Abstract 1226: Machine Learning-Based Prediction of In-Hospital Mortality in Cerebral Venous Sinus Thrombosis
Stroke: Vascular and Interventional Neurology · 18.10.2026 · 6 (Suppl_1)
DOI: 10.1161/svi270000_1226 ↗Kurz gefasst
Fachartikel in der Zeitschrift Stroke: Vascular and Interventional Neurology, erschienen am 18.10.2026. Verfasst von L. Cruz-Criollo und 11 weiteren Autor:innen.
Abstract (Originalsprache)
Machine Learning-Based Prediction of In-Hospital Mortality in Cerebral Venous Sinus Thrombosis Introduction: Cerebral venous thrombosis (CVT) typically affects younger patients and carries a favorable prognosis, but a minority die during hospitalization. The admission variables that best identify these patients remain incompletely characterized. We sought to identify predictors of mortality at hospital discharge.
Methods
We conducted a retrospective observational cohort study of patients diagnosed with acute CVT from the international CoLlabOraTion on Cerebral VENoUs Thrombosis Study (CLOT-VENUS) registry (2004-2024), including two comprehensive stroke centers in the USA and Mexico. Most relevant admission clinical, imaging, and laboratory results were collected. Patients were split into model training (n=257) and held-out test (n=137) sets. ( Figure 2 ) A gradient-boosted model (GBM) was optimized using Bayesian techniques along with cross-validation. This model was then compared to Lasso regression and stepwise logistic regression using the same resamples. The selection of predictors was refined through recursive feature elimination combined with permutation importance.
Results
Of 432 patients, 394 met the inclusion criteria. Median age was 40 years [IQR 28-55], and 65% were female. Key predictors identified through recursive feature elimination and model importance included ICH volume (100), Glasgow Coma Scale (88), age (30), hemoglobin (24), and leukocytes (7) for in-hospital mortality. ( Figure 1 ) In the validation model, GBM outperformed the other models, achieving an AUC of 0.89. In the testing validation set, it retained discrimination (AUC 0.87; 95% CI: 0.8 - 0.94) for in-hospital mortality. ( Figure 2 )
Conclusion
This study provides a straightforward, admission-based method for early identification of patients with acute CVT who are at risk of in-hospital mortality. By incorporating five objective clinical, laboratory, and imaging variables that are routinely available, making it practical for use at the bedside. CLOT-VENUS can aid in early clinical decision-making, help prioritize monitoring and escalation of care, and offer a standardized framework for risk enrichment and stratification in future clinical trials related to CVT. Figure 1. Variable importance and the score assignment plot (A) The relative importance of the five predictors in the final gradient-boosted model. (B-F) Points assigned by the score across the observed range for each predictor: ICH volume (B), age (C), Glasgow Coma Scale (D), white blood cell count (E), and hemoglobin (F). Figure 2. Model selection and validation of the mortality-at-discharge model. A - Model comparison, internal cross-validation (training set, n=257). B - Discrimination of the selected GBM, held-out test set (n=137)
Zitieren
APA
Cruz-Criollo, L., Rodriguez-Calienes, A., Brito, A., Nigenda, V. C., Galecio-Castillo, M., Abdelhakim, N., Rovegno-Salazar, S., Herrera-Calderon, R., Pompa, A. M., Kontowicz, E., Arauz, A., & Ortega-Gutierrez, S. (2026). Abstract 1226: Machine Learning-Based Prediction of In-Hospital Mortality in Cerebral Venous Sinus Thrombosis. Stroke: Vascular and Interventional Neurology, 6(Suppl_1). https://doi.org/10.1161/svi270000_1226
Vancouver
Cruz-Criollo L, Rodriguez-Calienes A, Brito A, Nigenda VC, Galecio-Castillo M, Abdelhakim N, et al. Abstract 1226: Machine Learning-Based Prediction of In-Hospital Mortality in Cerebral Venous Sinus Thrombosis. Stroke: Vascular and Interventional Neurology. 2026;6(Suppl_1). doi:10.1161/svi270000_1226
BibTeX
@article{CruzCriollo2026,
author = {Cruz-Criollo, L. and Rodriguez-Calienes, A. and Brito, A. and Nigenda, V. Cano and Galecio-Castillo, M. and Abdelhakim, N. and Rovegno-Salazar, S. and Herrera-Calderon, R. and Pompa, A. Mercado and Kontowicz, E. and Arauz, A. and Ortega-Gutierrez, S.},
title = {Abstract 1226: Machine Learning-Based Prediction of In-Hospital Mortality in Cerebral Venous Sinus Thrombosis},
journal = {Stroke: Vascular and Interventional Neurology},
year = {2026},
volume = {6},
number = {Suppl_1},
doi = {10.1161/svi270000_1226}
}
EUVIMED