Early recognition of persistent sepsis-induced coagulopathy burden in critically ill patients with sepsis: trajectory analysis and reduced compatible eICU evaluation

Background Sepsis-induced coagulopathy (SIC) is commonly assessed at a single early time point, but persistent SIC burden may reflect sustained thrombo-inflammatory and coagulation-organ dysfunction. Whether this dynamic phenotype can be recognized from first-day ICU data remains unclear. Methods We studied adult ICU patients with Sepsis-3-defined sepsis in MIMIC-IV. Standard SIC scores were calculated across 0–24 h, 24–48 h, and 48–72 h, and latent class trajectory modeling identified 72-h trajectory phenotypes. Outcomes and supportive care exposures were compared across phenotypes. A logistic model using only 0–24 h variables identified subsequent high-persistent versus non-high-persistent SIC; a multinomial model assessed three-class identification. Internal performance was evaluated in a held-out test set and by repeated cross-validation. A reduced compatible evaluation was performed in eICU. Results The complete trajectory cohort included 7,765 patients: low-stable ( n = 2,928), improving ( n = 1,041), and high-persistent ( n = 3,796). Twenty-eight-day mortality was 22.6%, 14.8%, and 32.7%, respectively. In the primary adjusted mortality model, improving was associated with lower mortality (OR 0.56; 95% CI, 0.46–0.69), whereas high-persistent was associated with higher mortality (OR 1.29; 95% CI, 1.14–1.45). The held-out first-day binary model for high-persistent versus non-high-persistent SIC had AUROC 0.897, AUPRC 0.903, sensitivity 0.832, and specificity 0.797. The multinomial model for low-stable, improving, and high-persistent SIC had accuracy 0.809, balanced accuracy 0.613, macro-F1 0.582, and macro-AUROC 0.870. In 20 repeated 5-fold cross-validations, the full binary model had mean AUROC 0.890 and AUPRC 0.897. The reduced compatible eICU evaluation included 955 complete cases and yielded AUROC 0.886 and AUPRC 0.942. Conclusions Early SIC trajectories identified resolving and persistent coagulopathy phenotypes. High-persistent SIC was associated with worse survival and higher vasopressor and renal replacement therapy exposure, while mechanical ventilation was most frequent in the improving group. First-day data identified subsequent high-persistent SIC with good internal discrimination, although part of this performance reflects overlap between early SIC components and trajectory definition. The model is an early phenotype-recognition tool requiring prospective validation, not a mortality model or treatment rule.

Authors

Publication Details

Journal
Thrombosis Journal
Published
2026-11-19
DOI
https://doi.org/10.1186/s12959-026-00943-9
Primary Topic
Sepsis Diagnosis and Treatment
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Early recognition of persistent sepsis-induced coagulopathy burden in critically ill patients with sepsis: trajectory analysis and reduced compatible eICU evaluation

Qiaoyi Wu, Hongcheng Zhang
Thrombosis Journal
Sepsis Diagnosis and Treatment
article

Early recognition of persistent sepsis-induced coagulopathy burden in critically ill patients with sepsis: trajectory analysis and reduced compatible eICU evaluation

Qiaoyi Wu, Hongcheng Zhang
article en

Abstract

Background Sepsis-induced coagulopathy (SIC) is commonly assessed at a single early time point, but persistent SIC burden may reflect sustained thrombo-inflammatory and coagulation-organ dysfunction. Whether this dynamic phenotype can be recognized from first-day ICU data remains unclear. Methods We studied adult ICU patients with Sepsis-3-defined sepsis in MIMIC-IV. Standard SIC scores were calculated across 0–24 h, 24–48 h, and 48–72 h, and latent class trajectory modeling identified 72-h trajectory phenotypes. Outcomes and supportive care exposures were compared across phenotypes. A logistic model using only 0–24 h variables identified subsequent high-persistent versus non-high-persistent SIC; a multinomial model assessed three-class identification. Internal performance was evaluated in a held-out test set and by repeated cross-validation. A reduced compatible evaluation was performed in eICU. Results The complete trajectory cohort included 7,765 patients: low-stable ( n = 2,928), improving ( n = 1,041), and high-persistent ( n = 3,796). Twenty-eight-day mortality was 22.6%, 14.8%, and 32.7%, respectively. In the primary adjusted mortality model, improving was associated with lower mortality (OR 0.56; 95% CI, 0.46–0.69), whereas high-persistent was associated with higher mortality (OR 1.29; 95% CI, 1.14–1.45). The held-out first-day binary model for high-persistent versus non-high-persistent SIC had AUROC 0.897, AUPRC 0.903, sensitivity 0.832, and specificity 0.797. The multinomial model for low-stable, improving, and high-persistent SIC had accuracy 0.809, balanced accuracy 0.613, macro-F1 0.582, and macro-AUROC 0.870. In 20 repeated 5-fold cross-validations, the full binary model had mean AUROC 0.890 and AUPRC 0.897. The reduced compatible eICU evaluation included 955 complete cases and yielded AUROC 0.886 and AUPRC 0.942. Conclusions Early SIC trajectories identified resolving and persistent coagulopathy phenotypes. High-persistent SIC was associated with worse survival and higher vasopressor and renal replacement therapy exposure, while mechanical ventilation was most frequent in the improving group. First-day data identified subsequent high-persistent SIC with good internal discrimination, although part of this performance reflects overlap between early SIC components and trajectory definition. The model is an early phenotype-recognition tool requiring prospective validation, not a mortality model or treatment rule.

Thrombosis Journal
Sepsis Diagnosis and Treatment
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon