Nutritional-inflammatory-metabolic indices associated with in-hospital mortality in acute kidney injury patients...

Composite indices derived from routine laboratory parameters can reflect nutritional, inflammatory, and metabolic status, yet their prognostic value in acute kidney injury (AKI) patients receiving continuous renal replacement therapy (CRRT) has not been systematically compared. This study aimed to evaluate and compare the prognostic value of six such indices and to integrate them into machine learning-based prediction models.

This retrospective cohort study enrolled 1,732 AKI patients who received CRRT between 2020 and 2025. Six composite indices-Albumin-to-Alkaline Phosphatase Ratio (AAPR), Prognostic Nutritional Index (PNI), Albumin-Bilirubin Score (ALBI), Blood Urea Nitrogen to Creatinine Ratio (BUN/Cr), Systemic Immune-Inflammation Index (SII), and Neutrophil-to-Lymphocyte Ratio (NLR)-were calculated from laboratory data within the first 24 h of ICU admission. Associations with in-hospital mortality were evaluated using multivariable Cox regression, restricted cubic spline (RCS) analysis, and subgroup analyses. Feature selection and six machine learning models with SHAP interpretability analysis were employed for risk prediction.

A total of 507 patients (29.3%) died during hospitalization. AAPR was the only index with a consistent protective association in the fully adjusted model (Q4 vs. Q1: HR 0.683, 95% CI 0.524-0.890; P for trend = 0.007). RCS analysis revealed an L-shaped relationship (P for non-linearity = 0.002) with an inflection point at 0.203; below this threshold, mortality risk decreased steeply with increasing AAPR, whereas above it the risk plateaued. PNI exhibited a U-shaped association (inflection point: 33.8). ALBI and BUN/Cr lost significance after full adjustment; SII and NLR showed no stable prognostic value. Subgroup analysis confirmed the robustness of AAPR, with age as the only significant effect modifier (P for interaction = 0.002). Gradient Boosting achieved the highest test-set AUC (0.728), and SHAP ranked AAPR as the most important derived index.

Among six nutritional-inflammatory-metabolic composite indices, AAPR showed the most robust prognostic value, with an L-shaped dose-response relationship identifying 0.203 as a risk stratification threshold. The U-shaped association of PNI cautions against equating higher values with better prognosis.