Early Hemodynamic Trajectory Phenotypes and Exploratory Albumin Exposure–Mortality Associations in Septic Shock: A Retrospective Cohort Study

PURPOSE: Identifying clinical phenotypes that may exhibit heterogeneous associations with treatment exposure is an important focus of sepsis research. Sepsis phenotyping mainly relies on static variables, which fail to fully capture early hemodynamic evolution. This study utilized longitudinal mean arterial pressure (MAP)/norepinephrine equivalent dose (NEQ) trajectories to characterize hemodynamic phenotypes in patients with septic shock and to explore phenotype-specific associations between albumin exposure and ICU mortality. METHODS: Adult patients diagnosed with septic shock within 24 h of ICU admission and with at least four MAP/NEQ measurements during the 72 h following initiation of the first vasopressor after ICU admission were identified from the MIMIC-IV database. The trajectory time origin was defined as initiation of the first vasopressor after ICU admission. Hourly MAP/NEQ values during the subsequent 72 h were used to identify phenotypes by group-based trajectory modeling. Associations between MAP/NEQ-trajectory phenotypes and ICU mortality were analyzed using doubly robust estimation. Subsequently, parametric G-formula was employed to estimate the cumulative mortality risk under hypothetical daily albumin infusion regimens within each phenotype. External validation was performed using the eICU-CRD. FINDINGS: A total of 3152 patients with septic shock were included in this study. Four distinct MAP/NEQ trajectory phenotypes were identified: phenotype A (low-level stable, 29.8%), phenotype B (ascending, 26.1%), phenotype C (descending, 24.0%), and phenotype D (high-level stable, 20.1%). Phenotype A was associated with the highest risk of ICU mortality compared with the other phenotypes (B: odds ratio [OR], 0.303; 95% CI, 0.225-0.408; C: OR, 0.606; 95% CI, 0.458-0.801; D: OR, 0.547; 95% CI, 0.391-0.766). G-formula analyses suggested that albumin exposure-mortality associations varied across phenotypes. Higher simulated daily albumin exposure categories were associated with lower model-estimated ICU mortality risk in phenotypes B, C, and D, whereas no significant association was observed in phenotype A (risk ratio [95% CI]: 0.89 [0.72-1.07]). Directionally similar model-estimated association patterns were observed in the external validation cohort. IMPLICATIONS: This study identified four MAP/NEQ trajectory phenotypes in patients with septic shock that exhibited differential associations with ICU mortality and albumin exposure. Since this study adopts a retrospective observational design, findings regarding albumin are exploratory and hypothesis-generating rather than proof of treatment benefit. Prospective studies and randomized trials are needed to determine whether MAP/NEQ trajectories have clinical utility for risk stratification or treatment decision-making.