C-Reactive Protein–Adjusted Prognostic Nutritional Index for Mortality Prediction in Older Adults in Intensive Care: Comparative ROC, Calibration, and Decision-Curve Analyses

Background: In elderly intensive care patients, malnutrition and inflammation together shape the risk of mortality. Traditional indices (PNI, GNRI) are widely used but show variable predictive value. In this study, our primary objective is to evaluate whether iPNI = PNI/(1+ln[CRP+1]), an index adjusted for inflammation, improved the discrimination and clinical utility of early mortality compared to PNI and GNRI. Secondary factors are the length of stay in the intensive care unit (ICU) and the use of critical organ support therapies. Methods: A single-center retrospective cohort (2018-2024; n=673) of adults aged ≥65 years admitted to the ICU. The primary outcome was defined as ICU mortality. The secondary endpoint was the need for ICU admission and mechanical ventilation (MV), vasopressor (VP), or renal replacement therapy (RRT). We performed DeLong tests, bootstrap internal validation, calibration (intercept, slope, Brier), decision curve analysis (10–40% threshold values), age/sex-adjusted logistic models, Spearman correlations for LOS, and ROC with CRP stratified ROC for PNI and GNRI. Youden cutoff values were determined exploratory a priori. Results: iPNI demonstrated the highest discrimination (AUC 0.961, 95% CI 0.940-0.982), outperforming PNI (0.901, 0.876-0.927) and GNRI (0.681, 0.639-0.723) (both p<0.001 vs. iPNI). The PNI+GNRI model did not outperform PNI alone. The advantages of iPNI persisted after age/sex adjustment; calibration was nearly ideal with the lowest Brier. Decision curves favored iPNI in the 10-40% range. iPNI had the strongest correlation with LOS (ρ=−0.412), but binary performance was limited for LOS ≥10 days (research purposes). Higher index values were associated with lower requirements for mechanical ventilation, VP, and RRT. Conclusion: In older ICU adults, inflammation-adjusted iPNI offers superior mortality discrimination with robust calibration and consistent clear benefit for early risk stratification. PNI remains more informative than GNRI. Given the single-center, retrospective design and sensitivity to CRP timing, external prospective validation is necessary before introducing it into routine practice.