Prognostic value of early creatinine trajectories in ischemic stroke patients: Insights from latent growth mixture...

Renal dysfunction is common in ischemic ischemic ischemic stroke patients, but the prognostic value of dynamic creatinine trajectories remains unclear.

Using the MIMIC-IV database, we included 1998 ischemic ischemic stroke patients who had at least three creatinine measurements within 72 h of ICU admission. Latent growth mixture modeling (LGMM) was applied to identify creatinine trajectories. The primary outcome was 90-day in-hospital mortality. Cox regression, Kaplan-Meier survival analysis, and subgroup analyses were performed.

Four distinct creatinine trajectories were identified: Class 1 (74.4 %) with stable low levels; Class 2 (20.9 %) with persistently moderate-to-high levels and a slight upward trend; Class 3 (2.2 %) with an initial increase followed by decline; and Class 4 (2.5 %) with high initial levels that decreased and then rose again. Significant differences in demographics, complications, and outcomes were observed across classes (all P < 0.05). In unadjusted and partially adjusted models, Classes 2-4 were associated with higher mortality, but after full adjustment only Class 2 remained significant (HR = 1.55, P < 0.001). Kaplan-Meier analysis confirmed significantly lower survival in non-Class 1 patients, most pronounced in Class 2. Subgroup analyses showed consistent results for Class 2 across sex, age, and comorbidity subgroups.

Early creatinine trajectories carry important prognostic value in ischemic stroke patients. The persistently moderate-to-high and gradually rising trajectory (Class 2) emerged as the strongest independent predictor of 90-day mortality, highlighting its potential role in risk stratification and clinical decision-making.