Table 6 Performance comparison of the three models developed.

From: An artificial neural network model for evaluating the risk of hyperuricaemia in type 2 diabetes mellitus

Characteristic

Baseline

Biochemical indicator

Baseline and biochemical indicators

Cutoff

0.588

0.516

0.488

AUC (95% CI)

0.704(0.679–0.730)

0.629(0.602–0.655)

0.744(0.721–0.768)

Accuracy (95% CI)

0.663(0.663–0.663)

0.589(0.589–0.589)

0.689(0.690–0.689)

Recall (95% CI)

0.615(0.583–0.646)

0.526(0.494–0.559)

0.625(0.591–0.658)

Specificity (95% CI)

0.721(0.689–0.753)

0.664(0.630–0.698)

0.749(0.720–0.778)

PLR (95% CI)

2.204(1.943–2.499)

1.566(1.392–1.763)

2.489(2.191–2.828)

NLR (95% CI)

0.535(0.487–0.587)

0.713(0.655–0.777)

0.501(0.454–0.553)

Precision (95% CI)

0.724(0.692–0.755)

0.651(0.616–0.685)

0.697(0.663–0.731)

NPV (95% CI)

0.611(0.579–0.644)

0.541(0.509–0.573)

0.684(0.654–0.713)

KAPPA (95% CI)

0.331(0.286–0.376)

0.187(0.141–0.233)

0.375(0.331–0.420)

F1-score (95% CI)

0.665(0.633–0.696)

0.582(0.548–0.616)

0.659(0.625–0.693)

  1. CI: confidence interval, AUC: area under the curve, PLR: positive likelihood ratio, NLR: negative likelihood ratio, NPV: negative predictive value.
Back to article page