Enabling ubiquitous data mining in intensive care : features selection and data...

• Autor(es): Santos, Manuel ; Portela, Filipe
• Data: 2011
• Identificador Persistente: http://hdl.handle.net/1822/15550
• Origem: RepositóriUM - Universidade do Minho
• Assunto(s): Ubiquitous data mining; Real-time intelligent decision support systems; Organ failure prediction; Clinical data mining; Intensive care environment
• Tipo de Documento: Artigo
• Idioma: Inglês

Descrição

Ubiquitous Data Mining and Intelligent Decision Support Systems are gaining interest by both computer science researchers and intensive care doctors. Previous work contributed with Data Mining models to predict organ failure and outcome of patients in order to support and guide the clinical decision based on the notion of critical events and the data collected from monitors in real-time. This paper addresses the study of the impact of the Modified Early Warning Score, a simple physiological score that may allow improvements in the quality and safety of management provided to surgical ward patients, in the prediction sensibility. The feature selection and data pre-processing are also detailed. Results show that for some variables associated to this score the impact is minimal.