Skip to main content

Table 4 Diagnostic performance of prediction models in the newly-diagnosed diabetic patients in the total validation group

From: Diabetic retinopathy risk prediction for fundus examination using sparse learning: a cross-sectional study

  AUC (95% CI) Accuracy (%) (95% CI) Sensitivity (%) (95% CI) Specificity (%) (95% CI) PPV (%) NPV (%)
LASSO* 0.90 (0.84-0.95) 89.2 (82.8-93.6) 75.0 (67.1-81.6) 89.6 (83.2-93.9) 16.7 99.2
LR-BS* 0.85 (0.79-0.91) 72.3 (64.2-79.2) 100.0 (96.8-100.0) 71.5 (63.4-78.5) 8.9 100.0
HbA1c 0.64 (0.55-0.72) 69.6 (61.4-76.8) 62.5 (54.2-70.8) 70.1 (62.0-77.3) 4.4 98.1
FPG 0.73 (0.65-0.80) 65.5 (57.2-73.1) 75.0 (67.1-81.6) 65.3 (57.0-72.8) 5.7 98.9
  1. *The LASSO and LR-BS models were trained in scenario 3.
  2. AUC Area under the receiver operating characteristic curve, CI Confidence interval, FPG Fasting plasma glucose, HbA1c Glycated hemoglobin, LASSO Least absolute shrinkage and selection operator, LR-BS Logistic regression with backward stepwise selection, NPV Negative predictive value, PPV Positive predictive value.