Abstract / Summary
Abstract Background: Risk stratification for intensive care unit patients with acute myocardial infarction requires early bedside variables that reflect circulatory stress, renal response, treatment intensity, and metabolic injury. This study evaluated an oliguria, renal function, hemodynamics, and acid-base abnormalities phenotype with early urine-output and fluid-balance trajectories for in-hospital mortality risk stratification. Methods: This retrospective cohort study analyzed 2097 adult acute myocardial infarction-coded intensive care unit stays from MIMIC-IV. Vital signs and laboratory predictors were defined from 6 hours before to 24 hours after intensive care unit admission. Input-output variables were summarized across 0-6, 6-12, 12-24, and 24-48 hours after admission. The primary outcome was in-hospital mortality. LightGBM models were evaluated with stratified 5-fold internal cross-validation. Robustness analyses included strict 48-hour landmarking, treatment-trajectory adjustment, net fluid-balance threshold testing, continuous renal replacement therapy exclusion, no-furosemide subgroup analysis, calibration assessment, and decision-curve analysis. Results: The cohort included 2097 intensive care unit stays and 285 in-hospital deaths (13.6%). Persistent/intermittent oliguria had the highest mortality among urine-output trajectory classes. The static phenotype model achieved an area under the receiver operating characteristic curve of 0.877, area under the precision-recall curve of 0.531, and Brier score of 0.094. Adding urine-fluid trajectories improved performance to 0.885, 0.552, and 0.089, respectively. Treatment-adjusted associations remained significant for persistent oliguria, oliguria burden, persistent positive fluid balance, and the high-risk urine-fluid phenotype. Conclusions: The oliguria, renal function, hemodynamics, and acid-base abnormalities phenotype combined with 0-48-hour urine-output and fluid-balance trajectories provided internally validated, interpretable mortality risk stratification in acute myocardial infarction-coded intensive care unit patients. External validation is required before clinical implementation.