PREDICTING HOSPITAL ADMISSION AND SURGERY BASED ON FRACTURE SEVERITY

According to World Health Organization, falls are the second leading cause of accidental injury deaths worldwide. In the United States alone, the medical costs and compensation for fall-related injuries are $70 billion annually (National Safety Council). Adjusted for inflation, the direct medical costs for all fall injuries are $31 billion annually of which hospital costs account for two-thirds of the total. The objective of this paper is to predict fall-related injuries that result in fractures that ultimately end up in hospital admission. In this study, we apply and compare Decision Tree, Gradient Boosted Tree (GBT), Xtreme Gradient Boosted Tree (XG Boost) and Neural Networks modeling methods to predict whether fall related injuries and fractures result in hospitalization. Neural networks had the best prediction followed by XG Boost and GBT methods. By being able to predict the injuries that need hospital admission, hospitals will be able to allocate resources more efficiently.