Table 3.
Prediction (AUROC) of each model at different hours in the sepsis studies
| Study | Model | Algorithm | Different hours |
||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| −48 | −24 | −12 | −10 | −8 | −6 | −5 | −4 | −3 | −2 | −1 | −0.25 | 0 | |||
| Delahanty et al. (2019) | RoS | Gradient boosting | 0.97 | 0.93 | |||||||||||
| Barton et al. (2019) | MLA | Gradient boosted trees | 0.83 | 0.84 | 0.88 | ||||||||||
| Kam and Kim (2017) | SepLSTM | long short-term memory | 0.93 | 0.94 | 0.96 | 0.99 | |||||||||
| Bloch et al. (2019) | SVM-RBF | SVM-RBF | 0.8141 | 0.8879 | 0.8807 | 0.8639 | 0.8675 | ||||||||
| Nemati et al. (2018) | Weilbull-Cox proportional hazards | Weilbull-Cox proportional hazards | 0.79 | 0.8 | 0.81 | 0.82 | |||||||||
| Lauritsen et al. (2020) | CNN-LSTM | CNN-LSTM | 0.752 | 0.792 | 0.842 | 0.879 | |||||||||
| Scherpf et al. (2019) | RNN | RNN | 0.76 | 0.79 | 0.81 | ||||||||||
Abbreviation:RoS: Risk of Sepsis; MLA: machine learning algorithm; LSTM: long short-term memory; SVM-RBF: support vector machines with radial basis function; CNN-LSTM: convolutional neural network-long short-term memory; RNN: recurrent neural network.