Table 2.
Model performance for predicting bacteremia according to forecasting time to event and time steps of the recurrent neural network model.
| Variables | AUROCa | AUPRCb | |
| Prior time to event (hour) | |||
|
|
0 (at event) | 0.98 | 0.17 |
|
|
8 prior | 0.96 | 0.18 |
|
|
16 prior | 0.95 | 0.17 |
|
|
24 prior | 0.93 | 0.15 |
| Time steps (days) | |||
|
|
1 | 0.98 | 0.14 |
|
|
2 | 0.98 | 0.15 |
|
|
4 | 0.98 | 0.15 |
|
|
6 | 0.98 | 0.15 |
|
|
8 | 0.97 | 0.16 |
|
|
10 | 0.97 | 0.174 |
|
|
12 | 0.98 | 0.165 |
aAUROC: area under the receiver operating characteristic curve.
bAUPRC: area under the precision-recall curve.