Association between inflammatory score, coagulopathy, and hemorrhagic progression in patients with traumatic intraparenchymal hemorrhage: an exploratory study with interaction and mediation models.

Abstract

The aim of this study was to incorporate biomarkers into an inflammatory score to improve risk prediction of coagulopathy and hemorrhagic progression of a contusion (HPC) in patients with traumatic intraparenchymal hemorrhage (tIPH), and to further explore the interaction and mediation effects within the chain of events. In this retrospective study, the medical records of patients with tIPH who received treatment at two centers from January 2019 to December 2021 were reviewed. Machine learning algorithms were applied for biomarker selection, and an inflammatory score was constructed. Multivariate logistic regression was used to assess the association between the inflammatory score, coagulopathy, and HPC. Measures on multiplicative and additive scales, as well as mediation effects, were subsequently estimated. Finally, by incorporating the inflammatory score, a hybrid model of HPC occurrence was established and validated. A total of 451 patients (median age 54 years [IQR 45-66 years]) with tIPH were included in this study. The inflammatory score was developed using a combination of parameters, including the mean platelet volume, lactate dehydrogenase level, pan-immune-inflammation value, hemoglobin-to-red blood cell distribution width ratio, and C-reactive protein-to-albumin ratio. The multivariate analysis confirmed that the inflammatory score was independently associated with both coagulopathy and HPC. Additionally, the effect of a high inflammatory score on HPC occurrence was partially mediated by coagulopathy, demonstrating both direct mediation and mediated interaction effects. As a key mediator, coagulopathy accounted for 9.6% of the positive associations. Furthermore, incorporating the inflammatory score into the hybrid model demonstrated significant incremental predictive value across the training, internal, and external test sets. The inflammatory score was significantly associated with HPC, and this relationship was partially mediated by coagulopathy, with a potential synergistic interaction observed. The hybrid model improved HPC risk prediction.