Table 3. Comparison of ROC AUC^a and average precision for machine learning models predicting GDM^b trained on the EHR^c data and VAL^d data, and validated against the VAL data.

Model	ROC AUC		Average precision		Intercept		Slope
	EHR-GDM	VAL-GDM	EHR-GDM	VAL-GDM	EHR-GDM	VAL-GDM	EHR-GDM	VAL-GDM
Logistic regression	0.817	0.817	0.451	0.450	0.093	−0.027	0.984	0.955
Random forest	0.797	0.801	0.418	0.419	−0.747	−0.618	0.553	0.638
XGBooste	0.780	0.782	0.389	0.393	−0.427	−0.507	0.619	0.608
EBMf	0.818	0.816	0.456	0.450	0.078	−0.047	0.975	0.940

^aROC AUC: area under the receiver operating characteristic curve.

^bGDM: gestational diabetes mellitus.

^cEHR: electronic health record.

^dVAL: validated.

^eXGBoost: Extreme Gradient Boosting.

^fEBM: Explainable Boosting Machine.