Dynamic Coagulation Deterioration and Cumulative Burden Preceding Noncoagulation Organ Dysfunction in Sepsis: A Two-Database Cohort Study

Background: Static coagulation measurements cannot distinguish transient fluctuation from sustained abnormality in sepsis. We assessed whether 0–24 h dynamic coagulation phenotypes preceded noncoagulation organ dysfunction during 24–72 h. Methods: We conducted a retrospective cohort study using the Medical Information Mart for Intensive Care IV (MIMIC-IV) for development and the eICU Collaborative Research Database (eICU-CRD) for component-level replication. Platelet count, international normalized ratio (INR) and partial thromboplastin time (PTT) were aligned to fixed 6 h grids. The two-marker coagulation deterioration index (CDI2) was the prespecified primary exposure; coagulation instability, recovery and cumulative coagulation burden (CCB) were prespecified secondary phenotypes. Modified Poisson models estimated adjusted risk ratios (aRRs) with 95% confidence intervals (CIs). Trajectories incorporating measurements after 24 h were excluded from temporally ordered association and prediction models. Results: Among 13,669 eligible MIMIC-IV sepsis stays, 2208 had complete five-organ outcome assessment; 1278 (57.9%) developed the composite. In the primary CDI2 analysis (n = 1002; 619 events), each MIMIC standard deviation was associated with the composite (aRR 1.051, 95% CI 1.007–1.098), but the strict treatment/device-truncated estimate attenuated (aRR 1.018, 95% CI 0.966–1.074). Prespecified secondary CCB2 was associated in observed care (n = 2185; aRR 1.076, 95% CI 1.033–1.122), remained associated after strict truncation (aRR 1.061, 95% CI 1.006–1.119), and was robust to measurement-intensity adjustment and selection weighting (aRR 1.089, 95% CI 1.009–1.175). CCB2 showed the largest associations with liver and renal worsening; these persisted after excluding cirrhosis and baseline liver dysfunction and were directionally reproduced for eICU kidney and liver components. Conclusions: The prespecified primary directional-deterioration signal was small and sensitive to treatment/device truncation. Prespecified secondary analyses consistently indicated that persistence of coagulation abnormality was more informative than short-term direction, particularly for hepatic and renal deterioration. These associative findings apply principally to patients with sufficient coagulation and organ monitoring and require prospective confirmation.