Principal objective
Predictors of failed mid-cavity VAD were assessed by calculating odds
ratios (ORs).
In the multivariate analysis, seven variables were shown to be
statistically significant independent risk factors for failed mid-cavity
VAD (Table 2).
ROC curves for the multivariate risk prediction were calculated. The
performance of the prediction model for the training set was compared to
that of its ability to predict failed mid-cavity VAD for the test set.
The model was reliable (Hosmer-Lemeshow test = 8.5; p = 0.39) and
accurate (concordance index = 0.74).
The equation corresponding to this model is as follows:
Odds =
exp [-2.5651
+ (0.5198 if caput succedaneum = medium-sized or 1.4332 if caput
succedaneum = bulky)
+ (0.2380 if duration of expulsive efforts = 11 to 20 or -0.6549 if
duration of expulsive efforts = 21 to 30 or 0.2611 if duration of
expulsive efforts ≥31)
+ (0.4668 if head-perineum distance = 51 to 60 or 1.8968 if
head-perineum distance ≥ 61)
+ (1.0722 if maternal height < 1.55 or -0.1937 if maternal
height = 1.55 to 1.64 or -0.4910 if maternal height ≥1.75)
+ (0.5438 if duration from five centimetres to complete cervical
dilation ≥ 3)
+ (0.6719 if term ≥ 42)
+ (0.3938 if fetal head position = occiput-posterior or
occiput-transverse)].
The odds for the predicted probability of mid-cavity VAD failure were
calculated as follows: probability = odds/(1 -odds).
Using the Youden index, the threshold to predict a failed mid-cavity VAD
was set at 0.163. For internal validation, a resampling of 5000 datasets
was performed. The mean concordance index was 0.75 (95 % confidence
interval = 0.71 - 0.79), the mean specificity was 67.4 (95 % confidence
interval = 64.3 – 70.6), and the mean sensitivity was 64.4 (95 %
confidence interval = 57.5 – 71.1).