A stepwise laboratory-based approach for early risk assessment of disseminated intravascular coagulation

Abstract

Background

Disseminated intravascular coagulation (DIC) is a life-threatening condition in which early identification is essential for timely clinical management. However, existing diagnostic approaches are often complex. This study aimed to develop and validate a laboratory-based prediction model for early DIC risk assessment, while additionally evaluating different D-dimer interpretation strategies in relation to current diagnostic recommendations.

Methods

This single-center retrospective observational study consecutively included adult patients with overt disseminated intravascular coagulation (DIC) diagnosed between January 2015 and December 2024, together with individuals without DIC. An independent cohort from January to December 2025 was used for external validation. Ten routinely available coagulation and platelet-related biomarkers measured within 24 h prior to DIC diagnosis were analyzed. The diagnostic performance of D-dimer alone was evaluated using both the Chinese expert consensus cut-off strategy and the 2025 International Society on Thrombosis and Haemostasis (ISTH) upper limit of normal (ULN)–based classification. A multivariable prediction model was subsequently developed using routinely available laboratory parameters.

Results

A total of 2750 participants were included in the main cohort, with an additional 363 individuals in the external validation cohort. D-dimer interpreted using the ISTH ULN-based classification demonstrated better overall diagnostic performance than fixed cut-off strategies, with the >3 × ULN threshold providing the most balanced sensitivity and specificity. The final model incorporated four laboratory parameters—prothrombin time (PT), activated partial thromboplastin time (APTT), D-dimer, and mean platelet volume (MPV)—and demonstrated strong discrimination, with areas under the receiver operating characteristic curve (AUCs) of 0.908 in the training cohort, 0.894 in the validation cohort, and 0.971 in the external validation cohort. Predicted risk showed a clear stepwise increase in DIC prevalence across risk strata.

Conclusions

This four-parameter laboratory-based model provides an interpretable and externally validated tool for early DIC risk assessment using routinely available laboratory tests. The model may support preliminary risk stratification and facilitate timely clinical evaluation in patients at risk for DIC.