JMIR Publications

ABSTRACT

Background:

Perioperative disseminated intravascular coagulation (DIC) is among the most lethal hemostatic emergencies in surgical and critical care medicine, complicating 25-35% of major trauma cases, up to 30% of hepatic transplantations, and approximately 34% of severe sepsis episodes, with case-fatality rates of 20-78% depending on surgical context. Beyond hemorrhage, established DIC drives acute kidney injury in 40-60%, acute respiratory distress syndrome in 30-50%, and multi-organ failure in a substantial proportion of affected patients.

Objective:

To synthesise current evidence on the epidemiology, molecular pathophysiology, diagnostic evolution, and therapeutic advances in perioperative DIC, and to evaluate the emerging role of Software as a Medical Device (SaMD) platforms powered by machine learning and explainable artificial intelligence in enabling continuous coagulation surveillance, pre-DIC detection, and precision-guided hemostatic intervention in surgical and critical care settings.

Methods:

This narrative review was conducted through a structured search of PubMed/MEDLINE, the Cochrane Library, EMBASE, and ClinicalTrials.gov (January 2000–April 2025) using MeSH terms and free-text combinations including "disseminated intravascular coagulation," "perioperative coagulopathy," "viscoelastic testing," "thromboelastography," "machine learning," "clinical decision support," and "Software as a Medical Device." Inclusion criteria prioritized adult perioperative and critical care populations and English-language publications. Pediatric-specific DIC, hereditary coagulation disorders, and hematological malignancy without a perioperative component were excluded. This review adheres to a narrative rather than systematic design, as the heterogeneity of evidence sources precludes PRISMA-level meta-analytic synthesis.

Results:

Viscoelastic-guided transfusion algorithms reduce allogeneic blood exposure by 15–30% and enable phenotype-specific management of fibrinolytic heterogeneity — the clinically decisive distinction between hyperfibrinolysis (TEG LY30 >3%, mandating tranexamic acid) and fibrinolytic shutdown (LY30 <0.8%, where antifibrinolytics are contraindicated). The SCARLET trial demonstrated a 5.8% absolute mortality reduction in the ISTH DIC subgroup with recombinant thrombomodulin. Machine learning models demonstrate AUROCs of 0.82–0.91 for perioperative hemostatic deterioration with detection lead times of 2–4 hours before overt ISTH DIC score positivity. Critical barriers to clinical translation include algorithmic opacity, training-set population shift, alert fatigue, and the absence of prospective randomized evidence for SaMD-guided hemostatic intervention as a primary mortality endpoint.

Conclusions:

The integration of continuous viscoelastic monitoring, pre-DIC biomarker surveillance, and explainable AI-enabled SaMD platforms offers a clinically coherent pathway from reactive coagulopathy management to proactive, precision-guided hemostatic stewardship — contingent on rigorous prospective validation and regulatory collaboration