Early Loop Diuretic Strategies and 30-Day Mortality in Patients With Sepsis and Acute Kidney Injury in the De-resuscitation Phase: A Retrospective Cohort Study

This ongoing retrospective cohort study investigates the association between early loop diuretic strategies and 30-day mortality in adults with sepsis and acute kidney injury (AKI), focusing on patients in the de-resuscitation phase. De-resuscitation involves reducing accumulated fluid after initial resuscitation while maintaining adequate organ perfusion. The recognition of AKI creates an important treatment decision: whether to maintain, increase, reduce, or discontinue an existing diuretic regimen, or to initiate or restart treatment. This study addresses these decisions through two separate comparisons. Surviving Sepsis Campaign guidelines The study uses MIMIC-IV version 3.1 and incorporates available hospitalization records without requiring patients to be in an intensive care unit at the index time. Eligible hospitalizations involve adults aged 18 years or older with sepsis and creatinine-defined AKI, including stages 1–3. AKI is identified by a serum creatinine increase of at least 0.3 mg/dL within 48 hours or an increase to at least 1.5 times the reference value within seven days. The reference is the lowest preceding creatinine measurement within the corresponding window. The index time, T0, is the availability time of the first qualifying inpatient creatinine result. Hospitalizations with documented renal replacement therapy before or at T0 are excluded. KDIGO AKI guideline The database definition of the study population uses cumulative fluid balance from hospital admission to T0 and available circulatory measurements. Missing fluid balance is addressed through 50 multiple imputations, with positive cumulative fluid balance required within each completed dataset. Blood pressure eligibility follows the cohort-source rules; a separately identified non-ICU extension includes hospitalizations with missing or out-of-range blood pressure measurements. These record-based criteria operationalize the intended clinical population rather than representing a prospectively adjudicated de-resuscitation phase. The first comparison evaluates four strategies for existing loop diuretic treatment: maintenance, dose escalation, dose reduction, and discontinuation. Classification is based on the first qualifying adjustment within 24 hours after T0. An existing regimen without a recorded adjustment during this window is classified as maintenance. The second comparison evaluates initiation or reinitiation of loop diuretics versus no recorded loop diuretic treatment from admission through the end of the exposure window. Initiation and reinitiation are analyzed separately from dose escalation. The primary outcome is death within 30 days of T0, including deaths occurring after discharge. The primary analysis uses a 24-hour landmark and includes hospitalizations in which patients remain alive and hospitalized beyond T0 plus 24 hours. Follow-up ends at the original T0 plus 30 days. Patient-clustered generalized estimating equations account for repeated hospitalizations, with 19 covariate terms in the four-strategy comparison and 17 in the initiation comparison. Marginal standardization provides adjusted mortality risks, risk differences, and risk ratios, with estimates pooled across imputations. Sensitivity analyses and survival curves assess the consistency of the findings. Supplementary XGBoost and SHAP analyses characterize mortality prediction and feature contributions. The expected outputs are estimates of the magnitude and uncertainty of mortality associations for each strategy. These findings are intended to inform clinical decision-making when AKI is recognized during de-resuscitation and to guide subsequent prospective research. The study remains ongoing. Data extraction and several analyses were performed before registration, and the investigators have reviewed relevant results; this registration documents the current study methods and remaining analytical work.