Table 1.
Brief review of ML models and patients groups for predicting hospital patients' LoS.
| References | Outcome: prediction type | ML models | Target group | Results |
|---|---|---|---|---|
| Mekhaldi et al. (2020) | Regression | RF Regressor, Gradient Boosting Regressor | General patients | GB performed better than RF; performance were checked by MSE, the R-squared and the Adjusted R-squared. |
| Daghistani et al. (2019) | Classification | RF, ANN, SVM, BN | Cardiac patients | RF model outperformed all other models: sensitivity (0.80), accuracy (0.80), and AUROC (0.94). |
| Tsai et al. (2016) | Regression | LR, ANN | Cardiac patients | LR model performed slightly better than ANN models, with a MAE value of 3.76 and 3.87 |
| Symum and Zayas-Castro (2020) | Classification | DT C5.0, linear SVM, KNN, RF, and multi-layered artificial neural net | Chronic disease (congestive heart failure, acute myocardial infarction, COPD, pneumonia, type 2 diabetes). | For all patient groups, LSVM (Lagrangian SVM) with wrapper feature selection performed well. |
| Tanuja et al. (2011) | Classification | Naive Bayes; KNN; DT classifiers; Multi-layer backpropagation | General patients | MLP and NB models had the best classification accuracy of around 85%, while KNN performed poorly with only 63.6% accuracy |
| Combes et al. (2014) | Regression and classification | Two based models: Classifier: RF, LMT (Logistical model tree), MP, DT (C4.5-J48), NBTree, REPTree, and SVM. Regression: LR, SV regression, MLP, IRM (Isotonic regression model), M5P, PRLM (Pace regression linear models) | Pediatric | Using 10-fold cross-validation, obtained the best performances in using logistic regression, and in continuous outcome SVM Regression showed a lower prediction error. |
| Etu et al. (2022) | Classification | LoR, GB, DT, and RF | COVID-19 Patients | The GB model outperformed the baseline classifier (LoR) and tree-based classifiers (DT and RF) with an accuracy of 85% and F1-score of 0.88 for predicting ED LoS |
| Alsinglawi et al. (2020a) | Regression | RF Regressor; GB Regressor; Stacking Regressor; DNN | Cardiovascular patients in the ICU | GB regressor outweighed the other proposed models, and showed a higher R-squared. |
| Kirchebner et al. (2020) | Classification | BT; KNN; SVM | Schizophrenic patients | Two factors have been identified as particularly influential for a prolonged forensic LoS, namely (attempted) homicide and the extent of the victim's injuries. |
| Thongpeth et al. (2021) | Regression | LR with three penalized linear (ridge, lasso, elastic net), and 4 ML model types: SVR, NN, RF, and XGBoost | Chronic disease | The RF model had the best predictive performance with the smallest prediction errors, while linear ridge regression had the poorest prediction performance with the largest prediction errors. |
LoR, logistic regression; LR, linear regression; RF, random forest; NB, Naive Bayes; ANN, artificial neural network; SVM, support vector machine; MLP, Multi-layer backpropagation; DT, decision tree; GB, Gradient Boosting; XGBoost, eXtreme Gradient Boosting; KNN, K-nearest neighbors; BN, Bayesian network; COPD, chronic obstructive pulmonary disease; MSE, mean square error; ICU, intensive care unit.