Comparative Prognostic Performance of HALP, PIV, and Naples Prognostic Score in Critically Ill Patients with Sepsis: A Retrospective Multicentre Cohort Study

Background: Sepsis is a life-threatening organ dysfunction caused by a dysregulated host response to infection (Sepsis-3 definition), associated with high mortality in intensive care unit (ICU) patients. Composite immune–nutritional indices derived from routine laboratory data have emerged as accessible prognostic tools; however, their comparative value in critically ill septic patients remains insufficiently characterised. This study aimed to compare the prognostic performance of the haemoglobin–albumin–lymphocyte–platelet (HALP) score, pan-immune-inflammation value (PIV), and Naples Prognostic Score (NPS) for predicting in-hospital mortality in ICU patients with sepsis as the primary outcome, and to assess their incremental predictive value as the secondary objective. Methods: In this retrospective, two-centre cohort study, 1020 consecutive eligible adult patients fulfilling Sepsis-3 criteria (suspected or confirmed infection with an acute increase in SOFA score ≥ 2 points) admitted to the ICUs of Necmettin Erbakan University Hospital and Beyhekim Training and Research Hospital between January 2016 and June 2025 were included. HALP was calculated as haemoglobin (g/L) × albumin (g/L) × lymphocyte count (×109/L) ÷ platelet count (×109/L); PIV as (neutrophil × platelet × monocyte) ÷ lymphocyte (all ×109/L). NPS was computed from serum albumin, neutrophil-to-lymphocyte ratio, and lymphocyte-to-monocyte ratio, with the total-cholesterol component imputed due to availability in only 31.7% of patients. Discriminative performance was evaluated by receiver operating characteristic (ROC) analysis, pairwise DeLong tests, bootstrap resampling (1000 iterations), Hosmer–Lemeshow calibration, and net reclassification improvement (NRI)/integrated discrimination improvement (IDI) analyses. Five pre-specified nested multivariable logistic regression models were constructed. Results: Of 1020 patients (median age 76 years, IQR 67–83; 59.8% male), 521 (51.1%) died during hospitalisation. HALP showed the highest discriminative ability among individual indices (AUC 0.626, 95% CI 0.594–0.658), while PIV was non-discriminatory (AUC 0.504, p = 0.78) and NPS showed limited performance (AUC 0.563, 95% CI 0.531–0.595). HALP remained an independent predictor of mortality after multivariable adjustment (OR 0.98, 95% CI 0.97–0.99, p = 0.002). NRI and IDI analyses showed no incremental value with NPS addition. Conclusions: HALP demonstrated modest but independently consistent discrimination for in-hospital mortality in ICU patients with sepsis, outperforming PIV and NPS. However, an AUC of 0.626 does not support standalone clinical use; external validation and comparison with established severity models are required before integration into risk stratification frameworks.