A novel approach for predicting the length of hospital stay with DBSCAN and supervised classification algorithms

Patient length of stay is the most commonly employed outcome measure for hospital resource consumption and to monitor the performance of the hospital. Predicting the patient's length of stay in a hospital is an important aspect for effective planning at various levels. It helps in efficient utilization of resources and facilities. So, there exist a strong demand to make accurate and robust models to predict length of stay. This paper analyzes various methods for length of stay prediction, its advantages and disadvantages and proposes a novel approach for predicting whether the length of stay of the patient is greater than one week. The approach uses DBSCAN clustering to create the training set for classification. The prediction models are compared using accuracy, precision and recall and found that using DBSCAN as a precursor to classification gives better results.