Machine-Learning-Derived Model for the Stratification of Cardiovascular risk in Patients with Ischemic Stroke

George Ntaios1, Dimitrios Sagris1, Athanasios Kallipolitis2

  • 1Department of Internal Medicine, University of Thessaly, Greece.

Insights

A new machine-learning model predicts cardiovascular risk in ischemic stroke patients using readily available data. This tool aids in stratifying risk and informing patient management strategies for better outcomes.

Area of Science:

  • Cardiology
  • Neurology
  • Data Science

Background:

  • Accurate cardiovascular risk stratification is crucial for managing patients post-ischemic stroke.
  • Identifying high-risk individuals can guide personalized treatment strategies and improve prognoses.

Purpose of the Study:

  • To develop and validate a machine-learning-derived prognostic model for predicting cardiovascular risk in ischemic stroke patients.
  • To identify key clinical and demographic factors associated with major adverse cardiovascular events after stroke.

Main Methods:

  • Utilized two prospective stroke registries for training, validation, and testing datasets.
  • Employed machine learning algorithms including XGBoost, Random Forest, and Support Vector Machines.
  • Assessed major adverse cardiovascular events (stroke, myocardial infarction, cardiovascular death) over a 2-year follow-up period.

Main Results:

  • The final model incorporated age, gender, atrial fibrillation, heart failure, hypertension, and medication history.
  • XGBoost classifier demonstrated the best performance.
  • Achieved an area under the curve of 0.648 in the validation dataset and 0.59 in the test dataset.

Conclusions:

  • An externally validated machine-learning model for cardiovascular risk estimation in ischemic stroke patients has been developed.
  • The model utilizes easily accessible parameters, facilitating its clinical application.
  • This prognostic tool can aid clinicians in risk stratification and management decisions for stroke survivors.