Prediction of 28-day mortality risk in patients with sepsis and underlying liver cirrhosis: the combined predictive value of INR and lactate and nomogram development

Sepsis with underlying liver cirrhosis is a high-risk disease combination with extremely high mortality in intensive care unit (ICU). Currently, there is a lack of precise prognostic assessment tools for this specific population, and traditional scoring systems have difficulty adequately capturing the complex pathological characteristics of the interaction between the “underlying chronic liver disease” and “acute sepsis insult”. This study aims to systematically explore the predictive value of the combined international normalized ratio (INR)–lactate (Lac) variable in patients with sepsis and underlying liver cirrhosis, and to construct a nomogram model for rapid assessment of 28-day mortality risk. This was a retrospective cohort study, with main data derived from the MIMIC-IV database. A total of 2069 ICU patients with sepsis and underlying liver cirrhosis were included. First, we analyzed the associations of international normalized ratio (INR) and lactate (Lac) (both individually and in combination) with 28-day mortality risk in patients using Cox proportional hazards models, and compared the predictive performance of the combined INR–Lac variable with that of individual indicators (INR, Lac) and the Sequential Organ Failure Assessment (SOFA) score. Subsequently, we applied a two-step method combining the “Boruta algorithm and least absolute shrinkage and selection operator (LASSO) regression” to screen for other core predictive variables. Based on the multivariate Cox model incorporating the combined INR–Lac variable and selected variables, we developed a nomogram. We evaluated the performance of the nomogram through area under the curve (AUC), C-index, calibration curves, and decision curve analysis (DCA). Additionally, we conducted external validation using the eICU-Critical Care Database (eICU-CRD) to assess the model’s generalizability. Both INR and lactate were independent predictors of 28-day mortality, with a synergistic effect. Multivariate analysis showed that compared with the “Low INR and Low Lac” group, the “High INR and High Lac” group had the highest 28-day mortality risk (HR = 2.822, 95% CI 2.214–3.599). The predictive efficacy of the combined INR–Lac variable (AUC = 0.717, 95% CI 0.692–0.743) was modestly better than that of single indicators and the SOFA score. The final constructed nomogram integrated seven core predictive factors including the combined INR–Lac variable, age, and SOFA, and maintained good predictive performance in the external cohort (AUC = 0.784, 95% CI 0.732–0.846). This study confirms the important predictive value of the combined INR and lactate index for 28-day mortality risk in patients with sepsis and underlying liver cirrhosis. The constructed nomogram, after rigorous internal and external validation, helps clinicians conduct rapid risk stratification, provides a basis for formulating individualized intervention strategies, and may assist in improving patient outcomes.