A Metabolism-Based Interpretable Machine Learning Prediction Model for Diabetic Retinopathy Risk: A Cross-Sectional Study in Chinese Patients with Type 2 Diabetes

Table 2

Discriminant evaluation of predictive models.

Model	Accuracy	ROCAUC (95% CI)	value^a	PRAUC (95% CI)	value^b
LR model 1	87.42%	0.73 (0.68, 0.74)	Ref	0.30 (0.24, 0.32)	Ref
XGBoost model 1	83.87%	0.64 (0.61, 0.72)	.023	0.26 (0.21, 0.39)	.312
LR model 2	87.10%	0.78 (0.73, 0.81)	.156	0.34 (0.27, 0.40)	.283
XGBoost model 2	88.39%	0.82 (0.75, 0.82)	.006	0.44 (0.31, 0.47)	<.001

Notes: features in model 1: sex, age, duration of type 2 diabetes, body mass index, systolic blood pressure, diastolic blood pressure, triglyceride, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, and total cholesterol. Features in model 2: duration of type 2 diabetes, age, systolic blood pressure, total cholesterol, alanine, citrulline, glutamate, ornithine, phenylalanine, threonine, tyrosine, C18 : 1, C18 : 1OH, and C18 : 2. ^aDelong test for area under the curve of receiver operating characteristic curve. ^bDelong test for area under the curve of precision recall curve. Abbreviations: ROC: receiver operating characteristic; AUC: area under the curve; CI; confidence interval; PR: precision recall; LR: logistic regression; XGBoost: extreme gradient boosting; Ref: reference.