Admission systemic inflammation response index is associated with 14-day adverse clinical outcomes in patients with pyogenic liver abscess: a single-center retrospective cohort study

Abstract

To assess the association between the systemic inflammation response index (SIRI) at admission and 14-day adverse clinical outcomes in patients with pyogenic liver abscess (PLA), and to evaluate its discriminatory performance. This single-center retrospective cohort study included 670 adults with PLA admitted to Linyi People’s Hospital between January 2021 and December 2024. SIRI was calculated from the first complete blood count obtained within 24 hours after admission according to the study data-collection criteria. The primary outcome was a composite of adverse clinical events within 14 days. A base logistic regression model and a model additionally including SIRI were fitted in 30 multiply imputed datasets, with complete-case and death-exclusion sensitivity analyses. Restricted cubic splines, receiver operating characteristic curves, and exploratory measures of incremental performance were evaluated. Of the 670 patients, 100 (14.9%) experienced an adverse clinical outcome, including four deaths. Median SIRI was higher in patients with adverse outcomes [8.95 (4.14-15.64) vs. 5.80 (2.93-11.64); P< 0.001]. In the multiply imputed linear model, the adjusted OR for SIRI was 1.0187 (95% CI, 1.0022-1.0354; P = 0.0262); the complete-case estimate was 1.0163 (95% CI, 0.9994-1.0334; P = 0.0588). After exclusion of four deaths, the adjusted OR was 1.0191 (95% CI, 1.0026-1.0360; P = 0.0233) among 666 patients with 96 nonfatal adverse outcomes. The spline overall association and nonlinearity P values were 0.0557 and 0.4080, respectively. The AUC for SIRI was 0.6072, lower than that for NLR (0.6782; paired DeLong P = 0.0015). Adding SIRI increased the apparent model AUC by 0.0116 (95% bootstrap interval, -0.0006 to 0.0351). Higher admission SIRI was associated with greater odds of 14-day adverse clinical outcomes in the multiply imputed linear model, but its standalone discrimination was limited. Sensitivity analyses and incremental performance estimates indicate uncertainty, and clinical utility requires further validation.