Shoulder dystocia (SD) is an unpredictable obstetric emergency associated with significant maternal & neonatal morbidity. Established risk factors include fetal macrosomia & GDM. Existing prediction models rely on information unavailable at the time of counselling. We aimed to develop antenatal and intrapartum prediction models for SD using routinely available clinical variables and to evaluate the contribution of GDM in a contemporary obstetric population.
We performed retrospective cohort study of 33,731 births across five SWSLHD hospitals (2019 - 2026). Two multivariable logistic regression models were developed: an antenatal model using variables available prior to labour and an intrapartum model incorporating factors available to delivery. Candidate predictors were screened using univariable logistic regression and refined using forward stepwise-selection with Akaike-Information-Criterion optimisation.
SD occurred in 225 deliveries (0.94% of vaginal births). Antenatal model retained fundal height, gestational age, BMI and smoking status, achieving modest discrimination (AUC 0.653). The intrapartum model retained 9 predictors and demonstrated improved discrimination (AUC 0.737). Despite its recognised association with SD, GDM was not retained in either model and was not an independent predictor after adjustment for other clinical variables.
Prediction of SD remains challenging, however incorporation of intrapartum factors improves predictive performance. Notably, GDM wasn’t an independent predictor of SD in this contemporary cohort. This may reflect intensive multidisciplinary management of diabetes in pregnancy within dedicated obstetric medicine services, reducing the impact of GDM on clinically significant SD. Further investigation is required to explore this association and determine whether evolving models of care are modifying traditional risk factors.