Clustering data mining models to identify patterns in weaning patient failures

Abstract

The weaning process of ventilated patients need to be carefully performed. This type of procedure is very common in Intensive Medicine. The procedure is well-defined and it is executed according the patient condition. During the weaning process, the patient can be in vary stages. At the end the extubation tentative can be considered as successful or not. Before the extubation, the patient is submitted to a set of tests in order to validate the procedure. When this procedure is wrong executed, it can provoke long term injuries to the patient. This work arises in order to avoid weaning failures by early detecting the procedure result. This work has as main goal identify possible patient patterns associated to weaning failures. In this context Clustering data mining was used to select and identify the features and the patterns associated to failures. As result an Index-Davies Bouldin of 0.51 was achieved and the most significant variables associated to a failure were identified. The physicians has now new and useful knowledge able to help to take a decision about weaning before it be initiated.