Article Text

DelIrium VULnerability in GEriatrics (DIVULGE) study: a protocol for a prospective observational study of electroencephalogram associations with incident postoperative delirium
1. Monique S Boord1,
2. Daniel H J Davis2,
3. Peter J Psaltis3,4,5,
4. Scott W Coussens1,
5. Daniel Feuerriegel6,
6. Marta I Garrido6,
7. Alice Bourke7 and
8. Hannah A D Keage1
1. 1Cognitive Ageing and Impairment Neurosciences Laboratory, Justice and Society, University of South Australia, Adelaide, South Australia, Australia
2. 2MRC Unit for Lifelong Health and Ageing, UCL, London, UK
3. 3Vascular Research Centre, Heart and Vascular Program, Lifelong Health Theme, South Australian Health and Medical Research Institute, Adelaide, South Australia, Australia
6. 6Melbourne School of Psychological Sciences, University of Melbourne, Melbourne, Victoria, Australia
7. 7Aged Care, Rehabilitation and Palliative Care (Medical), Northern Adelaide Local Health Network, Adelaide, South Australia, Australia
1. Correspondence to Monique S Boord; monique.boord{at}mymail.unisa.edu.au

## Abstract

Introduction Delirium is a neurocognitive disorder common in older adults in acute care settings. Those who develop delirium are at an increased risk of dementia, cognitive decline and death. Electroencephalography (EEG) during delirium in older adults is characterised by slowing and reduced functional connectivity, but markers of vulnerability are poorly described. We aim to identify EEG spectral power and event-related potential (ERP) markers of incident delirium in older adults to understand neural mechanisms of delirium vulnerability. Characterising delirium vulnerability will provide substantial theoretical advances and outcomes have the potential to be translated into delirium risk assessment tools.

Methods and analysis We will record EEG in 90 participants over 65 years of age prior to elective coronary artery bypass grafting (CABG) or transcatheter aortic valve implantation (TAVI). We will record 4-minutes of resting state (eyes open and eyes closed) and a 5-minute frequency auditory oddball paradigm. Outcome measures will include frequency band power, 1/f offset and slope, and ERP amplitude measures. Participants will undergo cognitive and EEG testing before their elective procedures and daily postoperative delirium assessments. Group allocation will be done retrospectively by linking preoperative EEG data according to postoperative delirium status (presence, severity, duration and subtype).

Ethics and dissemination This study is approved by the Human Research Ethics Committee of the Royal Adelaide Hospital, Central Adelaide Local Health Network and the University of South Australia Human Ethics Committee. Findings will be disseminated through peer-reviewed journal articles and presentations at national and international conferences.

Trial registration number ACTRN12618001114235 and ACTRN12618000799257.

• event-related potentials
• cognitive electrophysiology
• EEG
• geriatrics
• neurophysiology

This is an open access article distributed in accordance with the Creative Commons Attribution 4.0 Unported (CC BY 4.0) license, which permits others to copy, redistribute, remix, transform and build upon this work for any purpose, provided the original work is properly cited, a link to the licence is given, and indication of whether changes were made. See: https://creativecommons.org/licenses/by/4.0/.

## Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

### Strengths and limitations of this study

• Our prospective design measuring electroencephalography (EEG) before delirium will allow characterisation of neural mechanisms associated with delirium vulnerability.

• We will use state-of-the-art EEG analysis and visualisation methods to elucidate neural mechanisms.

• We will extend on previous studies by assessing at the delirium subtype level and by measuring event-related potentials.

• This study is limited geographically to Adelaide, South Australia, and is limited by our inability to balance subtype sample sizes.

## Introduction

### Measures

#### Cognitive function

Cognitive status is assessed before the procedure using the ACE-III, a global measure of cognitive function commonly used to screen for dementia.64 The ACE-III comprises five different cognitive domains, including attention, memory, language, verbal fluency and visuospatial ability, with a maximum score of 100. Higher scores indicate better cognitive function; specific subtotal scores include 18 points for attention, 26 points for memory, 14 points for verbal fluency, 26 points for language and 16 points for visuospatial ability.64 Cut-offs for dementia and mild cognitive impairment are characterised by scores lower than 82 and 88, respectively, and show high sensitivity (93%–100%) and specificity (96%–100%).65 66

#### Delirium assessment

Delirium presence, subtype, and severity will be captured using a comprehensive battery. To assess delirium in the intensive care unit, the Confusion Assessment Method (CAM) for the Intensive Care Unit (CAM-ICU) flowsheet will be used. The CAM-ICU features the four Diagnostic and Statistical Manual of Mental Disorders (DSM-III-R) delirium criteria: acute onset or fluctuating course, inattention, altered level of consciousness and disorganised thinking.67 It has high sensitivity (88% and 92%), specificity (92% and 100%) and interrater reliability (kappa.96), and is quickly administered.67 To assess delirium severity in ICU, the CAM-ICU 7 (CAM-ICU-7) will be used. The CAM-ICU-7 is a 7-point scale derived from CAM-ICU and Richmond Agitation Sedation Scale responses, and encompasses high internal consistency (Cronbach α: 0.85), and correlates well to the Delirium Rating Scale, Revised-98 (correlation coefficient: 0.64).68

Delirium assessment on the surgical wards will consist of the Memorial Delirium Assessment Scale (MDAS), which contains 10 items assessing: reduced level of consciousness, disorientation, short-term memory impairment, impaired digit span, reduced ability to maintain and shift attention, disorganised thinking, perceptual disturbance, delusions, decreased or increased psychomotor activity and sleep–wake cycle69. To help assess disturbances in arousal, the observational scale level of arousal (OSLA) is included in the ward assessments. The OSLA holds high sensitivity (0.87) and specificity (0.81), and characterises disturbances in arousal associated with delirium using four features: (1) eye opening, (2) eye contact, (3) posture and (4) movement.70 71 The MDAS (and OSLA) will ascertain a score of delirium severity and inform DSM-IV delirium presence or absence. The MDAS has high sensitivity (100%), specificity (95%), interrater reliability (κ: 0.92) and internal consistency (Cronbach α: 0.89).72 The short CAM is also collected, informed by MDAS and OSLA.

This comprehensive assessment will provide a dichotomous outcome for delirium (present/absent) along with the subtype, severity and duration of the delirium episode. Delirium will not be assessed by study staff on weekends. In the case of patients in hospital over the weekend, a chart-based review tool adapted from Inouye and colleagues will be used (74% sensitivity, 83% specificity and 0.41 interrater reliability κ)73. This method is not as extensive as the daily assessments during the working week, and may underestimate or inaccurately determine the presence of delirium.

#### EEG acquisition

We employ a 9-minute EEG recording using a 32-channel Ag/AgCI live electrode montage (Fp1, Fz, F3, F7, FT9, FC5, FC1, C3, T7, TP9, CP5, CP1, Pz, P3, P7, O1, Oz, O2, P4, P8, TP10, CP6, CP2, Cz, C4, T8, FT10, FC6, FC2, F8 and Fp2) positioned in an elastic cap according to the 10–10 system using Modified Combinatorial Nomenclature. EEG data is recorded using BrainVision Recorder (V.1.22.0001, Brain Products GmbH, Gilching, Germany) software at a sample rate of 1000 Hz, and is amplified by a LiveAmp amplifier (Brain Products GmbH, Gilching, Germany). We use actiCAPs (Brain Products GmbH, Gilching, Germany) with recording reference FCz, and ground Fpz electrode positions. Scalp electrode impedance will be kept below 10 kΩ before recording begins and if impedances drift above 25 kΩ during the recording, they will be interpolated.

The first 4 min of the recording constitutes the resting state period, comprised of 2 min eyes open and 2 min eyes closed. Immediately after, using Sennheiser Urbanite XL headphones, a 5-minute passive auditory oddball paradigm is employed, consisting of 300 stimuli of 150-millisecond stimulus duration and a 500-millisecond interstimulus interval; standard tones are presented at 600 Hz and deviant tones (23% of stimuli) at 1000 Hz. Sound density is set to −6 dBFS and the volume setting on the device is set to 86%. Participants are seated comfortably in front of a laptop placed on a table or available flat surface in the participant’s home. During the eyes open component of the resting state recording, participants are directed to look at a fixation point indicated by a cross in the centre of the laptop screen. During the oddball paradigm, participants are directed to watch a silent video on an iPad of passing traffic on a main road next to the University campus. Due to this being a clinically relevant protocol, we are unable to set individual auditory oddball parameters for participants, but we do check that they can hear the tones. Participants are shown the raw EEG signal to demonstrate common artefacts. They are instructed to relax, sit with their feet flat on the floor, and to avoid movement and excessive blinking.

### Data processing and analysis

Age and baseline cognitive function (two primary risk factors for delirium)36 will be used as covariates in our models. This will ensure that our EEG and ERP measures capture brain vulnerability to delirium independent of brain functional changes due to age and cognitive impairment. We will run sensitivity analyses covarying for procedure type (or stratified analyses if our numbers are too unbalanced), to ensure associations are not being driven by one patient group (CABG or TAVI). The EEG analysis approach combines measures of well-defined ERP components (eg, the MMN) and frequency bands (eg, theta) with data-driven approaches based on mass univariate analyses. Traditionally, most research has investigated the oscillatory (periodic) component of the EEG power spectra, but not the aperiodic background 1/f like component in which these oscillations are embedded.74 This aperiodic component has been found to change with ageing and cognitive state75 76 and contains features independent of oscillatory activity that appear to be physiologically relevant; failing to consider this aperiodic component may disguise physiologically relevant data.54 We will calculate traditional bandwith measures of the resting state data and will make these available.

### Power preprocessing

Resting state EEG data will be preprocessed in MATLAB (V.R2019a, The Mathworks, USA) using the EEGLAB toolbox V.v2019.1.77 We will remove bad or unused channels, and the data will be band-pass filtered from 1 Hz to 45 Hz. The data will be downsampled to 500 Hz and re-referenced to electrodes TP9 and TP10 before independent components analysis (ICA).78 ICLabel, an automated component classification method,79 will be applied to correct for ocular and muscle artefacts using an 80% threshold. Components identified as an 80% match to a previously identified artefact, that is, eye or muscle, will be removed.79 Bad or unused channels will be interpolated using clean data. Older adults have been shown to display a slower alpha frequency80 with alpha peaks around 8 Hz in some participants.81 Accordingly, in line with Tanabe and colleagues,33 we will set the lower limit of the alpha band to 6 Hz. From the aperiodic component, 1/f slope and offset features of the resting state EEG will be extracted using the open-source Python FOOOF (fitting oscillations and one over f) toolbox74 for comparison between groups. The FOOOF toolbox is available at https://githubcom/fooof-tools/fooof.

### ERP preprocessing

We will process ERP data in MATLAB V.R2019b with the ERPlab V.7.0.0 extension.82 Data will be re-referenced to electrodes TP9 and TP10. A 0.1 Hz high-pass filter, a 40 Hz low-pass filter and a 50 Hz notch filter will be applied and ICA will be performed on the filtered datasets. ICLabel will be used to remove bad components at a threshold of 80%. Bad channels will be interpolated using clean data. The data will then be low pass filtered at 20 Hz. Data will be epoched from −100 ms to +400 ms relative to auditory tone onset, and epochs containing amplitudes larger than ±100 μV will be excluded.

## Analytical approach

Averaged ERP data will be converted into three-dimensional spatiotemporal images for each participant and modelled using a mass-univariate general linear model implemented in statistical parametric mapping (SPM); the software is freely available.83 Statistical maps will be thresholded using family-wise error rate correction for multiple comparisons at a level of p <0.05, and clusters above the defined threshold will be examined. A cluster forming threshold of p <0.001 (uncorrected) will be used when the former (more conservative) approach does not reach significance. Only clusters p <0.05 cluster-corrected will be reported. This method allows for investigation of statistical effects across the entire dataset instead of using a priori time windows. Open-source MATLAB toolboxes Porthole and Stormcloud53 will be employed to visualise the scalp-time images created with SPM. Porthole and Stormcloud is available at https://githubcom/JeremyATaylor/Porthole.

We will also conduct traditional ERP analyses of the MMN and P3 component amplitude and latency to confirm our paradigm. We expect that components will display the typical age-related delays, where the MMN is found between 200 ms and 300 ms, and the P3 between 300 ms and 400 ms. Early components, which contribute to the MMN, including the P1 and N1, will be assessed in an exploratory manner. Standard analysis of covariance (ANCOVA) approaches will be used for the resting EEG and ERP component amplitude data. The independent variable will be delirium post procedure (delirium or no delirium) and the dependent variables will be the ERP amplitude and latency (MMN and P3), aperiodic offset and slope, and spectral power in theta, delta, alpha, beta and gamma frequency bands. Cognitive status (indexed by the ACE-III) and age will be modelled as covariates. We will categorise our electrodes into three regions: frontal, central and posterior, and divide our alpha by three (0.05/3) for the resting state EEG data. Independent samples t-tests will be used to assess differences across subtypes (hyperactive, hypoactive and mixed).

### Informed consent

This study is nested within two larger clinical trials, both of which have been approved for registration. Written consent is obtained from willing and eligible participants by study staff. Participants are reassured that participation is completely voluntary and that they can withdraw at any time, and that it will not affect the care provided to them while in hospital for their procedure.

### Data management

All identifying information are kept on a secure database (REDCap) accessible only by central study staff via institutional log in with individualised usernames and passwords. Information collected at recruitment is taken to secure storage in the laboratory immediately after.

### Risks

The study does not interfere with participants’ surgery or recovery and thus poses no additional risk to participants. EEG is completely non-invasive and poses no risk to participants. A small risk is posed to study staff when collecting data in participants’ homes, but this is mitigated by the use of a log-in safety application (HikerAlert), where if study staff fail to check in after a set amount of time, a nominated contact will receive an emergency message with their location.

### Ethical considerations

After cognitive testing, if a participant is found to score within the cut-offs for suspected dementia or mild cognitive impairment, results will be sent to their general practitioner for follow-up with the participants’ written consent (gained at baseline). While in hospital, if a participant is discovered to have delirium, study staff will alert a nurse or doctor to the assessment findings.

### Dissemination

It is anticipated that the results of this study will inform multiple publications, and will be presented at national and international conferences.

## Ethics statements

### Ethics approval

Ethical approval has been confirmed by the Human Research Ethics Committees of the Royal Adelaide Hospital (HREC/17/RAH/445 and HREC/17/RAH/391), the Central Adelaide Local Health Network (R20171020 and 20170916) and the University of South Australia (0000034053 and ET00013).

## Acknowledgments

The authors would like to acknowledge the participants for their involvement in the study, and the research assistants involved in the ongoing data collection for this study.