Not all creatinine rises are kidney injury, early renal congestion phenotypes in acute heart failure
E Vice, P Darko, B Otchere, E Hama, P Berchie, X Salazar, A Krishnamoorthy, B Demoss, R Singh, R Lougani, K Bauza, S Damle, E Molina, C MartiAbstract
Introduction
Worsening renal function (WRF) during hospitalization for acute decompensated heart failure (ADHF) is often interpreted as acute kidney injury (AKI) and may prompt premature de-escalation of diuretic therapy. However, creatinine rises in ADHF are biologically heterogeneous and may reflect intrinsic tubular injury or hemodynamic and venous congestion–related renal dysfunction during effective decongestion. Binary creatinine-based AKI definitions fail to capture this heterogeneity and may misclassify patients with hemoconcentration and electrolyte shifts reflecting plasma volume contraction. A data-driven approach using early laboratory trajectories and diuretic response may better define clinically meaningful cardiorenal phenotypes.
Purpose
To identify early renal–congestion phenotypes in ADHF using unsupervised learning and compare outcomes across trajectory-based patterns of creatinine change and decongestion.
Methods
We conducted a retrospective cohort study of adults hospitalized with ADHF in the MIMIC-IV database. Baseline laboratories were defined within 0–6 hours of admission, with renal–congestion trajectories derived from 6–48 hours. Phenotyping variables included creatinine and BUN/creatinine trajectories, chloride change, hemoconcentration, urine output–based renal flux, diuretic exposure, and indices of decongestion and diuretic response. Unsupervised clustering identified six renal–congestion phenotypes ranging from effective decongestion with stable renal function to failure to decongest with progressive renal dysfunction. Outcomes after 48 hours included ICU escalation, in-hospital mortality, and diuretic resistance. Multivariable regression adjusted for demographics, baseline kidney function, comorbidities, and initial hemodynamics.
Results
Among 31,369 patients (mean age 71.9 ± 13.7 years; 46.8% female), in-hospital mortality was 6.1%. Despite similar creatinine changes across phenotypes, mortality varied widely (2.4%–10.3%). Phenotypes with preserved urine output and biochemical evidence of effective decongestion had the lowest mortality and ICU escalation, while congestion-predominant and failure-to-decongest phenotypes had the highest risk. These differences persisted after adjustment for age and sex.
Conclusion
Early creatinine rises in ADHF are not biologically equivalent and should not be uniformly interpreted as intrinsic kidney injury. Trajectory-based phenotyping integrating renal flux and decongestion indices revealed distinct cardiorenal patterns with markedly different outcomes despite similar creatinine changes. Effective decongestion, even with creatinine elevation, was associated with favorable outcomes, whereas creatinine rises without decongestive signals portended worse prognosis. These findings challenge binary AKI definitions and support a physiology-informed approach to cardiorenal management in ADHF.For image description, please refer to the figure legend and surrounding text.For image description, please refer to the figure legend and surrounding text.