Identification of hemodynamic sub-phenotypes in septic shock using joint trajectory modeling of vasopressor requirements and fluid accumulation: a MIMIC-IV database study
- 1Department of Critical Care Medicine, Affiliated Cancer Hospital of Xinjiang Medical University
Abstract
Background: Patients with septic shock in critical care medicine often exhibit significant individual variability. Previous clinical assessments have largely relied on static severity scores at a single point in time to predict patient prognosis; however, this approach struggles to reflect the constantly changing clinical trajectory of patients during the resuscitation process. In recent years, research has gradually shifted from static assessments to dynamic observations. Dynamic trajectory observations may help characterize clinical heterogeneity and provide complementary prognostic information beyond static severity scores.
Methods: This study is a retrospective cohort study based on the MIMIC-IV database, which included a total of 2,909 patients with septic shock meeting the Sepsis-3 criteria. We used a latent class mixture model to observe the dynamic trajectories of norepinephrine equivalents and cumulative fluid balance during the first 72 h following the onset of shock. Landmark analysis was used to assess changes in patient mortality risk before and after 72 h, and the Net Reclassification Improvement (NRI) and Integrated Discrimination Improvement (IDI) indices were employed to evaluate the improvement in model performance following the incorporation of dynamic trajectories. Because trajectory classification required data from the full 72-h observation window, the resulting classes could not be determined at shock onset and were therefore treated as retrospective descriptions of early resuscitation patterns rather than real-time predictors.
Results: Ultimately, four trajectory subtypes with distinct hemodynamic characteristics were identified. Class 1 presented with severe initial metabolic acidosis and required high-dose vasoactive agents, but demonstrated rapid subsequent improvement and a favorable overall prognosis. Class 2 was characterized by a less favorable clinical course and had the highest observed mortality rate (57.3%). Class 3, which constituted the largest proportion, recovered relatively quickly, with a rapid reversal of circulatory status and the best clinical outcomes. Class 4 was characterized by relatively low early vasopressor requirements, while its mortality hazard was higher among patients who survived beyond 72 h. The possibility of impaired cardiac reserve may help explain this pattern, although cardiac function was not directly assessed in this study. The combined model showed a modest improvement in discrimination over the baseline SOFA model, with a bootstrap-corrected AUROC increase of 0.010. The continuous NRI was 0.286 and the IDI was 0.016. Among patients initially classified as moderate risk by the SOFA score, 114 were reclassified into the high-risk category by the combined model, with an observed mortality rate of 61.4% in this subgroup.
Conclusion: Joint trajectories of vasopressor requirements and fluid accumulation revealed heterogeneity in early resuscitation patterns among patients with septic shock. The combined model showed a modest improvement in discrimination beyond the baseline SOFA score, and trajectory information may provide complementary prognostic information. However, external validation and prospective evaluation are required to determine its clinical utility.
Related
- Hemodynamic Outcomes and Vasopressor Requirements with Remimazolam-Based Total Intravenous Anesthesia Versus Propofol-Sevoflurane Anesthesia in Robot-Assisted Direct-to-Implant Breast Reconstruction.
- Physiology-Guided Management of Acute Right Ventricular Failure and Pulmonary Hypertension in Critically Ill Patients: Hemodynamic Phenotypes, Ventilator Interactions, and Rescue Strategies
- Combined Triglyceride–High-density Lipoprotein Cholesterol Phenotypes and Cardiovascular Outcomes in Japanese Patients with Acute Coronary Syndrome Undergoing Percutaneous Coronary Intervention: Insights from Clinical Deep Data Accumulation System-PCI
- Early Vasopressor Initiation Before Complete Fluid Resuscitation in Adults With Septic Shock: A Systematic Review
- Cytokine and Vasopressor Trajectories During Continuous Renal Replacement Therapy with oXiris, CytoSorb, or Standard Haemofilters in Sepsis-Associated Acute Kidney Injury: An Exploratory Retrospective Cohort Study
BibTeX
@article{lin2026identification,
title = {{Identification of hemodynamic sub-phenotypes in septic shock using joint trajectory modeling of vasopressor requirements and fluid accumulation: a MIMIC-IV database study}},
author = {Dexin Lin and Lixin Yang},
journal = {Frontiers in Cardiovascular Medicine},
year = {2026},
month = nov,
doi = {10.3389/fcvm.2026.1994502},
url = {https://www.frontiersin.org/articles/10.3389/fcvm.2026.1994502/full},
note = {CC BY 4.0},
}
RIS
TY - JOUR
TI - Identification of hemodynamic sub-phenotypes in septic shock using joint trajectory modeling of vasopressor requirements and fluid accumulation: a MIMIC-IV database study
AU - Dexin Lin
AU - Lixin Yang
PY - 2026
DA - 2026/11/09
JO - Frontiers in Cardiovascular Medicine
DO - 10.3389/fcvm.2026.1994502
UR - https://www.frontiersin.org/articles/10.3389/fcvm.2026.1994502/full
AB - Background: Patients with septic shock in critical care medicine often exhibit significant individual variability. Previous clinical assessments have largely relied on static severity scores at a single point in time to predict patient prognosis; however, this approach struggles to reflect the constantly changing clinical trajectory of patients during the resuscitation process. In recent years, research has gradually shifted from static assessments to dynamic observations. Dynamic trajectory observations may help characterize clinical heterogeneity and provide complementary prognostic information beyond static severity scores. Methods: This study is a retrospective cohort study based on the MIMIC-IV database, which included a total of 2,909 patients with septic shock meeting the Sepsis-3 criteria. We used a latent class mixture model to observe the dynamic trajectories of norepinephrine equivalents and cumulative fluid balance during the first 72 h following the onset of shock. Landmark analysis was used to assess changes in patient mortality risk before and after 72 h, and the Net Reclassification Improvement (NRI) and Integrated Discrimination Improvement (IDI) indices were employed to evaluate the improvement in model performance following the incorporation of dynamic trajectories. Because trajectory classification required data from the full 72-h observation window, the resulting classes could not be determined at shock onset and were therefore treated as retrospective descriptions of early resuscitation patterns rather than real-time predictors. Results: Ultimately, four trajectory subtypes with distinct hemodynamic characteristics were identified. Class 1 presented with severe initial metabolic acidosis and required high-dose vasoactive agents, but demonstrated rapid subsequent improvement and a favorable overall prognosis. Class 2 was characterized by a less favorable clinical course and had the highest observed mortality rate (57.3%). Class 3, which constituted the largest proportion, recovered relatively quickly, with a rapid reversal of circulatory status and the best clinical outcomes. Class 4 was characterized by relatively low early vasopressor requirements, while its mortality hazard was higher among patients who survived beyond 72 h. The possibility of impaired cardiac reserve may help explain this pattern, although cardiac function was not directly assessed in this study. The combined model showed a modest improvement in discrimination over the baseline SOFA model, with a bootstrap-corrected AUROC increase of 0.010. The continuous NRI was 0.286 and the IDI was 0.016. Among patients initially classified as moderate risk by the SOFA score, 114 were reclassified into the high-risk category by the combined model, with an observed mortality rate of 61.4% in this subgroup. Conclusion: Joint trajectories of vasopressor requirements and fluid accumulation revealed heterogeneity in early resuscitation patterns among patients with septic shock. The combined model showed a modest improvement in discrimination beyond the baseline SOFA score, and trajectory information may provide complementary prognostic information. However, external validation and prospective evaluation are required to determine its clinical utility.
SN - 2297-055X
ER -