The possibilities of predicting late premature spontaneous births

Introduction. Late preterm spontaneous births (LPSB) constitute up to 75% of all preterm deliveries. However, current screening algorithms, demonstrate critically low sensitivity for predicting this specific subgroup, leaving most at-risk pregnancies undetected. Aim. To develop and evaluate a prognostic model for LPSB based on a pathophysiologically grounded grouping of risk factors, incorporating cervical length measurement during second-trimester ultrasound screening. Materials and methods. A retrospective analysis of 250 pregnant women (125 with LPSB, 125 with term births). Four predictors were assessed: adverse obstetric-gynecological history (AOGH), structural defects of the cervix (SDC), infectious-inflammatory conditions (IIC), and cervical canal length (CCL) measured transvaginally at 180–2066 weeks. The model was constructed using binary logistic regression. Predictive performance was evaluated via ROC analysis, sensitivity, specificity, and Youden’s index. R esults . Independent predictors of LPSB included: IIC (aOR = 28.284; p < 0.001), SDC (aOR = 3.320; p = 0.016), and CCL (aOR = 0.648 per 1 mm increase; p < 0.001). AOGH lost statistical significance in the multivariate model (p = 0.949). The model achieved an AUC of 0.956; at a cutoff of 0.563, both sensitivity and specificity reached 91.2%. Conclusions. The proposed model enables highly accurate prediction of LPSB as early as the second trimester, significantly outperforming existing clinical tools. An integrated online calculator facilitates implementation into routine antenatal care, promoting timely preventive interventions and reducing neonatal morbidity risks.