Track: Undergraduate Student Paper Competition

Abstract

Understanding the specific timings of ambulance arrivals during different shifts presents unique opportunities for ED throughput optimization. This research aims to determine whether there are significant differences in ambulance arrivals across various periods (i.e., morning, afternoon, evening, and night) and to use machine learning to forecast daily EMS arrivals. We performed a cohort study of patients presenting to an urban, academic ED between January 1, 2021 and April 18, 2022. Data comprised daily patient arrivals, demographic information, mode of arrival (either by walk-in or ambulance), chief complaints, and ED disposition, focusing exclusively on patients who arrived via ambulance. A Light Gradient Boosting Machine (LGBM) model was developed to forecast daily EMS arrivals based on different periods. The findings indicate distinct ambulance arrival patterns, with higher patient influx during the night shift. The LGBM model achieved moderate to high accuracy, particularly during the evening and night periods, with mean absolute percentage errors of approximately 21% and 19%, respectively. This study provides insights into ambulance arrival patterns, emphasizing the need for specialized care across different periods. These findings can inform the development of decision support systems that leverage advanced analytics to optimize resource allocation, enhance preparedness, and improve patient outcomes in EDs and health systems.