TITLE:
Innovative data mining approaches for outcome prediction of trauma patients
AUTHORS:
Eleni-Maria Theodoraki, Stylianos Katsaragakis, Christos Koukouvinos, Christina Parpoula
KEYWORDS:
Data Mining; Medical Data; Decision Trees; Classification Rules; Association Rules; Clusters; Confusion Matrix; ROC
JOURNAL NAME:
Journal of Biomedical Science and Engineering,
Vol.3 No.8,
August
23,
2010
ABSTRACT: Trauma is the most common cause of death to young people and many of these deaths are preventable [1]. The prediction of trauma patients outcome was a difficult problem to investigate till present times. In this study, prediction models are built and their capabilities to accurately predict the mortality are assessed. The analysis includes a comparison of data mining techniques using classification, clustering and association algorithms. Data were collected by Hellenic Trauma and Emergency Surgery Society from 30 Greek hospitals. Dataset contains records of 8544 patients suffering from severe injuries collected from the year 2005 to 2006. Factors include patients' demographic elements and several other variables registered from the time and place of accident until the hospital treatment and final outcome. Using this analysis the obtained results are compared in terms of sensitivity, specificity, positive predictive value and negative predictive value and the ROC curve depicts these methods performance.