Real-time survival prediction in emergency situations with unbalanced cardiac patient data

Cardiac disease is a major cause of morbidity and mortality worldwide. Currently, most researchers focus on identifying risk factors and treatment of cardiac disease. There has been little research on real-time prediction of patient survival in emergency situations with unbalanced data, which is critical to cardiac patient treatment. 2099 records were collected from cardiac patients at the Tel-Aviv Sourasky Medical Center. Using these records, a survival prediction model was built using empirical thresholding logistic regression with unbalanced cardiac patient data. This research (1) provided a simplified, highly efficient and flexible model to predict survival of patients with cardiac disease; (2) revealed important factors that influence survival prediction; and (3) discussed key points related to prediction with unbalanced medical data. The identified risk factors will help doctors concentrate on the most important factors for patient survival. This study provided novel technical and practical insights for patient survival analysis and prediction that traditionally suffers from the common unbalanced data problem.