Emergency department crowding during a 360-day period: associations with patient dropouts and early mortality in a high-volume tertiary emergency department

Abstract

Emergency department (ED) crowding is a major challenge for health systems globally, associated with delays in care, higher rates of patients leaving without being seen (LWBS), and potential adverse outcomes. This study assessed crowding using the National Emergency Department Overcrowding Score (NEDOCS) and examined its relationship with the number of patients who LWBS, and early mortality in the largest tertiary ED in Serbia. A retrospective analysis was conducted over 1 year (360 days with fully functional health information system) between April 2022 and April 2023, comprising 8,640 one-hour intervals. For each interval, NEDOCS was calculated using standardized input variables. Hourly data were collected on waiting times, LWBS, and mortality within 24 h of ED registration or hospital admission (early mortality). Descriptive statistics were used to characterize ED utilization. A Generalized Additive Model (GAM) was applied to predict LWBS as a function of NEDOCS. Spearman’s correlations tested associations between NEDOCS categories and outcomes. Across the study period, 178,679 ED visits were recorded, with 8.3% resulting in hospitalization; 3.0% classified as LWBS, and total early mortality rate of 0.43% within the first 24 h. Median longest waiting time per hourly interval was 2.25 h to the first physician check up, and 3.87 h to inpatient admission. Crowding was significantly higher during daytime ( p  < 0.001), with NEDOCS most frequently indicating “busy” or “very busy” conditions. Peak crowding occurred between 08:00–10:00. LWBS showed a significant non-linear association with NEDOCS, with the predicted number of LWBS patients increasing up to a NEDOCS value of approximately 129 and declining at higher score values. There was a significant positive correlation between NEDOCS category and LWBS ( p  < 0.001), while associations with mortality measures were not statistically significant. This high-resolution, year-long analysis reveals substantial ED crowding, pronounced morning peaks, and a significant non-linear association between NEDOCS-defined crowding and LWBS. No statistically significant association was observed between NEDOCS categories and early mortality. These findings suggest that routine crowding metrics may support real-time operational monitoring, but should be interpreted alongside triage, patient acuity, and additional process- and safety-related indicators. Not applicable.