Application of Fine–Gray Competing Risk Modeling to Predict Pediatric Care Escalation Pathways in Southeast Ethiopia
Year: 2026
Type: preprint
Study design: Observational Study
Abstract: Background Pediatric emergency conditions remain a primary driver of preventable childhood mortality in low-resource settings. The transition from initial stabilization to escalated care—termed the "Third Delay"—is a critical phase where dynamic competing risks (such as early in-hospital mortality) obscure true care escalation drivers. This study aimed to identify clinical determinants of care escalation while methodologically accounting for early mortality using a competing risks framework. Methods A retrospective facility-based study of 407 pediatric patients (2022–2025) was conducted at Asella Referral and Teaching Hospital (ARTH). Fine–Gray competing risk regression estimated Subdistribution Hazard Ratios (asHR) and Cumulative Incidence Functions (CIF), treating in-hospital mortality as a true competing event. Multivariable β-coefficients were scaled by 2 to construct a simplified bedside clinical risk score (0–22 points). Discrimination was validated via time-dependent ROC curve analysis and DeLong's test comparing continuous vs. integer scores. Results Of 407 pediatric emergency visits, 205 (50.4%) were discharged improved, 185 (45.5%) met the primary outcome of care escalation, 9 (2.2%) experienced in-hospital mortality prior to escalation, and 6 (1.5%) were referred out. Baseline malnutrition emerged as the single strongest independent accelerator of care escalation (asHR = 11.42; 95% CI: 3.10–42.50; p < 0.001). True emergency presentation (asHR = 6.85), pre-existing comorbidity (asHR = 6.50), altered mental status (asHR = 5.20), and delayed presentation > 3 days (asHR = 4.90) were independent predictors. The model demonstrated exceptional discrimination (AUC = 0.88; 95% CI: 0.821–0.908) with an optimal threshold of 85.0% sensitivity and 80.0% specificity. Stratification into Low (0–4), Moderate (5–9), and High (≥ 10) risk tiers yielded significant curve divergence (Gray’s test p < 0.001). Conclusion Care escalation in low-resource settings is heavily driven by underlying structural vulnerabilities—primarily baseline malnutrition. Implementing an integer-based risk score at initial triage enables rapid identification of high-risk children (≥ 10 points) and front-loads critical monitoring during the primary window of vulnerability.
Source: Research Square
Institutions: Arsi University, Adama Science and Technology University
Language: English
Subfield: Emergency Medicine
Field: Medicine
Domain: Health Sciences
Open Access status: green
OpenAlex ID: W7204495705