Session S42.1

Evaluating the Risk of a Rescue PCI after Thrombolysis Intervention: A Decision Tree Approach

V Lagani*, D Conforti, M Vatrano, R Ceravolo, VA Ciconte

CESIC - NEC Italia S.r.l.
Rende, Italy

Thrombolysis intervention is a common therapeutic practice used in order to dissolve the coronary atherosclerotic plate following a pharmacological approach. Anyway, in the case the thrombolytic drug fails, a Rescue Percutaneous Coronary Intervention (PCI) is needed in order to restore the normal coronary blood flow.
Thus, assessing the individual risk of thrombolysis failure is a crucial step before deciding to perform a thrombolytic intervention. Objective of the present study is to develop a simple set of rules able to support physicians in discriminating patients eligible for a thrombolysis intervention.
Our approach is based on the well known C4.5 algorithm, a decision tree algorithm that repeatedly partitions a set of data in order to provide easy – to – understand decision tree models. C4.5 algorithm parameters were optimized through a ten folds cross validation procedure.
A total of one hundred and two (102) subsequent patients, treated at the Pugliese – Ciaccio Hospital of Catanzaro (Italy), were studied. All the patients underwent a thrombolytic intervention after acute myocardial infarction (AMI); forty four patients needed a Rescue PCI after thrombolysis failure.
The results demonstrated the validity of our approach: the best obtained model showed a overall accuracy of 80.39%, with a Area Under the Curve (AUC) statistic of 0.816.
Beyond the results in terms of performances, our models pointed out some remarkable interactions between blood pressure values and clinical parameters related to patient’s metabolism, suggesting new interesting mutual influences among thrombolysis failure predictors.

(Abstract Control Number: 316)