Data Mining Medical Information: Should Artificial Neural Networks Be Used to Analyse Trauma Audit Data?

Data Mining Medical Information: Should Artificial Neural Networks Be Used to Analyse Trauma Audit Data?

Thomas Chesney (Nottingham University Business School, UK), Kay Penny (Napier University, UK), Peter Oakley (The University Hospital of North Staffordshire, UK), Simon Davies (University of Birmingham Research Park, UK), David Chesney (Freeman Hospital, UK), Nicola Maffulli (Keele University School of Medicine, UK) and John Templeton (Keele University School of Medicine, UK)
DOI: 10.4018/jhisi.2006040104
OnDemand PDF Download:
$37.50

Abstract

Trauma audit is intended to develop effective care for injured patients through process and outcome analysis, and dissemination of results. The system records injury details such as the patient’s sex and age, the mechanism of the injury, various measures of the severity of the injury, initial management and subsequent management interventions, and the outcome of the treatment including whether the patient lived or died. Ten years’ worth of trauma audit data from one hospital are modelled as an Artificial Neural Network (ANN) in order to compare the results with a more traditional logistic regression analysis. The output was set to be the probability that a patient will die. The ANN models and the logistic regression model achieve roughly the same predictive accuracy, although the ANNs are more difficult to interpret than the logistic regression model, and neither logistic regression nor the ANNs are particularly good at predicting death. For these reasons, ANNs are not seen as an appropriate tool to analyse trauma audit data. Results do suggest, however, the usefulness of using both traditional and non-traditional analysis techniques together and of including as many factors in the analysis as possible.

Complete Article List

Search this Journal:
Reset
Open Access Articles: Forthcoming
Volume 12: 4 Issues (2017): Forthcoming, Available for Pre-Order
Volume 11: 4 Issues (2016)
Volume 10: 4 Issues (2015)
Volume 9: 4 Issues (2014)
Volume 8: 4 Issues (2013)
Volume 7: 4 Issues (2012)
Volume 6: 4 Issues (2011)
Volume 5: 4 Issues (2010)
Volume 4: 4 Issues (2009)
Volume 3: 4 Issues (2008)
Volume 2: 4 Issues (2007)
Volume 1: 4 Issues (2006)
View Complete Journal Contents Listing