Table 2.
AKI prediction model for critical care settings.
| References | Modeling data sources | Data volume | Model performance AUC value | External verification | Machine learning methods | Model explanation | Diagnostic criteria | |||
|---|---|---|---|---|---|---|---|---|---|---|
| Data source | Data volume | Model performance AUC values | ||||||||
| Shawwa et al. (26) | Mayo Clinic | 98,472 | 0.69 | MIMIC-III | 51,801 | 0.656 | Gradient boosting model | Yes | KDIGO: Creatinine and urine volume | |
| Li et al. (16) | MIMIC-III | 14,470 | 0.779 | None | Naive Bayes Support vector machines Logistic regression Random forest Gradient boosting decision tree |
None | KDIGO: Creatinine and urine volume | |||
| Zimmerman et al. (14) | MIMIC-III | 23,950 | 0.783 | None | Logistic regression Random forest Artificial neural networks |
None | KDIGO: Creatinine | |||
| Zhang et al. (20) | MIMIC-III | 2,395 | 0.88 | The First Affiliated Hospital of Fujian Medical University, China | 499 | 0.78 | Extreme gradient boosting Adaptive boosting Random forest Logistic regression Multilayer perceptron |
None | KDIGO: Creatinine | |
| Liang et al. (21) | SHZJU- ICU and MIMIC-III | 58,492 | 0.86 | AmsterdamUMCdb | 15,341 | 0.86 | Multiple logistic regression Random forest Extreme gradient boosting Adaptive boosting Light gradient boosting machine Gradient boosting decision tree |
None | KDIGO: Creatinine and urine volume | |
| Sun et al. (15) | MIMIC-III | 14,469 | 0.83 | None | Logistic regression Random forest Naive Bayes Support vector machines |
None | KDIGO: Creatinine | |||
| Alfieri et al. (23) | eICU and MIMIC-III | 35,573 | 0.89 | None | Deep learning Logistic regression |
None | AKIN: Creatinine and urine output | |||
| Qian et al. (18) | MIMIC-III | 17,205 | 0.905 | None | Logistic regression Support vector machines Random forest Extreme gradient boosting Light gradient boosting machine Convolutional neural networks |
None | KDIGO: Creatinine and urine output | |||
| Wei et al. (19) | MIMIC-III | 25,711 | 0.926 | None | Extreme gradient boosting Logistic regression |
Yes | KDIGO: Creatinine and urine output | |||
| Fujarski et al. (22) | AmsterdamUMCdb | 23,106 | 0.883 | None | Categorical boosting Support vector machines |
KDIGO: Creatinine and urine output | ||||
| Sato et al. (17) | eICU | AKI I | 5,342 | 0.742 | None | Convolutional neural networks | None | KDIGO: Creatinine | ||
| AKI II | 1,450 | 0.844 | ||||||||
| Le et al. (25) | MIMIC-III | 12,347 | 0.86 | None | Convolutional neural networks Extreme gradient boosting |
None | KDIGO: Creatinine | |||
| Parreco et al. (24) | eICU | 151,098 | 0.834 ± 0.006 | None | Gradient boosting decision tree Logistic regression Deep learning |
None | KDIGO: Creatinine | |||
AUC, Area Under Curve; MIMIC-III, Medical Information Mart for Intensive Care III; AmsterdamUMCdb, The Amsterdam University Medical Centers Database; eICU, eICU Collaborative Research Database; SHZJU- ICU, The Second Affiliated Hospital of ZheJiang University School of Medicine, China; LightGBM, Light gradient boosting machine.KDIGO, Kidney Disease: Improving Global Outcomes; AKIN, Acute Kidney Injury Network.