Author(s)

A.E. Smith, V. Stankovski & D.M.Marsh

Abstract

A comparison of two alternative methodologies for estimation of length of hospital stay A.E. smith1, V. stankovski2 & D.M. ~ a r s h ' I Queen's University, Belfast, Northern Ireland 2 University Medical Centre, Ljubljana, Slovenia Abstract This study utilised a large audit database from a trauma unit, to model the length of stay of the patient in the unit, i.e. bed occupancy, for all those with upper and lower limb fractures. For this purpose, we applied two methodologies: a decision tree approach utilising Quinlan's See5 machine learning environment was compared with a modification of Cox's regression, a statistical approach and the two systems evaluated for classification accuracy. Data from 747 lower limb and 487 upper limb cases were analysed, with the model being extracted from key parameters, which were retrieved from the original database. We also used information from the medical personnel involved and indicators of significance from preliminary studies. Exactly the same model and para

Keywords