Multidimensional Radiological Assessment of Delirium in the Emergency Department

Abstract

Background: Delirium is a common, underdiagnosed neuropsychiatric syndrome in older adults, associated with high mortality and functional decline. Given its multifactorial nature and overlap with frailty, radiological markers may improve risk stratification in the emergency department (ED). Methods: We conducted a retrospective study on a small sample of 30 patients diagnosed with delirium in the emergency department who had recently undergone brain, thoracic, or abdominal CT scans for unrelated clinical indications. Using post-processing software, we analyzed radiological markers, including coronary artery calcifications (to estimate vascular age), cerebral atrophy (via the Global Cortical Atrophy scale), and cachexia (based on abdominal fat and psoas muscle volumetry). Results: Five domains were identified as significant predictors of 12-month mortality in univariate Cox regression: vascular age, delirium etiology, cerebral atrophy, delirium subtype (hyperactive vs. hypoactive), and cachexia. Based on these domains, we developed an exploratory 10-point delirium score. This score demonstrated acceptable diagnostic accuracy for mortality prediction (sensitivity 0.93, specificity 0.73, positive predictive value 0.77, negative predictive value 0.91) in this limited cohort. Conclusions: While preliminary and based on a small, retrospective sample of 30 patients, this multidimensional approach integrating clinical and radiological data may help improve risk stratification in elderly patients with delirium. Radiological phenotyping, particularly in terms of vascular aging and sarcopenia/cachexia, offers objective insights into patient frailty and could inform more personalized treatment pathways from the ED to safe discharge home, pending further validation.

Keywords: brain atrophy; cachexia; coronary calcium score; delirium; mortality.

Figure 1

ROC and univariate Cox regression of the 5 variables included in the delirium score.

Figure 2

Comparison of ROC curves of the 5 variables. Notes: “a” refers to the method used to calculate the standard error (typically under a nonparametric assumption). “b” refers to the null hypothesis tested for the AUC (i.e., true area = 0.5).

Figure 3

Importance and weight of the individual variables in the score calculation.

Figure 4

Survival based on point score: low score vs. intermediate scores vs. high score. Below is the figure statistical significance of the score in the Cox regression model.

Figure 5

ROC curve, sensitivity, and specificity of the score.

Figure 6

Statistical performance of the score in predicting mortality.