (0.697-0.742) | 0.301 | 0.676 (0.642-0.709) | 0.669 (0.647-0.690) | 0.449 (0.420-0.479) | 0.838 (0.818-0.856) | ||||||
|
|
XGBoost | 0.810 (0.790-0.830) | 0.290 | 0.793 (0.763-0.822) | 0.686 (0.665-0.707) | 0.503 (0.474-0.532) | 0.893 (0.876-0.908) | ||||||
|
|
LightGBM | 0.868 (0.851-0.885) | 0.303 | 0.828 (0.799-0.854) | 0.753 (0.733-0.773) | 0.573 (0.543-0.602) | 0.916 (0.901-0.929) | ||||||
|
|
MLP | 0.750 (0.728-0.771) | 0.339 | 0.633 (0.598-0.667) | 0.744 (0.724-0.764) | 0.497 (0.465-0.529) | 0.835 (0.817-0.853) | ||||||
|
|
RF | 0.706 (0.605-0.759) | 0.330 | 0.625 (0.561-0.637) | 0.761 (0.745-0.776) | 0.567 (0.533-0.599) | 0.776 (0.760-0.790) | ||||||
|
|
KNN | 0.725 (0.703-0.748) | 0.278 | 0.762 (0.730-0.792) | 0.558 (0.535-0.580) | 0.408 (0.382-0.434) | 0.854 (0.833-0.873) | ||||||
| External validation cohort | |||||||||||||
|
|
LRM | 0.772 (0.701-0.843) | 0.285 | 0.706 (0.612-0.790) | 0.740 (0.628-0.834) | 0.640 (0.532-0.739) | 0.794 (0.700-0.869) | ||||||
|
|
XGBoost | 0.698 (0.620-0.776) | 0.385 | 0.706 (0.612-0.790) | 0.636 (0.519-0.743) | 0.605 (0.490-0.712) | 0.733 (0.638-0.815) | ||||||
|
|
LightGBM | 0.746 (0.673-0.820) | 0.312 | 0.716 (0.621-0.798) | 0.740 (0.628-0.834) | 0.648 (0.539-0.747) | 0.796 (0.703-0.871) | ||||||
|
|
MLP | 0.770 (0.699-0.841) | 0.339 | 0.789 (0.700-0.861) | 0.701 (0.586-0.800) | 0.701 (0.586-0.800) | 0.789 (0.700-0.861) | ||||||
|
|
RF | 0.716 (0.639-0.792) | 0.337 | 0.606 (0.507-0.698) | 0.740 (0.628-0.834) | 0.570 (0.467-0.669) | 0.767 (0.664-0.852) | ||||||
|
|
KNN | 0.602 (0.519-0.685) | 0.188 | 0.312 (0.227-0.408) | 0.909 (0.822-0.963) | 0.483 (0.399-0.567) | 0.829 (0.679-0.928) | ||||||
aAKD: acute kidney disease.
bAUROC: area under the receiver operating characteristic curve.
cPPV: positive predictive value.
dNPV: negative predictive value.
eLRM: logistic regression model.
fXGBoost: Extreme Gradient Boosting.
gLightGBM: Light Gradient Boosting Machine.
hMLP: multilayer perceptron.
iRF: random forest.
jKNN: K-nearest neighbor.