Detecting Key Drivers for Long Length of Stay in Emergency Rooms

Abstract

Reducing length of stay (LOS) in the emergency department (ED) has been a challenge for hospitals for many years. Patient’s LOS is affected by many factors such as medical, demographics and ED operations (i.e., availability of staff and other resources). In order to reduce LOS, the hospital management first needs to find the main drivers for long LOS. Due to the large number of factors influencing LOS, finding a specific cohort of patients which are likely to have long LOS is not a trivial task. In this work, we use Association rules to find relations between medical/operational factors and long LOS. We suggest several techniques to remove redundant rules. We validate our approach with a data set that includes 100,000 visits within two years from one hospital in the United States. This validation derives several actionable insights.