Machine Learning-Based Prediction of Prolonged Length of Stay in Newborns

摘要

The ability to predict prolonged length of hospital stay for newborn children has clinical value as an indicator of newborn health status but also can assist in such health system resource considerations as improved utilization of hospital wards and beds. In this paper, we describe the application of machine learning-based prediction to a Healthcare Cost and Utilization Project dataset and report on the performance of various developed predictive models. Via only utilizing administrative data and minimal clinical data available near to the time of admission/birth, we are able to demonstrate high performing models. The use of HCUP data for building newborn prolonged length of stay models potentially applicable across health care providers is an important contribution, and additionally the models represent high-performing models in the field of published predictive models of newborn length of stay in general.