Comparative analysis of four nutritional scores in predicting delirium in ICU patients

The nutritional assessment indicators for critically ill patients are diverse, with limited research about comparing the predicting value of different nutritional assessment tools for delirium in the intensive care unit (ICU).

The study aimed to validate the relationship between malnutrition and ICU delirium and explore the optimal nutritional scores for predicting ICU delirium.

This study was based on the Medical Information Mart for Intensive Care IV (MIMIC-IV) database and included 319 ICU patients who met the inclusion and exclusion criteria. The study used four nutritional assessment tools: Geriatric Nutritional Risk Index (GNRI), Prognostic Nutritional Index (PNI), Triglycerides (TG) × Total Cholesterol (TC) × Body Weight (BW) Index (TCBI), and Controlling Nutritional Status (CONUT) score. Restricted cubic spline (RCS) modeling, single-factor logistic regression, and multivariate stepwise logistic regression were employed to elucidate the relationships between each nutritional score and delirium. Using area under the curve (AUC) evaluated the discriminatory ability of the adjusted models.

The RCS shows a strong linear connection between delirium and PNI (P for nonlinear = 0.66), as well as between delirium and CONUT score (P for nonlinear = 0.32). Multivariate logistic regression reveals that PNI (OR = 2.04, 95% CI: 1.05-4.03, p = 0.04) has the closest relationship with ICU delirium. The AUC of the PNI prediction model after adjusting covariates was 0.87 (95% confidence interval: 0.83-0.91, p < 0.05).

The study confirmed the association between poor nutritional status and increased risk of ICU delirium in patients. PNI demonstrated excellent independent predictive value for ICU delirium, warranting further clinical application and validation.