Development and validation of a simplified cumulative complication burden score (sccbs) for preterm birth prediction at hospital admission

Preterm birth remains the leading risk factor for of neonatal morbidity and mortality worldwide, affecting approximately 11% of all live births and long-term developmental impairment. Existing prediction models are often complex and require specialized testing, limiting their utility at the bedside. We aimed to develop and validate a simplified cumulative complication burden score (sCCBS) for predicting preterm birth risk at hospital admission using readily available clinical variables. We conducted a retrospective cohort study using electronic medical records from 66,357 deliveries at a tertiary hospital in Zhejiang Province, China (2014–2024). After excluding cesarean sections and records with incomplete data, 42,868 vaginal deliveries were included as the initial validation cohort. The sCCBS was developed using Least Absolute Shrinkage and Selection Operator (LASSO) regression and temporally validated (2014–2020 training, n = 31,372; 2021–2024 validation, n = 11,496). The score incorporates four admission variables: premature rupture of membranes (PROM, + 7 points), gestational diabetes mellitus (GDM, + 2 points), advanced maternal age (≥ 35 years, + 2 points), and non-local origin (+ 1 point). Model performance was assessed by discrimination (AUC), calibration (Brier score, calibration slope), and decision curve analysis. The sCCBS demonstrated good discrimination with an AUC of 0.714 (95% CI: 0.699–0.728) in the training set and 0.735 in the temporal validation set. At the optimal Youden cutoff (≥ 3), sensitivity was 62.4% and specificity was 74.8%. Calibration was moderate in the validation set (Brier score = 0.0392, calibration slope = 0.938, calibration intercept = 0.005; Hosmer–Lemeshow p = 0.000). Risk stratification identified three groups: low risk (score 0–1, 1.8% preterm rate), moderate risk (score 2–4, 3.5% preterm rate; RR = 2.05), and high risk (score 5–12, 11.9% preterm rate; RR = 6.76). Decision curve analysis demonstrated positive net benefit across threshold probabilities from 1% to approximately 11% (Fig. 4), with a net benefit of 0.016 at 5% and 0.003 at 10%. In a sensitivity analysis excluding PROM, the AUC decreased substantially from 0.714 (95% CI: 0.699–0.728) to 0.558 (95% CI: 0.528–0.551), confirming PROM as the dominant predictor while demonstrating that the remaining three predictors provided moderate risk stratification. The sCCBS is a simple, admission-based tool for predicting preterm birth risk using four routinely collected variables. The present study validated the score in vaginal deliveries as a methodologically rigorous first step. External validation in cesarean delivery cohorts, diverse geographic settings, and independent centers is required before broader potential evaluation can be claimed. The sCCBS is an admission-based preterm birth risk prediction tool requiring only four routinely collected variables at hospital admission. The score uses integer weights: PROM (+ 7), GDM (+ 2), advanced age (+ 2), and non-local origin (+ 1). In this initial validation in vaginal deliveries, the sCCBS achieved AUC = 0.714 (0.699–0.728) with moderate calibration (Brier score = 0.0392, slope = 0.938; H-L p = 0.000 in validation set). Three-tier risk stratification identified high-risk patients with a 6.76-fold increased risk of preterm birth. Excluding PROM in a sensitivity analysis reduced AUC to 0.558, confirming PROM as the dominant predictor while supporting the role of secondary predictors. The admission-based design supports applicability in broader populations, although applicability to cesarean deliveries, other institutions, and different populations remains to be established through independent external validation.