Hospital Specialization and Geographic Access as Determinants of Pediatric Emergency Mortality

Objectives: To examine individual, hospital, and community-level determinants of in-hospital mortality—defined as death during the emergency department visit or after hospitalization at the same hospital—among critically ill pediatric emergency patients in South Korea. Methods: We analyzed data from the National Emergency Department Information System, comprising 26,206 pediatric patients (aged ≤18 y, triage levels 1 to 2) at emergency medical centers in 2021, using multilevel logistic regression. The null model confirmed substantial between-hospital [intraclass correlation coefficient (ICC)=0.17] and between-community (ICC=0.01) variation, justifying the multilevel approach. Results: Overall, in-hospital mortality was 2.3%. At the individual level, compared with self-transport, patients arriving by 119 ambulance ( OR =2.73, 95% CI: 1.85-4.05) and private ambulance patients ( OR =3.14, 95% CI: 1.72-5.73) had higher odds. Intentional and unintentional injuries showed 2.06-fold and 2.97-fold higher odds versus noninjury, respectively. Consciousness level showed a dose-response association (coma: OR =228.64). Severe illness was associated with a 4.94-fold higher odds ratio. At the hospital level, higher critically ill proportions were associated with lower mortality (≥10%: OR =0.43), while hospitals with ≥45 combined pediatric and neonatal intensive care beds had higher odds ( OR =2.43), likely reflecting case-mix concentration. At the community level, metropolitan cities showed higher odds ( OR =1.64), while a higher Timely Relevance Index (TRI; percentage of residents accessing care within 60 min) was associated with lower mortality (≥95%: OR =0.55). Conclusions: Pediatric emergency mortality was associated with factors across all 3 levels. Risk-adjusted performance metrics are needed for tiered systems, and timely geographic access supports distributed initial stabilization capacity.