Resurgery Clusters in Intensive Medicine

Abstract

The field of critical care medicine is confronted every day with cases of surgical interventions. When Data Mining is properly applied in this field, it is possible through predictive models to identify if a patient, should or should not have surgery again upon the same problem. The goal of this work is to apply clustering techniques in collected data in order to categorize re-interventions in intensive care. By knowing the common characteristics of the re-intervention patients it will be possible to help the physician to predict a future resurgery. For this study various attributes were used related to the patient's health problems like heart problems or organ failure. For this study it was also considered important aspects such as age and what type of surgery the patient was submitted. Classes were created with the patients' age and the number of days after the first surgery. Another class was created where the type of surgery that the patient was operated upon was identified. This study comprised Davies Bouldin values between -0.977 and -0.416. The used variables, in addition to being provided by Hospital de Santo António in Porto, they are provided from the electronic medical record.