Diagnostic Performance and Utility of Quantitative EEG Analyses in Delirium: Confirmatory Results From a Large Retrospective Case-Control Study. (March 2019)

• Record Type: Journal Article
• Authors: Fleischmann, Robert; Tränkner, Steffi; Bathe-Peters, Rouven; Rönnefarth, Maria; Schmidt, Sein; Schreiber, Stephan J.; Brandt, Stephan A.
• Is Part Of: Clinical EEG and neuroscience. Volume 50:Number 2(2019)
• Page Start: 111
• Page End: 120
• Publication Date: 2019-03
• Subjects: delirium -- quantitative electroencephalography -- diagnostic performance -- validation -- reliability -- adults; Electroencephalography -- Periodicals; Neurosciences -- Periodicals; 616.8047547
• DOI: 10.1177/1550059418767584
• Languages: English
• ISSNs: 1550-0594

Abstract

Background . The lack of objective disease markers is a major cause of misdiagnosis and nonstandardized approaches in delirium. Recent studies conducted in well-selected patients and confined study environments suggest that quantitative electroencephalography (qEEG) can provide such markers. We hypothesize that qEEG helps remedy diagnostic uncertainty not only in well-defined study cohorts but also in a heterogeneous hospital population. Methods . In this retrospective case-control study, EEG power spectra of delirious patients and age-/gender-matched controls (n = 31 and n = 345, respectively) were fitted in a linear model to test their performance as binary classifiers. We subsequently evaluated the diagnostic performance of the best classifiers in control samples with normal EEGs (n = 534) and real-world samples including pathologic findings (n = 4294). Test reliability was estimated through split-half analyses. Results . We found that the combination of spectral power at F3-P4 at 2 Hz (area under the curve [AUC] = .994) and C3-O1 at 19 Hz (AUC = .993) provided a sensitivity of 100% and a specificity of 99% to identify delirious patients among normal controls. These classifiers also yielded a false positive rate as low as 5% and increased the pretest probability of being delirious by 57% in an unselected real-world sample. Split-half reliabilities were .98 and .99, respectively. Conclusion . This retrospective study yielded preliminary evidence that qEEG provides excellent diagnostic performance to identify delirious patients even outside confined study environments. It furthermore revealed reduced beta power as a novel specific finding in delirium and that a normal EEG excludes delirium. Prospective studies including parameters of pretest probability and delirium severity are required to elaborate on these promising findings.