Table 5.
Specific disease-related AKI prediction models.
| Reference | Modeling data sources | Data volume |
Model performance AUC value |
Comorbidity type | Machine learning methods | Model explanation | Diagnostic criteria | |
|---|---|---|---|---|---|---|---|---|
| Yue et al. (62) | MIMIC-III | 3,176 | 0.817 | Sepsis | Logistic regression K-nearest neighbor Support vector machines Decision trees Random forest Extreme gradient boosting Artificial neural networks |
None | KDIGO: Creatinine and urine volume | |
| Qu et al. (63) | Jinling Hospital, Nanjing, China | 324 | 0.919 | Acute pancreatitis | Support vector machines Random forest Classification regression tree Extreme gradient boosting Logistic regression |
None | KDIGO: Creatinine and urine volume | |
| Yang et al. (64) | Gezhouba Central Hospital of Sinopharm and Xianning Central Hospital, China | 424 | 0.902 | Acute pancreatitis | Random forest Support vector machines Extreme gradient boosting Artificial neural networks Decision trees |
None | KDIGO: Creatinine and urine volume | |
| Zhang et al. (65) | Zhongshan Hospital, Fudan University, Shanghai, China |
6,846 | 0.822/0.850 | Liver cancer Gallbladder cancer | Extreme gradient boosting LASSO regression |
None | KDIGO: Creatinine | |
| Scanlon et al. (66) | The Christie NHS Foundation Trust, UK | 48,865 | 30 days in advance | 0.881 | All types of solid tumors | Random forest | None | NHS-AKI: Creatinine |
| 1 day in advance | 0.947 | |||||||
| Park et al. (67) | Korea Central Cancer Registry (KCCR) in Seoul National University Hospital |
21,022 | Precision | 0.7892 | Respiratory tract cancer Gastrointestinal tract cancer Thymus cancer Hematologic malignancy Breast cancer Female genitourinary organ cancer, etc. |
Linear regression Ridge regression LASSO regression Least-angle regression Stochastic gradient descent Random forest Multivariate adaptive regression splines | None | KDIGO: Creatinine |
| Recall | 0.7506 | |||||||
| F value | 0.7576 | |||||||
| Li et al. (68) | Zhongshan Hospital, Fudan University, Shanghai, China | 6,459 | 0.823 | Esophageal cancer/Stomach cancer/Intestinal cancer | Bayesian network Naive Bayes Decision trees Logistic regression Random forest Support vector machines |
Yes | KDIGO: Creatinine | |
| Li et al. (69) | Zhongshan Hospital, Fudan University, Shanghai, China |
2,395 | 0.835 | Lymphoma/Leukemia/Multiple myeloma | Bayesian network Logistic regression |
Yes | KDIGO: Creatinine | |
| Tang et al. (70) | Severely burned patients from the Kunshan factory explosion in China on 8.2 | 157 | 0.920 | Burns | Extreme gradient boosting Logistic regression |
None | KDIGO: Creatinine and urine output | |
| Tran et al. (71) | University of California, Davis, US | 50 | Accuracy | 0.80–0.90 | Burns | K-nearest neighbor | None | KDIGO: Creatinine and urine volume |
| Rashidi et al. (72) | University of California, Davis, US | 50 | 0.87–0.92 | Burns | Logistic regression K-nearest neighbor Support vector machine Random forest Deep neural networks |
None | KDIGO: Creatinine and urine volume | |
KDIGO, Kidney Disease: Improving Global Outcomes; AKIN, Acute Kidney Injury Network; MIMIC-III, Medical Information Mart for Intensive Care III; NHS, National Health Service.