Prolonged length of stay in the Italian emergency departments: A competing risks frailty model

Background Emergency department (ED) overcrowding has increased worldwide in recent years. This paper provides quantitative information based on data collected immediately before the COVID-19 outbreak, which can serve as a baseline for developing systematic approaches to prolonged length of stay (LoS) in EDs. Materials and methods We analysed a random sample of 10,077 ED accesses drawn from the 503,797 eligible accesses recorded in 2019 in the 16 largest hospitals of Sicily, Italy. LoS was modelled using a cause-specific competing-risks Cox frailty model with Discharge and Hospitalization as competing endpoints. Two log-normal frailty components were included to account for unobserved heterogeneity between EDs and between patients. Non-proportionality was addressed by stratifying on pathology and by including time-varying effects for triage. Cumulative incidence functions (CIFs) were computed to estimate the probability of each endpoint and strata over time. Results Median LoS was 1.70 hours for Discharge and 3.77 hours for Hospitalization, and the gap widened markedly in the upper tail (90th percentile: 7.10 vs 34.95 hours). Discharge was associated with pathology, age, triage, sex and season; Hospitalization with pathology, age, triage and season. The hazard of Discharge decreased by 2% per additional year of age (HR = 0.98, 95% CI 0.977–0.983) and was 10% higher for accesses in summer (HR = 1.10, 95% CI 1.03–1.16). The hazard of Hospitalization increased by 0.4% per year of age (HR = 1.004, 95% CI 1.001–1.007). The effect of triage was time-dependent: at 10 hours of LoS, red-coded accesses had a 69% lower hazard of Discharge (HR = 0.31) and a 53% higher hazard of Hospitalization (HR = 1.53) than other codes. Circulatory and respiratory diseases and blood diseases showed the highest CIFs for Hospitalization and prolonged LoS for both endpoints, whereas skin and musculoskeletal diseases, injuries and poisoning, and nervous system diseases showed the highest CIFs for Discharge. Conclusions Prolonged LoS in Sicilian EDs was driven mainly by pathology, and the exit process differed systematically across disease groups: patients with circulatory and respiratory diseases and with blood diseases waited longest before both Discharge and Hospitalization.The large between-ED heterogeneity in the hazard of Hospitalization, indicates that admission delays reflect organisational as well as structural constraints. Interventions to reduce overcrowding — in particular any expansion of inpatient capacity — should therefore be targeted at the specific wards receiving these patients, rather than applied uniformly across the hospital.