Early serum creatinine trajectory phenotypes during the first 72 h after ICU admission and 1-year mortality in PCI-treated acute myocardial infarction: a retrospective cohort study

Background Static creatinine thresholds after percutaneous coronary intervention may miss early renal dynamics. We examined whether serum creatinine trajectory phenotypes during the first 72 h after intensive care unit admission were associated with mortality and renal outcomes in PCI-treated acute myocardial infarction. Methods We performed a retrospective cohort study using MIMIC-IV version 3.1. Adults with acute myocardial infarction who underwent percutaneous coronary intervention, had an intensive care unit admission, were not receiving chronic dialysis, and had at least two creatinine measurements in the first 72 h were included. Because a precise PCI timestamp was unavailable, trajectories were anchored to intensive care unit admission. Interpolated creatinine fold-change trajectories were clustered with Gaussian mixture models. The primary outcome was 1-year mortality after a 72-h landmark. Secondary outcomes were 28-day mortality, post-landmark renal replacement therapy, and renal nonrecovery. Cox models compared trajectory phenotype with KDIGO stage and a clinically adjusted model. Results After exclusion of 51 patients with apparent death before completion of the 72-h landmark window, the landmark analytic cohort included 1324 patients. Five phenotypes were analyzed: Stable/Recovery ( n = 382), Mild Rise ( n = 433), Moderate Rise ( n = 296), Severe Progressive ( n = 209), and Rapid Progressive ( n = 4). One-year mortality was 13.9%, 6.7%, 12.8%, 32.1%, and 25.0%, respectively. Renal nonrecovery was 8.6%, 3.7%, 8.1%, 33.5%, and 75.0%, and post-landmark renal replacement therapy occurred only in Severe Progressive and Rapid Progressive. Severe Progressive remained associated with higher 1-year mortality than Stable/Recovery after clinical adjustment (hazard ratio 1.56, 95% confidence interval 1.04–2.34) and after additional inclusion of KDIGO stage (hazard ratio 1.54, 95% confidence interval 1.01–2.33). In expanded sensitivity models incorporating measurement frequency, net fluid balance, AMI severity variables, mechanical circulatory support, and LVEF, the Severe/Rapid signal was attenuated, supporting cautious interpretation as exploratory risk enrichment rather than independence from KDIGO staging. Conclusions Early creatinine trajectory phenotypes after ICU admission identified progressive renal-response patterns with concentrated adverse renal events. Their present role is exploratory prognostic enrichment, with substantial overlap with AKI severity, illness severity, cardiac dysfunction, and measurement intensity, pending external validation.