Table 2.

Statistical outputs for VOC analysis for the prediction of BV and preterm delivery from vaginal swabs taken in pregnancy.

Classifier	Random forest	Gaussian process
Statistical output (95%CI) for prediction of bacterial vaginosis
AUC	0.92 (0.84–1)	0.94 (0.87–1)
Sensitivity	0.83 (0.65–0.94)	0.87 (0.69–0.96)
Specificity	0.97 (0.83–1)	0.93 (0.78–0.99)
Positive predictive value	0.96	0.93
Negative predictive value	0.85	0.88
p value (between groups)	< 0.001	< 0.001

	Delivered preterm	Delivered term	p value
First swab taken in pregnancy
Gestation when swab taken (weeks), Median (IQR)	17 + 4 (15 + 5 – 21 + 3)	16 + 6 (15 + 5 – 20 + 1)	0.3849
Number of weeks between swab taken and delivery, Median (IQR)	14 + 4 (10 + 1 – 17 + 1)	22 + 2 (19 + 1 – 24 + 0)	< 0.0001

Classifier	Random forest	Gaussian process
Statistical output (95%CI) for prediction of preterm delivery before 37 weeks
AUC	0.79 (0.72–0.85)	0.73 (0.66–0.80)
Sensitivity	0.66 (0.56–0.75)	0.67 (0.57–0.76)
Specificity	0.89 (0.82–0.95)	0.76 (0.66–0.84)
Positive predictive value	0.86	0.73
Negative predictive value	0.72	0.70
p value (between groups)	< 0.001	< 0.001

	Delivered preterm	Delivered term	p value
Swab taken closest to delivery
Gestation when swab taken (weeks), Median (IQR)	25 + 2 (22 + 6 – 27 + 1)	22 + 3 (18 + 5 – 25 + 5)	0.0019
Number of weeks between swab taken and delivery, Median (IQR)	7 + 2 (3 + 5 – 11 + 1)	17 + 1 (13 + 0 – 21 + 0)	< 0.0001

Classifier	Random forest	Gaussian process
Statistical output (95%CI) for prediction of preterm delivery before 37 weeks
AUC	0.84 (0.79–0.90)	0.77 (0.70–0.83)
Sensitivity	0.73 (0.64–0.81)	0.68 (0.58–0.76)
Specificity	0.9 (0.82–0.95)	0.76 (0.67–0.84)
Positive predictive value	0.88	0.74
Negative predictive value	0.77	0.70
p value (between groups)	< 0.001	< 0.001

This table illustrates the gestation when the first swab was taken per patient in pregnancy and when the swab was taken closest to delivery per patient.