# Development and validation of a novel nomogram for recurrent hemoptysis after bronchial artery embolization: a population-based cohort study

## TL;DR

This study creates a new tool to predict if patients will experience repeated hemoptysis after a bronchial artery embolization procedure, using clinical and radiological data.

## Contribution

A novel combined clinical and radiological nomogram is developed and validated for predicting hemoptysis recurrence after BAE.

## Key findings

- Model CR, combining clinical and radiological factors, showed the highest accuracy in predicting recurrence (AUC: 0.931 in training, 0.883 in validation).
- Calibration and statistical tests confirmed the strong fit and predictive power of Model CR.
- Decision curve analysis demonstrated that Model CR provides the most clinical benefit.

## Abstract

Recurrent hemoptysis after bronchial artery embolization (BAE) remains a significant clinical challenge. This study aims to develop and validate a predictive model to forecast hemoptysis recurrence post-BAE, enhancing clinical decision-making.

A retrospective analysis was conducted on 170 patients with hemoptysis from various causes who underwent their first BAE at three Chinese medical centers between January 2019 and December 2022. Data were split into training and validation groups (7:3 ratio). Independent predictors for recurrence were identified using the least absolute shrinkage and selection operator and multivariable logistic regression. Three models were developed: clinical (Model C), radiological (Model R), and combined (Model CR). The models’ performance was evaluated via receiver operating characteristic (ROC) curves, net reclassification improvement (NRI), integrated discrimination improvement (IDI), and calibration curves, to determine the best model. Decision curve analysis (DCA) was used to assess clinical benefits. A nomogram was then created using the optimal model.

Independent predictors for recurrence included clinical factors (hemoptysis volume, platelets, C-reactive protein) and radiological factors (fibrotic scarring, pleural thickening, bronchial artery diameter, number of arteries). Based on these, Model C and Model R were created, and Model CR incorporated all seven factors. Model CR outperformed the other models, with superior accuracy in both the training cohort (AUC: 0.931, 95% CI: 0.864–0.998) and validation cohort (AUC: 0.883, 95% CI: 0.792–0.974). Comparative tests (DeLong test, NRI, and IDI) showed Model CR had better predictive power. Calibration curves and the Hosmer-Lemeshow test confirmed good model fit (PH-L > 0.05). DCA revealed Model CR provided the most clinical benefit. A nomogram was developed from Model CR.

The nomogram based on clinical and radiological data shows strong predictive accuracy for hemoptysis recurrence after BAE, offering significant potential for clinical integration.