Search for a command to run...
<b>Background/Objectives</b>: Emergency department length of stay (ED LOS) is a key indicator reflecting emergency department crowding, patient safety, and healthcare resource efficiency. Among injured patients, ED LOS may be prolonged depending on injury severity and disposition pathways (admission and inter-hospital transfer). This nationwide study using the Korean National Emergency Department Information System (NEDIS) aimed to (1) describe the distribution and determinants of ED LOS among injured patients and (2) quantify the mediating effects of disposition (admission and transfer) on the association between injury severity measured by the International Classification of Diseases-based Injury Severity Score (ICISS) and ED LOS. <b>Methods</b>: We analyzed NEDIS injury-related ED visit records collected from the date of IRB approval through 12 January 2026. We conducted a retrospective observational study using NEDIS data. Of 1,048,575 injury-related ED visits, 1,035,484 visits with valid ED LOS and eligible records were included after excluding missing key variables and implausible time values. ED LOS was calculated in minutes using arrival and departure timestamps. Injury severity was assessed using ICISS (primary: based on 15 diagnoses; sensitivity: based on 20 diagnoses). Determinants of ED LOS were evaluated using gamma regression with a log link. Disposition was categorized as discharge, admission, and inter-hospital transfer; admission and transfer were modeled as binary mediators. Causal mediation analyses estimated the average causal mediation effect (ACME), average direct effect (ADE), total effect, and proportion mediated. Multiple sensitivity analyses (outlier handling, missing-data approaches, alternative log-linear modeling, and EMS arrival subgroup analyses) assessed robustness. <b>Results</b>: The median ED LOS was 150 min (IQR 90-260). ED LOS differed substantially by disposition: 120 min for discharged patients, 420 min for admitted patients, and 360 min for transferred patients. Overall, 17.9% of visits had an ED LOS ≥ 6 h, and prolonged stays were concentrated among admitted (≥6 h: 55.0%) and transferred (≥6 h: 45.0%) patients. In gamma regression, a 0.05 decrease in ICISS (greater severity) was associated with longer ED LOSs in the unadjusted model (Ratio 1.34) and remained significant in the fully adjusted model (Ratio 1.12, 95% CI 1.11-1.13). Admission and transfer were strong determinants of ED LOS in the final model (ratios of 2.35 and 2.05, respectively). In mediation analyses, admission mediated 36.8% of the severity-ED LOS association (ACME 0.085; ADE 0.146), and transfer mediated 14.3% (ACME 0.033; ADE 0.198). Findings were consistent across sensitivity analyses. <b>Conclusions</b>: In this nationwide cohort of injured patients, ED LOS showed a right-skewed distribution, with prolonged stays concentrated in admission and transfer pathways. Injury severity (ICISS) was independently associated with longer ED LOS, and a substantial proportion of this association was mediated through admission and transfer. Reducing ED LOS among severely injured patients likely requires not only streamlining diagnostic and treatment processes but also system-level interventions targeting output-stage bottlenecks, including inpatient bed operations/boarding management and transfer coordination.