Bayesian Network Modeling of Cardiovascular Risk in a High-Risk Population

• Activity Number: 569
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistical Learning and Data Mining
• Author(s): Peter Salzman*+ and James P. Corsetti and Charles E. Sparks
• Companies: University of Rochester Medical Center and University of Rochester Medical Center and University of Rochester Medical Center
• Keywords: bayesian network ; classification ; graphical model

Abstract

Recent work challenges long-held beliefs that high-density lipoprotein (HDL) measured by cholesterol content is protective against coronary artery disease (CAD). Many of these challenges are based on the notion that HDL is transformed from being anti-atherogenic to pro-athergenic by chronic low-grade inflammation and oxidative stress in the body. Understanding this transformation could potentially help in developing modalities to reduce CAD morbidity and mortality. To this end, we used Bayesian network model to generate easily interpretable graphical representations that map influences that variables have on each other to better understand pathophysiologic mechanisms underlying CAD. Analyzing the posterior distribution we found independent mechanisms that cause CAD. The results generate novel hypothesis related to understanding and treating CAD.