Development and Evaluation of Nomogram- and Decision Tree-Based Risk Prediction Models for Severe Ovarian Hyperstimulation Syndrome in Patients with High Ovarian Reserve

The nomogram-based model shows strong predictive performance for severe OHSS in patients with high ovarian reserve, potentially aiding in individualized risk assessment.

Summary and findings

This study aimed to compare nomogram- and decision tree-based models for predicting severe ovarian hyperstimulation syndrome (OHSS) in patients with high ovarian reserve. A total of 803 patients were included, with 614 used for model development and validation. The nomogram showed an area under the curve (AUC) of 0.884 in the training cohort.

Abstract

<title>Abstract</title> <p>Objective To compare the predictive performance of nomogram- and decision tree-based models for severe ovarian hyperstimulation syndrome (OHSS) in patients with high ovarian reserve, and to develop a visual clinical risk assessment tool. Methods This retrospective study included 803 patients with high ovarian reserve who underwent assisted reproduction using a long-acting follicle-phase protocol at the Reproductive Medicine Department of the Affiliated Hospital of Zunyi Medical University between January 2020 and October 2025. Of these, 614 cases (80 with severe OHSS, 534 without) were used for model development and internal validation, and 189 for external validation. Independent risk factors were identified through univariate analysis and least absolute shrinkage and selection operator (Lasso) regression. Nomogram and decision tree models were constructed separately. Model performance was evaluated using receiver operating characteristic (ROC) curves, calibration plots, and decision curve analysis (DCA). External validation was conducted to assess generalizability. A web-based risk calculator was developed based on the optimal model. Results Seven independent predictors were identified: anti-Müllerian hormone (AMH), antral follicle count (AFC), basal luteinizing hormone/follicle-stimulating hormone ratio (bLH/bFSH), estradiol (E₂) level on day 5 of gonadotropin (Gn) stimulation, E₂ level on the day of human chorionic gonadotropin (hCG) trigger, number of follicles ≥ 10 mm on hCG day, and oocyte yield. In the training cohort, the area under the curve (AUC) for the nomogram was 0.884 (95% CI: 0.841–0.927), compared to 0.865 (95% CI: 0.814–0.916) for the decision tree. After bootstrap internal validation, the AUCs were 0.880 (95% CI: 0.873–0.884) and 0.812 (95% CI: 0.717–0.890), respectively. Calibration curves indicated better performance for the nomogram, and DCA demonstrated higher net clinical benefit across a wider range of threshold probabilities. In external validation, the nomogram achieved an AUC of 0.874 (95% CI: 0.821–0.928), while the decision tree yielded 0.795 (95% CI: 0.702–0.888). A web-based risk calculator (http://fxygu.one.sxyuean.cn/) was successfully developed based on the nomogram. Conclusion The nomogram-based model demonstrated superior predictive performance, reproducibility, and stability in patients with high ovarian reserve. Combined with the web-based calculator, it may facilitate individualized risk stratification and precision management in clinical practice.</p>

Elsewhere in the HCG (Human Chorionic Gonadotropin) corpus