Composite socio-environmental risk score for cardiovascular assessment: An explainable machine learning approach

Area of Science:

• Cardiology
• Public Health
• Data Science

Background:

• Cardiovascular disease (CVD) is the leading global cause of mortality.
• Socio-environmental factors significantly influence CVD morbidity and mortality.
• Understanding these factors is crucial for improving risk assessment and interventions.

Purpose of the Study:

• To develop and evaluate a composite socio-environmental (SE) cardiovascular risk score.
• To assess its predictive power for major adverse cardiovascular events (MACE).
• To incorporate traditional and novel socio-environmental risk factors into cardiovascular risk assessment.

Main Methods:

• A Survival Random Forest (RSF) model was employed to create the SE cardiovascular risk score.
• 22 census-tract level variables from the CLARIFY registry (62,438 patients) were utilized.
• Cox Proportional Hazard and SHapley Additive exPlanations (SHAP) models were applied for validation and variable importance.

Main Results:

• The study included 62,438 individuals with common comorbidities like hypertension and diabetes.
• The SE-MACE risk score demonstrated a robust association with MACE (HR, 1.21 [95 % CI, 1.11-1.32]), independent of clinical factors and CAC score.
• Significant predictors included smoking prevalence, insurance status, and median household income, with clear risk stratification observed across SE-MACE score quartiles.

Conclusions:

• The SE-MACE risk score effectively integrates socio-environmental determinants into cardiovascular risk prediction.
• It identifies individuals at higher risk for MACE, supporting holistic assessment frameworks.
• Further validation in diverse populations is recommended.