Table 4.
Comparison of our results with the related literature.
| Author/ year (reference) | Domain | Applied model | Evaluation sample/ number of features attribute | Evaluation Method | Performance evaluation |
|---|---|---|---|---|---|
| Adawiyah, 202029 | Preterm birth | CBR | 20 test data/18 | Retrospective/internal validation | Accuracy: 90% |
| Jaskari, 202036 |
Neonatal Mortality Morbidity |
LR, LDA, QDA, KNN, SVM, 3 different Gaussian process, RF | 977 cases/10 | Retrospective/internal validation |
AUC (RF): 0.922 F-Score (RF): 0.47 |
| Cooper, 201826 | Neonatal mortality | 14 machine learning algorithms & LR | 3552 cases/68 | Retrospective/internal validation |
Best results in 14 machine learning algorithms AUC (development): 0.91 AUC (test): 0.87 |
| Beluzos, 202037 | Neonatal mortality | XGBoost, LR, RF | 698 cases/23 | Retrospective/internal validation |
Accuracy (RF): 93% AUC (RF): 0.965 |
| Coimbra, 201627 | LOS | CBR | 284 cases/13 | Retrospective/ internal validation | Accuracy: 84.9% |
| Safdari, 201613 | Neonatal mortality | Fuzzy logic | 200 records/14 | Retrospective/internal validation |
Accuracy: 90% Specificity: 83% Sensitivity: 97% |
| Gunaratnam, 201318 | Mortality- preterm birth | Decision tree (C5.0) & ANN | 32,760 cases/13 | Retrospective/internal validation | Best results for Decision tree (mortality) sensitivity: 62.24%, precision: 99.95% (for preterm birth) sensitivity: 79.32%, precision: 91.97% |
| Pepler, 20129 | Neonatal mortality-LOS | LR | 1578 cases/10 | Retrospective/internal validation |
AUC: 0.85 Accuracy: 86.4% R2: 0.70 |
| Rodríguez, 200828 | Pediatric mortality | Combined CBR, ANN, and Fuzzy logic | 99 cases/33 | Prospective/external validation | Accuracy: 89.89% |
| Sheikhtaheri, 202135 | Neonatal mortality | ANN, decision tree, SVM, Bayesian Network, and Ensemble models | 92 cases/17 | Prospective/external validation |
Accuracy: 86% Precision: 96% Specificity: 83% Sensitivity: 86% F-score: 0.91 AUC: 0.92 |
| Our results | Neonatal survival -LOS | CBR | Retrospective evaluation on the 336 cases, prospective evaluation on 92 cases/17 for survival, 13 for LOS | Retrospective and Prospective/internal and external validation |
Retrospective on the balanced dataset: F-score for survival: 0.986, RSME for LOS: 4.78 Prospective on the balanced dataset: F-score for survival: 0.933, RSME for LOS: 3.27 |
CBR case-based reasoning, LR logistic regression, LDA linear discriminant analysis, QDA quadratic discriminant analysis, KNN K-nearest neighborhood, SVM support vector machine, RF random forest, ANN artificial neural network, AUC area under curve, RMSE root mean square error.