Table 6.

Prediction model for AKI associated with specific nephrotoxin exposure.

Reference	Modeling data sources	Data volume	Model performance AUC value		Type of exposure	Machine learning methods	Model explanation	Diagnostic criteria
Huang et al. (80)	American College of Cardiology (ACC) National Cardiovascular Data Registry (NCDR)	947,091	0.752		Contrast agents	Extreme gradient boosting	None	AKIN: Creatinine
Sun et al. (82)	Changzhou No. 2 People's Hospital of Nanjing Medical University, China	1,459	0.85		Contrast agents	Decision tree Support vector machines Random forest nearest neighbor Naive Bayes Gradient boosting machine	None	KDIGO: Creatinine and urine volume
Niimi et al. (81)	Japan Cardiovascular Database-Keio Interhospital Cardiovascular Studies (JCD-KiCS)	22,958	0.838			Logistic regression Extreme gradient boosting	None	KDIGO: Creatinine
Huang et al. (83)	American College of Cardiology (ACC) National Cardiovascular Data Registry (NCDR)	2,076,694	Contrast agent 0.3 mg	0.777	Contrast agents	Generalized additive model	None	Modified definition
			Contrast agent 0.5 mg	0.839
			Contrast agent 1.0 mg	0.870
Ibrahim et al. (84)	Massachusetts General Hospital in Boston, Massachusetts, USA	889	0.79		Contrast agents	LASSO regression	None	KDIGO: Creatinine
Okawa et al. (85)	Fujita Health University Hospital, Japan	1,014	0.76		Cisplatin	Neural networks Gradient boosting decision tree	None	KDIGO: Creatinine
Yang et al. (86)	Center for Medicare and Medicaid Services, US	17,694	0.72		SGLT2 inhibitors	Random forest Resilient network LASSO regression	None	Unknown

KDIGO, Kidney Disease: Improving Global Outcomes; AKIN, Acute Kidney Injury Network; SGLT2, sodium-dependent glucose transporters 2.