Prognostic Performance of the Blood Urea Nitrogen-to-Albumin Ratio and Prognostic Nutritional Index for 30-Day Mortality in Patients with Delirium.

BACKGROUND

Delirium is a common and serious syndrome in hospitalized patients and is associated with substantial short-term mortality. Simple laboratory-based markers may help identify patients at increased risk after delirium diagnosis.

AIMS: To compare the prognostic performance of the Prognostic Nutritional Index (PNI) and the Blood Urea Nitrogen-to-Albumin Ratio (BAR) for 30-day mortality in hospitalized patients with delirium and to evaluate their incremental contribution beyond selected clinical and laboratory variables.

METHODS

This study included 115 hospitalized patients diagnosed with delirium after psychiatric consultation. Thirty-day vital status and documented medical causes of death were determined using both hospital records and the national electronic health record system. Parsimonious multivariable logistic regression models were constructed using age, platelet count, red cell distribution width (RDW), and creatinine as common covariates, with PNI and BAR entered separately. Sequential sensitivity analyses additionally considered infection and malignancy. Predictive discrimination was assessed using receiver operating characteristic analyses and area under the curve (AUC) comparisons.

RESULTS

Forty-two patients (36.5%) died from medical causes within 30 days. Non-survivors had lower PNI values than survivors (p=.001), whereas BAR was numerically higher but did not differ significantly in the unadjusted comparison (p=.095). In the primary multivariable models, higher BAR was associated with increased odds of 30-day mortality (p=.004), whereas PNI was not significantly associated with mortality (p=.074). BAR remained significant when infection and malignancy were added separately but was attenuated when both were included simultaneously (p=.063). In unadjusted ROC analysis, PNI showed a numerically higher AUC than BAR (0.690 vs 0.594; p=.125). Addition of BAR to the common base model produced a larger numerical increase in AUC than addition of PNI (+0.041 vs +0.020), although the BAR- and PNI-containing models did not differ significantly (p=.596).

CONCLUSIONS

BAR showed a stronger adjusted association with 30-day mortality than PNI and provided a numerically greater incremental improvement in model discrimination. However, direct comparisons did not demonstrate statistically significant superiority of BAR over PNI, and its association was attenuated after simultaneous adjustment for infection and malignancy. These findings support further evaluation of BAR as a simple prognostic marker in larger prospective cohorts.