Creating a guide to identify patients who may leave without being seen: A machine learning approach

Julia Sarty, Eleanor Fitzpatrick, Katrina Hurley, Peter Vanberkel, Majid Taghavi,
Creating a guide to identify patients who may leave without being seen: A machine learning approach,
Operations Research, Data Analytics and Logistics,
Volume 45,
2025,
200486,
ISSN 3050-7847,
https://doi.org/10.1016/j.ordal.2025.200486.
(https://www.sciencedirect.com/science/article/pii/S3050784725000200)
Abstract: Patients and their caregivers who seek care in an Emergency Department (ED) may ultimately choose to leave without being seen by a physician. This occurrence is labeled “left without being seen” (LWBS) and can account for up to 15% of all patients who come to an ED. Patients who LWBS do not receive the care they seek in the ED and may experience clinical deterioration related to delayed diagnosis or treatment. Identifying which patients are more likely to become LWBS patients (and intervening) could prevent adverse outcomes related to LWBS. This paper aims to create a paper-based guide to identify patients at risk of LWBS proactively. The emphasis is on creating a guide that can be easily used by staff in real time with readily available data. The most important features for predicting LWBS are determined using descriptive statistics and SHAP value analysis. The typical ranges for those features are then analyzed with a trained machine-learning model to determine feature combinations that lead to LWBS. Threshold values for these feature combinations are then determined to form the foundation for the guide designed to fit a single piece of paper. The guide was developed using data from the Pediatric Emergency Department at IWK Health in Halifax, Nova Scotia, Canada.
Keywords: Left without being seen; Machine learning; Emergency department; Pediatric; SHAP values; Real-time decision making