Comparison of nutritional inflammatory indicators for predicting outcomes in patients with sepsis: a retrospective study based on clinical and MIMIC data.

Abstract

Background: This study aimed to investigate the utility of the prognostic nutrition index (PNI) and the Naples prognostic score (NPS) in predicting the prognosis of patients with sepsis.

Methods: A total of 149 clinical patients from Binhai County People's Hospital and 200 patients from the MIMIC-IV database were independently analyzed. Baseline data, including infection sources, causative organisms, and therapeutic interventions (vasoactive drugs, steroids, and RRT), were collected within the first 24 h of ICU admission. Kaplan-Meier curves and hierarchical Cox proportional hazard models (adjusting for SOFA, NLR, and comorbidities) were constructed to identify independent prognostic factors. Model performance was comprehensively evaluated using the C-index, Likelihood Ratio Test, and calibration curves.

Results: Both cohorts revealed that patients with low PNI (L-PNI < 32.1) and high NPS (H-NPS) had a significantly increased risk of 28-day mortality. Non-survivors exhibited a higher prevalence of pulmonary infections and higher requirements for vasoactive drugs and RRT (all p < 0.05). Analysis of the independent MIMIC-IV cohort further confirmed that NPS was an independent risk factor while PNI was an independent protective factor, even after adjusting for SOFA scores and individual inflammatory components (VIF < 5). The combined ROC model (PNI + NPS) demonstrated excellent discriminative power with an AUC of 0.805 in the clinical cohort and 0.806 in the MIMIC cohort. Calibration curves showed high agreement between predicted and actual survival probabilities (Hosmer-Lemeshow p > 0.05).

Conclusions: PNI and NPS are independent risk factors for 28-day mortality in patients with sepsis. PNI combined with NPS provides greater predictive value for the survival and prognosis of patients with sepsis.

Clinical trial: Not applicable.