Diuretic efficiency ratio adds little beyond urine output for kidney risk in ICU heart failure

Critically ill adults given early intravenous loop diuretics in the intensive care unit are often assessed by how much urine they produce relative to the drug dose they receive, a measure known as diuretic efficiency. A new retrospective analysis of the MIMIC-IV critical care database suggests that while this dose-indexed ratio does carry prognostic information about subsequent kidney injury, it adds essentially nothing beyond what urine output and diuretic dose already reveal when each is modeled separately.

The study, published in BMC Cardiovascular Disorders, was a 12-hour landmark analysis of MIMIC-IV version 3.1. The researchers included adults with an acute or acute-on-chronic heart-failure discharge code in any diagnosis position who received an intravenous furosemide or bumetanide bolus within 24 hours of ICU admission. Diuretic efficiency was defined as cumulative urine output from the landmark time through the following 12 hours, divided by the cumulative loop-diuretic dose and expressed as milliliters per 40 mg of furosemide equivalent. Worsening renal function was defined strictly as a creatinine increase of at least 0.3 mg/dL occurring after the landmark time and through 48 hours.

The primary cohort comprised 2,206 patients, of whom 2,167 had evaluable creatinine measurements. Among these, 462 patients, or 21.3 percent, developed worsening renal function within the observation window. The team built a series of logistic models containing strictly landmark-available clinical predictors alone, then compared this baseline against models that added urine output and diuretic dose separately, or added the diuretic efficiency ratio instead.

Optimism-corrected discrimination, measured by the area under the receiver operating characteristic curve, was 0.667 for the clinical model alone, rising to 0.705 when urine output and dose were added separately and 0.703 when the diuretic efficiency ratio was added instead. The empirical intervals around these estimates overlapped substantially. Compared with the clinical model, adding the ratio changed the corrected AUC by 0.036 and improved the Brier score by 0.007, indicating a modest gain in both discrimination and calibration.

The critical comparison, however, was between the ratio model and the model containing urine output and dose separately. Here the paired corrected AUC difference was -0.002, with an interval spanning -0.008 to 0.006, meaning the ratio did not outperform its two components modeled independently. In an exploratory subgroup restricted to patients with heart failure as the principal diagnosis, which included 286 patients and 51 events, the conditional paired-bootstrap AUC increment was just 0.002, with an interval from -0.020 to 0.022 that included the null value.

These findings address a long-standing ambiguity in diuretic monitoring. The results indicate that in this population, whatever prognostic signal the ratio carries is already contained in urine output and dose when both are entered into the model separately, so the ratio offers no incremental predictive value over its components.

The authors emphasize that lower record-derived early diuretic efficiency was associated with subsequent creatinine-defined worsening renal function in this retrospectively defined, heterogeneous ICU population. But the association observed in a database cohort does not establish clinical utility, and the study’s conclusions are bounded by its design. The cohort was identified through discharge codes rather than clinically adjudicated heart failure, and the analysis relied on routinely recorded data rather than standardized urine-output and creatinine assessment protocols.

Missing predictors were multiply imputed without using the outcome as an imputation predictor, and both the outcome-free imputation and the model fitting were repeated within 500 bootstrap resamples, a procedure intended to make the performance estimates more robust. The optimism correction and empirical intervals reflect the internal validation approach, which speaks to stability within the dataset but not to performance in external populations.

The study’s authors frame the work as a direct comparison that clarifies the prognostic information captured by the diuretic efficiency ratio and provides a benchmark for future research, rather than as a recommendation for practice. They call for prospective evaluation using standardized urine-output and creatinine assessment in clinically defined heart-failure populations, noting that the retrospective, code-based nature of the MIMIC-IV analysis limits how far the findings can be generalized to bedside decision-making.

For now, the message for clinicians and researchers is that the popular dose-indexed ratio may be less informative than its widespread use implies. In this large ICU cohort, simply measuring urine output and recording the diuretic dose separately provided equivalent prognostic information about which patients would go on to develop worsening renal function, without the need to compute a ratio at all.

Subject of Research: Prognostic value of early diuretic efficiency for worsening renal function in critically ill adults with heart failure

Article Title: Early loop-diuretic efficiency and subsequent worsening renal function in critically Ill adults with a coded acute heart-failure diagnosis: a 12-hour landmark analysis of MIMIC-IV

Article References: Wang, J., Deng, N., Huang, Y., Liu, Z., & Zeng, H. (2026). Early loop-diuretic efficiency and subsequent worsening renal function in critically Ill adults with a coded acute heart-failure diagnosis: a 12-hour landmark analysis of MIMIC-IV. BMC Cardiovascular Disorders. https://doi.org/10.1186/s12872-026-06913-3

Image Credits: AI Generated

DOI: 10.1186/s12872-026-06913-3

Keywords: diuretic efficiency, loop diuretics, worsening renal function, MIMIC-IV, intensive care, acute heart failure, creatinine, urine output, risk prediction, landmark analysis, furosemide, retrospective study