Association of the atherogenic index of plasma with in-hospital mortality in patients with sepsis-induced coagulopathy.

BACKGROUND

Few inexpensive early parameters are available to describe sepsis-induced coagulopathy (SIC) in practice. The atherogenic index of plasma (AIP = log₁₀[TG/HDL-C]) integrates triglyceride excess and HDL-C depletion—metabolic shifts implicated in endothelial injury and thrombosis. We evaluated the association between the AIP and in-hospital mortality in patients with SIC.

METHODS

We performed a retrospective cohort study of patients with SIC admitted to the ICU of the First Affiliated Hospital of Wenzhou Medical University (2017–2023). The primary outcome was in-hospital mortality. Multivariable logistic models adjusted sequentially for age/sex and Boruta/LASSO-selected covariates (age, SOFA score, albumin, sodium, potassium, ventilation, and renal diseases) plus major comorbidities. Dose–response was assessed by restricted cubic splines. Sensitivity analyses were performed using different SIC criteria and alternative adjustment strategies to verify robustness. Machine learning (CatBoost, RF, LR, and MLP) classifiers were compared for discrimination and calibration.

RESULTS

1096 patients were included. Higher baseline AIP was associated with increased mortality across models (per-SD OR 1.29, 95% CI 1.09–1.53, fully adjusted). Restricted cubic splines supported an approximately linear association. The logistic model reached an AUROC of 0.71 in the test set, with a positive net benefit on decision curve analysis. Additional machine-learning models showed comparable discrimination in the test set (AUROC around 0.70). SHAP analysis indicated that ventilation contributed most strongly to model predictions, with AIP also emerging as an important prognostic feature.

CONCLUSIONS

A higher AIP is linked to increasing in-hospital mortality in SIC patients, which aligns with the metabolic–vascular disturbances that the AIP captures. When used as a complementary measure, the AIP may aid early clinical appraisal without implying stand-alone prognostication.