Ultrasonography predictive nomogram for intrauterine device embedment in peri- and post-menopausal women: a retrospective cohort study

To develop an ultrasonography predictive model for intrauterine device (IUD) embedment in peri- and post-menopausal women, facilitating optimal surgical approach selection for removal. A retrospective cohort study of 998 peri- and post-menopausal women undergoing hysteroscopic IUD removal (2019–2021) at a tertiary obstetrics/gynecology hospital. Predictors were derived via Least Absolute Shrinkage and Selection Operator (LASSO) regression. Two models were constructed and compared: Model 1 (5 predictors, including pelvic X-ray) and Model 2 (4 predictors, excluding X-ray for potentially resource-saving). Performance was assessed using area under the ROC curve (AUC) and calibration metrics. Model 1 achieved moderate predictive accuracy (AUC: 0.72). Model 2, optimized for clinical utility, demonstrated comparable performance (AUC: 0.71) without X-ray data. Key predictors included ultrasound-identified embedment, prior removal failure, uterine width and thickness. Both models showed acceptable discrimination and calibration. Pelvic X-ray does not significantly improve IUD embedment prediction in this cohort. The simplified ultrasound-based nomogram may serve as an adjunctive tool for preoperative risk stratification and support hysteroscopic removal decisions when combined with clinical judgment. Further multi-center external validation is warranted before widespread clinical application, given the single-center retrospective design.