Early echocardiographic elevated filling-pressure phenotype and in-hospital mortality in critically ill adults with preserved left ventricular ejection fraction: a retrospective cohort study.

backgroundCritical care echocardiography is increasingly used for bedside hemodynamic assessment, but the prognostic relevance of large-scale, early diastolic phenotyping in unselected ICU populations remains uncertain. We evaluated whether an early echocardiography-derived elevated filling-pressure phenotype, operationally defined from structured ASE/EACVI-aligned variables, was associated with in-hospital mortality in critically ill adults with preserved left ventricular ejection fraction (LVEF).

methodsWe performed a retrospective cohort study using the MIMIC-IV database linked to the MIMIC-IV-ECHO structured echocardiography subset. Adult ICU stays were screened for echocardiographic variables recorded within 48 h before or after ICU admission. The primary analytic cohort included stays with best available LVEF ≥ 50% and at least two analyzable diastolic criteria. A reconstructed phenotype was operationally defined from variables aligned to the 2016 ASE/EACVI framework: septal or lateral e' velocity, average E/e' ratio, tricuspid regurgitation velocity, and left atrial volume index (LAVI). The phenotype was assigned when more than half of available criteria were abnormal. Because the components capture both relaxation and filling-pressure or atrial-remodelling physiology, we describe the construct as a reconstructed elevated filling-pressure phenotype rather than as guideline-adjudicated diastolic dysfunction. The primary outcome was in-hospital mortality. Multivariable logistic regression adjusted for age, sex, SOFA score, Charlson comorbidity index, best available LVEF, atrial fibrillation/flutter, heart failure, and chronic kidney disease/end-stage renal disease. Pre-specified sensitivity analyses included exclusion of populations known to confound diastolic interpretation (severe mitral stenosis or regurgitation, mitral annular calcification, constrictive pericarditis, non-cardiac pulmonary hypertension, left ventricular assist device, and heart transplantation), additional adjustment for ICU loading conditions (mechanical ventilation, vasoactive support, fluid balance, rhythm at echo), per-criterion mortality associations, and stricter thresholds aligned with the 2025 ASE update.

resultsAmong 33,076 adult ICU stays, 9,481 had early structured echocardiographic data and 9,176 had available LVEF. Preserved LVEF was present in 6,350 stays. The primary analytic cohort included 4,087 stays, of which 1,422 (34.8%) met the reconstructed phenotype definition. In-hospital mortality was higher in the phenotype-positive group than in the phenotype-negative group (16.3% vs. 10.1%). The reconstructed phenotype was associated with mortality in the unadjusted model (odds ratio [OR] 1.73, 95% confidence interval [CI] 1.43-2.09; p < 0.001) and remained independently associated with mortality after multivariable adjustment (OR 1.39, 95% CI 1.11-1.74; p = 0.004). In per-criterion analyses, tricuspid regurgitation velocity > 2.8 m/s (adjusted OR 1.47, 95% CI 1.25-1.73; p < 0.001) and LAVI > 34 mL/m² (adjusted OR 1.35, 95% CI 1.09-1.67; p = 0.006) were independently associated with mortality, whereas isolated tissue Doppler-based criteria (e' below threshold or average E/e' > 14) were not. Median ventilation duration, ICU length of stay, and hospital length of stay were all longer in the phenotype-positive group.

conclusionsAn early structured echocardiography-derived phenotype that captures elevated filling pressures and atrial remodelling was independently associated with in-hospital mortality in critically ill adults with preserved LVEF. The mortality signal was driven primarily by tricuspid regurgitation velocity and LAVI rather than by tissue Doppler-based relaxation indices, suggesting that the prognostic information reflects elevated filling pressure and chronic congestion burden rather than narrowly defined diastolic dysfunction. The findings are association-only and require prospective external validation before clinical use.