Abstract
BACKGROUND: To evaluate the relationship between short-term longitudinal serum creatinine (Scr) trajectories and 28-day mortality in patients with acute kidney injury (AKI) on chronic kidney disease (CKD). METHODS: The data sources of this study were the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. Group-Based Trajectory Modeling was used to classify the trajectories of Scr indices within 96 h after Intensive Care Unit admission. Kaplan-Meier survival curves were used to analyze the 28-day survival probabilities of patients with different Scr trajectories. Multivariate Cox proportional hazards models were applied to explore the association between different Scr trajectories and 28-day mortality. Receiver operating characteristic curves were employed to assess the predictive performance of the predictive model (model 3) for 28-day mortality. Decision curve analysis (DCA) was conducted to explore the clinical net benefit of the predictive model (model 3). Subgroup analysis was conducted to explore the robustness of the relationship. RESULTS: A total of 7,852 patients with AKI on CKD were included in this study. Through GBTM analysis, five distinct trajectory groups were identified: Group G1 (n = 1762, 22.4%) with Scr levels maintained at approximately 1 and showing a stable trend (Scr low-stable group); Group G2 (n = 915, 11.7%) with baseline Scr levels above 8 and showing a decreasing trend (Scr high-decreasing group); Group G3 (n = 2017, 25.7%) with Scr levels maintained at approximately 1.3-1.5 and showing a stable trend (Scr moderate-low stable group); Group G4 (n = 1707, 21.7%) with Scr levels maintained at approximately 2 and showing a stable trend (Scr moderate-stable group); and Group G5 (n = 1451, 18.5%) with Scr levels maintained at approximately 4 and showing a stable trend (Scr moderate-high stable group). The 28-day survival probabilities in Groups G1, G2, and G3 were higher than that in Groups G4 and G5. Compared with Group G1, the risk of 28-day mortality increased by 0.819-fold in Group G2, by 0.454-fold in Group G4, and by 0.860-fold in Group G5. The area under the curve (AUC) of model 3 was 0.795 (95% CI 0.782-0.807). DCA results showed that when the threshold probability ranged from 5 to 85%, the net benefit of model 3 was significantly higher than that of the "treat all" and "treat none" models. Group G4 and Group G5 were associated with increased 28-day mortality in both the male and female subgroups. CONCLUSIONS: The results showed that AKI on CKD patients with Scr high-decreasing trend, Scr moderate-stable trend, and Scr moderate-high stable trend had a higher risk of 28-day mortality.