Skip to main content
. 2023 Dec 7;2:e46717. doi: 10.2196/46717

Table 4.

Summary of machine learning (ML) methods.

Study, year ML methods Best models Best performance metrics External validation
Inselman et al [27], 2022 GLMNeta, RFb, and GBMc GBM

  • AUCd=0.74

 No
Hurst et al [25], 2022 Lasso, RF, and XGBooste XGBoost

  • 30-d AUC=0.761

  • 90-d AUC=0.752

  • 180-d AUC=0.739

 No
Hogan et al [28], 2022 Cox proportional hazard, LRf, and ANNg ANN

  • AUC=0.636

 No
Zein et al [29], 2021 LR, RF, and GBDTh GBDT

  • Nonsevere AUC=0.71

  • Hospitalization AUC=0.85

  • EDi AUC=0.88

 No
Sills et al [30], 2021 AutoML, RF, and LR AutoML

  • AUC=0.914

 No
Hozawa et al [31], 2021 XGBoost XGBoost

  • AUC=0.656

 No
Lisspers et al [32], 2021 XGBoost, LGBMj, RNNk, and LR (Lasso, Ridge, and Elastic Net) XGBoost

  • AUC=0.90

 No
Ananth et al [23], 2021 LR, DTl, and ANN LR

  • AUC=0.802

 No
Tong et al [33], 2021 WEKAm and XGBoost XGBoost

  • AUC=0.902

 Yes
Mehrish et al [24], 2021 GLMn, correlation models, and LR LR

  • AUC=0.78

 No
Xiang et al [4], 2020 LR, MLPo, and LSTMp with an attention mechanism LSTM with an attention mechanism

  • AUC=0.7003

 No
Cobian et al [34], 2020 LR, RF, and LSTM LR with L1 (Ridge)

  • AUC=0.7697

 No
Luo et al [35], 2020 WEKA and XGBoost XGBoost

  • AUC=0.859

 No
Roe et al [22], 2020 XGBoost, NNq, LR, and KNNr XGBoost

  • AUC=0.75

 No
Luo et al [26], 2020 WEKA and XGBoost XGBoost

  • AUC=0.820

 Yes
Wu et al [21], 2018 LSTM LSTM

  • Binary classification F1-score=0.8508

  • Multiclass classification F1-score=0.4976

 Yes
Patel et al [11], 2018 DT, Lasso, RF, and GBDT GBDT

  • AUC=0.84

 No

aGLMNet: Lasso and Elastic-Net Regularized Generalized Linear Models.

bRF: Random Forest.

cGBM: gradient boosting method.

dAUC: area under the curve.

eXGBoost: extreme gradient boosting.

fLR: logistic regression.

gANN: artificial neural network.

hGBDT: gradient boosting decision tree.

iED: emergency department.

jLGBM: light gradient boosting method.

kRNN: recurrent neural network.

lDT: decision tree.

mWEKA: Waikato Environment for Knowledge Analysis.

nGLM: Generalized Linear Model.

oMLP: multilayers perceptron.

pLSTM: long short-term memory.

qNN: neural network.

rKNN: K-nearest neighbor.