Critically ill patients often carry more than one leaking heart valve, yet most cohort studies grade the tricuspid, mitral and aortic valves separately, as clinical guidelines do. A new exploratory study asks whether a single number that adds the three regurgitation grades together can capture the same prognostic information as the individual grades. The answer, drawn from a large intensive care database, is nuanced: the combined score tracks one-year mortality, but it adds almost nothing that a model keeping the valves apart does not already contain.
The research, published in Echo Research & Practice, was a retrospective cohort analysis using MIMIC-IV version 3.1 linked with MIMIC-IV-ECHO, two publicly available critical care databases hosted on PhysioNet. The investigators, led by Hasan Burak Isleyen, included adults who had an intensive care unit admission and a clinically obtained transthoracic echocardiogram within 72 hours of that admission. Only patients whose index echocardiogram carried structured, complete ordinal grading for tricuspid regurgitation, mitral regurgitation and aortic regurgitation were eligible, and one admission per patient was retained. The final cohort contained 5,362 patients, and the unit of analysis was the patient-level index ICU-echo episode.
The exposure was deliberately simple. Each of the three regurgitant lesions was mapped to an ordinal grade from 0 to 4, and the Multi-Valvular Burden (MVB) score was defined as the arithmetic sum of the three grades, producing a possible range of 0 to 12. The authors stress that the score is not outcome-derived and does not imply equal physiology across valves; it is a transparent summary of cumulative regurgitant severity. Primary analyses treated MVB both continuously and as categories of 0, 1, 2, 3 and 4 or higher. The primary endpoint was one-year all-cause mortality measured from the index echocardiogram, with follow-up administratively censored at 365 days.
The burden of disease in the cohort was substantial. Of the 5,362 patients, 1,979, or 36.9 percent, died within one year. Unadjusted survival fell steadily across burden categories: mortality was 24.6 percent among patients with an MVB score of 0 and rose to 47.4 percent among those scoring 4 or higher. The high-burden group contained 2,257 patients, and the score distribution extended across the full range, with four patients reaching the theoretical maximum of 12.
After adjustment for a rich clinical model that included age, sex, sequential organ failure assessment score, acute physiology score III, Charlson comorbidity index, prior myocardial infarction, congestive heart failure, stroke, chronic pulmonary disease, liver disease, diabetes, chronic kidney disease, malignancy, vasopressor use, mechanical ventilation, renal replacement therapy, creatinine, hemoglobin and mean arterial pressure, the continuous MVB score remained associated with one-year mortality, with a hazard ratio of 1.07 per point (95 percent confidence interval 1.05 to 1.09; P less than 0.001). Patients with an MVB of 4 or higher had higher mortality than those with lower burden, with a hazard ratio of 1.35 (95 percent confidence interval 1.22 to 1.48; P less than 0.001). The adjusted Cox models included 5,320 patients; 42 were excluded because one or more clinical covariates were missing.
Two statistical caveats shaped how these estimates should be read. Schoenfeld residual testing showed a small but statistically significant departure from proportional hazards for the continuous MVB term in the primary model (P = 0.032), so the hazard ratios are reported as average associations over the 365-day follow-up rather than strictly time-invariant effects. Restricted cubic spline analysis with three knots found no material non-linearity in the MVB-mortality relationship (P for non-linearity = 0.43), supporting the use of a single linear term.
When the three valve grades were modeled separately, tricuspid regurgitation carried the clearest signal (hazard ratio 1.09 per grade, 95 percent confidence interval 1.05 to 1.14; P less than 0.001), followed by mitral regurgitation (hazard ratio 1.06 per grade, 95 percent confidence interval 1.02 to 1.11; P = 0.002), while aortic regurgitation was weaker and not statistically significant (hazard ratio 1.05, 95 percent confidence interval 0.99 to 1.10; P = 0.099). Yet the model retaining all three individual grades showed virtually identical discrimination to the MVB model: a C-index of 0.7585 versus 0.7582. The clinical model alone achieved a C-index of 0.7543, so adding MVB produced an absolute gain of just 0.0039, even though the likelihood-ratio improvement was statistically significant (chi-square 50.61; P less than 0.001). Because MVB is the exact sum of its three components, models containing both were structurally collinear and were interpreted only as penalized sensitivity analyses in the supplement.
The high-burden threshold of 4 or more turned out to have a distributional basis in this cohort rather than a clinical one. Although a single severe lesion could theoretically push a patient over the threshold, only 51 of the 2,257 high-burden patients reached it that way; the remaining 2,206 had regurgitation of any grade in at least two valves. When the high-burden definition was tightened to require at least two affected valves, the association with mortality was nearly unchanged (hazard ratio 1.34, 95 percent confidence interval 1.22 to 1.47; P less than 0.001). The categorical analysis therefore genuinely described multivalve involvement, but the authors caution that this does not validate the threshold as a clinical decision point.
Sensitivity analyses probed whether stenotic lesions, ejection fraction or right ventricular function explained the finding. Using database-derived proxies for aortic and mitral stenosis, with explicit indicators for unavailable source data so that missing measurements were not treated as evidence of absent stenosis, MVB remained associated with mortality (hazard ratio 1.07 per point, 95 percent confidence interval 1.05 to 1.09). Notably, the cleaned aortic stenosis proxy itself was not independently associated with mortality, whereas the unavailability of aortic stenosis source data was (hazard ratio 1.25, 95 percent confidence interval 1.03 to 1.50; P = 0.022), a pattern the authors say shows that measurement availability can encode care processes and should not be given a biological interpretation. In a subset of 4,180 patients with left ventricular ejection fraction and structured right ventricular function data, MVB remained associated with mortality (hazard ratio 1.05 per point, 95 percent confidence interval 1.03 to 1.08), alongside lower ejection fraction and explicitly documented right ventricular dysfunction (hazard ratio 1.40, 95 percent confidence interval 1.23 to 1.58). Adjusted subgroup testing found no interaction for severity strata defined by organ failure scores or comorbidity burden, but the MVB association was attenuated among patients receiving vasopressors and mechanical ventilation, with statistically significant interaction terms, suggesting effect-size modification by acute organ-support status.
The authors are explicit about what the score is and is not. MVB is an exploratory summary marker rather than an outcome-derived or mechanistic weighting system, and the near-identical discrimination to the separate-valve model, together with the very modest incremental gain, argue for viewing it as a compact descriptive marker of broader cardiovascular burden within systemic critical illness. Because the endpoint was all-cause mortality in critically ill patients, deaths may reflect sepsis, respiratory failure, malignancy, neurological complications or multiorgan dysfunction rather than a cardiovascular pathway, so the data cannot establish whether cumulative regurgitation itself contributes causally to mortality. The limitations are equally clear: this was a single-center retrospective database study of patients selected for clinically indicated echocardiography, regurgitation grades were structured clinical measurements rather than blinded core-lab adjudications, the score assigns equal weight to physiologically heterogeneous lesions, several model terms departed from proportional hazards, and quantitative measures such as TAPSE and natriuretic peptides were available only in small subsets. Residual confounding, treatment indication, transient hemodynamic effects and selection associated with obtaining echocardiography cannot be excluded. Within those bounds, the study offers researchers a reproducible shorthand for describing regurgitant burden in risk-adjusted cohort analyses, while firmly declining to promote it as a stand-alone clinical decision tool.
Subject of Research: An exploratory MIMIC-IV cohort study of a multi-valvular burden score and one-year mortality in ICU patients undergoing echocardiography
Article Title: Multi-valvular burden score and one-year mortality in ICU patients undergoing echocardiography: an exploratory cohort study
Article References: Isleyen, H. B., Yavuz, S. T., Bulut, S., Kizkapan, F., Alioglu, C., Sozen, A. A., & Khanmohammadi, M. (2026). Multi-valvular burden score and one-year mortality in ICU patients undergoing echocardiography: an exploratory cohort study. Echo Research & Practice, 13(1), Article 36. https://doi.org/10.1186/s44156-026-00139-x
Image Credits: AI Generated
DOI: 10.1186/s44156-026-00139-x
Keywords: multi-valvular burden score, ICU echocardiography, valvular regurgitation, one-year mortality, MIMIC-IV, retrospective cohort, Cox proportional hazards, tricuspid regurgitation, mitral regurgitation, aortic regurgitation, risk prediction, critical care
News Source: Phoebe Ingram. (October 28, 2026). Research report: Multi-valvular burden score and one-year mortality in ICU patients undergoing echocardiography: an exploratory cohort study. Scienmag.
