Association of PIV value with early mortality in ICU patients with sepsis-associated acute kidney injury from the MIMIC IV database.

Abstract

Sepsis is a severe systemic inflammatory response, and sepsis-associated acute kidney injury (SA-AKI) is one of its most common complications. The pan-immune inflammation value (PIV), a novel inflammatory index, is designed to comprehensively reflect the status of systemic immune and inflammatory responses. However, the relationship between PIV and short-term clinical outcomes in SA-AKI patients remains unclear. This study was a retrospective analysis of SA-AKI patients from the MIMIC-IV database. The Boruta algorithm was used to identify key features predicting short-term mortality in SA-AKI patients. The relationships between ln (PIV) and all-cause mortality at 28 days and 90 days were assessed via multivariate Cox proportional hazards regression, subgroup analysis, sensitivity analysis, restricted cubic spline (RCS) modelling, and Kaplan‒Meier (K-M) survival analysis. A total of 4369 patients were included in the study, of whom 57.0% were male. Boruta analysis indicated that ln (PIV) was an important clinical feature. The results of multivariable Cox regression analysis revealed a positive correlation between ln (PIV) and mortality risk at both 28 days and 90 days (HR [95% CI] = 1.057 [1.009, 1.106], P = 0.019; HR [95% CI] = 1.075 [1.032, 1.120], P < 0.001). The RCS model revealed a nonlinear relationship between ln (PIV) and mortality at 28 and 90 days, with a critical threshold of 6.72. Above this threshold, a higher ln (PIV) was associated with increased mortality risk at both time points; sensitivity analyses confirmed that this association remained significant after specific patients were excluded. Subgroup analyses revealed that ln (PIV) significantly affected short-term mortality in diabetic patients (P < 0.05). Ln (PIV) is closely associated with short-term mortality in ICU patients with SA-AKI, suggesting its potential application in early risk assessment and clinical intervention.