Predicting the risk of pulmonary embolism in patients with tuberculosis using machine learning algorithms

Area of Science:

• Medical Informatics
• Pulmonology
• Machine Learning

Background:

• Pulmonary embolism (PE) poses a significant risk to patients with tuberculosis (TB).
• Accurate risk assessment is crucial for timely intervention and improved patient outcomes.
• Current methods for PE risk stratification in TB patients require enhancement.

Purpose of the Study:

• To develop and validate machine learning models for predicting pulmonary embolism risk in tuberculosis patients.
• To identify key clinical variables associated with PE risk in this population.
• To provide a tool for early identification and management of PE in TB patients.

Main Methods:

• Utilized Recursive Feature Elimination (RFE) for variable selection from a development cohort.
• Constructed predictive models using logistic regression, random forest, XGBoost, LightGBM, and SVM.
• Evaluated model performance using nested cross-validation, AUC, SHAP, and external validation.

Main Results:

• The random forest (RF) model demonstrated superior performance, achieving an AUC of 0.839 in the development cohort.
• Key predictors included D-dimer, smoking status, dyspnea, age, sex, diabetes, and platelet count.
• The RF model maintained high performance in external validation (AUC: 0.906 ± 0.041).

Conclusions:

• The random forest model is a robust and effective tool for predicting pulmonary embolism risk in tuberculosis patients.
• This model facilitates early risk stratification and supports clinical decision-making for PE management in TB.
• The identified predictors offer insights into the multifactorial nature of PE risk in this patient group.