Related Experiment Video
Updated: Oct 11, 2025
A Machine Learning Approach to Design an Efficient Selective Screening of Mild Cognitive Impairment
Published on: January 11, 2020
Advantages of score-based delirium detection compared to a clinical delirium assessment-a retrospective, monocentric
Markus Jäckel1,2, Nico Aicher1,2, Xavier Bemtgen1,2
1Faculty of Medicine, Department of Cardiology and Angiology I, Heart Center Freiburg University, University of Freiburg, Freiburg, Germany.
Purpose:
Delirium is an underdiagnosed complication on intensive care units (ICU). We hypothesized that a score-based delirium detection using the Nudesc score identifies more patients compared to a traditional diagnosis of delirium by ICU physicians.
Methods:
In this retrospective study, all patients treated on a general medical ICU with 30 beds in a university hospital in 2019 were analyzed. Primary outcome was a documented physician diagnosis of delirium, or a delirium score ≥2 using the Nudesc.
Results:
In 205/943 included patients (21.7%), delirium was diagnosed by ICU physicians compared to 438/943 (46.4%; ratio 2.1) by Nudesc≥2. Both assessments were independent predictors of ICU stay (p<0.01). The physician diagnosis however was no independent predictor of mortality (OR 0.98 (0.57-1.72); p = 0.989), in contrast to the score-based diagnosis (OR 2.31 (1.30-4.10); p = 0.004). Subgroup analysis showed that physicians underdiagnosed delirium in case of hypoactive delirium and delirium in patients with female gender and in patients with an age below 60 years.
Conclusion:
Delirium in patients with hypoactive delirium, female patients and those below 60 years was underdiagnosed by physicians. The score-based delirium diagnosis detected delirium more frequently and correlated with ICU mortality and stay.
