Predicting radiology service times for enhancing emergency department management

Abstract

Emergency departments (EDs) are increasingly challenged by overcrowding, resource shortages, and rising demand for care, which compromise operational efficiency and service quality. In response, machine learning (ML) is emerging as a powerful tool for ED management, offering predictive models to enhance real-time decision-making and optimize workflows. This research aims to develop an ML-based system to predict service times for X-ray examinations in real-time – the most frequently conducted diagnostics in EDs. Using a dataset of 50,070 x-ray exams from a medium-sized ED, the model incorporates patient characteristics, radiology conditions, and ED status to estimate service times from prescription to report release. A comparative analysis of ML techniques identified Gradient Boosting as the most accurate approach. Additionally, feature importance and SHAP analysis revealed key factors influencing X-ray service times. The developed system has the potential to provide ED managers with early warnings of potential delays or critical situations in the radiology unit, enabling proactive interventions and improving patient management.