|
| Authors | Objective of the study | Predicted outcome |
|
| Lauritsen et al. [28] | It was developed for the early identification of intense fundamental diseases. | Area under receiver operating curve (AUROC) with mean values: 0.92 |
| Ellis et al. [29] | Classified the patients having a risk of developing dependence, the risk of overdose, etc. | AUROC: 92% |
| Hong et al. [30] | Developed and evaluated an FHIR-based EHR phenotype framework. | F1- micro: 0.9466 |
| Nori et al. [31] | Worked on the accuracy of a predictive model for dementia. | AUC: 94.4%
F1 score: 54.1% |
| Santos et al. [32] | To improve the quality of patient healthcare. | F-measure: 90% |
| Lindberg et al. [33] | To utilize tree-based AI techniques to decide indicators of inpatient falls. | AUROC: 0.89
Random forest: 0.90
Boosting: 0.89 |
| Steele et al. [34] | Compared machine learning models and traditional models to identify novel predictive variables. | Net Cox regression model: 95% |
| Rajkomar et al. [35] | Demonstrated that neural network identified relevant information. | AUROC: 0.90 |
| Wang. et al. [36] | Developed neural network for prediction of the cancer risk. | AUC: 0.922
Sensitivity: 0.837
Specificity: 0.867
PVV value: 0.532 |
| Park and Choi [37] | Demonstrated the deep learning model to assist the clinicians to evaluate infection in patients. | AUROC: 0.97
Average PR: 0.17 |
| Rasmy et al. [38] | To predict the risk of heart failure in diabetic patients and pancreatic cancer. | AUROC: 85.87% |
| Wang. et al. [39] | Explored the use of features representing patient-level EHR. | Regression: 18.70%
Enhanced regression: 9.69% |
| Meng et al. [40] | Performed bidirectional symbol learning. | PRAUC: 0.76 |
| Gligic et al. [41] | To evaluate the SRML mortality predictor framework and record the parameters of each model. | Accuracy: 81.30%
AUC: 81.38% |
| Omoregbe et al. [42] | Developed the method that worked on named entity identification. | F1- score: 94.6 |
| Priyanga et al. [43] | It is focused on assessing the symptoms of tropical diseases in Nigeria. | Mean SUS: 80.4% |
| Kavitha and Hanumanthappa [44] | Proposed a novel half and half repetitive neural organization (RNN) vgt6–calculated disarray-based whale streamlining to anticipate the coronary illness. | Accuracy: 99%
Specificity: 98%
Precision: 96%
F-measures: 0.9892
AUC: 98%
Prediction time: 9.23 sec |
| Gupta and Gupta [45] | Their primary goal was to predict cancer from the hybrid algorithm. | Accuracy: 90% |
| Hamedan et al. [46] | Developed a fuzzy logic-based expert system for the diagnosis and prediction of chronic kidney diseases. | Accuracy: 92.13%
Sensitivity: 95.37%
Specificity: 88.88%
AUC: 0.92
Kappa coefficient: 0.84 |
| Roopa and Harish [47] | Distinguished the area of clots in the guilty party corridor utilizing the data fluffy organization. | Accuracy: 92.30% |
| Shawwa et al. [48] | Developed to validate the model for predicting acute kidney injury in the ICU using patient data. | AUC of mayo clinic cohort set: 0.690
MIMIC III: 0.656 |
| Shoenbill et al. [49] | To identify the patterns and predictions of lifestyle modification in EHR. | Ensemble model AUROC: 0.831 |
| Hernandez-Boussard et al. [50] | Their goal was to decide if customary RWE guarantees information legitimacy in EHR. | Recall and precision: EHR-S: 51.7% and 98.3%
EHR-U: 95.5% and 95.3% |
| Masino et al. [51] | To build up a model that perceived the baby's sepsis at any rate 4 hours before the clinical acknowledgment. | AUC: 0.85–0.87 |
| Afshar et al. [52] | Their motivation was to prepare and approve an NLP classifier for distinguishing patients with liquor issues. | AUC: 0.78
Sensitivity: 56.0%
Specificity: 88.9% |
| Li et al. [53] | Their purpose was to investigate the utility of an AI approach for patient risk satisfaction. | AUROC: 0.84
Specificity: 93.9%
Sensitivity: 63.8%
Precision: 90%
F score: 73.9% |
| Hung et al. [54] | Their aim was to apply a deep neural network to achieve high modeling power. | AUC: 0.920
Sensitivity: 92.5%
Specificity: 79.8% |
| Samad et al. [55] | They used AI to more precisely anticipate endurance after echocardiography. | Accuracy: 96%
AUC: 0.82 |
| Li et al. [56] | They intended to explore the viability of BERT-based models for biomedical. | F1 score: 93.82% |
|
