Delirium is an acute clinical condition characterized by a wide range of symptoms affecting multiple domains. Patients with delirium often present with attention deficits, confusion, hallucinations, agitation, or disturbances in sleep-wake cycle. Typically, delirium occurs during critical illness, such as after surgery or stroke, with rapid onset and fluctuating symptoms [1]. Poststroke delirium (PSD) is one of the most common complications after stroke, with reported prevalence rates reaching up to 40% in older patients [1-3]. Both delirium in general and PSD in particular are associated with longer hospitalization, higher mortality, accelerated cognitive decline, and lower quality of life [1,4-11]. There is a broad consensus of a multifactorial origin [1,12,13], with predisposing factors such as dementia, age, comorbidities, and precipitating risk factors, including infection, stroke, or surgery [1,12,13]. The presence of multiple and fluctuating symptoms is diagnostically challenging, especially in patients with stroke with communication difficulties (eg, aphasia), and requires clinical expertise and standardized screening tools, such as the well-established ICDSC (Intensive Care Delirium Screening Checklist) [14,15]. Despite the high prevalence and the nondeclining mortality rate over the past decades [2,7], objective diagnostic tools and prognostic biomarkers (ie, fluid or proteomic biomarkers), as well as protective markers for long-term cognitive outcome, have not yet been implemented in clinical practice and remain the focus of ongoing studies [5,16-18].

In general, examining delirium patients is a challenging field, which limits the number of available studies. Patients with delirium, such as patients with PSD, represent a vulnerable and critically ill population. Especially during the acute (within 5 d poststroke) and the postacute phase (5 d to 12 wk poststroke), patients may experience secondary deteriorations or be scheduled for invasive diagnostic procedures or surgeries, making the conduction of studies difficult. Additionally, the fluctuating nature of delirium symptoms complicates the timing and consistency of study assessments. Furthermore, during the acute state of delirium, especially cases involving high arousal or hallucinating symptoms, cognitive assessment or magnetic resonance imaging (MRI) assessment becomes impossible, as patients cannot comply with task instructions. This results in a lack of studies and underscores the need for further research in this area. Another challenge in this field concerns the study design, varying with regard to concurrent versus sequential study designs, as well as the choice of an appropriate control group.

Previous studies have reported a higher risk of PSD in total anterior circulation infarcts compared to partial anterior, posterior, and lacunar infarcts [19-21]. Cortical networks in the anterior circulation, such as fronto-parietal networks, are known to be involved in higher-level cognitive functions, including executive functions, attention, and memory. These very cognitive domains are affected in PSD. Memory functions, which represent a critical component of human cognition, are frequently affected and represent a highly disabling cognitive domain in stroke survivors [22]. For example, memory impairments have been significantly associated with poor quality of life of patients with stroke and their caregivers, for example, [22,23], demonstrating the high relevance to daily life and potential burden. Moreover, memory assessment is a well-established component of both neuropsychological testing and in functional magnetic resonance imaging (fMRI) studies, making it an ideal focus for our study.

In the challenging field of cognitive assessment in patients with delirium, the majority of studies focus on behavioral measures including the assessment of multiple cognitive domains such as attention, language, memory, executive functions, and visuospatial construction. Previous studies have demonstrated a significantly poorer cognitive outcome (eg, global cognition and executive functions) in patients with postdelirium compared to healthy controls [24-26].

Additionally, advanced techniques such as MRI have been implemented to provide further insights. Structural MRI studies, which do not require a certain level of patients’ compliance, have reported abnormalities in white matter tracts of patients with postdelirium, which are known to be involved in memory and attention [1,13,27]. Since delirium is known to be a disorder of disrupted brain networks [1,12,16,28,29], fMRI studies, which were predominantly conducted as resting-state studies in patients with postdelirium, have demonstrated aberrant connectivity in dorsal prefrontal and subcortical regions compared to healthy controls, which may contribute to the emergence of delirium symptoms [13,28,30]. These findings in network dysfunctions are restricted to the default mode network, and memory functions may not be detected during resting-state fMRI [31]. An essential requirement for studying memory functions and related activity or connectivity dysfunctions in delirium is a memory task-based fMRI study. Unfortunately, these are exceedingly limited.

To date, only 1 fMRI study has been conducted, using a well-known N-Back working memory task [32]. While reporting no association between delirium and brain region activity, several methodological limitations and concerns have been raised, such as the infeasibility of the working memory task (especially at the time of discharge), the infeasible MRI protocol, as well as the absence of a control group. Of note, connectivity analyses have not been conducted. Further research is therefore needed, alongside the development of an adequate and reliable task for working memory assessment during fMRI.

An additional challenge in studying PSD and cognitive outcomes is the unresolved distinction between PSD-related and pure stroke−related cognitive impairments. In detail, cognitive deficits associated with PSD and other types of delirium, for example, attention or memory deficits, have also been reported in patients with stroke without PSD [23]. Previous imaging studies have shown reduced brain activation as well as abnormal connectivity within fronto-parietal networks in patients without PSD compared to healthy controls [31,33,34]. Thus, the origin of cognitive impairments in patients with PSD (ie, distinction between “stroke-related” vs “PSD-related” impairments) remains unclear.

To summarize, there is a lack of reliable and feasible fMRI tasks for assessing memory functions in patients with PSD. The few existing task-based fMRI studies have either been conducted without control groups or have only compared groups of patients with PSD to healthy controls, limiting the ability to disentangle the distinct cognitive and neural contributions of PSD versus stroke. Therefore, there is a significant gap in the literature regarding appropriate methods to study PSD and its cognitive outcomes, emphasizing the need for comprehensive, longitudinal investigations that combine neuropsychological assessments with functional neuroimaging.

To address this gap, we present a protocol for a comprehensive longitudinal proof-of-concept study that aims to systematically and comprehensively examine poststroke cognitive impairment (PSCI) in a critically ill stroke population, both with and without PSD (control group). Assessments will be conducted during the acute (within 5 d post stroke) and postacute phases (5‐14 d post stroke), and at a 3-month follow-up using a multicomponent neuropsychological assessment and task-based fMRI. Generating reliable feasibility and preliminary data through this approach will provide essential groundwork for future multicenter studies.

The specific objectives of this study are to (1) compare the global PSCI of patients with and without PSD at a 3-month follow-up, (2) compare domain-specific cognitive outcome measures of patients with and without PSD at a 3-month follow-up, (3) generate preliminary data regarding the differentiation between PSD-related and stroke-related cognitive impairment using behavioral and fMRI connectivity measures, and (4) contribute toward developing protective and prognostic tools for the early detection and treatment of PSD.

This study was funded in December 2023. The recruitment of patients was initiated in June 2024 and will continue until September 2025. As of May 2025, 15 patients with stroke have been included. Data acquisition will continue until December 2025. Data analysis will commence at the beginning of 2026, and study findings are expected to be published in summer 2026.

Here, we present the protocol of a longitudinal study (RECOVER) designed to investigate both short-term and long-term effects of PSD on cognitive impairment and recovery in patients with stroke. PSCI is assessed using a multimodal approach that includes a multicomponent digital neuropsychological assessment and task-based fMRI at 3 time points: during the acute (prior PSD development) phase, the postacute phase, and at the 3-month follow-up.

Prior behavioral studies showed significantly poorer cognitive outcome (eg, in global cognition or memory decline) in both patients with stroke with PSD and without PSD compared to healthy controls [1,24,33]. Thus, a clear distinction between “stroke-related” and “PSD-related” cognitive impairments has not been established so far.

Brain neuroimaging studies using structural MRI reported dysconnectivity within the cortical-subcortical systems, for example, reticulo-thalamo-cortical network [1,13,79] in patients with delirium. fMRI studies using resting state have further revealed disrupted connectivity in dorsal-prefrontal and subcortical networks in patients with delirium (compared to healthy controls). These alterations may be linked to core delirium symptoms, such as disorientation, altered arousal, attention deficits, or disturbances in the sleep-wake cycle [1,13,16,28,30,80,81].

Memory is a fundamental aspect of human cognition and a frequently affected domain in patients with stroke with PSD and without PSD [22]. However, memory impairments may not be detectable during resting-state fMRI [31]. Unfortunately, task-based fMRI studies that allow for the direct assessment of memory functions are currently very limited. To date, only 1 fMRI study has investigated this challenging area, reporting no significant association between delirium and brain region activity [32]. However, this study raises methodological concerns due to the absence of a control group and the use of an infeasible working memory task. Further research is therefore needed, alongside the development of an adequate and reliable assessment of working memory during fMRI.

Finally, there remains a notable gap in the literature regarding appropriate methods for studying PSD and related cognitive outcomes. This gap has been addressed by this longitudinal study protocol, which includes 3 measurement points and 2 groups of patients with stroke, PSD, and non-PSD (control), to assess PSCI. We followed a multicomponent assessment strategy for the neuropsychological assessment and introduced a suitable digital tablet app tailored for both patients with stroke with PSD and without PSD. This partly tablet−based neuropsychological assessment enables the generation of additional neuropsychological parameters, such as time in air or time on surface [42,74], which may help to detect patients with PSD at an early stage.

In addition, we present an extended (ie, verbal and visual items) and modified version of a well-known working memory task (Sternberg task) [58-61] to examine patients with stroke with PSD and without PSD. This study design enables the direct assessment of memory functions at both the behavior level via neuropsychological testing and advanced parameters derived from digitalized tests, and at the brain level via task-based fMRI, where patients actively perform the working memory task during scanning. Therefore, we can map the neural correlates of working memory functions and conduct brain-behavior correlation analyses. Additionally, this comprehensive longitudinal design, involving 2 patient groups (PSD and non-PSD), may allow us to differentiate between stroke-related and PSD-related cognitive impairments. Moreover, we aim to contribute preliminary insights toward developing protective and prognostic tools that support the early detection and treatment of PSD, ultimately helping to improve long-term cognitive recovery in patients with stroke.

The findings and experience gained from this proof-of-concept protocol will serve as early, preliminary investigation and provide essential groundwork for the design and implementation of future multicenter studies with a larger sample size. This protocol represents a first step toward expanding previous knowledge and improving the understanding of the contribution of PSD to long-term cognitive impairment after stroke.

These early findings from the proof-of-concept study hold promising clinical implications for the diagnosis and treatment of PSD, with further research required to validate these results on a larger scale. A more profound understanding of the cognitive dysfunctions associated with PSD could significantly influence clinical decision-making and rehabilitation strategies. Recognizing PSD as a distinct clinical condition characterized by specific patterns of cognitive impairment could help clinicians differentiate it from typical poststroke cognitive dysfunction. Specifically, the early identification of PSD-related cognitive deficits may enable the timely initiation of targeted, high-intensity cognitive rehabilitation, including domain-specific interventions such as those focused on attention, memory, and executive function. This approach has the potential to guide the development of individualized care plans tailored to patients’ needs, prevent long-term cognitive decline, and promote optimal cognitive recovery.

However, this study protocol has several limitations. First, our study population consisted of patients with stroke with mild-to-moderate symptoms (NIHSS ≤15) and excluded patients with stroke with more severe stroke symptoms, particularly those with a GCS score ≤8, as well as patients with aphasia, which limits the generalizability of our findings to the broader population with stroke. However, patients with global aphasia are often unable to comprehend test instructions or follow simple requests (eg, “close your eyes”), making the assessment of cognitive functions, such as memory functions, extremely challenging, time-consuming, and often frustrating for the patient. Patients with a GCS score ≤ 8 have typically significantly reduced levels of consciousness, making a reliable cognitive assessment not possible. Second, the sample size is 40. Although this proof-of-concept study is underpowered, it is essential for gathering preliminary data and assessing feasibility. Insights gained will provide critical groundwork for designing larger, adequately powered studies in the future. Another important limitation of our study protocol is the variable timing of assessments, especially for S2, which is scheduled within a broad window of 5-14 days post stroke. These flexible time windows were intentionally selected to accommodate clinical realities and to minimize dropout rates and missing data. Regarding S2, this variability is driven by clinical realities, including patient transfers on short notice, fluctuating medical conditions, limited MRI availability, and the unpredictable course of delirium recovery. Unfortunately, the MRI assessment cannot take place during the acute phase of a delirium but only after partial recovery from delirium, a common problem in behavioral and neuroimaging studies on delirium or PSD. While this flexible scheduling approach enhances the feasibility of conducting comprehensive neuropsychological and imaging assessments in a population with acute stroke, it introduces potential confounding effects related to differences in the timing of measurements across patients. To mitigate this issue, we will incorporate the number of days post stroke as a covariate in our statistical analyses. Future large-scale studies should consider using more standardized normalization strategies to handle timing variability at S2, such as time normalization or statistical correction for nonlinear recovery effects in mixed effects models.

Another concern in our multimodal longitudinal study is the increased risk of participant dropout compared to cross-sectional studies. This risk is particularly pronounced among patients with PSD, given the association between PSD, lower quality of life, and higher mortality rates, which may lead to a higher than anticipated dropout rate. To address this, we have implemented several mitigation strategies. First, as described above, we have established predefined time windows to accommodate clinical realities within this vulnerable study population while prioritizing patients’ well-being. Assessments are adjusted within predefined time windows based on the patient’s condition, balancing protocol adherence with clinical care and comfort. Second, both neuropsychological and MRI assessments include opportunities for breaks or pauses to manage patient fatigue or distress. Third, when the patient’s condition or transfer prevents in-hospital assessment, we offer follow-up visits at patients’ homes or rehabilitation centers. Collectively, these measures enhance feasibility and data completeness while ensuring patient safety and comfort.

This study protocol involves important ethical considerations due to the vulnerability of patients with acute stroke, particularly those experiencing PSD. Ensuring ethical responsibility and patient safety is of the utmost importance. In an effort to minimize the burden on participants, both neuropsychological testing and MRI sessions are kept as brief as possible and can be paused or rescheduled based on the patient’s condition. Testing is immediately halted if any signs of distress or fatigue arise. Throughout the entirety of the procedures, the presence and attentiveness of clinical staff are maintained. The inclusion of cognitively vulnerable patients is ethically justified, as this group is often underrepresented in research. However, their participation is carefully balanced with stringent safeguards to prevent any undue burden on their health or recovery. The presented protocol is critical to provide a comprehensive framework for assessing cognitive outcomes tailored to the challenges of evaluating older patients with stroke with and without PSD.

While this was beyond the scope of this proof-of-concept study, future studies should be initiated as multicenter trials to enhance the generalizability and robustness of the findings. Increasing the sample size will be essential to achieve adequate group sizes and improve statistical power. Additionally, balancing groups according to key variables, such as lesion location or stroke severity, would strengthen the validity of comparisons.

Another important aspect is the consideration of the amount of cognitive rehabilitation on PSCI. Unfortunately, this was beyond the scope of this study, and this protocol includes only documentation of whether rehabilitation was received and its duration. A detailed evaluation of the effect of cognitive rehabilitation impact on long-term outcomes should be addressed in future studies. Another consideration would be extended follow-ups, for example, after 6 and 12 months, to better capture cognitive trajectories.

This protocol represents a proof-of-concept study designed to systematically and comprehensively examine cognitive functions in critically ill patients with acute stroke with and without delirium. It combines comprehensive neuropsychological testing with a multimodal MRI protocol, including digitalized assessments and task-based functional MRI. By addressing a significant clinical gap and using innovative methodologies, this study lays essential groundwork toward larger, adequately powered studies.