Abstract
Background
Delirium in the ICU is a major and potentially modifiable risk factor for subsequent neurocognitive decline. However, the cerebral dysfunctions underlying its behavioural manifestations remain poorly understood, limiting targeted clinical management. These dysfunctions may reflect impairments in low-level perceptual processes (stimulus detection), higher-order integrative processes (stimulus discrimination), or both. The P3 event-related potential (ERP), a positive deflection in the electroencephalogram occurring approximately 300 ms after stimulus presentation, reflects neural processes involved in attention and cognitive integration and provides an objective measure of auditory detection (P3a) and discrimination (P3b).
Methods
In this prospective single-centre observational study conducted between March and October 2021 in the Critical Care Unit of Purpan University Hospital, Toulouse, France, we used a multidimensional P3 ERP battery incorporating auditory paradigms designed to assess stimulus detection (P3a) and discrimination (P3b). The battery included both non-verbal stimuli (local–global paradigm) and verbal stimuli (subject’s own name and arithmetic paradigms). Critically ill COVID-19 patients with delirium (n=18) and without delirium (n=20) were prospectively recruited. The primary outcome was the presence of P3 subcomponents across paradigms. Group differences were assessed using spatial–temporal cluster-based permutation testing, with statistical significance defined at cluster-level P≤0.05.
Results
In patients without delirium, both detection-related (P3a) and discrimination-related (P3b) responses were present across verbal and non-verbal paradigms. In contrast, patients with delirium demonstrated preserved P3a responses to non-verbal auditory stimuli but absence of P3b responses, indicating impaired auditory discrimination (cluster-level P≤0.05). Source modelling suggested reduced activation of frontal cortical generators in patients with delirium compared with those without delirium.
Conclusions
In critically ill patients with COVID-19, delirium was associated with preserved automatic auditory detection but impaired higher-order auditory discrimination. This dissociation provides neurophysiological evidence of selective impairment in integrative cognitive processing during ICU delirium and contributes to improved mechanistic understanding of delirium-related brain dysfunction.
Keywords: cognitive dysfunction, event-related potentials, ICU-related delirium, P3, P3a, P3b
Delirium is an acute confusional state with attentional and cognitive impairment.1 Converging evidence indicates that delirium is a strong predictor of long-term cognitive decline, which may persist for months or even years after an ICU stay.2 This highlights the need for a more comprehensive characterisation of the potentially debilitating set of cognitive dysfunctions associated with delirium, in particular to better understand their underlying cerebral mechanisms.
Theoretical models suggest that delirium is associated with increased distractibility and impaired orienting responses.3 However, attentional deficits may involve different levels of the cognitive system and distinct component processes. A recent study reported that critically ill COVID-19 patients with delirium exhibit language processing impairments, specifically marked by altered semantic and lexical priming effects.4 These findings suggest that delirium may involve disruptions in the integration of, or access to, linguistic information, in line with broader cognitive disturbances observed in this condition. Nevertheless, it remains unclear whether this phenomenon results from an impairment in automatic stimulus detection, reflecting low-level auditory processing,5 or deficits in stimulus integration and decision-making, which rely on higher-order cognitive mechanisms.6, 7, 8 Clarifying this distinction is essential to understanding the nature of cognitive alterations in delirium. To address this question, we used a hierarchical, multidimensional electrophysiological battery specifically designed to probe the detection and discrimination of salient non-verbal auditory stimuli (tones) and verbal stimuli, by measuring the P3 event-related potential (ERP) and its early and late subcomponents known as P3a and P3b, respectively.6 Indeed, given the reports of aberrant functional interactions between remote cortical regions,9,10 we formulated two explicit, theory-driven hypotheses grounded in contemporary models of attention and cognitive control: (1) ICU-related delirium would be associated with significant alterations in the P3b typically linked with higher-order cognitive processes that support the active discrimination and decision-making of auditory stimuli, whereas (2) the P3a component, which is typically associated with automatic attentional shifts toward novelty, might remain relatively preserved for simple non-verbal auditory stimuli. Ultimately, identifying these specific electrophysiological signatures may help guide personalised and targeted therapeutic interventions to reduce the long-term cognitive burden associated with ICU stays.
This observational study aimed to characterise how ICU-related delirium affects auditory detection and discrimination, with the presence of P3 subcomponents across paradigms serving as the primary outcome.
Methods
Study design and participants
As this was one of the first exploratory studies on P3 subcomponents in delirium, no a priori sample size calculation was performed. Data from critically ill patients were collected in this prospective clinical and EEG observational study from March 2021 to October 2021 in the Critical Care Unit of the Purpan University Hospital, Toulouse, France (Fig. 1).
Fig 1.
Study flowchart. CAM-ICU, Confusion Assessment Method for ICU; DEL–, patients without delirium; DEL+, patients with delirium; ERP, event-related potentials. ∗All patients between 18 and 80 yr of age who were COVID-19 positive and admitted to our ICU from March 2021 to October 2021 were considered for eligibility.
Inclusion criteria were adult critically ill patients aged between 18 and 80 yr with COVID-19. Exclusion criteria included a past medical history of psychiatric disorders, preexisting significant cognitive deficits (short Informant Questionnaire on Cognitive Decline in the Elderly [short IQCODE] ≥3.6), blindness, deafness, and non-French speakers.
Ethical approval and reporting standards
This minimal-risk observational study was approved by the Ethics Committee of the University Hospital of Toulouse, France, in January 2021 (Ref. RC 31/20/0441, NCT04785157). Patients’ families provided informed consent for their relative’s participation, and all investigations conformed to the Declaration of Helsinki and French regulatory requirements.
The study was conducted and reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement11 and the Committee on Best Practices in Data Analysis and Sharing for EEG/MEG (COBIDAS-MEEG) guidelines.12
Behavioural assessment
After complete discontinuation of sedative agents, patients were screened for delirium throughout their ICU stay using twice-daily assessments with the Confusion Assessment Method for the ICU (CAM-ICU).3 Based on this score, patients were classified as having (DEL+) or not having delirium (DEL–). In addition, patients’ level of arousal was assessed using the Richmond Agitation–Sedation Scale (RASS) to allow identification of hypoactive, hyperactive, or mixed delirium phenotypes. Delirium and coma duration, and the number of days with acute brain dysfunction (i.e. number of days spent in coma, delirium, or both), were also collected.
Auditory paradigms
A specifically designed auditory P3 ERP battery was used (Fig. 2; Supplementary material 1). This battery encompasses three auditory protocols: the local–global (Fig. 2a), the subject’s own name (SON, Fig. 2b) and the arithmetic (Fig. 2c) paradigms. It was previously validated in a group of 14 healthy volunteers (median age 27.5 yr [22–29]; Supplementary Fig. 1). The local–global paradigm, derived from the auditory oddball task, crosses two types of auditory regularities (intra-trial and inter-trial) of complex sinusoidal tones (chords) and tests neural responses to their respective violations (i.e. the ability to detect local violations or to discriminate global violations).13 The SON paradigm involves presenting sequences of equiprobable verbal stimuli, the subject’s own first name (SON), and other unfamiliar first names (OFN), and tests the ability to discriminate a self-referential stimulus.14 The arithmetic paradigm is based on simple additions (from 1+1 to 5+5) with unpredictable correct (50% of trials) or incorrect verbal results, and tests the ability to discriminate incorrect additions.15,16 All auditory stimuli were individually created using NaturalReader 14 (16 bits, 44 100 Hz), equalised to the same dB-A weighting level (around 65 dB-A) and presented binaurally.17 All stimuli were presented under passive listening conditions. The description of the P3 waves and their cognitive meaning in line with the literature6 are available in supplementary material (Supplementary Table 1).
Fig 2.
Multidimensional auditory battery. (a) Local–global paradigm (adapted from Bekinschtein and colleagues13). This paradigm crosses two types of auditory regularities to test neural responses to their respective violations. First is intra-trial local regularity. Each trial contains five sounds. Although the first four are always identical, the fifth can be identical (local standard, blue) or different (local deviant, orange). Second is inter-trial global regularity. These trials are included in blocks structured by an inter-trial global regularity: in half of the blocks, the global regularity fits with the local regularity (80% of trials have five identical sounds), whereas in the other half of blocks, global and local regularities are in opposition (80% of trials have a distinct fifth sound). Each sound lasted 50 ms. The inter-stimulus interval (ISI) between each sound within trial was fixed at 100 ms. The ISI between trials was variable and lasted from 1400 to 1600 ms. The duration of this auditory task was approximately 37 min. (b) Subject’s own name (SON) paradigm. For each subject, 12 sequences (S) of 42 names (one subject’s own name [SON, orange] + six other unfamiliar first names [OFN, blue]) were presented six times in a pseudo-random order. All first names were presented with the same probability (1/7 ≈ 14.3% for each name). The ISI ranged from 400 to 500 ms. Each sequence was spoken with one voice randomly selected from three. The duration of a sequence of 42 names was approximately 60 s. Each sequence was separated by 4 s. The duration of this experiment was approximately 12 min. (c) Arithmetic paradigm. For each subject, 100 arithmetic facts were presented auditorily in two distinct blocks (B) of 50 trials. Each trial consisted of simple addition (selected among 25 additions from 1+1 to 5+5). All the 25 additions were pseudo-randomly presented twice within each block, either followed by the arithmetic correct result (ACR, blue) (e.g. ‘2+3=5’, 50% of trials) or the arithmetic incorrect result (AIR, orange) (e.g. ‘5+1=4’, 50% of trials). The trial duration between the onset of the arithmetic fact and the end of the result was fixed to 3000 ms. The ISI between each trial was fixed to 2000 ms. The duration of the arithmetic paradigm was ∼9 min.
Electroencephalographic acquisition and preprocessing
EEG preprocessing and analysis were performed blinded to delirium status. All sedative medications were discontinued in advance of EEG recording (i.e. ≥24 h for propofol, ketamine, clonidine, or dexmedetomidine, and ≥72 h for benzodiazepines) such that all patients had a RASS score >–3 at the time of recording. High-density EEG data were recorded using a sampling rate of 500 Hz with a 128-electrode geodesic sensor net (EGI®; Philips, Amsterdam, Netherlands) referenced to the vertex (Cz). Data were first visually inspected to exclude bad channels using a zero-phase FIR filter with a 50 Hz notch, and then bandpass filtered between 0.5 and 25 Hz. Eye channels were recreated through subtraction of the channels above and below the eye, Cz was interpolated using the spherical spline method, and all channels were re-referenced according to an average reference. Independent component analysis (ICA) using fastICA was performed to automatically detect and remove blink-related components from the signal by capturing electrooculogram artifacts of all participants. Trials were subsequently segmented into epochs ranging from –200 to +1000 ms relative to stimulus onset for the SON and arithmetic paradigms (baseline correction –200 to 0 ms), and from –800 to +800 ms relative to the fifth sound for the local–global paradigm (baseline correction –800 to 0 ms). To further clean the data, an automatic rejection procedure was applied, whereby contaminated trials were excluded based on trial-wise assessment of individual sensor thresholds (epochs were rejected when peak-to-peak amplitudes exceeded 100 μV for EEG channels and 200 μV for electrooculography channels).18 Individual averages, and then grand average for each group, of responses to target and non-target stimuli were computed for each auditory paradigm: local deviant (LD) and local standard (LS), global deviant (GD) and global standard (GS), SON and OFN, arithmetic incorrect result (AIR) and arithmetic correct result (ACR). To achieve the same signal-to-noise ratio required for averages comparisons, the number of control and target stimuli analysed was identical. For illustrative purpose, the global field power (GFP) has been computed at the group level for target stimuli by averaging the individual GFP (Supplementary Fig. 2).19 All these analyses were performed using MNE-Python version 0.19.2 (https://mne.tools).
Cortical source reconstruction
The cortical generators of the grand-average responses to target stimuli in the DEL– and DEL+ groups were reconstructed using Brainstorm software.20 Source estimations were performed on the preprocessed evoked EEG data. The reconstruction was based on the MNI anatomical template ‘ICBM 152’. The OpenMEEG BEM method was used for computing the forward model. An identity matrix was used as the noise covariance. The minimum norm imaging method with the current density map measures and the constrained model were used for source reconstruction.21 The current density maps were then normalised according to a Z-score transformation of brain maps. For cortical sources visualisation, their reconstructions were time-averaged within the P3 dedicated significant time window centred on the maximum brain electrical spatial variability (i.e. maximum GFP) of each group in response to target stimuli. For the presentation of figures, source activity cutoff was set at 35% of the maximum range of the scale, with a minimum size of 15 vertices (Supplementary Figs 2 and 3).
Statistical analysis
Clinical data
The normality of the clinical data was tested using the Shapiro–Wilk test. Quantitative variables were expressed as median (25th–75th percentile) as appropriate. Qualitative variables were expressed as number (percentage). The comparison of continuous variables between the DEL– and DEL+ groups was performed using the Mann–Whitney U-test. Categorical variables were compared using the χ2 test. Stepwise logistic regression was performed in order to identify the explanatory variables associated with delirium. The variables included in the model were selected in accordance with our hypotheses and established factors known to influence the occurrence of delirium. Statistical analysis was performed using MedCalc software (version 12.6.1; MedCalc Software bvba, Ostend, Belgium; 2013). A P-value <0.05 was considered statistically significant.
Electrophysiological data
Our primary outcome was the presence of a differential P3 cerebral response, and secondary outcomes were their amplitudes and the associated cortical activation patterns.
Averages of the cerebral response were computed for each individual. Effects between target and non-target stimuli were tested at the group level: ‘local effect’ (LD vs LS), ‘global effect’ (GD vs GS), ‘SON effect’ (SON vs OFN), and ‘arithmetic effect’ (AIR vs ACR). To test the significance of the effect, we used cluster-based permutation tests (spatiotemporal clustering) with two-tailed t-tests and 10 000 permutations.22 The cluster-level alpha was set at 0.05, with a cluster forming threshold of 0.05. Cluster-based permutation approach is statistically appropriate and has been used in numerous studies to evaluate the contrast between two experimental conditions.23, 24, 25, 26
Maximum amplitudes (from baseline) of the P3 component were calculated on individual averaged traces in the dedicated statistically significant temporal window (defined on the DEL– grand-averaged ERPs).14
For each patient, we quantified the area of cortical source activation within the P3-specific time window across predefined frontal, temporal, and parietal regions based on Brainstorm parcellation with constrained scout size parameters.
Comparisons of amplitude and area between the DEL– and DEL+ groups were performed using the Mann–Whitney U-test.
Data availability
The anonymised raw data supporting the conclusions of this article will be made available by the authors without undue reservation. All codes are available upon request from the corresponding author.
Results
Clinical data
Eighty-nine critically ill patients with COVID-19 were prospectively assessed for eligibility, of whom 48 were enrolled in the study. Their characteristics are presented in Table 1.
Table 1.
Clinical and demographic characteristics. ADL, activities of daily living; DEL–, patients without delirium; DEL+, patients with delirium; IADL, instrumental activities of daily living; SAPS II, Simplified Acute Physiology Score II; SOFA, Sequential Organ Failure Assessment. Data are expressed as median (25th–75th percentile) and number (%), and compared between groups using the Mann–Whitney U-test and χ2 test.
| DEL– (n=28) | DEL+ (n=20) | P-value | |
|---|---|---|---|
| Age (yr) | 64.5 (56.5–71) | 66.5 (54–72.5) | 0.83 |
| Sex: male | 23 (82) | 19 (95) | 0.38 |
| BMI (kg m–2) | 27.7 (25.4–30.9) | 28.6 (26.1–30.8) | 0.65 |
| Short IQCODE | 3.3 (3.0–3.3) | 3.2 (3.1–3.4) | 0.15 |
| Past medical history | |||
| Neurological medical history | 4 (14.3) | 2 (10) | 1 |
| Stroke/epilepsy/aneurysm/migraine | 2/0/1/1 | 1/1/0/0 | |
| Anxious depression | 4 (14.3) | 3 (15) | 1 |
| Cardiovascular diseases | 17 (60.7) | 11 (55) | 0.77 |
| Respiratory disease | 7 (25) | 4 (20) | 0.74 |
| Chronic renal disease | 2 (7.1) | 1 (5) | 1 |
| Diabetes | 12 (42.9) | 11 (55) | 0.56 |
| Alcohol abuse | 0 (0) | 3 (15) | 0.07 |
| Charlson Comorbidity Index | 3 (2–4) | 3 (1–6) | 0.42 |
| ADL (/6) | 6 (6–6) | 6 (6–6) | 0.40 |
| IADL (/8) | 8 (8–8) | 8 (8–8) | 0.25 |
| ICU stay | |||
| SAPS II | 35 (30.5–42) | 40 (30.5–49.5) | 0.20 |
| SOFA | 4 (3.5–5) | 5 (4–8) | 0.13 |
| PaO2/FiO2 on admission (mm Hg) | 96 (83.5–127) | 80 (71.5–107.5) | 0.02 |
| Mechanical ventilation | |||
| Invasive/duration (days) | 9 (32)/17 (6.7–33.2) | 18 (90)/17 (8–25) | <0.001 |
| Noninvasive/duration (days) | 20 (71)/5 (4–9.5) | 18 (90)/4 (2–6) | 0.16 |
| High-flow oxygen therapy/duration (days) | 28 (100)/5 (3.5–8.5) | 19 (95)/4 (2–6) | 0.42 |
| Prone positioning (awake or sedated) | 19 (67.9) | 15 (75) | 0.75 |
| Sedative treatments | |||
| Midazolam | 8 (28.6) | 18 (90) | <0.001 |
| Duration (days) | 10.5 (4–23.5) | 12 (7–17) | |
| Propofol | 7 (25) | 16 (80) | <0.001 |
| Duration (days) | 11 (4.7–14.5) | 8.5 (3–13) | |
| Opioids | 9 (32.1) | 18 (90) | <0.001 |
| Duration (days) | 10 (5–25.7) | 16 (8–23) | |
| Ketamine | 3 (10.7) | 9 (45) | 0.016 |
| Duration (days) | 10 (4–19) | 5 (2–8.5) | |
| Alpha-2 agonists | 7 (25) | 18 (90) | <0.001 |
| Duration (days) | 4 (4–6.5) | 7.5 (4–12) | |
| Vasopressors | 8 (28.6) | 13 (65) | 0.019 |
| Steroids | 27 (96.4) | 18 (90) | 0.56 |
| Length of stay (days) | |||
| ICU | 8 (5–15) | 21.5 (13.5–39) | <0.001 |
| Hospital | 15 (11.5–30) | 35 (26.5–81.7) | 0.003 |
| Family visitation (in person or virtual) | 26 (92.9) | 17 (85) | 0.64 |
Use of midazolam during sedation was independently associated with delirium, with an odds ratio of 22.5 (95% confidence interval [CI] 4.2–120.2) and an area under ROC curve of 0.81 (95% CI 0.67–0.91) (p<0.001). The categorical variable ‘invasive mechanical ventilation’ during ICU stay was not retained in the model, nor were the continuous variables ‘SOFA score’ or ‘PaO2/FiO2’ on admission. Characteristics of brain dysfunction in the 20 patients with delirium are provided in Supplementary Table 2.
Electrophysiological data
Five patients were excluded for insufficient EEG quality and five did not complete the full battery (Fig. 1). The final cohort consisted of 38 patients (age 61 [54.3–72.3] yr, 35 (92%) male), of whom 18 experienced delirium. The duration of delirium was 8 (5–10.8) days.
Local effect
In both patient groups (DEL– and DEL+), the local deviance consistently elicited an early negative response corresponding to the typical mismatch negativity (MMN)27 followed by a positive P3a component extending from 150 to 310 ms after stimulus onset (Fig. 3a and b). This local effect was statistically significant in DEL– (p=0.001) and DEL+ (p=0.01) groups.
Fig 3.
Grand average of event-related potentials (ERPs). Grand average of ERPs obtained using the multidimensional cognitive electrophysiological battery in 20 patients without delirium (DEL–) and 18 patients with delirium (DEL+) in response to the local–global auditory task (a and b: local effect; c and d: global effect), subject’s own name (e and f: SON), and arithmetic (g and h) paradigms. Topographic maps of significant clusters of P3 effects after spatiotemporal cluster permutation testing are indicated by white electrodes on each averaged map within the timeframe of significance (a: 70–300 ms, b: 75–310 ms, c: 370–800 ms, e: 145–410 ms and 570–1000 ms, and g: 580–1000 ms). Significance threshold: ∗alpha-cluster ≤0.05, P≤0.05 for comparison between conditions. Colour coding: purple curves for target condition, blue curves for control condition. ACR, arithmetic correct result; AIR, arithmetic incorrect result; GD, global deviant sound; GS, global standard tone; LD, local deviant tone; LS, local standard sound; OFN, other unfamiliar first names; P3a: early component of the P3 ERP; P3b: late component of the P3 ERP; SON, subject’s own name.
Global effect
In patients without delirium, a centroparietal positive deflection to global deviance corresponding to a P3b component was observed from 370 to 800 ms after stimulus onset (P=0.049; Fig. 3c). In patients with delirium, no global effect was observed (P=0.32; Fig. 3d).
Subject’s own name effect
In patients without delirium, a centroparietal P3 component to the SON condition extending from 570 to 1000 ms after stimulus onset was identified. This resulted in a statistically significant SON effect (P=0.001; Fig. 3e). For patients with delirium, no SON effect was observed (P=0.27; Fig. 3f).
Arithmetic effect
In patients without delirium, a positive P3 component in response to AIR was observed from 580 to 1000 ms (Fig. 3g). This arithmetic effect, with a centroparietal distribution, was statistically significant (P=0.006). In patients with delirium, no arithmetic effect was detected (P=0.64; Fig. 3h).
Amplitudes of the P3 event-related potentials
The P3 amplitudes at group level in response to LD tones were not significantly different between DEL– and DEL+ groups (2 [1–2] vs 1 [0.4–1.6] μV, respectively; P=0.09). The P3 amplitudes in response to GD, SON, and AIR stimuli were higher in patients without delirium than in patients with delirium (GD: 2 [1–3] vs 1 [0.4–1.7] μV, P=0.03; SON: 3 [2–3.3] vs 2 [1.6–2.5] μV, P=0.03); and AIR: 2.3 [1.8–3] vs 2 [0–2.3] μV, P=0.04).
Cortical sources of the P3 event-related potentials
The frontal cortical generators of the P3 component in response to LD, SON, and AIR were statistically reduced in patients with delirium (Table 2 and Supplementary Figs 2 and 3 for illustrative purpose).
Table 2.
Frontal, temporal, and parietal activation areas of cortical generators. Frontal, temporal, and parietal activation areas of cortical generators in response to target stimuli within their P3 dedicated temporal window. Data are expressed as median (25th–75th percentile) and compared between groups with the Mann–Whitney U-test. AIR, arithmetic incorrect result; DEL–, patients without delirium; DEL+, patients with delirium; GD, global deviant; LD, local deviant; SON, subject’s own name.
| Activation areas of cortical generators in response to: | DEL– (n=20) | DEL+ (n=18) | P-value |
|---|---|---|---|
| LD tone | |||
|
15.8 (0.6–28.7) | 0 (0–2.3) | 0.02 |
|
4.3 (0–26.3) | 0 (0–2.6) | 0.2 |
|
0 (0–0) | 0 (0–0) | 0.93 |
| GD tone | |||
|
14.4 (2.6–67) | 14 (0–39.8) | 0.42 |
|
3.8 (0–16.2) | 0 (0–23.4) | 0.87 |
|
0 (0–6.5) | 0 (0–14.3) | 0.65 |
| SON | |||
|
16.2 (9.2–91.8) | 0 (0–20.1) | 0.04 |
|
2.6 (0–57.5) | 0 (0–5.6) | 0.19 |
|
3.7 (0–12.8) | 4.2 (2.4–11.9) | 0.66 |
| AIR | |||
|
13.6 (2.9–45) | 2.2 (0–18.1) | 0.03 |
|
3.1 (0–12.7) | 0 (0–8.8) | 0.42 |
|
0 (0–10.2) | 0 (0–6.5) | 0.78 |
Discussion
In this observational EEG study, we prospectively characterised the cerebral dysfunction underlying cognitive impairment in ICU-related delirium by probing neural processes supporting the detection and discrimination of auditory stimuli. Patients with delirium showed preserved P3a responses to local deviance but lacked P3b responses to global deviance, their own name, and incorrect arithmetic results.
Automatic detection of auditory stimuli is preserved in critically ill patients with delirium
The P3a response, signalling detection of local deviance in simple auditory stimuli during the local–global task, was observed in all patient groups. P3a is widely considered to be a correlate of detection processes that accompany the reorienting of involuntary attention having been triggered by processes indexed by the MMN.13,27, 28, 29 Accordingly, P3a is indicative of an automatic, pre-attentional, non-conscious, encapsulated mode of processing5 as it is largely resistant to attentional effects.13 Finally, our results suggest that automatic echoic memory is preserved in patients with delirium.
Discrimination of auditory stimuli associated with decision-making is impaired in critically ill patients with delirium
We identified that patients with delirium had a significantly impaired ability to discriminate violations of global auditory regularity of simple auditory stimuli. Previous studies have suggested that the global effect is associated with higher-order cortical processing, such as active discrimination and memory and context updating.13 The absence of global effect in patients with delirium could be a surrogate marker of voluntary attention deficit that could hamper the spotlight required for the active maintenance of previous stimuli in working memory required to detect long-time scale auditory irregularities.
The significantly impaired ability of patients with delirium to discriminate and solve simple arithmetic facts corroborates this hypothesis. The posterior P3b response to incorrect results—typically indexing the categorical decision between correct and incorrect solutions15—was absent in patients with delirium. This may reflect reduced attentional resources for working memory, impaired abstract number representation, or both. Indeed, each verbal term of the arithmetic problem requires an effortful recruitment of attentional resources to be integrated in time and projected into the future. Overall, frontal dysfunction, responsible for a reduction in information integration, could explain the comprehension and conceptual reasoning difficulties observed in patients with delirium.30
Interestingly, the discrimination of self-related, ecologically valid words was also impaired in critically ill patients with delirium. Strikingly, the brain responses evoked by hearing one’s own name (SON) in patients with delirium were similar to those elicited by hearing other first names (OFN). This suggests that the SON was processed similarly to an OFN, as if only low-level stimulus detection mechanisms remained accessible. Notably, the persistence of a SON effect has already been demonstrated in various states of reduced consciousness, such as sleep31 or coma,32 supporting the notion of a relatively automatic and preconscious processing of self-related words. In this context, further data will be needed to better characterise impairments in self-recognition and related difficulties in retrieving elements of autobiographical memory.
Results overview and clinical relevance of these findings
Integrative mechanisms required for higher-order cognitive processes of decision-making were selectively impaired in patients with delirium. Our results from cortical source reconstruction suggest that the cognitive processes involved in discriminating target stimuli may require global neural integration across distributed brain regions, with predominant engagement of the frontal cortex. Our results align with previous reports of large-scale cortical recruitment during the generation of late positive components.7,33 Specifically, the brain’s differential response to GD stimuli involved a broadly distributed network, including bilateral frontal, parietal, hippocampal, cingulate, and temporo-occipital regions, consistent with higher-order cognitive processes (see Supplementary Table 1 for key references). Delirium has been characterised as a disruption of interactions within functional brain networks between distant cortical modules (such as the dorsolateral prefrontal and posterior cingulate cortices that are known to play a role in self-processing).9,10 Thus, the mental state of delirium would be associated with global neural network changes rather than specific regional ones.34,35 From a cognitive theoretical perspective, we suggest that delirium is related to a deficit in the active mobilisation of attentional resources involving integrative mechanisms required to broadcast the initial information. From a therapeutic point of view, a neuropsychological approach considering cognitive remediation therapies encompassing multisensory stimulation could be of major interest by facilitating stimulus processing and compensating for attentional deficits. Moreover, the role of neuromodulation techniques in promoting brain communication in patients with delirium has not been studied, although these new therapeutic approaches seem to be valuable for patients with prolonged disorders of consciousness.36
Limitations
Our results must be interpreted with caution. Firstly, compared with patients without delirium, a greater proportion of patients with delirium (90%) required invasive mechanical ventilation with benzodiazepine-based sedation. This may represent a confounding factor or a causal contributor. Multivariate analysis identified benzodiazepine sedation as an independent factor significantly associated with delirium, consistent with prior reports.37,38 Secondly, we acknowledge that we cannot determine whether the observed electrophysiological signatures reflect a specific feature of COVID-19-associated delirium or a more general marker of ICU-related delirium. There is some evidence that the inflammatory process seen in SARS-CoV-2 infection could preferentially and directly target the frontal lobes.39,40 Future studies should focus on data analysis from ICU patients without SARS-CoV-2 infection to further investigate this issue. Thirdly, our research could have benefited from an additional active auditory paradigm (e.g. task of counting one’s own name) in order to definitely reveal the cognitive disability of patients with delirium to sustain volitional attention in response to specific instructions. Fourthly, the relatively small sample size may limit the generalisability of our findings and increases the risk of type II errors. However, the methodological framework and sample size are consistent with previous EEG-based exploratory studies conducted in similar clinical contexts, such as in disorders of consciousness, where recruitment and data acquisition are particularly challenging, and sample sizes ranging from 10 to 20 patients are commonly accepted.14,23,26 We acknowledge the need for replication in larger cohorts to confirm these findings and enhance external validity. Finally, no statement can be made regarding residual cognitive function at the individual level. Actually, negative results at the individual level are difficult to interpret as they may be linked to the known lack of sensitivity of electrophysiological analyses. However, brain signatures extracted from group level analyses allow the characterisation of dedicated cognitive dysfunctions.
Conclusions and future directions
We report, in critically ill COVID-19 patients with delirium, preservation of low-level cognitive processing of automatic detection and a higher-order cognitive dysfunction of discrimination of auditory stimuli. This selective impairment may relate to a deficit in the active mobilisation of attentional resources, particularly involving frontal cortical areas. We suggest that our findings pave the way for future research focusing on the long-term neurocognitive impact of discrimination deficits.
Authors’ contributions
Conceptualisation: FF, WB, SS, FP, AC
Methodology: FF, WB, SS, FP, AC
Software: AC (development of stimulation software and electrophysiological analysis scripts)
Investigation (data acquisition): FF, WB, BS
Formal analysis (statistical analyses): FF, LH
Resources/clinical management of patients: BR, SS
Writing – original draft: FF, SS, FP
Writing – review and editing: all authors
Supervision: SS, FP
Approval of the final manuscript: all authors
Funding
Partially supported by the French National Research Agency (ANR) (‘Research and Action on Coronavirus Disease 2019’ grant to SS).
Declaration of interests
The authors report no competing interests.
Acknowledgements
We thank the medical and nursing teams from the ICU at Purpan Hospital, Toulouse, for their support throughout data acquisition. We would like to thank all the patients and their relatives involved in this study.
Handling Editor: Joana Berger
Footnotes
Supplementary data to this article can be found online at https://doi.org/10.1016/j.bjao.2025.100517.
Appendix A. Supplementary data
The following is the Supplementary data to this article:
References
- 1.Pandharipande P.P., Ely E.W., Arora R.C., et al. The intensive care delirium research agenda: a multinational, interprofessional perspective. Intensive Care Med. 2017;43:1329–1339. doi: 10.1007/s00134-017-4860-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2.Pandharipande P.P., Girard T.D., Jackson J.C., et al. Long-term cognitive impairment after critical illness. N Engl J Med. 2013;369:1306–1316. doi: 10.1056/NEJMoa1301372. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Ely E.W., Inouye S.K., Bernard G.R., et al. Delirium in mechanically ventilated patients: validity and reliability of the confusion assessment method for the intensive care unit (CAM-ICU) JAMA. 2001;286:2703–2710. doi: 10.1001/jama.286.21.2703. [DOI] [PubMed] [Google Scholar]
- 4.Ferre F., Buffieres W., Heine L., et al. Impairment of central language processing in critically ill coronavirus disease 2019 patients with delirium. Brain Commun. 2023;5 doi: 10.1093/braincomms/fcad073. fcad073. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Naatanen R., Tervaniemi M., Sussman E., Paavilainen P., Winkler I. "Primitive intelligence" in the auditory cortex. Trends Neurosci. 2001;24:283–288. doi: 10.1016/s0166-2236(00)01790-2. [DOI] [PubMed] [Google Scholar]
- 6.Polich J. Updating P300: an integrative theory of P3a and P3b. Clin Neurophysiol. 2007;118:2128–2148. doi: 10.1016/j.clinph.2007.04.019. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Rohaut B., Naccache L. Disentangling conscious from unconscious cognitive processing with event-related EEG potentials. Rev Neurol (Paris) 2017;173:521–528. doi: 10.1016/j.neurol.2017.08.001. [DOI] [PubMed] [Google Scholar]
- 8.Verleger R. Effects of relevance and response frequency on P3b amplitudes: review of findings and comparison of hypotheses about the process reflected by P3b. Psychophysiology. 2020;57 doi: 10.1111/psyp.13542. [DOI] [PubMed] [Google Scholar]
- 9.van Montfort S.J.T., van Dellen E., van den Bosch A.M.R., et al. Resting-state fMRI reveals network disintegration during delirium. Neuroimage Clin. 2018;20:35–41. doi: 10.1016/j.nicl.2018.06.024. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Oh J., Shin J.E., Yang K.H., et al. Cortical and subcortical changes in resting-state functional connectivity before and during an episode of postoperative delirium. Aust N Z J Psychiatry. 2019;53:794–806. doi: 10.1177/0004867419848826. [DOI] [PubMed] [Google Scholar]
- 11.von Elm E., Altman D.G., Egger M., et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. PLoS Med. 2007;4 doi: 10.1371/journal.pmed.0040296. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Pernet C., Garrido M.I., Gramfort A., et al. Issues and recommendations from the OHBM COBIDAS MEEG committee for reproducible EEG and MEG research. Nat Neurosci. 2020;23:1473–1483. doi: 10.1038/s41593-020-00709-0. [DOI] [PubMed] [Google Scholar]
- 13.Bekinschtein T.A., Dehaene S., Rohaut B., Tadel F., Cohen L., Naccache L. Neural signature of the conscious processing of auditory regularities. Proc Natl Acad Sci U S A. 2009;106:1672–1677. doi: 10.1073/pnas.0809667106. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Perrin F., Schnakers C., Schabus M., et al. Brain response to one's own name in vegetative state, minimally conscious state, and locked-in syndrome. Arch Neurol. 2006;63:562–569. doi: 10.1001/archneur.63.4.562. [DOI] [PubMed] [Google Scholar]
- 15.Dickson D.S., Wicha N.Y.Y. P300 amplitude and latency reflect arithmetic skill: An ERP study of the problem size effect. Biol Psychol. 2019;148 doi: 10.1016/j.biopsycho.2019.107745. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Nunez-Pena M.I., Honrubia-Serrano M.L. P600 related to rule violation in an arithmetic task. Brain Res Cogn Brain Res. 2004;18:130–141. doi: 10.1016/j.cogbrainres.2003.09.010. [DOI] [PubMed] [Google Scholar]
- 17.Corneyllie A., Perrin F., Heine L. Sharing an open stimulation system for auditory EEG experiments using Python, Raspberry Pi, and HifiBerry. eNeuro. 2021;8 doi: 10.1523/ENEURO.0524-20.2021. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Jas M., Larson E., Engemann D.A., et al. A reproducible MEG/EEG group study with the MNE software: recommendations, quality assessments, and good practices. Front Neurosci. 2018;12:530. doi: 10.3389/fnins.2018.00530. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Lehmann D., Skrandies W. Reference-free identification of components of checkerboard-evoked multichannel potential fields. Electroencephalogr Clin Neurophysiol. 1980;48:609–621. doi: 10.1016/0013-4694(80)90419-8. [DOI] [PubMed] [Google Scholar]
- 20.Tadel F., Baillet S., Mosher J.C., Pantazis D., Leahy R.M. Brainstorm: a user-friendly application for MEG/EEG analysis. Comput Intell Neurosci. 2011;2011 doi: 10.1155/2011/879716. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21.Dale A.M., Liu A.K., Fischl B.R., et al. Dynamic statistical parametric mapping: combining fMRI and MEG for high-resolution imaging of cortical activity. Neuron. 2000;26:55–67. doi: 10.1016/s0896-6273(00)81138-1. [DOI] [PubMed] [Google Scholar]
- 22.Maris E., Oostenveld R. Nonparametric statistical testing of EEG- and MEG-data. J Neurosci Methods. 2007;164:177–190. doi: 10.1016/j.jneumeth.2007.03.024. [DOI] [PubMed] [Google Scholar]
- 23.Beukema S., Gonzalez-Lara L.E., Finoia P., et al. A hierarchy of event-related potential markers of auditory processing in disorders of consciousness. Neuroimage Clin. 2016;12:359–371. doi: 10.1016/j.nicl.2016.08.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Cruse D., Beukema S., Chennu S., Malins J.G., Owen A.M., McRae K. The reliability of the N400 in single subjects: implications for patients with disorders of consciousness. Neuroimage Clin. 2014;4:788–799. doi: 10.1016/j.nicl.2014.05.001. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Heine L., Corneyllie A., Gobert F., Luaute J., Lavandier M., Perrin F. Virtually spatialized sounds enhance auditory processing in healthy participants and patients with a disorder of consciousness. Sci Rep. 2021;11 doi: 10.1038/s41598-021-93151-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Sergent C., Faugeras F., Rohaut B., et al. Multidimensional cognitive evaluation of patients with disorders of consciousness using EEG: a proof of concept study. Neuroimage Clin. 2017;13:455–469. doi: 10.1016/j.nicl.2016.12.004. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Naatanen R., Paavilainen P., Alho K., Reinikainen K., Sams M. The mismatch negativity to intensity changes in an auditory stimulus sequence. Electroencephalogr Clin Neurophysiol Suppl. 1987;40:125–131. [PubMed] [Google Scholar]
- 28.Sutton S., Braren M., Zubin J., John E.R. Evoked-potential correlates of stimulus uncertainty. Science. 1965;150:1187–1188. doi: 10.1126/science.150.3700.1187. [DOI] [PubMed] [Google Scholar]
- 29.Squires N.K., Squires K.C., Hillyard S.A. Two varieties of long-latency positive waves evoked by unpredictable auditory stimuli in man. Electroencephalogr Clin Neurophysiol. 1975;38:387–401. doi: 10.1016/0013-4694(75)90263-1. [DOI] [PubMed] [Google Scholar]
- 30.Hart R.P., Levenson J.L., Sessler C.N., Best A.M., Schwartz S.M., Rutherford L.E. Validation of a cognitive test for delirium in medical ICU patients. Psychosomatics. 1996;37:533–546. doi: 10.1016/S0033-3182(96)71517-7. [DOI] [PubMed] [Google Scholar]
- 31.Perrin F., Garcia-Larrea L., Mauguiere F., Bastuji H. A differential brain response to the subject's own name persists during sleep. Clin Neurophysiol. 1999;110:2153–2164. doi: 10.1016/s1388-2457(99)00177-7. [DOI] [PubMed] [Google Scholar]
- 32.Ferre F., Heine L., Naboulsi E., et al. Self-processing in coma, unresponsive wakefulness syndrome and minimally conscious state. Front Hum Neurosci. 2023;17 doi: 10.3389/fnhum.2023.1145253. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Sergent C., Baillet S., Dehaene S. Timing of the brain events underlying access to consciousness during the attentional blink. Nat Neurosci. 2005;8:1391–1400. doi: 10.1038/nn1549. [DOI] [PubMed] [Google Scholar]
- 34.Numan T., Slooter A.J.C., van der Kooi A.W., et al. Functional connectivity and network analysis during hypoactive delirium and recovery from anesthesia. Clin Neurophysiol. 2017;128:914–924. doi: 10.1016/j.clinph.2017.02.022. [DOI] [PubMed] [Google Scholar]
- 35.van Dellen E., van der Kooi A.W., Numan T., et al. Decreased functional connectivity and disturbed directionality of information flow in the electroencephalography of intensive care unit patients with delirium after cardiac surgery. Anesthesiology. 2014;121:328–335. doi: 10.1097/ALN.0000000000000329. [DOI] [PubMed] [Google Scholar]
- 36.Thibaut A., Schiff N., Giacino J., Laureys S., Gosseries O. Therapeutic interventions in patients with prolonged disorders of consciousness. Lancet Neurol. 2019;18:600–614. doi: 10.1016/S1474-4422(19)30031-6. [DOI] [PubMed] [Google Scholar]
- 37.Hayhurst C.J., Pandharipande P.P., Hughes C.G. Intensive care unit delirium: a review of diagnosis, prevention, and treatment. Anesthesiology. 2016;125:1229–1241. doi: 10.1097/ALN.0000000000001378. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 38.Pun B.T., Badenes R., Heras La Calle G., et al. Prevalence and risk factors for delirium in critically ill patients with COVID-19 (COVID-D): a multicentre cohort study. Lancet Respir Med. 2021;9:239–250. doi: 10.1016/S2213-2600(20)30552-X. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 39.Cani I., Barone V., D'Angelo R., et al. Frontal encephalopathy related to hyperinflammation in COVID-19. J Neurol. 2021;268:16–19. doi: 10.1007/s00415-020-10057-5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 40.Toniolo S., Di Lorenzo F., Scarioni M., Frederiksen K.S., Nobili F. Is the frontal lobe the primary target of SARS-CoV-2? J Alzheimers Dis. 2021;81:75–81. doi: 10.3233/JAD-210008. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The anonymised raw data supporting the conclusions of this article will be made available by the authors without undue reservation. All codes are available upon request from the corresponding author.