Table 2.

Performance metrics of final models tested on independent test set.

	Model 1 (8 features)c	Model 2 (15 features)d	Model 3 (23 features)e
Accuracya	0.86 (0.85–0.88)	0.87 (0.85–0.88)	0.87 (0.85–0.88)
Sensitivitya	0.25	0.40	0.43
Specificitya	0.91	0.90	0.89
Positive predictive valuea	0.16	0.22	0.22
Negative predictive valuea	0.95	0.96	0.96
AUROCb (95% confidence interval)	0.69 (0.65–0.73)	0.77 (0.73–0.80)	0.78 (0.74–0.81)

^aCalculated at a decision threshold of 0.5.

^bArea under receiver operating characteristic curve.

^cFeatures—Household social class, household equivalised income, PCG age, gestational age, PCG BMI, PCG alcohol intake, maternal depression score, PCG highest education.

^dFeatures—Household social class, household equivalised income, PCG age, gestational age, PCG BMI, PCG alcohol intake, maternal depression score, PCG highest education, number bedrooms in home, English native language of PCG, birthweight category, accommodation ownership, number in household smoking during pregnancy, community connectedness, PCG self-rated health.

^eHousehold social class, household equivalised income, PCG age, gestational age, PCG BMI, PCG alcohol intake, maternal depression score, PCG highest education, number bedrooms in home, English native language of PCG, birthweight category, accommodation ownership, number in household smoking during pregnancy, community connectedness, PCG stress score, hours spent on learning activities, books in the home, hours of sleep for PCG, quality of attachment score, PCG employment status, verbal interaction between PCG and baby, level of support received by PCG, gestational diabetes diet.