Abstract
BACKGROUND: The MEESSI-AHF score is widely recommended for 30-day risk stratification in acute heart failure (AHF), but its ability to predict in-hospital mortality - an endpoint with immediate implications for triage and resource allocation - has not been evaluated in Vietnam. OBJECTIVE: This study externally validated MEESSI-AHF for in-hospital mortality in Vietnamese patients hospitalized with AHF, assessing both discrimination and calibration. METHODS: A prospective cohort of adults hospitalized with AHF (Sep 2024 - Mar 2025; n = 241) was analyzed. The endpoint was in-hospital mortality; performance was assessed by AUC and calibration (CITL, slope, Brier, HL g=6, plots), with minimal recalibration (intercept/slope) if needed. RESULTS: In-hospital mortality was 10.8% (26/241). MEESSI-AHF scores were higher in non-survivors than survivors (median 0.49 vs - 2.17; p<0.001) and showed excellent discrimination (AUC 0.886, 95% CI 0.832 - 0.940). Observed mortality increased stepwise across six pre-specified strata (p for trend < 0.001). Overall calibration suggested over-prediction (CITL -1.201, 95% CI -1.665 to -0.737; slope 1.118, 95% CI 0.734 - 1.502), with Brier score 0.092 (Brier skill 0.042) and poor HL fit (g=6: χ²=28.44, p < 0.001), mainly in the highest-risk stratum. Relative to low-intermediate risk, mortality was higher in high-risk (OR 17.55, 95% CI 1.89 - 162.51) and very-high-risk groups (OR 57.12, 95% CI 7.49 - 435.83). CONCLUSION: In Vietnamese patients hospitalized with AHF, MEESSI-AHF shows excellent discrimination for in-hospital mortality but systematically overpredicts risk; minimal recalibration corrects this and supports its use for bedside triage with local adjustment.