Prevalence of Poststroke Cognitive Impairment
South London Stroke Register 1995–2010
Background and Purpose—Stroke is a common long-term condition with an increasing incidence as the population ages. This study evaluates temporal changes in the prevalence of cognitive impairment after first-ever stroke stratified by sociodemography, vascular risk factors, and stroke subtypes, up to 15 years after stroke.
Methods—Data were collected between 1995 and 2010 (n=4212) from the community-based South London Stroke Register covering an inner-city multiethnic population of 271 817 inhabitants. Patients were assessed for cognitive function using Abbreviated Mental Test or Mini-Mental State Examination at the onset, 3 months, and annually thereafter. All estimates were age adjusted to the European standard.
Results—The overall prevalence of cognitive impairment 3 months after stroke and at annual follow-up remained relatively unchanged at 22% (24% [95% CI, 21.2–27.8] at 3 months; 22% [17.4–26.8] at 5 years to 21% [3.6–63.8] at 14 years). In multivariate analyses, the poststroke prevalence ratio of cognitive impairment increased with older age (2% [1–3] for each year of age), ethnicity (2.2 [1.65–2.89]-fold higher among black group) and socioeconomic status (42% [8–86] increased among manual workers). A significant, progressive trend of cognitive impairment was observed among patients with small vessel occlusion and lacunar infarction (average annual percentage change: 10% [7.9–12.8] and 2% [0.3–2.7], respectively, up to 5 years after stroke).
Conclusions—The prevalence of cognitive impairment after stroke remains persistently high over time, with variations being predominantly explained by sociodemographic characteristics. Given population growth and ageing demographics, effective preventive strategies and poststroke surveillance are needed to manage survivors with cognitive impairment.
Stroke is the second leading cause of death after ischemic heart disease globally and a major cause of long-term disability.1 The nonstroke literature indicates that both cognitive and physical disabilities have significant impacts not only on patients, but also on families and informal caregivers.2
Patients who have had a stroke have an increased likelihood of cognitive impairment than people who have not had a stroke.3 The short-term prevalence of poststroke dementia, including cognitive impairment, has been reported in many studies.4–7 Most studies have used various standardized diagnostic measures, Diagnostic and Statistical Manuals of Mental Disorders IV, or a Mini-Mental State Examination (MMSE) score of <24 as an outcome. In a systematic review of 73 articles including 21 hospital-based and 8 population-based cohorts (7511 patients),8 it was reported that the overall rates of dementia in the first year after first-ever stroke were highly heterogeneous. Ninety-three percent of the variance was explained by study methods and case mix, with rates of poststroke dementia ranging from 7% in population cohorts (2 studies, 1045 patients) in which prestroke dementia has been excluded, to 41% in hospital-based studies (4 studies, 409 patients) in which recurrent stroke and prestroke dementia had not been excluded. However, most studies included in this review had short follow-up durations and included small numbers of patients. Furthermore, the systematic review showed the limitations of the available longitudinal studies to identify accurate prevalence in the overall population as well as in high-risk groups9 of cognitive impairment and dementia in poststroke survivors.
The objective of the current study is to evaluate temporal changes and trends in the prevalence of cognitive impairment after first-ever stroke by sociodemography, past medical history of vascular risk factors, and stroke subtypes up to 15 years of follow-up in population sample, the South London Stroke Register (SLSR) (1995–2010).
Materials and Methods
The SLSR is a prospective population-based stroke register set up in January 1995, recording all first-ever strokes in patients of all ages for an inner area of South London based on 22 electoral wards in Lambeth and Southwark. Data collected between 1995 and 2010 were used in this analysis. The total source population of the SLSR area was 271 817 individuals, self-reported as 63% white, 28% black (9% black Caribbean, 15% black African, and 4% black other), and 9% of other ethnic group in the 2001 census.10
Standardized criteria were applied to ensure completeness of case ascertainment, including multiple overlapping sources of notification. All patients with a suspected diagnosis of stroke or transient ischemic attack documented in different hospital and community-based information sources were investigated for study eligibility. Patients admitted to hospitals serving the study area were identified by regular reviews of acute wards admitting stroke patients, weekly checks of brain imaging referrals, and monthly reviews of bereavement officer and bed manager records. Additionally, national data on patients admitted to any hospital in England and Wales with a diagnosis of stroke were also screened for additional patients. To identify patients not admitted to hospital, all general practitioners within and on the borders of the study area were contacted regularly and asked to notify the SLSR of stroke patients. Regular communication with general practitioners was achieved by telephone contact and quarterly newsletters. Referral of nonhospitalized stroke patients to a neurovascular outpatient clinic (from 2003) or domiciliary visit to patients by the study team was also available to general practitioners. Community therapists were contacted every 3 months. Death certificates were checked regularly. Completeness of case ascertainment has been estimated at 88% by a multinomial-logit capture-recapture model using the methods described elsewhere.10,11
Special trained study nurses and field workers collected all data prospectively. A study clinician verified stroke diagnosis. The following sociodemographic characteristics were collected at initial assessment: self-definition of ethnic origin (census question), stratified into white, black (black Caribbean, black African, and black other), and other ethnic group. Socioeconomic status was categorized as nonmanual, manual, and economically inactive (retired and no information on previous employment), according to the patient’s current or most recent employment using the UK General Register Office occupational codes.
Stroke was defined according to World Health Organization criteria. Pathological stroke subtypes were classified using neuroradiology or necropsy results into ischemic stroke, primary intracerebral hemorrhage or subarachnoid hemorrhage. Ischemic strokes were further investigated according to 2 stroke subtypes: The Oxfordshire Community Stroke Project12 classification and the modified Trial of Org 10172 in Acute Stroke Treatment classification. The Oxfordshire Community Stroke Project contains 4 subtypes according to clinical features: lacunar infarct, total anterior circulation infarcts, partial anterior circulation infarcts, and posterior circulation infarcts. Modified Trial of Org 10172 in Acute Stroke Treatment denotes 5 pathogenetic subtypes: large artery atherosclerosis, cardioembolism, small vessel occlusion (SVO), other determined pathogenesis and no pathogenesis identified. Data with modified Trial of Org 10172 in Acute Stroke Treatment classification were collected only after the year 1999.
Past medical history of vascular risk factors for stroke (either self-reported or from medical notes) were collected including smoking, hypertension, diabetes mellitus, atrial fibrillation, ischemic heart disease (history of angina pectoris or myocardial infarction), and transient ischemic attack. Before stroke treatment of these risk factors were also collected (antihypertensives, anticoagulants, antiplatelets, diabetes control treatments, and cholesterol-lowering medications).
Barthel Index13 scores with a cutoff of 15 was used before and after stroke to identify patients with moderate-to-severe disability and Glasgow Coma Score14 dichotomized to <13 or ≥13 was used to measure stroke severity at onset. Other case-mix indicators used for initial stroke severity included urinary incontinence, aphasia, dysphagia, visual field defects, visuospatial neglect, dysphasia, dysarthria, hemiparesis, and cerebellar symptoms.
Cognitive state was assessed using the MMSE15 or Abbreviated Mental Test16 in the acute phase as well as at follow-ups. Before January 1, 2000, all assessments were conducted using the MMSE; after January 1, 2000, the Abbreviated Mental Test was administered. Subjects were defined as cognitively impaired according to predefined cutoff points (MMSE, 24 or Abbreviated Mental Test, 8).10 Subjects who could not undergo the cognitive test, because of severe aphasia or dysphasia or dysarthria, deafness, or visual impairment were excluded from the study. Patients were assessed at the stroke onset, 3 months, and annually after stroke. All follow-up assessments included in the present study were completed by August 31, 2010.
The prevalence of cognitive impairment was stratified by sociodemography, past medical history of vascular risk factor, and stroke subtypes. Prevalence ratios (PRs) between patient groups were used for comparison. All estimates were applied to the standard European population using the direct method with 4 age groups 0 to 64, 65 to 74, 75 to 84, and 85+. Ninety-five percent confidence intervals of age-standardized rates and ratios were calculated using the percentile bootstrap technique with 10 000 replications. Multivariate analyses, using a European age-standardized Poisson regression model with a robust error variance, were used to assess whether changes in prevalence ratios at each time point could be explained by differences in sociodemographic, past medical history, case-mix stroke severity, or stroke subtype.
Trends in the prevalence rates of cognitive impairment were analyzed using joinpoint regression. The joinpoint permutation test was used to select the optimal model that best fitted the data and tests of significance use a Monte Carlo permutation method. The average annual percentage change was used to quantify the change in prevalence rates over time. Kaplan–Meier estimates were used to model survival and to measure the cumulative survival and 95% CI at 1, 5, 10, and 15 years after stroke according to the cognitive status at 3 month after stroke. Differences in survival according to cognitive status were compared with the multivariate Cox proportional hazards model. The assumption for proportionality of the mortality hazards was assessed by visual inspection of cumulative hazard logarithm plots. All multivariate models were adjusted for sociodemographic, past medical history, disability, stroke severity, and Oxfordshire Community Stroke Project stoke subtype.
To assess the robustness of the results, we conducted sensitivity analyses with imputation of missing data as described in our previous study for stroke outcomes on SLSR (1995–2005).10 Similarly, when assessing trends over time of all imputation methods, although overall rates were altered, the trends over time closely followed those in the observed and complete case analysis. The observed data analysis was used for the present study. Statistical analyses and graphics were performed using STATA17 and JoinPoint program.18
All patients and their relatives gave written informed consent to participate in the study, and over the study period very few patients have declined to be registered. The design of the study was approved by the ethics committees of Guy’s and St Thomas’ NHS Foundation Trust, King’s College Hospital Foundation Trust, St George’s University Hospital, National Hospital for Nervous Diseases, and Westminster Hospital.
A total of 4212 patients with their first-ever stroke between January 1, 1995 and December 31, 2010 were registered in the SLSR, of whom 1101 were dead within 3 months, 534 were not eligible because of late registration, and 570 subjects were unable to undergo a cognitive assessment because of severe verbal, visual, or hearing impairment. Of the 2007 remaining subjects, a cognitive test was conducted in 1618 at 3 months and 389 were lost to follow-up by 3 months. Three months poststroke characteristics of these patients including sociodemographic, past medical history, case mix, and stroke subtypes are presented in Table.
The cognitive impairment rate in stroke survivors was strongly associated with age at all time points and it progressively increased after 5 years of stroke for patients aged 65 to 85 years old (Figure 1). Among stroke survivors, the age-standardized prevalence of cognitive impairment ranged from 24% (95% CI, 21.2–27.8) at 3 months, 22% (95% CI, 17.4–26.8) at 5 years, and 18% (95% CI, 8.9–26.6) at 10 years to 21% (95% CI, 3.6–63.8) at 14 years after stroke. On average, the overall prevalence of cognitive impairment 3 months after stroke and annually between 1995 and 2010 was similar and remained relatively unchanged at around 22% up to 15 years after stroke (Figure 1). No significant differences were found between males and females, and no significant trends of cognitive impairment rates over the period 1995–2010 were observed (Figures 1 and 2).
In analyses adjusted for sociodemography, prestroke risk factors, case mix, and stroke subtypes, the age-standardized prevalence ratio after stroke increased on average by 2% for each year of chronological age (PR=1.02; 95% CI, 0.01–0.03 at 3 months, PR=1.01; 95% CI, 0.1–0.03 at 3 years, and PR=1.03; 95% CI, 0.01–0.05 at 5 years).
Higher cognitive impairment rates were observed within first 7 days, particularly for stroke survivors aged >65. At 3 months, cognitive function improved but then remained stable with little variation up to 15 years (Figure 1). Among patients who were cognitively impaired in the acute stage (up to 7 days after stroke) function was regained by nearly a quarter of patients at 3 months (24%; 95% CI, 20.8, 26.6). After 3 months, cognitive impairment was significantly higher among the black ethnic group (26%) compared with the white ethnic group (17%) and the manual socioeconomic group (24%) compared with the nonmanual (20%) (Figure 2). Black-to-white and manual-to-nonmanual prevalence ratios of cognitive impairment were higher up to 7 years after stroke. PR of cognitive impairment among high-risk groups are presented in Figure 3. In a multivariate analysis controlling for sociodemography, prestroke risk factors, case mix, and stroke subtypes, black-to-white adjusted PR were 2.2, 95% CI, 1.65–2.89 at 3 months, 2.3, 95% CI, 1.64–3.19 at 3 years, and 1.6, 95% CI, 1.02–2.38 at 7 years; and manual-to-nonmanual adjusted PR were 1.42, 95% CI, 1.08–1.86 at 3 months, 1.6, 95% CI, 1.11–2.23 at 3 years, and 1.7, 95% CI, 1.05–2.76 at 7 years.
Cognitive impairment rates were also high in total anterior circulation infarct (49%) and large artery atherosclerosis (40%) subtypes up to 5 years after stroke (Figure 2). Cognitive impairment in lacunar infarct and SVO strokes was increasingly prevalent during the first 3 to 4 years after stroke and decreased slightly thereafter (Figure 2). Using joinpoint regression, the estimated average annual percentage change of cognitive impairment was even greater for SVO compared with lacunar infarct up to 5 years after stroke (10% [95% CI, 7.9–12.8] for SVO and 2% [95% CI, 0.3–2.7] for lacunar infarct). Similarly, a progressive trend of cognitive impairment was also observed for patients with no prestroke vascular risk factor with an average annual percentage change of 6% (95% CI, 5.3–7.5) in the first 5 years after stroke. However, a negative trend of cognitive impairment was observed in patients with prestroke vascular risk factors (average annual percentage change of −6% (95% CI, −11.0 to −0.3).
Cumulative survival up to 15 years after stroke stratified by 3 months cognitive status is illustrated in Figure 4. Patients cognitively impaired at 3 months had the worst survival, with 53% (95% CI, 48.2–56.8), 37% (95% CI, 32.4–40.7), and 34% (95% CI, 29.6–37.7) surviving up to 5, 10, and 15 years, respectively, after stroke. After fitting the Cox proportional hazards model adjusted for sociodemographic, previous vascular risk factors, case mix, and stroke subtypes, stroke survivors with cognitive impairment at 3 months experienced a 53% increased risk of death compared with those with no known cognitive impairment at the same time point (hazard ratio: 1.53, 95% CI, 1.30–1.80). Moreover, the prevalence ratio of disability at each year after stroke was on average twice as high for cognitively impaired patients. Using the Poisson regression model with robust error variance adjusted for sociodemographic, prestroke vascular risk factors, case mix, and stroke subtypes the relative risk of disability progression among patients with cognitive impairment at 3 months were 2.4 (95% CI, 1.93–3.08) at 1 year, 1.9 (95% CI, 1.38–2.60) at 3 years, and 1.8 (95% CI, 1.27–2.55) at 5 years.
This study estimates the prevalence of cognitive impairment up to 15 years after stroke using practical and commonly administered psychometric screening questionnaires for moderate cognitive deficits or dementia in stroke.19–21 It not only provides population estimates, to our knowledge for the first time, on the longer term cognitive outcomes in a diverse inner-city population, but it also highlights the long-term impacts of stroke in subgroups of the population and illustrates the risk of cognitive impairment long term. It is rare that population-based studies estimate this range of outcomes in such a prospective manner, with up to 15 years of follow-up.8 Furthermore, the use of these year-on-year prevalence estimates provides more precise estimates of temporal changes of poststroke cognitive impairment based on population observations.
The overall prevalence of poststroke cognitive impairment seemed to be stable after 3 months after stroke and during the 1995–2010 period. A major observation is that there are some differences in the cognitive impairment rates among groups when stratified by sociodemography, past medical history of vascular risk factors, and stroke subtypes. These differences in prevalence of cognitive impairment by group and by time after stroke bear witness to the heterogeneity of this condition. Although age could be linked to accumulated lifetime exposures affecting cognitive function and socioeconomic status could be a proxy for education level, the effect of ethnicity is largely unexplained. Given population growth and ageing demographics, these could prove to be some of the main challenges of our time.
A prevalence of cognitive impairment of almost half the patients up to 5 years after stroke was found in patients having a total anterior circulation infarct stroke. However, a stepwise progression of cognitive impairment frequencies was observed among stroke survivors with SVO and lacunar infarct stroke, which may represent progressive vascular dementia associated with stroke.3 Other studies have shown that the progressive cognitive decline of these groups could be related to vascular dementia (where the prevalence of cognitive impairment could double every 5 years)22,23 or Alzheimer disease.3 A similar but unexplained progressive pattern was observed among patients with no known prestroke vascular risk factors.
This population-based study has produced estimates of cognitive status over time in stroke survivors, but with no comparison to the nonstroke population. The group-specific estimates and temporal trends confirm the dynamic process and heterogeneous nature of poststroke cognitive impairment. Further longitudinal analyses of predictors in various sociodemographic, vascular risk factors, stroke subtype, and case-mix groups are thus needed to develop a useful predictive tool for patient management. Furthermore, the psychometric screening tools used in this study may underestimate the impact of cognitive impairments, particularly mild cognitive impairment. It has been shown that MMSE or Abbreviated Mental Test are insensitive to mild cognitive impairment and executive function.19,24,25 Although these are limitations in detecting mild cases, this study has shown high prevalence of cognitive impairments. Similar large long-term studies of poststroke cognitive function using sensitive tools to detect all cases, including mild cognitive impairments, will be of benefit to reconfirm the burden of this condition.
The loss to follow-up rates in this study, once deaths are accounted for, are <20% at each time point except at 3 months. One might have expected the highest participation rate at 3 months. However, some of subjects were unable to complete the cognitive assessments because of language difficulties, visual or hearing impairment, and a proportion of patients are registered retrospectively for whom a 3-month assessment is not possible. Owing to late registrations and latter improvement in some patients who were originally severely impaired, there were subjects who had a cognitive test done at their annual assessments, but not at 3 months.
This loss to follow-up may introduce bias, yet estimates from analyses of the patients with complete data did not differ significantly from those presented here. Loss to follow-up may be an issue in certain sociodemographic groups, although we have not been able to identify such groups in this analysis. The healthier participants and those from higher socioeconomic groups may be more likely to engage in research follow-up. In other cohort and stroke register studies, loss to follow-up rates are not often presented. Inner-city populations are mobile, with large numbers of migrant families. Although we acknowledge this as a potential factor in loss to follow-up, efforts were made for all patients’ changes of address to be recorded from hospital, general practice, or family sources.
To conclude, cognitive impairment is one of the indicators of long-term impact of stroke. The overall prevalence of cognitive impairment remains persistently high over time in stroke patients after their first stroke. Variations in poststroke cognitive impairment rates could be explained by subgroup differences and temporal changes after stroke. Age, ethnicity, socioeconomic characteristics, and stroke subtypes are the most predominant factors that are associated with poststroke cognitive impairment. Given the predicted global population growth and ageing demographics, this study shows that effective preventive intervention and strategic planning are needed for health systems to identify and to manage stroke survivors with cognitive impairment.
We wish to thank all the patients and their families and the healthcare professionals involved. Particular thanks to all the fieldworkers and the whole team who have collected data since 1995 for the South London Stroke Register. We are also very thankful for the anonymous reviewers for their helpful comments.
Sources of Funding
The study was funded by the Northern & Yorkshire NHS R&D Programme in Cardiovascular Disease and Stroke, Guy’s and St Thomas’ Hospital Charity, Stanley Thomas Johnson Foundation, The Stroke Association, Department of Health HQIP grant, National Institute for Health Research Programme Grant (RP-PG-0407-10184). The authors acknowledge financial support from the National Institute for Health Research (NIHR) Biomedical Research Centre based at Guy’s and St Thomas’ NHS Foundation Trust and King’s College, London. The views expressed in this article are those of the authors and not necessarily those of the National Health Service, the NIHR, or the Department of Health.
- Received August 8, 2012.
- Revision received September 21, 2012.
- Accepted September 26, 2012.
- © 2013 American Heart Association, Inc.
- Truelsen T,
- Begg S,
- Mathers C
- Allan LM,
- Rowan EN,
- Firbank MJ,
- Thomas AJ,
- Parry SW,
- Polvikoski TM,
- et al
- Béjot Y,
- Aboa-Eboulé C,
- Durier J,
- Rouaud O,
- Jacquin A,
- Ponavoy E,
- et al
- Patel M,
- Coshall C,
- Rudd AG,
- Wolfe CD
- Tatemichi TK,
- Foulkes MA,
- Mohr JP,
- Hewitt JR,
- Hier DB,
- Price TR,
- et al
- Tilling K,
- Sterne JA,
- Wolfe CD
- Hodkinson HM
- 17.↵StataCorp. Stata: Release 11.2. Statistical software. College Station, TX: StataCorp; 2011.
- 18.↵Surveillance Research Program National Cancer Institute. JoinPoint regression program version 3.5. Bethesda, MD: Statistical Methodology and Applications Branch and Data Modeling Branch; 2011.
- Kelly-Hayes M,
- Robertson JT,
- Broderick JP,
- Duncan PW,
- Hershey LA,
- Roth EJ,
- et al
- Ganguli M
- Gorelick PB,
- Scuteri A,
- Black SE,
- Decarli C,
- Greenberg SM,
- Iadecola C,
- et al
- Pendlebury ST,
- Cuthbertson FC,
- Welch SJ,
- Mehta Z,
- Rothwell PM
- Nys GM,
- van Zandvoort MJ,
- de Kort PL,
- Jansen BP,
- Kappelle LJ,
- de Haan EH