Domain-specific cognitive impairment 6 months after stroke: The value of early cognitive screening

Background: Cognitive screening following stroke is widely recommended, yet few studies have considered the prognostic value of acute domain-specific function for longer-term cognitive outcome. Identifying which post-stroke cognitive impairments more commonly occur, recover, and persist, and which impairments hold prognostic value, could inform care planning, and resource allocation. Aims: This study aimed to determine the prevalence of domain-specific impairment acutely and at 6 months, assess the proportion of change in cognitive performance, and examine the prognostic value of acute domain-specific cognitive screening. Methods: A prospective stroke cohort completed the Oxford Cognitive Screen acutely (⩽2 weeks) and 6 months post-stroke. We determined the prevalence of acute and 6-month domain-specific impairment and proportion of change in performance from acute to 6 months. Hierarchical multivariable regression was used to predict global and domain-specific cognitive impairment at 6 months adjusted for demographic/vascular factors, stroke severity, and lesion volume. Results: A total of 430 stroke survivors (mean/SD age 73.9/12.5 years, 46.5% female, median/interquartile range (IQR) National Institute of Health Stroke Scale (NIHSS) 5/2–10) completed 6-month follow-up. Acutely, domain-specific impairments were highly prevalent ranging from 26.7% (n = 112) in praxis to 46.8% (n = 183) in attention. At 6 months, the proportion of domain-specific recovery was highest in praxis (n = 73, 71%) and lowest in language (n = 89, 46%) and memory (n = 82, 48%). Severity of 6-month cognitive impairment was best predicted by the addition of acute cognitive impairment (adj R2 = 0.298, p < 0.0001) over demographic and clinical factors alone (adj R2 = 0.105, p < 0.0001). Acute cognitive function was the strongest predictor of 6-month cognitive performance (p < 0.0001). Acute domain-specific impairments in memory (p < 0.0001), language (p < 0.0001), and praxis (p < 0.0001) significantly predicted overall severity of cognitive impairment at 6 months. Conclusion: Post-stroke cognitive impairment is highly prevalent across all domains acutely, while impairments in language, memory, and attention predominate at 6 months. Early domain-specific screening can provide valuable prognostic information for longer-term cognitive outcomes.


Introduction
Though the prevalence of post-stroke cognitive impairment varies depending on the timing, assessment, and inclusion criteria, most stroke survivors experience at least one cognitive deficit. 1,2Identifying which cognitive impairments more commonly occur, recover, and persist will aid in intervention planning and setting rehabilitation goals.
4][5][6] The Oxford Cognitive Screen (OCS) 7,8 is a short, domain-specific cognitive screen which compromises between the benefits of global tests and neuropsychological batteries.The OCS offers superior patient inclusivity and sensitivity than brief screens (e.g. the Mini-Mental State Examination (MMSE) and Montreal Cognitive Assessment (MoCA)). 1,9t also assesses multiple domains much like a neuropsychological battery, which is in line with clinical guidance.Acute domain impairments are highly prevalent after stroke; 1 however, the prevalence and trajectories of domainspecific impairment beyond the acute stage are relatively unknown.In addition, while the OCS has recently been shown to provide value information regarding functional outcomes, 10 it has not been established whether the short screen had prognostic value in determining longer-term cognitive outcome.
Very few studies have considered acute post-stroke cognitive performance in longer-term outcome prediction models, and instead, the focus tends to be on clinical and neuroimaging markers. 114][15][16][17] Acute MoCA scores predict long-term cognitive and functional outcome, and mortality. 13However, the MoCA is not always suitable for stroke as it assumes cognitive functions (e.g.speech, vision) are intact.Consequently, common post-stroke impairments (e.g.aphasia, neglect) can confound assessment. 9,12ne study used a multi-domain neuropsychological battery to investigate the prognostic value of early cognitive assessment 18,19 and found acute performance to predict 6-month cognitive and functional outcome better than demographic or clinical variables. 19These findings support the utility of cognitive data; however, a large neuropsychological battery (>1 h) is typically not feasible in routine acute clinical practice due to time requirements and increased burden on the patient, as well as staff availability and expertise.
This study aimed to use the OCS to (1) determine the prevalence of domain-specific cognitive impairments acutely and at 6 months post-stroke, (2) assess change in function across timepoints, and (3) examine the predictive value of early domain-specific screening.Importantly, this study does not aim to build optimal prognostic models employing all potential predictors (e.g.genetic risk) but instead aims to determine the value added by considering cognitive data.Ultimately, this could lead to practical prognostic models based on routinely collected and easily accessible data, promoting informed and efficient clinical decision-making.

Methods
This study considered existing data from the OCS-Tablet and OCS-Recovery studies (National Research Ethics Committee (UK), references 14/LO/0648 and 18/ SC/0550, in accordance with the Declaration of Helsinki).Patients (⩾18 years old) were included if they could provide written/witnessed informed consent, could concentrate for 20 min, and had sufficient English language comprehension.

Cognitive assessment
The OCS was used for acute and follow-up cognitive screening. 7The OCS covers a broad range of cognitive domains and was designed specifically for use in acute stroke taking 15-20 min to complete.Subtests are categorized into six domains: language (picture naming, semantic understanding, and sentence reading), attention (egocentric and allocentric attention; broken hearts test), executive function (trail-making test), memory (orientation, verbal, and episodic memory; delayed recall/recognition), praxis (meaningless gesture imitation), and number processing (calculations and number writing).Each task has a threshold cut-off to indicate impairment based on published normative data. 7,8Subtests were binarized into impaired or unimpaired based on normative scores for each subtest.Domains were considered impaired if there is at least one subtest in the domain (e.g.language impaired if reading, naming, or semantics impaired).Subtests range from one to three across domains.The OCS was administered by trained neuropsychologists and occupational therapists.The OCS is licensed through Oxford University Innovations free of charge for publicly funded research and clinical use.Further information regarding administration of the OCS is detailed in supplemental methods.

Statistical analysis
Descriptive statistics were used to summarize the prevalence of cognitive impairment/recovery as cognitive data were limited to binarized outcome.Associations between International Journal of Stroke, 19 (3)   impairments within and across timepoints were quantified using tetrachoric correlation. 20Sensitivity/specificity of acute impairments for predicting 6-month outcomes was calculated using receiver operating curves (ROCs).
The predictive value of early cognitive screening was quantified using hierarchical multivariable regression.Three models were produced progressing in terms of cognitive detail with cognitive performance at follow-up considered as the dependent variable.Model 1 examined the relationship between proportion of cognitive subtests impaired acutely and at follow-up.Model 2 examined acute domain-specific function and proportion of subtests impaired at follow-up.To ensure model results are robust predictors rather than the products of model overfitting, each resultant model was cross-validated.Each full model's resultant significant covariates were iteratively (n = 1000) fitted to a random data subset (70%) and tested in the remaining 30% (testing data).If models yield abovechance fits in testing data, this suggests that the identified covariates are robust predictors of patient outcomes.
Model 3 examined each domain individually at acute and follow-up.Demographic/clinical factors were entered in the first block, and acute cognition was added in the second block.Demographic covariates included age, sex, and years of education.Clinical variables included stroke severity (the National Institute of Health Stroke Scale (NIHSS)), lesion volume, recurrent stroke, atrial fibrillation, hypertension, diabetes, smoking, and days from stroke onset to cognitive assessment.Lesion masks were constructed using a standard protocol. 21Missing NIHSS data were amended through multiple imputation.Analyses with p-values less than 0.05 were considered significant (Bonferroni-corrected where appropriate).All analyses were performed in R version 4.0.5.

Results
Demographics and clinical characteristics of the cohort are outlined in Table 1.The OCS was administered a mean of 4.39 (4.46 SD) days and again at a mean 6.65 (1.06 SD) months after stroke.Patient demographics did not statistically differ between those who were and were not reassessed; however, the attrition group comprised more severe stroke and cognitive impairments (Supplemental Tables 1 and 2).
Cross-validation indicated that the results of Model 1 (Block 2) were robust with 100% of tested models yielding above-chance fits within the new testing data (average testing model R 2 = 0.271 (SD = 0.066), range = 0.025-0.513).

Acute domain impairments predict withinand cross-domain outcomes
Hierarchical regressions were conducted to identify acute factors predictive of individual domain impairments at follow-up.Each base model with only conventional predictors improved with the addition of domain-specific cognition (Supplemental Figure 2).Each follow-up impairment was best predicted by the same domain impairment acutely except attention (strongest predictor was age: OR = 1.0673, p = 0.0006) and number impairment, where no predictor remained significant after correction.Full results for all domain-specific regression models are presented in Supplemental Tables 4-10.

Discussion
Nearly every patient was impaired on at least one cognitive subtest initially following stroke, while over twothirds were impaired at follow-up.Attention and language impairments were most prevalent acutely with the addition of memory impairments at follow-up.The impairment prevalence decreased within all domains from acute to follow-up, and most participants who were unimpaired acutely remained unimpaired at follow-up.However, newly acquired impairments were found across all domains at follow-up.The severity of acute domain-specific cognitive impairment identified with the OCS was the strongest predictor of follow-up cognitive function, over and above conventional demographic and clinical factors.Acute impairments in memory, language, and praxis were particularly important in predicting the severity of follow-up cognitive impairment.
7][28] Our findings of higher prevalence at both timepoints are likely due to this study's representative sample including potential pre-stroke cognitive decline, recurrent strokes, severe aphasia, 29 and the timing of acute assessment (average 4 days post-stroke).In addition, the OCS has higher sensitivity to stroke-specific deficits compared to other tools. 1,9The increased prevalence of multi-domain impairments found in this cohort is in line with Nys et al. 18 who reported that impaired patients averaged three domain deficits using a neuropsychological battery.Similarly, Jokinen et al. 30 found that 50% were impaired in three or more domains.The frequency of multi-domain impairments is reflected in clinical stroke guidance emphasizing the need for multi-domain screening. 3Furthermore, global screens that fail to assess stroke-specific deficits, such as visuospatial neglect, may omit information valuable to rehabilitation planning. 1,9he frequency of domain-specific deficits reported by previous studies varies.For example, both Hurford et al. 25 and Leśniak et al. 24 found attention to be the most commonly affected domain, though their attention measures included different executive demands.Different measures and domain definitions may explain variability across studies.In our cohort, attention and language were the most frequently impaired domains, which are hallmark deficits of lateralized stroke. 31,32A high prevalence of spatial attention impairments was also found in previous studies. 33Both neglect types were prevalent in left-and right-sided stroke, highlighting that allocentric and rightsided neglect are not uncommon. 34Comparatively high language impairment may be explained by our inclusion of aphasic patients.Aphasia is present in 30% of acute stroke patients 35 and excluding these patients risks  International Journal of Stroke, 19 (3)   biasing post-stroke cognitive profiles.In addition, though number processing, memory, executive function, and praxis deficits were less frequent, impaired subtests still occurred in 27-42% acutely and 20-32% at follow-up.This is consistent with the studies highlighting executive function, processing speed, and episodic memory as more commonly reported deficits months after stroke. 23,25,30verall, cognitive impairments occur in multiple domains, frequently affecting complex abilities in which attention and language have a major role.Impairment prevalence decreased across all domains from acute to follow-up, aligning with previous research. 18,19,25The highest proportion of recovery was in praxis and number processing, while language and memory impairments were most persistent.This aligns with the study by Hurford et al. 25 where memory deficits did not change between assessments but contradicts the study by Turunen et al., 26 which found the greatest rates of recovery from baseline to 6 months in executive functions and visual memory.Persistent and acquired memory deficits could be due to memory being associated with pre-stroke neurodegeneration, which would not be expected to change over a relatively short period.Similarly, persistent language deficits were consistent with reports of long-term language impairments at 1-year post-stroke, 33 and of language recovery decreasing beyond 6 months post-stroke. 36Another important finding is the proportion of acquired impairments observed, ranging 11-20%, and consistent with Nys et al. 19 who reported new impairments at 6-month follow-up in all five domains assessed.The presence of varying domain outcomes in this sample mirrors previous varying overall trends toward recovery, stability, and decline. 37,38In general, the risk of post-stroke cognitive decline is dependent on the combination of pre-existing cerebral vulnerability/ reserve and the impact of stroke. 39he key finding of this study is that routine early domain-specific screening with the OCS can predict 6-month cognitive function and provide domain-specific profiling.Critically, the aim of this study was not to identify an optimal predictive model, but to determine if acute cognition could provide useful prognostic information.Acute cognitive functioning was strongly associated with both severity of cognitive impairment at follow-up and domain-specific impairment, explaining ~30% more outcome variance than conventional demographic and clinical factors.More specifically, memory, language, and praxis deficits acutely were highly significant predictors of the severity of cognitive impairment at follow-up.Age emerged as the most consistent demographic risk factor for severity of cognitive impairment at follow-up and was the strongest predictor of a 6-month attention impairment.International Journal of Stroke, 19 (3)   domain-specific cognitive function as measured by the OCS are important when considering neurorehabilitation and cognitive trajectories post-stroke.These findings support the current recommendation for routine early domainspecific cognitive screening post-stroke and demonstrate the prognostic value of early cognitive markers, both of which have been highlighted as priorities for stroke research. 4,6,40his study has strengths, including a representative sample, a balance of minor and major stroke, and longitudinal domain-specific cognitive screening.However, there are also limitations.Though efforts were made to maximize inclusivity, all studies requiring a sufficient comprehension to provide (witnessed) informed consent will exclude those with most severe stroke and the poorest cognitive outcomes.A proportion of attrition was due to death (19%), again representing a potential source of bias toward underrepresentation of more severe stroke.Nevertheless, attrition in a representative stroke sample is generally unavoidable and a relatively large sample size was retained.
In summary, early domain-specific cognitive profiling with the OCS provides valuable prognostic information with respect to longer-term cognitive functioning.The OCS is currently used routinely in clinical settings and these findings suggest it could be employed for prognostic purposes within a stroke care pathway, prompting further follow-up with a more detailed neuropsychological battery.Cognitive impairment was highly prevalent initially after stroke, though after 6 months, prevalence of impairment had decreased across all domains demonstrating a general trend toward recovery.However, persistent impairments were also found across all domains, particularly memory and language deficits.Each post-stroke cognitive profile is unique and highlighting different strengths and weaknesses in performance early allows for more accurate information to be communicated to the patient, 22 more tailored discharge care packages and appropriate allocation of rehabilitation resources.

Figure 2 .
Figure 2. Association between acute and follow-up domain impairments and acute impairments predictive of follow-up impairments: (a) Associations between acute domain impairments.(b) Associations between follow-up impairments.(c) Acute impairments predicting follow-up impairments.For (a and b), tetrachoric correlation coefficient is shown with color indicating correlation coefficients.For (c), color indicates AUC; acute: x-axis, follow-up: y-axis.

Figure 3 .
Figure 3. Proportion of change in impairment across timepoints.(a) Unimpaired: proportion of individuals who remained unimpaired, recovered: were impaired, but recovered by follow-up, persistent: remained impaired, and acquired (red): were unimpaired initially, but were impaired at follow-up.Across all domains, 52.2% were unimpaired, 21.3% recovered, 16.6% had persistent impairment, and 9.9% acquired impairment.(b) Improved: fewer domains impaired at follow-up, stable: no change, and declined: more domains impaired at follow-up.Solid line represents the regression with the considered clinical/demographic covariates.

Table 1 .
Cohort demographics and clinical characteristics.
a Vascular supply is a crude characterization of primary territory affected.bNIHSS recorded for n = 320 (74%) due to not being recorded before 2014.cPremorbid modified Rankin Scale (mRS) reported for n = 248 (58%).dBarthel index was reported for n = 141 (33%).eDependence was categorized as requiring formal support.(Continued)

Table 2 .
Prevalence of post-stroke cognitive impairment acutely and at follow-up.
Domain impairment prevalence at acute and 6-month assessments.Any domain: an impairment on any subtest in any domain; multi-domain: impairment in ⩾ 1 domain.Supplemental Table2summarizes patients not followed-up at 6 months.aPrevalence of impairment on each subtest for only patients who were assessed at both acute and 6 months for each subtest is detailed in Supplemental Table3b.All domains were assessed at both timepoints in n = 322/430 (74.8%).International Journal of Stroke,19(3)

Table 3 .
Regression results for clinical/demographic and acute cognitive factors predictive of follow-up impairment.
Collectively, these findings indicate that both age and acute R 2 = 0.345, Adjusted R 2 = 0.309, F = 9.612, 167 and 310 df, p < 0.0001 OCS: Oxford Cognitive Screen; NIHSS: National Institute of Health Stroke Scale.a Remained significant after Bonferroni correction for multiple comparisons.*Significance p < 0.05.