Background: Pediatric heart failure (PHF) remains a major contributor to morbidity and mortality, yet standardized diagnostic and prognostic frameworks–particularly those leveraging left ventricular ejection fraction (LVEF)–are not well-established. This study evaluates clinical profiles, therapeutic interventions, and mortality outcomes across LVEF thresholds while identifying an optimal cutoff to refine risk stratification in PHF. Methods: This multicenter retrospective cohort study analyzed 1449 PHF patients (aged 1–18 years) across 30 tertiary centers (2013–2022). LVEF stratification employed conventional thresholds (50%, 55%) and an ROC-optimized cutoff (53%, derived via Youden index maximization). The primary outcome was in-hospital all-cause mortality. Multivariable logistic regression models, adjusted for clinical covariates, evaluated mortality predictors. The discriminative performance of LVEF thresholds was compared using area under the curve (AUC) analysis. Results: Distinct clinical profiles, etiologies, and treatments were observed across LVEF strata (50% vs. 55%; p < 0.05). A data-driven optimized LVEF threshold of 53% was identified for mortality prediction, demonstrating superior diagnostic accuracy with enhanced sensitivity and specificity across age groups. Multivariate analysis revealed LVEF ≥ 55% as protective (OR = 0.81, 95% CI: 0.68–0.96, p = 0.003), while ≥50% was non-significant (OR = 0.91, 95% CI: 0.74–1.12, p = 0.06). Elevated BNP (OR = 2.78, p < 0.001) and NT-proBNP (OR = 2.34, p < 0.001) strongly correlated with mortality risk. Age and sex showed no significant association with outcomes. Conclusion: In conclusion, an LVEF of 53% emerged as the optimal pediatric threshold for mortality prediction, outperforming conventional cutoffs of 50% and 55%. The integration of LVEF with biomarkers (BNP/NT-proBNP) provides a robust prognostic framework, underscoring the necessity for pediatric-specific LVEF criteria and multidimensional risk assessment in PHF management.