Mechanical thrombectomy (MT) is a highly effective treatment for major ischemic stroke due to large vessel occlusion (LVO) and is now the standard of care for acute stroke treatment. However, patients who have suffered a major stroke are a group at high risk for developing complications during hospital stay (Shaw et al., 2019). A common complication in acute stroke patients is delirium, a syndrome characterized by an altered level of consciousness, disorganised thoughts or disorientation, psychomotor disturbances and circadian dysrhythmia, typically with acute onset and fluctuations over time. Delirium in stroke patients has been reported to be associated with worse outcomes (Shi et al., 2012).

Since treatment options are limited, delirium prevention is the goal for hospitalized patients. While interdisciplinary multicomponent interventions have proven effective for delirium prevention in the hospital setting (Burton et al., 2021), identification of individuals at risk of developing delirium may help to design more efficient and targeted preventive measures and enable timely diagnosis of delirium. Etiology of delirium is multifactorial and reports on risk factors vary throughout the literature with sometimes conflicting conclusions. Patient characteristics, including older age and premorbid cognitive decline; stroke characteristics, e.g., stroke location and severity; and further factors, such as infections, have all been reported to be associated with development of delirium in stroke patients (Rhee et al., 2022; Oldenbeuving et al., 2011; Hahn et al., 2023). Although clinical screening instruments for delirium in hospitalized patients have been developed, questions have been raised about their validity in patients with neurological deficits (Vater et al., 2024). Therefore, identification of risk factors and development of predictive tools may help to prevent and diagnose delirium in patients hospitalized due to acute stroke. The few predictive instruments that have been specifically developed for stroke patients largely remain to be externally validated (Drozdowska et al., 2021). Moreover, existing studies originate from mixed stroke cohorts, which predominantly consist of non-major stroke patients and include also patients with hemorrhagic stroke (Oldenbeuving et al., 2014). However, patients with LVO treated by MT constitute a distinct subgroup of stroke patients; data on delirium in these patients is sparse. These patients are poorly represented in published analyses of delirium, as mixed stroke cohorts contain few cases of MT or analyses were conducted before the era of MT as standard therapy for LVO (Shaw et al., 2019).

We investigate the association of patient-, stroke- and treatment characteristics with the development of delirium in a cohort of patients with LVO who were treated by MT. Additionally, we explore the role of easily obtainable imaging biomarkers of cerebral atrophy and white matter lesions (WML) arising from native computed tomography (CT) as a potential independent risk factor for developing delirium. Finally, we analyze whether delirium is an independent predictor for functional outcome three months after MT of LVO.

We here show that delirium is a complication associated with worse functional outcome in patients with LVO who were treated by MT and present data on conventional and novel predictive factors for development of delirium in these patients, which help to identify patients at risk and implement specific preventive measures to improve clinical outcome.

With regard to functional outcome, our results are in line with previous analyses of delirium in patients with acute stroke. These reported worse functional outcome in mixed stroke cohorts, mostly focusing on mortality and functional dependence at hospital dicharge (Shi et al., 2012; Rollo et al., 2022; Pasińska et al., 2019). Patients with LVO treated by MT were largely underrepresented in former studies of delirium in acute stroke and yet, depict a relevant share of the disease. We expand former findings by showing that also in stroke patients with LVO treated by MT delirium is associated with worse functional outcome at discharge. This finding persists three months after stroke and, notably, is independent of stroke severity. We consider this remarkable, especially since functional outcome in patients with major ischemic stroke due to LVO is mainly driven by severity of the disease. In contrast to former studies of delirium in stroke patients, we did not observe increased mortality in patients with LVO who developed delirium, which is likely due to high overall mortality rates attributable to other complications and severity of the disease. Increased hospital stay duration has been reported in patients with delirium in mixed stroke cohorts (Shi et al., 2012). This was also true for our patients treated by MT. This is notable, because in this cohort, hospital discharge is often determined by stroke-associated neurological deficits affecting independence in daily activities and self-care.

Our analyses identify patient-, stroke- and treatment characteristics that are associated with development of delirium in patients treated by MT. These have direct clinical implications by enabling identification of patients at risk for delirium to target preventive measures or contribute to early diagnosis. As expected, older age and infectious complications were independently associated with delirium following MT. These are known risk factors for delirium in hospitalized patients and, as shown previously, in patients with acute stroke (Oldenbeuving et al., 2011). We observed male sex to be independently associated with development of delirium following MT. This is especially interesting and merits further investigation, since prospective validated prediction models do not describe sex as a risk factor for delirium (Inouye et al., 2014; Zaal et al., 2015). Potentially, a difference in motor subtypes of delirium could contribute to our observation. Supporting this, a recent meta-analysis reports that hypoactive cases of delirium were more likely to be female (Ghezzi et al., 2022). Consequently, this might lead to increased diagnosis of delirium in men in observational routine care data, especially since persisting neurological deficits further impede diagnosis of delirium following MT of LVO (Vater et al., 2024).

Premorbid cognitive decline and cerebral atrophy have been repeatedly reported to be associated with development of delirium in patients with acute stroke. Investigation of the predictive capacity of cerebral atrophy for delirium in these patients, independent of conventional risk factors, has yielded conflicting results (Oldenbeuving et al., 2011; Rhee et al., 2022; Czyzycki et al., 2021; Qu et al., 2018). We show that a pathologic MTA score may be an easily accessible biomarker in native CT imaging that is independently associated with a 2-fold increase of odds for development of delirium following MT. Pathological MTA scores are associated with cognitive decline and Alzheimer’s disease (Scheltens et al., 1992). However and interestingly, premorbid clinical diagnosis of dementia was neither more frequent in patients with delirium nor adding predictive value to multiple regression modelling of delirium in our cohort. Increased awareness among stroke-unit personnel for the need of delirium prevention in patients with known dementia and consecutively better preventive measures may have potentially contributed to absence of increased delirium rates in these patients. However, it is more likely that this observation may be due to underdiagnosis of dementia or early phases of cognitive decline, not yet diagnosed as dementia, but apparent as brain atrophy. Similarly, underreporting of dementia diagnosis when obtaining medical history from a third party, as is often the case for patients with major ischemic stroke treated by MT, may contribute to our observation. Consequentially, pathological MTA scores, as an objective CT imaging biomarker associated with cognitive decline may provide additional predictive benefit instead.

With regard to treatment characteristics, the use of general anesthesia during MT was the only procedural feature independently associated with development of delirium. We report an almost 2.5-fold increase of odds to develop delirium following MT under general anesthesia. This finding is new in patients with LVO treated by MT and is clearly supported by systemic reviews reporting strong evidence for mechanical ventilation as a precipitating factor for delirium in the intensive care unit (Zaal et al., 2015). The role of anesthesia type during MT in functional outcome is controversially discussed (Sarraj et al., 2023; Chabanne et al., 2023; Campbell et al., 2023). Although further investigation is warranted, our finding demonstrates that delirium is a potential complication associated with use of general anesthesia during MT and argues against its across-the-board use in the management of LVO MT. Future studies should examine, whether patients developing delirium after MT benefit less from general anesthesia during MT, and, whether risk stratification for benefit of general anesthesia during MT could be improved by including independent risk factors for developing delirium.

Only few predictive tools exist to identify individuals with acute stroke at high risk of developing delirium (Drozdowska et al., 2021). Even though internal validation of these tools yielded comparatively high predictive capacity within the small mixed stroke cohorts they were derived from, it is questionable whether such tools are applicable to patients with LVO treated by MT. Such developed tools only provided limited predictive power in our cohort of MT-treated patients. We suggest that this is partly due to high weights placed on stroke severity in published predictive tools. This appears to enable identification of individuals at risk in mixed stroke cohorts, yet might be of limited significance in cohorts of patients with major stroke due to LVO. In line with this, we observed no significant difference in NIHSS on admission in patients with and without delirium in our cohort. Additional factors associated with delirium in MT-treated patients may enhance risk stratification in this distinct subgroup of acute stroke patients and should be considered in risk assessment for developing delirium in these patients.

Despite analyzing a broad dataset of patient-, stroke- and treatment characteristics, our study approach has several limitations. We observed a comparatively low incidence of delirium (8.2%) in our study cohort. The literature is inconsistent with a broad range regarding delirium incidence following acute stroke, yet observational data from routine care, as is our dataset, might be prone to underdiagnosis of delirium. Especially hypoactive phenotypes of the disease might be underdiagnosed. The implementation of screening for delirium on a regular basis, as has recently been added to national guidelines of acute stroke care (Ringleb et al., 2022), might help to overcome this potential source of bias in the future and allow for even more accurate analyses of delirium from observational data. Furthermore, reduced sensitivity for diagnosis of delirium has been shown for ICD-10 diagnoses, e.g., compared to discharge reports (Chuen et al., 2022; Thomas et al., 2012), which might further explain the comparatively low delirium incidence in our dataset. Due to the nature of our monocentric study cohort, our findings are also limited with regard to transferability and generalizability. Our analysis might also have limited generalizability to populations with an ethnical distribution differing from the western European one. Future studies should confirm and further examine our findings.

Our analysis is based on a well-characterized large cohort of patients with major stroke due to LVO treated by MT. Therefore, we were able to investigate a comprehensive set of stroke- and treatment characteristics, potentially associated with delirium. As delirium has previously only been analyzed in mixed stroke cohorts with few cases treated by MT, our findings are new and highly relevant for acute care of these patients. Furthermore, we identify risk factors and an imaging biomarker in native CT that are easy to obtain in clinical practice. This is an important advantage with regard to transferability into clinical practice for identification of individuals at high risk of developing delirium following MT.

We demonstrate that previously developed tools for identifying stroke patients at high risk for delirium have only limited predictive capacity in a cohort of patients treated by MT. Besides conventional risk factors such as age and infectious complications, additional risk factors for delirium exist that were not included in tools for risk stratification in mixed stroke cohorts until now. A pathologic MTA score, rated in native CT, may be an easy to use imaging biomarker identifying patients at increased risk for delirium following MT as is the use of general anesthesia during the MT procedure. Prevention of delirium is our best therapeutic strategy, as therapy options, once incident, are limited. Identification of individuals at high risk may help to intensify preventive measures and allow for a targeted resource allocation. The fact that delirium is associated with worse functional outcome, even in our cohort of severely diseased patients with LVO, justify and stress the importance of increased efforts to predict and prevent this complication.