Table 2 Area under the receiver operating curve (AUC) and area under the precision recall curve (AUPRC) for the models on the three problems. The best performance is in bold
From: Attention-based neural networks for clinical prediction modelling on electronic health records
| Â | Mortality | Readmission | Dementia | |||
|---|---|---|---|---|---|---|
AUC (%) | AUPRC (%) | AUC (%) | AUPRC (%) | AUC (%) | AUPRC (%) | |
LASSO | 92.0 ± 0.33 | 33.8 ± 0.98 | 67.0 ± 0.79 | 18 ± 1.0 | 87.5 ± 0.89 | 11.8 ± 2.0 |
XGBoost | 93.8 ± 0.26 | 37.9 ± 0.98 | 67.5 ± 0.79 | 18.4 ± 1.0 | 87.2 ± 0.92 | 11.6 ± 2.0 |
RETAIN | 93.6 ± 0.27 | 34.2 ± 0.98 | 66.5 ± 0.78 | 16.9 ± 1.0 | 85.6 ± 1.06 | 9.6 ± 2.0 |
GNN | 93.5 ± 0.26 | 36.7 ± 0.98 | 65.7 ± 0.79 | 16.9 ± 1.0 | 86.5 ± 0.92 | 8.9 ± 2.0 |
Transformer | 94.1 ± 0.25 | 37.8 ± 0.98 | 67.1 ± 0.77 | 17.7 ± 1.0 | 87.0 ± 0.91 | 10.0 ± 2.0 |
SARD | 94.5 ± 0.24 | 41.1 ± 0.98 | 68.1 ± 0.77 | 18.5 ± 1.0 | 87.4 ± 0.90 | 11.3 ± 2.0 |