Frontiers in Medicine
Identification and validation of early vital sign trajectory subphenotypes for risk stratification in sepsis-associated liver injury: a multicenter retrospective cohort study
By Weihang Wang, Jing Ma, Bingqing Wang and 8 others.
Abstract
BackgroundSepsis-associated liver injury (SALI) is a serious clinical syndrome with heterogeneity in outcome among affected individuals. Physiological changes that are critical to outcome prediction are not considered in the existing system of stratification based on single-point observations. This could compromise the effectiveness of individualized treatment.ObjectiveTo develop and validate dynamic clinical subphenotypes of SALI using early vital sign trajectories.MethodsThis retrospective cohort study included 1,159 patients from MIMIC-IV as the derivation cohort and 168 patients from the Second Xiangya Hospital as the external validation cohort. Early vital sign trajectories within 12 h of ICU admission were summarized as trajectory-derived features and clustered using K-means. Candidate solutions (k = 2–6) were assessed using internal clustering metrics and bootstrap stability. A four-cluster solution was retained as an exploratory partition, considering trajectory interpretability and subgroup sizes. Clinical characteristics, organ support, and mortality outcomes were compared across the identified subphenotypes.ResultsFour exploratory SALI subphenotypes with distinct early vital sign trajectories were identified, ranging from relatively stable profiles to marked hemodynamic and physiological stress. Subphenotype 3 exhibited the most stable vital signs and was associated with lower adjusted 28-day (HR 0.74, 95% CI 0.56–0.97) and 90-day mortality (HR 0.76, 95% CI 0.59–0.97) compared with the highest-risk Subphenotype 4. These prognostic differences were consistently replicated in the external cohort, supporting short- to intermediate-term prognostic generalizability.ConclusionEarly vital sign trajectories identified exploratory SALI groups associated with differences in 28-day and 90-day mortality. Their reproducibility, incremental prognostic value, and clinical utility require further evaluation.





