Multimethod survival analysis for identifying predictors and forecasting mortality in a heart patient cohort study

Abstract

This study presents a multi-method survival analysis of 125 cardiac patients from IIMCT-Pakistan Railway Hospital in Rawalpindi, Pakistan. Parametric accelerated failure-time modeling identified the Weibull distribution as optimal for describing time-to-event data. Semi-parametric analyses, including Cox proportional hazards and Bayesian Cox regression, consistently identified hypertension, ischemic heart disease, and smoking as significant predictors of elevated mortality risk. Higher systolic blood pressure demonstrated a protective effect. Kaplan-Meier analysis revealed steadily declining survival rates up to 300 days with no significant gender differences. The random survival forest model achieved robust predictive accuracy, identifying ischemic heart disease, smoking, and age as the most influential predictors. Our multi-methodological approach demonstrates the value of integrating parametric, semi-parametric, Bayesian, and machine learning techniques for comprehensive risk assessment in cardiac patient cohorts, offering potential for enhanced clinical risk stratification and personalized prognosis.