Additive prognostic value of the systemic inflammatory response index to conventional severity scores in patients with respiratory failure: a retrospective analysis based on the MIMIC-IV database

The systemic inflammatory response index (SIRI) can effectively appraise the equilibrium between inflammatory response and immune status. However, its predictive value for poor prognosis in patients with respiratory failure (RF) remains unclear. We aimed to clarify the association between SIRI and all-cause mortality (ACM) in RF and evaluate its incremental prognostic value over traditional assessment tools. Patients with RF were selected from the Medical Information Mart for Intensive Care IV and evenly divided into three groups based on their SIRI values. The primary endpoint was the 30-day ACM among patients with RF. The Cox proportional hazards model and restricted cubic spline (RCS) model were used to explore the association between SIRI and ACM in this population. To assess incremental prognostic value, SIRI was added to a baseline logistic regression model that included the Sequential Organ Failure Assessment (SOFA) score. Discrimination was assessed using the Concordance statistic and receiver operating characteristic curve; calibration was assessed using the calibration curve and Brier score; model fit was assessed using the likelihood ratio test; and risk reclassification was assessed using continuous net reclassification improvement (NRI) and integrated discrimination improvement (IDI). Kaplan–Meier curves were also plotted by combined stratification of SOFA score and SIRI tertiles. The research encompassed 4,267 patients with RF, with 53.8% of the participants being male. The 30-day mortality rate was 32.3%. The Cox analysis revealed a significant association between SIRI and ACM in patients with RF (hazard ratio [HR], 1.10 [95% confidence interval (CI) 1.067–1.133]; P < 0.001). The RCS model showed a J-shaped nonlinear association between SIRI and ACM in this population ( P < 0.001). In the incremental analysis, incorporating SIRI into the baseline model significantly improved the prediction of 30-day ACM (continuous NRI = 0.1528, [95% CI 0.0842–0.2214], P < 0.001; IDI = 0.0036, [95% CI 0.0013–0.0059], P < 0.05). For 90-day ACM, SIRI likewise provided significant incremental predictive value. SIRI is independently associated with ACM in patients with RF, exhibiting a J-shaped relationship. SIRI can serve as a simple supplementary indicator to traditional severity scores for early risk stratification in patients with RF. Not applicable.