Risk stratification and model performance test
Next, based on the nomogram, the
restricted cubic spline curve showed the risk escalated continuously
with the increasing scores obtained from the nomogram, which supports
the reliability of the model(Supplementary Figure 4). With the ROC
curve, the optimal cut-off value of risk score(124.76) was calculated
based on the maximum Youden index(Supplementary Figure 5). Then the
cut-off value categorized the training population into the low risk
group (163 twin pregnancies with risk score ≤ 124.76) and the high risk
groups (601 twin pregnancies with risk score > 124.76),
respectively(OR = 17.21, 95% CI(10.30-28.76), P
<0.001)(Supplementary Table 2). A sensitivity of 73.91%,
specificity of 85.86%, false positive rate of 14.14% and negative
predictive value of 96.01% were reached by the nomogram model. Through
the comparison of the survival curves of the high and low risk groups,
we observed the probability of SPTB in the high risk group was
significantly higher than that in the low risk group (HR = 2.59, 95% CI
(2.03-3.30), P <0.001), and the gestational age at delivery
was significantly earlier in the high risk group (Supplementary Figure
6).