A forecast model proposal for the risk of atrial fibrillation developing in postoperative period in patients with coronary artery bypass

Data mining is used to extract large amounts of hidden, valuable, usable information from large amounts of data and provide strategic decision support. In addition, data mining has created a new perspective on the use of health data and has become a method that continues to grow rapidly. AF is one of the most common complications after open heart and vascular surgery. The aim of the study is to develop a system that will support the physicians in the preoperative period for taking preventive measures for patients who are at risk of developing AF. In the study the archive files of patients with CABG (Coronary artery bypass graft) operation at the Ordu State Hospital Cardiovascular Surgery Clinic between the dates of 01 January 2015 and 31 June 2016, preoperative and postoperative pre-determinants of AF, gender, hypertension, ECG P wave elongation total drainage quantity, weight, preoperative, EF heart valve diseases, preoperetif metoprolol use, age, number of grafts, PHT, DM, laboratory (CRP, urea, creatinine, AST, GGT hemoglobin, hematocrit, Ca, K) values were analyzed retrospectively. After the operation patients were divided into two groups, AF and non-AF. Patients who had only CABG operation included in the study. Patients with chronic AF, at the same time, did not include other operative structures. In this context, the decision trees, support vector machine (SVM) and artificial neural network (ANN) and the possibility of atrial fibrillation classification after Coronary artery bypass graft (CABG and the comparison of these methods are discussed. At the end of the study in the analyze of the decision trees the sensitivity values were % one hundred and specificity values were % 94.11. Sensitivity value in support vector machine SVM was % 83.33, specificity value was % 94.11; Sensitivity analyzes performed in artificial neural network (ANN) as % one hundred, specificity % 94.11 was calculated. When the ROC curves of these methods are compared, the best classification is made with artificial neural network (ANN) and decision trees . It has been shown that the study could contribute to the prevention and precautionary measures of the prevalence of atrial fibrillation AF in cases with high probability of AF.