Healthcare Data Mining: Predicting Hospital Length of Stay (PHLOS)

摘要

A model to predict the Length of Stay (LOS) for hospitalized patients can be an effective tool for measuring the consumption of hospital resources. Such a model will enable early interventions to prevent complications and prolonged LOS and also enable more efficient utilization of manpower and facilities in hospitals. In this paper, the authors propose an approach for Predicting Hospital Length of Stay (PHLOS) using a multi-tiered data mining approach. In their aproach, the authors form training sets, using groups of similar claims identified by k-means clustering and perfom classification using ten different classifiers. The authors provide a combined measure of performance to statistically evaluate and rank the classifiers for different levels of clustering. They consistently found that using clustering as a precursor to form the training set gives better prediction results as compared to non-clustering based training sets. The authors have also found the accuracies to be consistently higher than some reported in the current literature for predicting individual patient LOS. Binning the LOS to three groups of short, medium and long stays, their method identifies patients who need aggressive or moderate early interventions to prevent prolonged stays. The classification techniques used in this study are interpretable, enabling them to examine the details of the classification rules learned from the data. As a result, this study provides insight into the underlying factors that influence hospital length of stay. They also examine the authors 'prediction results for three randomly selected conditions with domain expert insights.