Identification of factors related to complications in myocardial revascularization surgery: an approach with multi-target association rules networks

Abstract

Myocardial revascularization surgery is one of the recommended approaches for the treatment of chronic coronary disease. Several complications related to mortality, sequelae, length of stay, and hospital costs are also associated with this procedure. Death rates and complications depend on the characteristics of each patient. Knowing the factors related to hospital mortality and complications is paramount to improving outcomes. Association Rules Mining may support the discovery of those factors. In this work we propose a new approach, called Multi-target Association Rules Network (MTARN), to analyze association rules based on networks with a simultaneous focus on two parameters. The use of association rules networks aids the analysis of a high number of association rules and the multi-target strategy allows a complete exploration, explaining which factors directly influence the analyzed set. We evaluated our approach in conjunction with domain experts and compared it to two other approaches: Decision Trees and Filtered-ARNs, a single target approach based on networks for pattern visualization. The results indicates that MTARNs optimize the discovery of knowledge directly linked to complication and death factors in patients undergoing coronary artery bypass grafting. These parameters may be used in the construction of an intelligent monitoring system to aid myocardial revascularization patients.