Hostname: page-component-8448b6f56d-c47g7 Total loading time: 0 Render date: 2024-04-24T20:23:51.064Z Has data issue: false hasContentIssue false

Longitudinal associations between stroke and psychosis: a population-based study

Published online by Cambridge University Press:  05 June 2023

Alvin Richards-Belle
Research Department of Clinical, Educational and Health Psychology, University College London, London, UK Division of Psychiatry, University College London, London, UK
Norman Poole
Department of Neuropsychiatry, South West London and St George's Mental Health NHS Trust, London, UK
David P.J. Osborn
Division of Psychiatry, University College London, London, UK
Vaughan Bell*
Research Department of Clinical, Educational and Health Psychology, University College London, London, UK South London and Maudsley NHS Foundation Trust, London, UK
Corresponding author: Vaughan Bell; Email:
Rights & Permissions [Opens in a new window]



The co-occurrence of stroke and psychosis is a serious neuropsychiatric condition but little is known about the course of this comorbidity. We aimed to estimate longitudinal associations between stroke and psychosis over 10 years.


A 10-year population-based study using data from the English Longitudinal Study of Ageing. A structured health assessment recorded (i) first-occurrence stroke and (ii) psychosis, at each wave. Each were considered exposures and outcomes in separate analyses. Logistic and Cox proportional hazards regression and Kaplan–Meier methods were used. Models were adjusted for demographic and health behaviour covariates, with missing covariates imputed using random forest multiple imputation.


Of 19 808 participants, 24 reported both stroke and psychosis (median Wave 1 age 63, 71% female, 50% lowest quintile of net financial wealth) at any point during follow-up. By 10 years, the probability of an incident first stroke in participants with psychosis was 21.4% [95% confidence interval (CI) 12.1–29.6] compared to 8.3% (95% CI 7.8–8.8) in those without psychosis (absolute difference: 13.1%; 95% CI 20.8–4.3, log rank p < 0.001; fully-adjusted hazard ratio (HR): 3.57; 95% CI 2.18–5.84). The probability of reporting incident psychosis in participants with stroke was 2.3% (95% CI 1.4–3.2) compared to 0.9% (95% CI 0.7–1.1) in those without (absolute difference: 1.4%; 95% CI 0.7–2.1, log rank p < 0.001; fully-adjusted HR: 4.98; 95% CI 2.55–9.72).


Stroke is an independent predictor of psychosis (and vice versa), after adjustment for potential confounders.

Original Article
Creative Commons
Creative Common License - CCCreative Common License - BY
This is an Open Access article, distributed under the terms of the Creative Commons Attribution licence (, which permits unrestricted re-use, distribution and reproduction, provided the original article is properly cited.
Copyright © The Author(s), 2023. Published by Cambridge University Press


Stroke and psychosis are both severe, disabling conditions, independently associated with a high risk of morbidity and mortality, meaning that their co-occurrence is considered to be a particularly serious neuropsychiatric condition (John et al., Reference John, McGregor, Jones, Lee, Walters, Owen and Lloyd2018; Stroke Association, 2020). Clinically, people affected by both conditions are less likely to receive either adequate stroke or psychiatric care (Kisely, Campbell, & Wang, Reference Kisely, Campbell and Wang2009; Willers, Sunnerhagen, Lekander, & von Euler, Reference Willers, Sunnerhagen, Lekander and von Euler2018) indicating a clear need to develop the evidence-base and clinical pathways. However, research on stroke-psychosis comorbidity is limited to cohort studies focusing on stroke risk in people with schizophrenia (Li, Fan, Tang, & Cheng, Reference Li, Fan, Tang and Cheng2014) and bipolar disorder (Yuan et al., Reference Yuan, Xiao, Zhou, Rao, Huang, Nie and Xu2022) diagnoses, while the majority of publications on psychosis after stroke are single case studies (Stangeland, Orgeta, & Bell, Reference Stangeland, Orgeta and Bell2018).

Although psychosis has been frequently described as a ‘rare’ complication of stroke (Nemani & Gurin, Reference Nemani and Gurin2021), a meta-analysis of post-stroke psychosis reported the prevalence of delusion and hallucination to be 4.67% and 5.05% respectively (Stangeland et al., Reference Stangeland, Orgeta and Bell2018). In terms of stroke occurring after psychosis (‘post-psychosis stroke’), a meta-analysis by Li et al. (Reference Li, Fan, Tang and Cheng2014) reported that individuals with schizophrenia have an increased risk of stroke (relative risk 1.50; 95% confidence interval [CI] 1.25–1.80) and an increased risk of stroke mortality (relative risk 1.65; 95% CI 1.31–2.08). These findings are echoed in individuals with bipolar disorder, where a recent meta-analysis by Yuan et al. (Reference Yuan, Xiao, Zhou, Rao, Huang, Nie and Xu2022) reported increased stroke incidence (hazard ratio (HR) 1.43; 95% CI 1.09–2.18) and stroke mortality (HR 1.35; 95% CI 1.26–1.45). A recent cross-sectional study that combined data from national representative psychiatric epidemiology studies from four countries (UK, USA, Chile and Colombia) reported that 3.81% of people (approximately 1 in 26) with probable psychosis have stroke, while 3.15% of people with stroke (approximately 1 in 32) have probable psychosis (Bell, Tamayo-Agudelo, Revill, Okai, & Poole, Reference Bell, Tamayo-Agudelo, Revill, Okai and Poole2023). These studies indicate a higher level of co-occurrence than has been assumed in, previous, admittedly ad-hoc estimates, of ‘rare’ occurrence in the literature.

To date, however, estimates of risk and co-occurrence have been taken from studies that have either solely focused on stroke incidence in a psychosis population, or psychosis incidence in a stroke population. Methodologically, they have either been cross-sectional or cohort studies with limited waves of follow-up. This has meant it has been impossible to estimate how stroke-first or psychosis-first presentations differ in terms of risk for stroke-psychosis comorbidity – something important to estimate within the same population – or the extent to which the risk for comorbidity changes over time. Initial evidence suggests that risk of psychosis after stroke may peak at approximately six months post-stroke (Stangeland et al., Reference Stangeland, Orgeta and Bell2018). Conversely, cardiovascular risk factors compound over time in people with psychosis (Morgan et al., Reference Morgan, Waterreus, Ambrosi, Badcock, Cox, Watts and Jablensky2021), potentially leading to a late rather than ‘early peak’ risk profile. Indeed, Fleetwood et al., reported this pattern in individuals with schizophrenia and bipolar disorder where stroke risk markedly increased at five year follow-up but remained similar between 30-day and 1-year follow-up (Fleetwood et al., Reference Fleetwood, Wild, Smith, Mercer, Licence, Sudlow and Jackson2021).

One challenge, however, is that few studies measure both psychosis and stroke within the same cohort, something needed for a direct comparison between time to onset. One exception is the English Longitudinal Study of Ageing (ELSA; Steptoe, Breeze, Banks, & Nazroo, Reference Steptoe, Breeze, Banks and Nazroo2013) a representative longitudinal study of adults aged 50 years and older that collects economic, social, psychological, cognitive, health, biological and genetic data which was initiated in 2002 and is ongoing. We used data from ELSA to estimate associations between stroke and psychosis and the incidence of stroke after psychosis and psychosis after stroke over 10 years of follow-up.


Study design

This population-based longitudinal study used data from the ELSA. ELSA is an ongoing, prospective, observational, longitudinal study that began in 2002 and which includes a representative sample of adults aged 50 years and over in England and their cohabiting partners of any age (partners were not selected based on representativeness) (Steptoe et al., Reference Steptoe, Breeze, Banks and Nazroo2013). ELSA received ethical approval from the London Multi-Centre Research Ethics Committee (REC) (waves 1–3, references: MREC/01/2/91, MREC/04/2006, 05/MRE02/63), National Hospital for Neurology and Neurosurgery & Institute of Neurology Joint REC (wave 4, reference: 07/H0716/48), Berkshire REC (wave 5, reference: 09/H0505/124), NRES Committee South Central – Berkshire (waves 6–9, references: 11/SC/0374, 13/SC/0532, 15/SC/0526, 17/SC/0588). All participants provided informed consent. ELSA data are publicly accessible; data were accessed via UK Data Service (project ID: 222747). The authors assert that all procedures contributing to this work comply with the ethical standards of the relevant national and institutional committees on human experimentation and with the Helsinki Declaration of 1975, as revised in 2008.


Full details of ELSA methods are described in Steptoe et al. (Reference Steptoe, Breeze, Banks and Nazroo2013). ELSA comprises ‘core members’ and their partners as participants. Core members, intended to be nationally representative, were sourced from their prior participation in the Health Survey for England (HSE), a national survey monitoring health and care trends (NHS Digital, 2022). The original core members, selected from the HSE carried out in 1998, 1999 and 2001, were eligible for ELSA if they agreed to follow-up during their HSE participation, were born prior to 1952, and lived in private accommodation. Currently, core members and their partners are followed-up every two years. Refresher samples of adults aged ⩾50 years (and their partners) are added to the study periodically to ensure continuing representativeness. To ensure maximal sample size, we included ELSA participants (core members and partners) who took part in at least one interview data collection wave between 2002 and 2019 (ELSA waves 1–9). Trained interviewers carried out fieldwork interviews in participants’ homes (or institutions, if that was where a participant was now residing). Where necessary, interviews were carried out with proxies (e.g. if the participant lacked mental capacity or was a hospital inpatient during data collection).

Outcomes and exposures

Stroke and psychosis were each considered as main outcomes and as exposures in separate analyses. At each wave, participants were asked if a doctor had ever diagnosed them with ‘a stroke (cerebral vascular disease)’ as part of a structured health assessment. Free-text responses indicative of stroke were coded as stroke by the interviewer, where necessary. Analysis of the relationship between self-report and medically verified strokes in ELSA's sister study, the Health and Retirement Study, replicated known risk factor associations and indicated misreporting was non-systematic, indicating that participant-report can be used to study stroke risk (Glymour & Avendano, Reference Glymour and Avendano2009). However, due to the longitudinal nature of the study, we added a second validation step by checking that earlier wave self-reported strokes were also self-reported in later waves. If there was a discrepancy between an earlier self-reported stroke and a later self-report that indicated misdiagnosis or misreporting, then their stroke data for the previous wave was recoded to no stroke. Data on age at first stroke and number of stroke recurrences (0, 1, 2, or 3+) were also reported; age at first stroke was calculated as the youngest reported stroke age, and number of stroke recurrences calculated as the sum of total number of reported stroke recurrences, across all waves. Strokes (‘stroke/cerebral haemorrhage/cerebral thrombosis’) reported by participants in the HSE, prior to their ELSA participation, were also included in this study (this data was available only for participants who joined ELSA from HSE survey years between 1998 and 2006).

At each ELSA wave, participants were also asked if a doctor had ever diagnosed them with ‘any emotional, nervous or psychiatric problems.’ Participants responding affirmatively to this screening question were then probed to identify which of the following problems they had; hallucinations, anxiety, depression, emotional problems, schizophrenia, psychosis, mood swings, and/or manic depression. Psychosis was defined in this study as either a report of ‘hallucinations’, ‘schizophrenia’ or ‘psychosis’.


In addition to unadjusted analyses, two adjusted models were reported. Age, sex, ethnicity and baseline socioeconomic status are independent predictors of stroke (Avan et al., Reference Avan, Digaleh, Di Napoli, Stranges, Behrouz, Shojaeianbabaei and Azarpazhooh2019; Stansbury, Jia, Williams, Vogel, & Duncan, Reference Stansbury, Jia, Williams, Vogel and Duncan2005) and psychosis (Castillejos, Martín-Pérez, & Moreno-Küstner, Reference Castillejos, Martín-Pérez and Moreno-Küstner2018; Jongsma, Turner, Kirkbride, & Jones, Reference Jongsma, Turner, Kirkbride and Jones2019; Solmi et al., Reference Solmi, Radua, Olivola, Croce, Soardo, Salazar de Pablo and Fusar-Poli2022) and so were included in model 1 that adjusted for demographic characteristics; namely, age at wave 1 (calculated by subtracting the year of birth from 2002), sex, ethnicity (white compared with other ethnicities) and baseline net financial wealth. Where a participant had multiple ethnicities recorded across waves, the most frequent was used (or the most recent, if frequencies were equal). Quintile of net financial wealth, defined as participants’ gross financial wealth minus debt, was included as a proxy for socioeconomic status, as is common in studies of older adults and consistent with a previous study using ELSA data (Poole & Steptoe, Reference Poole and Steptoe2018).

The second model (model 2) additionally adjusted for baseline health behaviours – cigarette smoking status, level of vigorous physical activity, and level of alcohol consumption – due to evidence for their pre-onset association with stroke (Lee, Folsom, & Blair, Reference Lee, Folsom and Blair2003; Patra et al., Reference Patra, Taylor, Irving, Roerecke, Baliunas, Mohapatra and Rehm2010; Peters, Huxley, & Woodward, Reference Peters, Huxley and Woodward2013) and psychosis (Degenhardt et al., Reference Degenhardt, Saha, Lim, Aguilar-Gaxiola, Al-Hamzawi and Alonso2018; Mustonen et al., Reference Mustonen, Ahokas, Nordström, Murray, Mäki, Jääskeläinen and Niemelä2018; Stubbs et al., Reference Stubbs, Koyanagi, Schuch, Firth, Rosenbaum, Gaughran and Vancampfort2017). Smoking status referred to whether the participant currently smoked at the time of data collection. Frequency of vigorous physical activity, in reference to sports and activities in daily life, was considered according to the following levels: ‘hardly ever or never’, ‘1–3 times a month’, ‘once a week’ or ‘more than once a week’. Frequency of alcohol consumption referred to the last 12 months; different responses scales were used across waves, these were recoded to ‘daily/almost daily’, ‘1–4 times/week’, ‘monthly’, ‘rarely/special occasions only’ or ‘not at all’ for consistency.

In addition to outcomes/exposures and covariates, participants were described in terms of ever-reporting depression or anxiety across waves 1–9, study participation (i.e. number of waves participated, whether took part in all waves) and whether participants died prior to wave 6 (mortality status was well characterised only up to this time-point).


Data were anchored to the first wave that participants took part in (e.g. if a participant joined ELSA at wave 4, then wave 4 was taken as their baseline). Means with standard deviations (s.d.), medians with interquartile ranges (IQR), and/or counts and proportions were used for descriptive statistics, as appropriate. Follow-up times were converted from number of waves to years for interpretability.

Logistic regression was used to investigate the associations between stroke and psychosis, with unadjusted and adjusted (see Covariates) odds ratios (OR) and associated 95% CIs, reported after 4 and 10 years of follow-up. Kaplan–Meier probability estimates (with 95% CIs) and survival curves were used to summarise and illustrate the cumulative incidence of first-reported stroke in participants with and without psychosis up to 10 years. The log-rank test was used to test differences in being stroke-free at the end of follow-up between the groups. Cox proportional hazards regression was used to estimate the association between psychosis and the hazard of stroke; unadjusted and adjusted (see Covariates) HR with 95% CIs were reported. In participants reporting psychosis, the start of follow-up was defined as the wave that they reported psychosis for the first time. For those not reporting psychosis, the start of follow-up was the first wave that they participated in. The end of follow-up was defined as either the wave that a stroke was first reported (in those reporting a stroke) or the last wave that the participant took part in (in those not reporting stroke). Follow-up time was set to zero if first stroke was reported prior to psychosis and if stroke and psychosis were first reported during the same wave. The proportional hazards assumption was checked visually and by review of the Schoenfeld global test. The same methods were used to investigate the effect of stroke on incidence/hazard of psychosis. All adjusted survival models included an individual ‘frailty term’ (Hougaard, Reference Hougaard1995) to account for individual random effects.

Adjusted results are reported following multiple imputation as well as from complete-case analysis. Random forest multiple imputation, implemented using the missForest package in R, was used to impute values for missing baseline covariates (outcomes were not imputed). This method was used as it allows for simultaneous imputation of numeric and categorical variables, it does not rely on distributional assumptions – allowing for complex interactions and non-linear relations between variables, and has been shown to outperform several other imputation methods (Stekhoven & Bühlmann, Reference Stekhoven and Bühlmann2012). Baseline, auxiliary and outcome variables were included in the imputation model and we reviewed levels and patterns of missing data to assess plausibility of the missing at random assumption (Sterne et al., Reference Sterne, White, Carlin, Spratt, Royston, Kenward and Carpenter2009).

Analyses were conducted using R version 4.2.1 (R Core Team, 2020) and the full code and output for the analysis is available in the format of a Jupyter Notebook (Rule et al., Reference Rule, Birmingham, Zuniga, Altintas, Huang, Knight and Rose2019), a document that combines both code and the output in a form that can be re-run and reproduced. All code and output is available on the online archive:



Demographics and descriptive statistics for the sample are shown in Table 1. Of 19 808 participants who took part in at least one wave between 2002 and 2019, a total of 1279 (6.5%) reported stroke and 150 (0.8%) reported psychosis at least once. A total of 45 participants had a record of a stroke recoded to ‘no stroke’ due to the participant stating their previous report was either a ‘misdiagnosis’ (n = 17) or that they ‘never had’ the stroke (n = 28), 19 of which later went on to report an undisputed stroke in a later wave and ultimately contributed stroke data to this study. The level of missing baseline data was low, with most variables either complete or with less than 5% missing values (Table 1) – except for level of alcohol use (10% missing), for which we deemed the missing at random assumption plausible when conditioned on other variables included in the imputation model (online Supplementary Table 1).

Table 1. Demographics and descriptive statistics for sample

a Median (IQR).

Twenty-four participants reported both stroke and psychosis at any point across waves – equating to 1.9% of the stroke population and 16.0% of the psychosis population. Participants reporting both stroke and psychosis tended to be older than those reporting only psychosis but younger than those reporting only stroke. Almost three quarters (71%) of the stroke and psychosis group were female and half (50%) were in the lowest quintile of net financial wealth (double the proportion of those reporting just stroke or no stroke/psychosis at all). This group also had the lowest levels of baseline vigorous physical activity and highest level of alcohol use in the last 12 months, whilst the psychosis only group had the highest proportion of cigarette smokers. A high proportion of participants reporting psychosis also reported depression and anxiety, whereas the stroke only group were more similar to those without stroke or psychosis in this regard (online Supplementary Table 2). Compared to those reporting stroke only, first stroke tended to occur earlier in participants who reported both stroke and psychosis (median age, 66 v. 69) and the mean total number of stroke recurrences was higher (2.2 v. 1.1) (online Supplementary Table 2).

The rate of follow-up across waves was high – participants with stroke and psychosis or psychosis only participated in a median of six (out of nine) waves, compared with five waves for those reporting stroke only (lower likely due to the higher mortality rate in the latter group) (online Supplementary Table 3).

Stroke risk after psychosis

After 4 years of follow-up, the odds of stroke in people reporting psychosis were higher than in those without psychosis (unadjusted OR, 3.99, 95% CI 2.39–6.34) (Table 2). Following multiple imputation to impute missing covariates (missing data is illustrated in Table 1) and adjustment, the ORs increased, and was highest in model 1 (adjusted for age, sex, ethnicity, and net financial wealth) (OR 5.40, 95% CI 3.15–8.83). By 10 years, the ORs had reduced slightly, but psychosis was still significantly associated with an increased risk of stroke – resulting in an OR of 3.12 (95% CI 1.92–4.87) in the model additionally adjusting for smoking status, level of vigorous physical activity and level of alcohol use (model 2).

Table 2. Odds ratios estimates for stroke risk after psychosis and psychosis risk after stroke

OR, Odds ratio. All models were significant at the p < 0.001 level. Model 1: adjusted for age, sex, ethnicity, quintile of net financial wealth + participant (frailty). Model 2: adjusted for age, sex, ethnicity, quintile of net financial wealth, smoking status, level of alcohol use in the past 12 months, level of vigorous physical activity + participant (frailty). Adjusted models are reported following multiple imputation of missing baseline covariates.

The cumulative incidence of first-reported stroke, stratified by psychosis status and up to 10 years, is shown in Fig. 1. By 10 years, the probability of reporting an incident first stroke in the psychosis group was 21.4% (95% CI 12.1–29.6) compared to 8.3% (95% CI 7.8–8.8) in those without psychosis (absolute difference: 13.1%, 95% CI 20.8–4.3, log rank test, p < 0.001), equating to an overall unadjusted HR of 2.87 (95% CI 1.91–4.29) (Table 3). When adjusted for baseline demographic covariates, the HR increased to 3.84 (95% CI 2.43–6.07) and when additionally adjusted for baseline health behaviour covariates (smoking status, level of vigorous physical activity and level of alcohol use) the HR reduced slightly to 3.57, with a similar 95% CI (2.18–5.84).

Fig. 1. The cumulative incidence of stroke in participants reporting and not reporting psychosis. Log rank test p = <0.001.

Table 3. Hazard ratios estimates for stroke risk after psychosis and psychosis risk after stroke

1HR, Hazard ratio. All models were significant at the p < 0.001 level. Model 1: adjusted for age, sex, ethnicity, quintile of net financial wealth + participant (frailty). Model 2: adjusted for age, sex, ethnicity, quintile of net financial wealth, smoking status, level of alcohol use in the past 12 months, level of vigorous physical activity + participant (frailty). Adjusted models are reported following multiple imputation of missing baseline covariates.

Psychosis risk after stroke

After 4 years of follow-up, and following adjustment for baseline demographic covariates, stroke was associated with increased odds of reporting psychosis compared to people without a stroke (adjusted OR, 4.94, 95% CI 2.87–8.12). When additionally adjusted for baseline health behaviour covariates, the OR reduced to 4.15 (95% CI 2.42–6.83). By 10 years, the ORs had reduced, but continued to indicate that stroke was associated with psychosis (fully adjusted OR, 2.81, 95% CI 1.73–4.40) (Table 2).

The cumulative incidence of first-reported psychosis, stratified by stroke status and up to 10 years, is shown in Fig. 2. By 10 years, the probability of reporting incident psychosis in the stroke group was 2.3% (95% CI 1.4–3.2) compared to 0.9% (95% CI 0.7–1.1) in those without stroke (absolute difference: 1.4%, 95% CI 0.7–2.1, log rank test p = <0.001), equating to an unadjusted HR of 3.08 (95% CI 1.99–4.77) (Table 3). When adjusted for baseline demographic covariates, the HR increased to 4.91 (95% CI 2.65–9.08), and remained similar when additionally adjusting for baseline health behaviour covariates (4.98, 95% CI 2.55–9.72).

Fig. 2. The cumulative incidence of psychosis in those reporting and not reporting stroke. Log rank test p = <0.001.

ORs and HRs from complete-case analysis were broadly similar to those obtained following multiple imputation but tended to result in slightly smaller estimates (online Supplementary Tables 4–5). The Schoenfeld global test suggested that the proportional hazards assumption was met in all cox regression models (online Supplementary Table 6).


We report increased odds and hazards of stroke in people with psychosis (and vice versa) over 10 years of follow-up in a representative longitudinal study of adults aged 50 years and over in England. Stroke was an independent predictor of psychosis, and psychosis an independent predictor of stroke, over and above the effects of covariates in adjusted analyses. Stroke-first individuals had a higher prevalence of psychosis at baseline and a higher risk of subsequent psychosis, and psychosis-first individuals showed a similar pattern with regard to later stroke occurrence. However, the risks were not symmetrical: there was a greater risk of psychosis after stroke (fully adjusted HR, 4.98; 95% CI 2.55–9.72) than stroke after psychosis (fully adjusted HR, 3.57; 95% CI 2.18–5.84). These results contribute to our understanding of stroke-psychosis comorbidity, indicating that each raises the risk of the other and, in the context of an ageing population, it is likely the prevalence of this serious comorbidity will increase over time.

These results provide additional evidence for bidirectional causality between stroke and psychosis, although specific causal mechanisms have, to date, been better characterised for those who develop stroke after psychosis than vice versa. For stroke after psychosis, antipsychotic medication is a well-established risk factor, likely due to metabolic side-effects increasing stroke risk (Zivkovic, Koh, Kaza, & Jackson, Reference Zivkovic, Koh, Kaza and Jackson2019). In addition, psychosis may lead to changes in several key health behaviours such as increased smoking, alcohol use, or reduced physical activity levels that worsen cardiovascular health (Rodrigues, Wiener, Stranges, Ryan, & Anderson, Reference Rodrigues, Wiener, Stranges, Ryan and Anderson2021). Indeed, cardiovascular risk factors are particularly prevalent among people with diagnoses of major psychotic disorders and become more prevalent as illness duration increases (Mitchell, Vancampfort, De Herdt, Yu, & De Hert, Reference Mitchell, Vancampfort, De Herdt, Yu and De Hert2013). Causal mechanisms for psychosis after stroke are less well-evidenced. Right-hemisphere lesions have long-been hypothesised as important causal factors although concerns about the extent to which accurate diagnosis can be made in those with language-impairing left-hemisphere strokes remain (Stangeland et al., Reference Stangeland, Orgeta and Bell2018). Other suggested risk factors, albeit based on relatively small studies, include more extensive cerebral impairment (Rabins, Starkstein, & Robinson, Reference Rabins, Starkstein and Robinson1991), a longer period of institutional care (Buijck, Zuidema, Spruit-van Eijk, Geurts, & Koopmans, Reference Buijck, Zuidema, Spruit-van Eijk, Geurts and Koopmans2012), and night-time behavioural disturbances (van Almenkerk, Depla, Smalbrugge, Eefsting, & Hertogh, Reference van Almenkerk, Depla, Smalbrugge, Eefsting and Hertogh2012). The relatively small number of studies in this area highlights a need for more extensive studies to identify predictors, and ideally, modifiable risk factors for psychosis after stroke.

This study has several important strengths, including using stratified sampling to be broadly representative of the population of England aged 50 years and over, stroke reporting having been validated by comparison with medical diagnosis in the Health and Retirement study (Glymour & Avendano, Reference Glymour and Avendano2009), and the ability to examine the occurrence of psychosis in stroke-first participants, and stroke in psychosis-first participants, within the same cohort.

However, we note several important study characteristics that are important when interpreting this study. There was a slightly lower reported population prevalence of psychosis in people without stroke reported in this study (0.9%) compared to a meta-analytic estimate of 1.7% in adults in the same 50 years-and-older age range (Volkert, Schulz, Härter, Wlodarczyk, & Andreas, Reference Volkert, Schulz, Härter, Wlodarczyk and Andreas2013). In this study, psychosis was measured by a two-stage screening process: individuals were asked if they had been diagnosed with mental health problems, were given a list of conditions if they concurred, and were coded as having psychosis if they indicated ‘hallucinations’, ‘schizophrenia’ or ‘psychosis’. Unlike previous studies that measured psychosis with standardised diagnostic interviews (Bell et al., Reference Bell, Tamayo-Agudelo, Revill, Okai and Poole2023) or relied on clinical records of diagnoses (e.g. Fleetwood et al., Reference Fleetwood, Wild, Smith, Mercer, Licence, Sudlow and Jackson2021), it is possible that this form of self-report is more likely to under-report psychosis due to stigma, lack of diagnosis, or a combination. Indeed, community rates of psychosis in active case ascertainment studies are 2–3 times higher than the rates of diagnosed psychosis (Anderson et al., Reference Anderson, Norman, MacDougall, Edwards, Palaniyappan, Lau and Kurdyak2019; Hogerzeil, van Hemert, Rosendaal, Susser, & Hoek, Reference Hogerzeil, van Hemert, Rosendaal, Susser and Hoek2014), suggesting a significant proportion of undiagnosed psychotic disorders that would be missed even if self-reporting of diagnosis perfectly reflected rates of diagnosis. In addition, due to the episodic nature of psychosis, we were not able to include additional data integrity measures that we did for stroke – namely, removing any cases of stroke where the person reported a stroke diagnosis at one time point, but did not report it at later time points.

We also note a prevalence of psychosis in stroke population of 1.9% and a prevalence of stroke in psychosis population of 16.0%. This compares to recent estimates of a prevalence of psychosis in the stroke population of 3.15% (international) and 1.1% (United Kingdom) and a prevalence of stroke in the psychosis population 3.81% (international) and 3.03% (United Kingdom) (Bell et al., Reference Bell, Tamayo-Agudelo, Revill, Okai and Poole2023). There may be several factors contributing to these differences. One may be the relatively small number of stroke and psychosis cases identified, meaning that population estimates may be affected by relatively small differences in case numbers. The under-reporting of psychosis, as discussed above, is likely to be a factor and may particularly contribute to the much higher estimated prevalence of stroke in psychosis than psychosis in stroke. However, it is also worth noting that the previous estimates are drawn from whole population studies and this study is based on the ELSA cohort that includes adults aged 50 and older only and it is currently not clear to what extent prevalence may differ by age cohort. Younger adults with strokes make up 10–15% of the total stroke population (Smajlović, Reference Smajlović2015) with the prevalence of strokes rapidly increasing in this age cohort (George, Tong, & Bowman, Reference George, Tong and Bowman2017). Given the typically younger age of onset of psychosis, it is possible that younger individuals with stroke-psychosis may have distinct characteristics from older adults with the same and this remains a topic for further investigation.

We also note that follow-up of the ELSA cohort was completed every two years and this may have missed cases of stroke or psychosis in individuals who developed either condition in the time since their last follow-up but before they died. In addition, it is not clear the extent to which stroke-psychosis comorbidity may have led to individuals being lost to follow-up, leading to selection bias due to increased attrition in this group, or potential interactions where, for example, stroke-related aphasia may have affected the diagnosis or reporting of psychosis (Stangeland et al., Reference Stangeland, Orgeta and Bell2018). We used the wave that stroke or psychosis were reported as a proxy for when these conditions first occurred; analyses using dates of onset instead would have been more precise, but these were not available.

One potentially curious feature of the analysis was that risk slightly increased after additionally controlling for several potential demographic and health-related confounders. This effect was largely down to frequency of alcohol consumption where those reporting more frequent consumption of alcohol showed lower risk of stroke and psychosis. This finding, that increased frequency of alcohol consumption is associated with better health in the ELSA cohort, has been the focus of dedicated studies (Frisher et al., Reference Frisher, Mendonça, Shelton, Pikhart, de Oliveira and Holdsworth2015; Kojima et al., Reference Kojima, Jivraj, Iliffe, Falcaro, Liljas and Walters2019) that have indicated that these effects may be explained both by poorer baseline health in non-drinkers and generally lower levels of alcohol consumption by volume in adults as they age reflecting a higher prevalence of ‘light, frequent drinkers’.

There are some additional factors we were not able to examine in this study which are likely to be important in understanding stroke-psychosis comorbidity. Data for antipsychotic prescribing were not available at baseline and there is evidence that certain antipsychotic medications may increase the risk of mortality and further stroke morbidity in those with stroke and psychosis (Nielsen, Banner, & Jensen, Reference Nielsen, Banner and Jensen2021). In addition, data on stroke type, severity and/or location were not available, meaning it was not possible to examine the bidirectional association between stroke characteristics and psychosis. As our focus was on stroke-psychosis comorbidity, we did not study the incidence of other cardiovascular conditions and diseases (e.g. hypertension, myocardial infarction) which also play a role in stroke risk and could represent competing risks. Whether there are shared genetic pathways to stroke and psychosis also remains unexplored.

In conclusion, we report that there is a bidirectional increase in risk for stroke-psychosis comorbidity. There was a greater risk of psychosis after stroke than stroke after psychosis. However, due to the likely under-reporting of psychosis, there is a pressing need for studies that use gold standard stroke and psychosis diagnostic methods within the same large-scale multi-year follow-up study to examine this issue in more detail.

Supplementary material

The supplementary material for this article can be found at


ELSA data are publicly accessible; data were accessed via UK Data Service (project ID: 222747). Data cannot be shared by the authors of this article but can be obtained directly from the UK Data Service (


The ELSA was developed by a team of researchers based at University College London, NatCen Social Research, the Institute for Fiscal Studies, the University of Manchester and the University of East Anglia. The data were collected by NatCen Social Research. The funding is currently provided by the National Institute on Aging in the US, and a consortium of UK government departments coordinated by the National Institute for Health Research. Funding has also been received by the Economic and Social Research Council.

Financial support

ARB is funded by the Wellcome Trust through a PhD Fellowship in Mental Health Science (218497/Z/19/Z). This research was funded in whole or in part by the Wellcome Trust. For the purpose of Open Access, the author has applied a CC BY public copyright licence to any Author Accepted Manuscript version arising from this submission.

DPJO is supported by the University College London Hospitals NIHR Biomedical Research Centre and the NIHR North Thames Applied Research Collaboration. This funder had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The views expressed in this article are those of the authors and not necessarily those of the NHS, the NIHR, or the Department of Health and Social Care.

Conflict of interest

There are no conflicts of interest.


Anderson, K. K., Norman, R., MacDougall, A. G., Edwards, J., Palaniyappan, L., Lau, C., & Kurdyak, P. (2019). Estimating the incidence of first-episode psychosis using population-based health administrative data to inform early psychosis intervention services. Psychological Medicine, 49(12), 20912099. ScholarPubMed
Avan, A., Digaleh, H., Di Napoli, M., Stranges, S., Behrouz, R., Shojaeianbabaei, G., … Azarpazhooh, M. R. (2019). Socioeconomic status and stroke incidence, prevalence, mortality, and worldwide burden: An ecological analysis from the Global Burden of Disease Study 2017. BMC Medicine, 17(1), 191. ScholarPubMed
Bell, V., Tamayo-Agudelo, W., Revill, G., Okai, D., & Poole, N. (2023). Association between stroke and psychosis across four nationally representative psychiatric epidemiological studies. BJPsych Open, 9(3), E71. ScholarPubMed
Buijck, B. I., Zuidema, S. U., Spruit-van Eijk, M., Geurts, A. C., & Koopmans, R. T. (2012). Neuropsychiatric symptoms in geriatric patients admitted to skilled nursing facilities in nursing homes for rehabilitation after stroke: A longitudinal multicenter study. International Journal of Geriatric Psychiatry, 27(7), 734741. ScholarPubMed
Castillejos, M. C., Martín-Pérez, C., & Moreno-Küstner, B. (2018). A systematic review and meta-analysis of the incidence of psychotic disorders: The distribution of rates and the influence of gender, urbanicity, immigration and socio-economic level. Psychological Medicine, 48(13), 21012115. ScholarPubMed
Degenhardt, L., Saha, S., Lim, C. C. W., Aguilar-Gaxiola, S., Al-Hamzawi, A., & Alonso, J., … Collaborators, on behalf of the W. W. M. H. S. (2018). The associations between psychotic experiences and substance use and substance use disorders: Findings from the World Health Organization World Mental Health surveys. Addiction, 113(5), 924934. ScholarPubMed
Fleetwood, K., Wild, S. H., Smith, D. J., Mercer, S. W., Licence, K., Sudlow, C. L. M., … Jackson, C. A. (2021). Association of severe mental illness with stroke outcomes and process-of-care quality indicators: Nationwide cohort study. The British Journal of Psychiatry, 221(1), 394401. Scholar
Frisher, M., Mendonça, M., Shelton, N., Pikhart, H., de Oliveira, C., & Holdsworth, C. (2015). Is alcohol consumption in older adults associated with poor self-rated health? Cross-sectional and longitudinal analyses from the English Longitudinal Study of Ageing. BMC Public Health, 15(1), 703. ScholarPubMed
George, M. G., Tong, X., & Bowman, B. A. (2017). Prevalence of cardiovascular risk factors and strokes in younger adults. JAMA Neurology, 74(6), 695703. ScholarPubMed
Glymour, M. M., & Avendano, M. (2009). Can self-reported strokes be used to study stroke incidence and risk factors? Stroke, 40(3), 873879. ScholarPubMed
Hogerzeil, S. J., van Hemert, A. M., Rosendaal, F. R., Susser, E., & Hoek, H. W. (2014). Direct comparison of first-contact versus longitudinal register-based case finding in the same population: Early evidence that the incidence of schizophrenia may be three times higher than commonly reported. Psychological Medicine, 44(16), 34813490. ScholarPubMed
Hougaard, P. (1995). Frailty models for survival data. Lifetime Data Analysis, 1(3), 255273. ScholarPubMed
John, A., McGregor, J., Jones, I., Lee, S. C., Walters, J. T. R., Owen, M. J., … Lloyd, K. (2018). Premature mortality among people with severe mental illness—New evidence from linked primary care data. Schizophrenia Research, 199, 154162. ScholarPubMed
Jongsma, H. E., Turner, C., Kirkbride, J. B., & Jones, P. B. (2019). International incidence of psychotic disorders, 2002–17: A systematic review and meta-analysis. The Lancet Public Health, 4(5), e229e244. ScholarPubMed
Kisely, S., Campbell, L. A., & Wang, Y. (2009). Treatment of ischaemic heart disease and stroke in individuals with psychosis under universal healthcare. The British Journal of Psychiatry, 195(6), 545550. ScholarPubMed
Kojima, G., Jivraj, S., Iliffe, S., Falcaro, M., Liljas, A., & Walters, K. (2019). Alcohol consumption and risk of incident frailty: The English longitudinal study of aging. Journal of the American Medical Directors Association, 20(6), 725729. ScholarPubMed
Lee, C. D., Folsom, A. R., & Blair, S. N. (2003). Physical activity and stroke risk. Stroke, 34(10), 24752481. ScholarPubMed
Li, M., Fan, Y.-L., Tang, Z.-Y., & Cheng, X.-S. (2014). Schizophrenia and risk of stroke: A meta-analysis of cohort studies. International Journal of Cardiology, 173(3), 588590. ScholarPubMed
Mitchell, A. J., Vancampfort, D., De Herdt, A., Yu, W., & De Hert, M. (2013). Is the prevalence of metabolic syndrome and metabolic abnormalities increased in early schizophrenia? A comparative meta-analysis of first episode, untreated and treated patients. Schizophrenia Bulletin, 39(2), 295305. ScholarPubMed
Morgan, V. A., Waterreus, A., Ambrosi, T., Badcock, J. C., Cox, K., Watts, G. F., … Jablensky, A. (2021). Mental health recovery and physical health outcomes in psychotic illness: Longitudinal data from the Western Australian survey of high impact psychosis catchments. Australian & New Zealand Journal of Psychiatry, 55(7), 711728. ScholarPubMed
Mustonen, A., Ahokas, T., Nordström, T., Murray, G. K., Mäki, P., Jääskeläinen, E., … Niemelä, S. (2018). Smokin’ hot: Adolescent smoking and the risk of psychosis. Acta Psychiatrica Scandinavica, 138(1), 514. ScholarPubMed
Nemani, K., & Gurin, L. (2021). Neuropsychiatric complications after stroke. Seminars in Neurology, 41(1), 85100. ScholarPubMed
NHS Digital. (2022). Health Survey for England. Retrieved May 9, 2022, from NHS Digital website: Scholar
Nielsen, R. E., Banner, J., & Jensen, S. E. (2021). Cardiovascular disease in patients with severe mental illness. Nature Reviews Cardiology, 18(2), 136145. ScholarPubMed
Patra, J., Taylor, B., Irving, H., Roerecke, M., Baliunas, D., Mohapatra, S., & Rehm, J. (2010). Alcohol consumption and the risk of morbidity and mortality for different stroke types—A systematic review and meta-analysis. BMC Public Health, 10(1), 258. ScholarPubMed
Peters, S. A. E., Huxley, R. R., & Woodward, M. (2013). Smoking as a risk factor for stroke in women compared with men: A systematic review and meta-analysis of 81 cohorts, including 3980359 individuals and 42401 strokes. Stroke, 44(10), 28212828. Scholar
Poole, L., & Steptoe, A. (2018). Depressive symptoms predict incident chronic disease burden 10years later: Findings from the English Longitudinal Study of Ageing (ELSA). Journal of Psychosomatic Research, 113, 3036. Scholar
Rabins, P. V., Starkstein, S. E., & Robinson, R. G. (1991). Risk factors for developing atypical (schizophreniform) psychosis following stroke. The Journal of Neuropsychiatry and Clinical Neurosciences, 3(1), 69. ScholarPubMed
R Core Team. (2020). R: A language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing. Retrieved from Scholar
Rodrigues, M., Wiener, J. C., Stranges, S., Ryan, B. L., & Anderson, K. K. (2021). The risk of physical multimorbidity in people with psychotic disorders: A systematic review and meta-analysis. Journal of Psychosomatic Research, 140, 110315. ScholarPubMed
Rule, A., Birmingham, A., Zuniga, C., Altintas, I., Huang, S.-C., Knight, R., … Rose, P. W. (2019). Ten simple rules for writing and sharing computational analyses in Jupyter Notebooks. PLOS Computational Biology, 15(7), e1007007. ScholarPubMed
Smajlović, D. (2015). Strokes in young adults: Epidemiology and prevention. Vascular Health and Risk Management, 11, 157164. ScholarPubMed
Solmi, M., Radua, J., Olivola, M., Croce, E., Soardo, L., Salazar de Pablo, G., … Fusar-Poli, P. (2022). Age at onset of mental disorders worldwide: Large-scale meta-analysis of 192 epidemiological studies. Molecular Psychiatry, 27(1), 281295. ScholarPubMed
Stangeland, H., Orgeta, V., & Bell, V. (2018). Poststroke psychosis: A systematic review. Journal of Neurology, Neurosurgery & Psychiatry, 89(8), 879885. ScholarPubMed
Stansbury, J. P., Jia, H., Williams, L. S., Vogel, W. B., & Duncan, P. W. (2005). Ethnic disparities in stroke. Stroke, 36(2), 374386. ScholarPubMed
Stekhoven, D. J., & Bühlmann, P. (2012). MissForest—Non-parametric missing value imputation for mixed-type data. Bioinformatics (Oxford, England), 28(1), 112118. ScholarPubMed
Steptoe, A., Breeze, E., Banks, J., & Nazroo, J. (2013). Cohort profile: The English longitudinal study of ageing. International Journal of Epidemiology, 42(6), 16401648. ScholarPubMed
Sterne, J. A. C., White, I. R., Carlin, J. B., Spratt, M., Royston, P., Kenward, M. G., … Carpenter, J. R. (2009). Multiple imputation for missing data in epidemiological and clinical research: Potential and pitfalls. BMJ, 338, b2393. ScholarPubMed
Stroke Association. (2020, October 7). Stroke statistics. Retrieved May 3, 2022, from Stroke Association website: Scholar
Stubbs, B., Koyanagi, A., Schuch, F., Firth, J., Rosenbaum, S., Gaughran, F., … Vancampfort, D. (2017). Physical activity levels and psychosis: A mediation analysis of factors influencing physical activity target achievement among 204 186 people across 46 low- and middle-income countries. Schizophrenia Bulletin, 43(3), 536545. ScholarPubMed
van Almenkerk, S., Depla, M. F. I. A., Smalbrugge, M., Eefsting, J. A., & Hertogh, C. M. P. M. (2012). Institutionalized stroke patients: Status of functioning of an under researched population. Journal of the American Medical Directors Association, 13(7), 634639. ScholarPubMed
Volkert, J., Schulz, H., Härter, M., Wlodarczyk, O., & Andreas, S. (2013). The prevalence of mental disorders in older people in Western countries – a meta-analysis. Ageing Research Reviews, 12(1), 339353. ScholarPubMed
Willers, C., Sunnerhagen, K. S., Lekander, I., & von Euler, M. (2018). The association of pre-stroke psychosis and post-stroke levels of health, resource utilization, and care process: A register-based study. Frontiers in Neurology, 9, 1042. ScholarPubMed
Yuan, M., Xiao, Z.-L., Zhou, H.-Y., Rao, W., Huang, G., Nie, H.-B., … Xu, R.-S. (2022). Bipolar disorder and the risk for stroke incidence and mortality: A meta-analysis. Neurological Sciences, 43(1), 467476. ScholarPubMed
Zivkovic, S., Koh, C. H., Kaza, N., & Jackson, C. A. (2019). Antipsychotic drug use and risk of stroke and myocardial infarction: A systematic review and meta-analysis. BMC Psychiatry, 19(1), 189. ScholarPubMed
Figure 0

Table 1. Demographics and descriptive statistics for sample

Figure 1

Table 2. Odds ratios estimates for stroke risk after psychosis and psychosis risk after stroke

Figure 2

Fig. 1. The cumulative incidence of stroke in participants reporting and not reporting psychosis. Log rank test p = <0.001.

Figure 3

Table 3. Hazard ratios estimates for stroke risk after psychosis and psychosis risk after stroke

Figure 4

Fig. 2. The cumulative incidence of psychosis in those reporting and not reporting stroke. Log rank test p = <0.001.

Supplementary material: File

Richards-Belle et al. supplementary material
Download undefined(File)
File 41.8 KB