Table 5 Model performance on the independent test set and independent cross-cultural/ethnic test set.

From: An explainable machine learning-based clinical decision support system for prediction of gestational diabetes mellitus

Model

Test population

AUC-PR

AUC-ROC

Sensitivity

Specificity

Balanced accuracy

1

Non-white (45 cases)

0.572

0.717

0.6

0.8

0.7

1

White (110 cases)

0.485

0.792

0.733

0.768

0.751

2

Non-white (45 cases)

0.263

0.643

0.3

0.686

0.493

2

White (110 cases)

0.208

0.659

0.6

0.6

0.6

3

Non-white (45 cases)

0.293

0.677

0.3

0.829

0.564

3

White (110 cases)

0.199

0.656

0.533

0.674

0.604

  1. Model 1: feature-agnostic model. Features: family history of diabetes mellitus, weight, white cell count, fasting glucose, insulin.
  2. Model 2: clinical routine model. Features: gestational age, maternal age, family history of diabetes mellitus, weight, white cell count.
  3. Model 3: remotely usable model. Features: gestational age, maternal age, family history of diabetes mellitus, weight.
Back to article page