Delirium: characteristics and monitoring: Catching the wave

Abstract

This thesis starts by sharpening the diagnosis of delirium and proposing a new reference in Chapter 2 titled: The Delirium Interview. Composed after in-depth discussions with delirium experts from diverse medical fields, the Delirium Interview offers a logistically feasible approach method standardized for the heterogenous delirium population including intubated patients and lightly sedated patients. We compared live evaluation of patients with the written information provided by the trained researcher to validate if delirium experts could diagnose delirium without being physically present at the patient's bedside. This chapter showed that the Delirium Interview has a high sensitivity 89% (95% confidence interval [CI]: 71%–98%) and specificity of 82% (95% CI: 71%–90%). After our delirium reference was validated, we focused on the EEG profile of delirium. Chapter 3 used a random forest (RF) model that was based on 98 EEG features selected from previous literature to which measures suitable for quantifying characteristics of time series data were added - such as skewness - ensuring a comprehensive perspective. The model showed good performance (AUC: 0.76, 95%CI: 0.71-0.80) for delirium but could not differentiate postoperative and non-postoperative delirium cases, suggesting that the two have similar EEG profiles. Lastly, we looked at brain changes after major surgery and postoperative delirium in chapter 4. Three months post-surgery we observed an increased global fMRI connectivity, however, in patients with postoperative delirium, functional connectivity decreased, indicating long-lasting brain network effects. A decrease of global connectivity strength was associated with significant cognitive decline three months after surgery. To investigate the effects and mechanisms of cognitive impairments linked to delirium, employing an objective tool to monitor brain function on a large scale is pivotal. Chapter 5 introduces an automated EEG algorithm using wave shape analysis to detect Polymorphic delta activity in patients with acute encephalopathy presenting as delirium. Together with the wave shape analysis technique for PDA, an automated artefact detection and rejection algorithm was developed. The algorithm was built into a monitor and fragile electrodes were replaced by an easy to apply self-adhesive patch to develop a device (DeltaScan) that can provide an output score within a minute. In Chapter 6 we validated DeltaScan in the DeltaStudy: one of the largest validation studies for a delirium assessment tool including both ICU and non-ICU patients. The results show that the DeltaScan has a high success rate, good performance and can accurately differentiate delirium from other psychiatric disorders. Chapter 7 examines the reproducibility and fluctuation of acute encephalopathy measurements using the DeltaScan, compared to clinical delirium symptoms. A substantial reproducibility was found and delirium was most accurately classified in the morning, compared to afternoon and evening measurements. In conclusion, this thesis offers a comprehensive exploration of delirium, covering its diagnosis, characteristics, and impact. Thereafter it presents an innovative method for delirium monitoring: polymorphic delta activity in single-channel EEG as a key advancement with reproducible results in this multifaceted condition.