Independent Predictors of In-Hospital Mortality in Critically Ill Children Receiving Continuous Kidney Replacement Therapy

Abstract

Objective: Mortality among critically ill children requiring continuous kidney replacement therapy (CKRT) remains high, yet independent predictors are incompletely defined. The aim of this study was to identify independent predictors of in-hospital mortality and describe outcomes according to CKRT indication groups in a tertiary pediatric intensive care unit (PICU). Materials and Methods: We conducted a retrospective cohort study of children aged 1 month to 18 years who received CKRT in a tertiary PICU between January 2024 and January 2026. CKRT indications were classified into four mutually exclusive groups (acute kidney injury [AKI] alone, fluid overload [FO] alone, AKI+FO, and non-renal) according to the WE-ROCK/ESPNIC framework. Multivariable logistic regression with bootstrap validation was performed. Results: The study included 81 patients (median age, 34 months; 55.6% male). In-hospital mortality was 33.3% (27/81). The predominant CKRT indication was FO alone (33.3%), followed by AKI+FO (32.1%), non-renal indications (17.3%), and AKI alone (17.3%). Crude mortality ranged from 25.9% (FO alone) to 42.9% (non-renal; p=0.639). In multivariable analysis, the PRISM III score (aOR 1.30 per point, 95% CI 1.15–1.75, p<0.001), pSOFA score at CKRT initiation (aOR 1.43 per point, 95% CI 1.22–1.90, p<0.001), and FO>10% (aOR 3.35, 95% CI 1.36–8.46, p=0.012) were independent predictors of mortality. The model demonstrated high discrimination (AUROC 0.929; 95% bootstrap CI 0.905–0.929; optimism-corrected AUROC 0.920) and good calibration (Hosmer–Lemeshow p=0.918). Conclusion: Three readily available bedside variables—PRISM III, pSOFA, and fluid overload >10%—demonstrated high discriminative performance for in-hospital mortality in critically ill children receiving CKRT, supporting their potential use as a mortality risk stratification tool pending prospective multi-center validation.