# Development and validation of a nomogram for predicting bleeding risk in patients with pulmonary embolism

## TL;DR

This study created a tool to predict bleeding risk in pulmonary embolism patients, aiming to improve anticoagulation treatment decisions.

## Contribution

A novel PE-specific bleeding risk prediction model was developed and validated for personalized anticoagulation management.

## Key findings

- The model identified six key predictors for bleeding risk, including prior bleeding history and renal insufficiency.
- The model showed strong discrimination with an AUC of 0.756 in development and 0.729 in validation.
- Using the model's optimal threshold reduced major bleeding by 42% compared to standard care.

## Abstract

Bleeding during anticoagulation therapy represents a critical challenge in pulmonary embolism (PE) management, this study aimed to develop and validate a PE-specific bleeding risk prediction model.

This retrospective cohort study utilized a clinical research big data platform, including 5,632 hospitalized PE patients (January 2013–December 2024). Significant bleeding within 6 months served as the primary outcome. After excluding variables with >20% missingness, 29 predictors were analyzed. The cohort was randomly split into development (n = 3,942) and validation sets (n = 1,690). LASSO regression identified key predictors, with multivariable logistic regression constructing the final model. Performance was assessed via AUC-ROC, calibration plots, and decision curve analysis (DCA).

The final model identified six predictors: prior bleeding history, renal insufficiency, red blood cell count, systolic pressure, cerebral infarction, and creatinine. The model demonstrated robust discrimination (development AUC: 0.756, 95%CI: 0.729–0.784; validation AUC: 0.729, 95%CI: 0.685–0.773) and calibration (validation slope: 0.810). DCA confirmed significant net benefit at 5–35% thresholds, with 30% as the optimal cut-off. At this threshold, the model reduced major bleeding by 42% versus standard care.

This novel PE-specific bleeding risk tool provides clinically actionable stratification, enabling personalized anticoagulation intensity adjustment. Implementation may reduce hemorrhage-related morbidity while optimizing resource utilization.