Predictive Models for Hospital Bed Management Using Data Mining Techniques

Abstract

It is clear that the failures found in hospital management are usually related to the lack of information and insufficient resources management. The use of Data Mining (DM) can contribute to overcome these limitations in order to identify relevant data on patient’s management and providing important information for managers to support their decisions.

Throughout this study, were induced DM models capable to make predictions in a real environment using real data. For this, was adopted the Cross-Industry Standard Process for Data Mining (CRISP-DM) methodology. Three distinct techniques were considered: Decision Trees (DT), Naïve Bayes (NB) and Support Vector Machine (SVM) to perform classification tasks. This work explored the possibility to predict the number of patient discharges using only the number of discharges veirifed in the past. The models developed are able to predict the number of patient discharges per week with acuity values ranging from ≈82.69% to ≈94.23%. The use of these models can improve the efficiency of the administration of hospital beds. An accurate forecasting of discharges allows a better estimate of the beds available for the coming weeks.