Identification of vital sign trajectory phenotypes and treatment response heterogeneity in critically ill patients with ischemic stroke: A multicenter study with external validation

Critically ill patients with ischemic stroke exhibit heterogeneous hemodynamic patterns, yet previous trajectory-based studies have focused on single vital sign parameters. We conducted a multicenter retrospective cohort study to identify vital sign trajectory phenotypes, evaluate their prognostic value compared with traditional severity scores, and explore treatment response heterogeneity. Using group-based multi-trajectory modeling of six vital signs (systolic blood pressure, diastolic blood pressure, mean arterial pressure, heart rate, respiratory rate, and oxygen saturation) during the first 12 h after intensive care unit admission, we analyzed 3100 patients from MIMIC-IV for development and 3951 patients from eICU and Chinese Critical Care Database for external validation. Three distinct phenotypes were identified: tachycardic-tachypneic (25.6%), hypertensive-stable (37.5%), and low-diastolic quiescent (36.8%), with in-hospital mortality rates of 29.7%, 5.9%, and 14.9%, respectively. After multivariable adjustment, the tachycardic-tachypneic phenotype demonstrated significantly elevated in-hospital mortality risk compared with hypertensive-stable (HR 3.55, 95% CI 2.23-5.64), with consistent associations for 28-day (HR 4.38, 95% CI 2.97-6.46) and one-year mortality (HR 4.50, 95% CI 3.35-6.04). Phenotype-based classification outperformed SOFA, SAPS II, and Charlson index for one-year mortality prediction. Exploratory analysis revealed phenotype-specific dose-mortality associations for normal saline volumes, with tachycardic-tachypneic and hypertensive-stable phenotypes showing lower predicted mortality at higher Day 1 vol (2357-3500 mL and 1929-3500 mL), while low-diastolic quiescent showed lower predicted mortality with restricted Day 2 fluids (0-714 mL). Propofol similarly showed differential dose-mortality associations across phenotypes. These findings support trajectory-based phenotyping for early risk stratification and may inform precision-guided therapeutic approaches in critically ill ischemic stroke patients.