Prevalence of depression and cognitive impairment and their inter-relationship and association with quality of life among older stroke survivors: the findings of a national survey in China

Abstract

Background

Post-stroke depression (PSD) and post-stroke cognitive impairment (PSCI) are prevalent neuropsychiatric problems that are associated with high disability burden and low quality of life (QoL). This study explored the PSD-PSCI network, along with the interaction and association with QoL among Chinese older stroke survivors.

Methods

Data from the 2017–2018 wave of the Chinese Longitudinal Healthy Longevity Survey were obtained to investigate the inter-relationship between PSD and PSCI among older stroke survivors. Central and bridge symptoms within the PSD-PSCI network and their association with QoL were explored. Depressive symptoms, cognitive impairment and QoL were measured using the 10-item Center for Epidemiologic Studies Depression Scale (CESD-10), Mini-Mental State Examination and the WHO QoL-brief version, respectively.

Results

The prevalence of PSD and PSCI among older stroke survivors was 31.5% and 22.1%, respectively. In the PSD-PSCI network, ‘feeling blue/depressed’ (CESD3, strength: 1.117) and ‘Attention and calculation’ (At_C, strength: 0.972) were the most influential symptoms, while ‘Naming’ (Nam, bridge strength: 0.175) was the most significant bridge symptom. Notably, ‘Sleep disturbances’ (CESD10) had the strongest association with lower QoL.

Conclusion

This study revealed that both PSD and PSCI were prevalent among older stroke survivors. The key central and bridge symptoms in the PSD-PSCI network, along with those symptoms that negatively impact on QoL, should be prioritised in targeted interventions to enhance treatment outcomes in this population.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Post-stroke depression (PSD) and post-stroke cognitive impairment (PSCI) are prevalent and have a significant impact on the quality of life (QoL) of stroke survivors. However, the interrelationship between these conditions and their impact on QoL, especially in older stroke survivors in China, remains unclear.

WHAT THIS STUDY ADDS

• This study conducted a network approach to determine central symptoms, such as ‘Feeling blue/depressed’ and ‘Attention and calculation’ and bridge symptoms like ‘Naming’, which connect PSD and PSCI. We also found that ‘Sleep disturbances’ most strongly correlated with QoL.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• The findings emphasize the significance of targeting specific symptoms in clinical interventions to improve the management of PSD and PSCI in stroke survivors. This study may guide future research and the development of individualised therapeutic approaches to enhance rehabilitation and QoL.

Introduction

Stroke is one of the leading causes of mortality and long-term neurological disability among older adults globally.¹ As reported by the Global Burden of Disease Study,² the worldwide incidence and prevalence of stroke in 2019 were 12.2 million and 101 million, respectively, with 6.55 million fatalities and 5.7% of overall Disability-Adjusted Life Years, resulting from strokes. In China, the stroke prevalence reached 28.76 million in 2019.³ Sequelae of strokes, including physical dysfunction (eg, mobility challenges) and neuropsychiatric disorders (eg, post-stroke depression (PSD) and cognitive impairment), impose a heavy burden on the survivors, caregivers and healthcare systems.

Depression and cognitive decline are prevalent and challenging post-stroke consequences. A meta-analysis of 77 studies with 27 401 participants revealed a pooled PSD prevalence of 27% (95% CI 25% to 30%) among survivors at any stroke stage.⁴ Survivors commonly experience PSD within three to 6 months poststroke.⁵ They often exhibit increased disability and cognitive decline, decreased quality of life (QoL) and elevated mortality risk compared with their non-depressed counterparts. On the other hand, post-stroke cognitive impairment (PSCI) has an impact on up to two-thirds of stroke survivors in the initial year after stroke, with little cognitive improvement, contributing to mortality, disability and low QoL.⁶

There is a bidirectional relationship between PSD and PSCI, with survivors with PSD having significantly more PSCI compared with those without, and vice versa.⁶ A longitudinal stroke cohort study found co-occurrence of PSD and PSCI, with lower predictive markers such as serotonin and IL-6, and left-sided stroke being associated with PSCI in the subacute phase of the cerebrovascular event.⁷ Within PSCI, memory, non-verbal problem-solving, attention and psychomotor speed are the main impairments found in PSD, while dysphasia tends to increase the risk of PSD.⁸ PSD and PSCI not only impact on QoL but also contribute to caregiver burden. Understanding their inter-relationship is crucial to address the burden on families and healthcare systems. However, most studies have assessed PSD and PSCI using total or mean scores of standard rating scales, despite the multifaceted nature of these conditions. Thus, there has been insufficient exploration of individual PSD and PSCI symptoms in this population.

Network analysis represents a novel method for exploring the structure of psychopathology. Unlike traditional methods like regression and factor analysis, network approach provides a theoretical framework for conceptualising psychiatric disorders as systems of interlinked symptoms, enabling visualisation, analysis and examination. This approach emphasizes prioritising interventions by identifying central and bridge symptoms that substantially influence other symptoms within or across communities. Interventions targeting central symptoms may reduce overall network activation, thus potentially preventing the development of disorders. Furthermore, targeting bridge symptoms may prevent the spread of comorbidity between mental health problems. A previous network analysis study of PSD and PSCI identified ‘mood’, ‘concentration’ and ‘executive function’ as central symptoms and ‘psychomotor functioning’ and ‘attention’ as key bridge symptoms within the PSD-PSCI network using Patient Health Questionnaire-8 (PHQ-8) for PSD and the NIH Toolbox Cognitive Battery (NIHTCB) for PSCI.⁹ However, the small study sample size (N=202) might have led to insufficient statistical power. Limited research has explored the inter-relationship between PSD and PSCI, leaving a critical knowledge gap. A network approach to investigate the structures of PSD-PSCI networks and their associations with QoL among older stroke survivors is needed to develop effective treatments and preventive strategies for stroke survivors.

Thus, our study aimed to explore the inter-relationship across PSD and PSCI among older survivors in China using data from the Chinese Longitudinal Healthy Longevity Survey (CLHLS) and to assess the interactions between depressive symptoms and QoL.

Methods

Study design and participants

Initiated in 1998 by Peking University, the CLHLS is a nationwide community-based study conducted over eight waves (1998, 2000, 2002, 2005, 2008–2009, 2011–2012, 2014 and 2017–2018).¹⁰ The detailed methodologies have been published elsewhere.¹¹ In this study, we performed a secondary analysis using data from the 2017–2018 CLHLS (figure 1). Participants were recruited from those aged 65 years and above, with a history of stroke, who had completed assessments on depression and cognition and were able to understand Chinese. Individuals with severe cognitive impairment (eg, dementia) were excluded.

Figure 1. Flowchart of the study sample inclusion process. CESD-10, 10-item Center for Epidemiologic Studies Depression Scale; CLHLS, Chinese Longitudinal Healthy Longevity Survey; MMSE, Mini-Mental State Examination; WHOQOL-BREF, WHO Quality of Life-brief version.

The CLHLS data from the 2017 to 2018 wave are publicly available. Information on the data can be accessed at https://opendata.pku.edu.cn/file. xhtml?fileId=10 356&version=2.1. The CLHLS granted authorisation for the use of data prior to the commencement of this study.

Measures and assessments

Basic sociodemographic information, such as age, sex, marital status, educational attainment and yearly household income, was obtained. Depressive symptoms over the past two weeks were evaluated with the validated Chinese version of the Center for Epidemiologic Studies Depression Scale (CESD-10)¹² which consists of 10 items: (1) Feeling bothered; (2) Difficulty with concentrating; (3) Feeling blue/depressed; (4) Everything was an effort; (5) Hopelessness; (6) Feeling nervous/fearful; (7) Lack of happiness; (8) Loneliness; (9) Inability to get going; (10) Sleep disturbances. Of them, items 5, 7 and 10 required reverse scoring. The CESD-10 is based on a 5-point Likert scale, scoring the frequency of depressive symptoms from 0 (‘rarely or none of the time; less than 1 day’) to 3 (‘most or all of the time; 5–7 days’), with total scores spanning from 0 to 30. A greater total CESD-10 score denotes enhanced severity of depressive symptoms, with an overall CESD score of more than 10 is regarded as ‘having depression’.

PSCI symptoms were evaluated utilising the validated Chinese version of the Mini-Mental State Examination (MMSE),¹³ with six domains, including orientation, registration, attention and calculation, recall, language and naming. The overall score of MMSE spans from 0 to 30, with a lower score indicating more severe PSCI symptoms. The cut-off values of the MMSE total scores are characterised as ≥24 (normal cognitive functioning), 18–23 (mild cognitive impairment) and <18 (moderate to severe cognitive impairment).¹⁴ In this study, PSCI was determined to be a total MMSE score <24.

Global QoL (QoL hereafter) was measured with the overall scores of the first two items of the WHO Quality of Life-Brief Version (WHOQOL-BREF, validated Chinese version).¹⁵ The WHOQOL-BREF is rated using a 5-point Likert scale with items being scored from ‘1’ (very poor) to ‘5’ (every good), where elevated total scores reflect improved QoL.

Statistical analysis

Item redundancy

Items displaying less than 25% of significant polychoric correlations between two particular variables were considered redundant and excluded.¹⁶ Redundant items between CESD-10 and MMSE were determined utilising the function ‘goldbricker’ from the package ‘networktools’ (V.1.5.2).

Network structure

Network estimation and visualisation

Network approach was performed using the R program (V.4.3.2).¹⁷ Nodes represented depressive symptoms or cognitive domains, with edges showing correlations. The Gaussian Graphical Model was applied for partial correlation estimation.¹⁸ A graphic least absolute shrinkage and selection operator (GLASSO) with a tuning parameter of 0.5 was applied to minimise weak correlations.¹⁹ Additionally, Extended Bayesian Information Criterion model selection was conducted based on GLASSO balanced model fit and complexity.²⁰ The network was estimated and visualised with the ‘qgraph’ (V.1.9.8), ‘ggplot2’ (V.3.4.4) and ‘bootnet’ (V.1.5.6) R-packages.

Node centrality and predictability

To identify the most influential symptoms in the PSD-PSCI network, strength, which sums up all the absolute edge weights that a node has a direct connection to, was applied as a reliable centrality index since both the unique and shared variances are considered. A higher strength indicates a symptom more central to the entire network. The package ‘qgraph’ (V.1.9.8) was performed to examine the most central symptoms. Furthermore, we defined node predictability in this study as the degree to which a particular node’s variance was accounted for by all other nodes in the network model and estimated it with the ‘mgm’ package (V.1.2–14).

Bridge centrality

To explore the interrelation between the communities of depression and cognitive function, bridge strength, defined as the summed strength a node relates to other nodes from its neighbouring communities, was applied to determine the bridge symptoms between PSD and PSCI communities. Higher bridge strength values indicate a stronger association between a node in one community and all other nodes in the other community. The package ‘networktools’ (V.1.5.2) was performed to examine the most influential bridge symptoms.

Network accuracy and stability

To examine accuracy and stability, we implemented a three-step process: (1) edge-weights accuracy estimation through bootstrapped 95% CIs; (2) centrality stability via case-dropping bootstrap and the Correlation Stability Coefficient (CS-coefficient) where a value greater than 0.25 signifies a stable result from the observed network model, although a value greater than 0.5 is more preferable²¹; (3) bootstrapped difference tests for differences of edge weights and centrality indices (eg, strength) to identify whether they were significant. Network stability and accuracy of the PSD-PSCI network were assessed utilising the ‘bootnet’ package.

Rationale of parameter choices

In this study, the parameters for the network analysis were chosen according to their relevance to guarantee the reliability of the PSD-PSCI network. The sparseness of the PSD-PSCI network generated through GLASSO depended on how we set the value of the tuning parameter (λ) with a higher λ value selected to remove more edges that reflect weak or zero correlations, thus further influencing the structure of the PSD-PSCI network.²⁰ The default λ value of 0.5 was adopted in this analysis, as it could effectively minimise spurious edges while emphasising stronger associations, thus enhancing the accuracy of the network structure.²⁰ Strength was chosen as both node and bridge centrality index since it could directly capture the overall connectedness of a node within a symptom community or across different symptom communities by aggregating the absolute values of its edge weights, which could reflect more reliability of a node’s true connectedness than other centrality indices including expected influence, betweenness and closeness.²² These parameter choices were integral to constructing a robust PSD-PSCI network and capturing the intricate interactions underlying post-stroke mental health.

Results

Participant characteristics

Altogether, 913 older stroke survivors were included in this study using data from the 2017–2018 CLHLS. The prevalence of PSD (CESD-10 total score ≥10) was 31.5% (n=288; 95% CI 28.6% to 34.7%), and the prevalence of PSCI (MMSE total score <24) was 22.1% (n=202; 95% CI 19.5% to 25.0%). Sociodemographic characteristics of stroke survivors are summarised in table 1. The mean age of the study population was 80.17 years (SD=8.77), and 52.2% (n=477) were men. The means (SDs) of CESD-10, MMSE and QoL total scores were 8.01 (4.69), 24.88 (2.97) and 4.99 (1.48), respectively.

Table 1. Sociodemographic characteristics of older stroke survivors.

Variables	Total	Number of cases	Percentage
Male gender	913	477	52.2
Married	907	494	54.5
Education level	807
Uneducated		232	28.7
Educated		575	71.3
Annual household income from the last year	867
Less than 30 000 (RMB)		313	36.1
30 000 and above (RMB)		554	63.9
	Total	Mean	SD
Age (years)	913	80.17	8.77
CESD-10 total	913	8.01	4.69
MMSE total	913	24.88	2.97
QoL total	913	4.99	1.48

Notes: Bolded values: <0.05.

CESD-10, 10-item Center for Epidemiologic Studies Depression Scale; MMSE, Mini-Mental State Examination; QoL, quality of life.

Network structure of PSD and PSCI

No item redundancy was determined, so all CESD-10 and MMSE items were included. Figure 2 shows the network structure of PSD and PSCI. The node with the highest strength value was CESD3 (‘Feeling blue/depressed’, strength: 1.117), followed by At_C (‘Attention and calculation’, strength: 0.972). The mean predictability was 0.277, suggesting that 27.7% of the variance in each node could be explained by its neighbouring nodes. Online supplemental table S1 shows the descriptive statistics and network centrality indices for each item in CESD-10 and MMSE.

Figure 2. Network structure of depressive symptoms and cognitive functions in older stroke survivors.

Figure 3 shows the network structure of bridge symptoms between PSD and PSCI. The node demonstrating the greatest bridge strength value was Nam (‘Naming’, bridge strength=0.175). Figure 4 indicates that CESD10 (‘Sleep disturbances’, average edge weight=−0.200) exhibited the most significant negative correlation with QoL.

Figure 3. Bridge symptoms of the depressive symptoms and cognitive functions network in older stroke survivors.

Figure 4. Flow network for quality of life, depressive symptoms and cognitive functions in older stroke survivors.

Online supplemental figure S1 presents the network accuracy and stability. The CS-coefficients for strength and bridge strength were 0.75 and 0.284, respectively, indicating that the network structure would remain stable, even when 75% (strength) and 28.4% (bridge strength) of the sample were dropped. Online supplemental figure S2 shows narrow bootstrapped 95% CIs of edge weights, suggesting both stability and accuracy. Online supplemental tables S3 and S4 show significant edge weight and node strength differences after bootstrapped difference tests, confirming the reliability of the PSD-PSCI network.

Discussion

This was the first study to employ network analysis to investigate the interplay between PSD and PSCI among Chinese older stroke survivors, and the association with QoL, utilising the CLHLS data collected during the 2017–2018 wave. We found that ‘Feeling blue/depressed’ and ‘Attention and calculation’ were the most central symptoms, while ‘Naming’ was the key bridge symptom. Additionally, ‘Sleep disturbances’ had the most negative association with QoL.

Prevalence of PSD and PSCI

The prevalence of PSD was 31.5% (95% CI 28.6% to 34.7%) among Chinese older stroke survivors in the CLHLS, consistent with a pooled PSD prevalence of 33.0% (95% CI 29.0% to 36.0%) reported in a previous systematic review.²³ However, a global meta-analysis reported an overall PSD prevalence at any given time as 27.0% (95% CI 25.0% to 30.0%), with prevalence rates of 24.0% (95% CI 21.0% to 28.0%) as determined by clinical interview and 29.0% (95% CI 25.0% to 32.0%) derived from assessment tools.⁴ Variations in PSD prevalence rates between studies could result from various socioeconomic factors (eg, educational level, economic status), activities of daily living (ADL) and night-time sleep duration, and clinical factors (eg, stroke severity/disability, PSCI and stroke lesion location).²⁴

The reported prevalence of PSCI was found to be 22.1% (95% CI 19.5% to 25.0%), which is consistent with a British study based on data collected from the South London Stroke Register (1995–2010). This study found the pooled PSCI prevalence at three months following a stroke was 24.0% (95% CI 21.2% to 27.8%), 22.0% (95% CI 17.4% to 26.8%) at five years and 21.0% (95% CI 3.6% to 63.8%) at 14 years after stroke.²⁵ In contrast, a Chinese cross-sectional study reported that the PSCI prevalence across the entire study sample was 80.9% (95% CI 77.8 to 84.1%),²⁶ which is much higher than our findings. Numerous factors might explain the wide variation in PSCI prevalence such as variations in study locations, duration after the stroke occurred, stroke subtypes, assessment techniques and diagnostic standards for cognitive impairment. Of note, the probability of stroke survivors developing PSCI with complications was three times higher than in those without complications. Furthermore, PSCI has an adverse long-term impact on the QoL of survivors.^{26 27} Therefore, it is essential to address and manage PSCI effectively to improve the long-term prognosis and well-being of stroke survivors.

Interplay between PSD and PSCI and their impact on QoL

The inter-relationship among PSD, PSCI symptoms and QoL is determined by the multifaceted interactions of post-stroke changes, including biological mechanisms, psychological responses and social contexts. Biologically, stroke-induced dysregulation and hyperactivation of the hypothalamic–pituitary–adrenal axis, driven by proinflammatory processes and hippocampal damage, may lead to the development of PSD and PSCI through altered corticosteroid receptor function and stress regulation.²⁸ Recent evidence from a scoping review regarding the impact of cortisol on cognitive and emotional disturbances following a stroke found that survivors with elevated cortisol levels tended to experience PSD and PSCI later in life, suggesting that cortisol is a potential predictive biomarker for these conditions.²⁹ Psychologically, mood and emotional disturbances post-stroke, such as depression, anxiety, anger proneness and fatigue, can exacerbate the emotional burden of stroke survivors, further impairing their cognitive functioning and diminishing their overall QoL.³⁰ Furthermore, social support is considered as a key determinant of stroke outcomes, with higher levels of social support being connected to a decreased risk of PSD and PSCI.³¹ Educational attainment can also be another key determinant, with lower educational levels being related to a heightened likelihood of PSCI and severity of long-term cognitive impairment being modulated by educational background.³² Therefore, understanding the inter-relationship between PSD and PSCI and their impact on QoL is crucial for developing preventive strategies and targeted clinical interventions. By addressing these interconnected factors, it is possible to optimise long-term outcomes and promote the overall well-being of stroke survivors.

In the PSD-PSCI network model, ‘Feeling blue/depressed’ (CESD3) and ‘Attention and calculation’ (At_C) were the most central symptoms in older stroke survivors, which are partially aligned with the findings of a recent study that identified both these symptoms as the most influential in older adults, according to the 2017–2018 CLHLS.³³ ‘Feeling blue/depressed’ represented the most influential symptom in our network model, which is supported by a recent study showing that depressed mood was the most influential centrality within the PSD symptom community using the PHQ-8 among mild-to-moderate stroke survivors.⁹ ‘Feeling blue/depressed’ represents depressed mood, which can adversely affect rehabilitation, cognitive and motor recovery and increases the risk of recurrent neurovascular events.²⁴ It is linked to lesions in the frontal or anterior cerebral areas, or in the basal ganglia, as well as limited social support and substantial disability following a stroke.²⁴ ‘Feeling blue/depressed’ can influence other PSD symptoms such as insomnia and fatigue, leading to worsening depressed mood within the PSD symptom network.³⁴ Additionally, stroke survivors commonly experience anxiety, reduced QoL, speech and language dysfunction, feelings of despair and poor adherence to treatment, all of which can result in feeling blue/depressed.³⁵ Clinically, selective serotonin reuptake inhibitors (SSRIs, eg, fluoxetine and paroxetine) are typically first-line treatment for PSD, which are safe, tolerable and efficacious in reducing depressive symptoms (eg, feeling blue or depressed).³⁶

‘Attention and calculation’ was another key central symptom, which aligns with a recent study, indicating that concentration and executive function were the most central symptoms within the PSD-PSCI symptom communities using the PHQ-8 and NIHTCB to assess the survivors with mild-to-moderate stroke.⁹ The ‘Attention and calculation’ domain of the MMSE specifically evaluates the ability of stroke survivors to concentrate and solve basic calculations. Attention is a component of concentration, while calculation is part of executive function in cognition.³⁷ Stroke-related cognitive impairment is more weighted towards attention-executive deficits than other types of dysfunction (eg, memory dysfunction).³⁸ Therefore, cognitive dysfunction in the attention-calculation domain is critical and can lead to an increase in ADL difficulties and functional dependency on others, which may increase the risks of PSD.

Apart from central symptoms, ‘Naming’ (Nam) was recognised as the most significant bridge symptom within the PSD-PSCI network, which is consistent with a previous CLHLS study showing that ‘Naming’ and ‘Difficulty with concentrating’ were also bridge symptoms in older adults.³³ ‘Naming’ was a bridge symptom showing the greatest bridge strength. In this study, the Naming domain referred to aspects of executive function in the MMSE as participants were tasked with naming as many edible items as possible within 60 s. Stroke survivors with naming difficulties (eg, aphasia) and incorrect assessment of their naming abilities tend to exhibit more significant PSCI on tasks that partially depend on semantic knowledge, thus indicating more extensive deficits in language and executive function.³⁹ Executive and working memory dysfunction, which are identified in patterns of PSCI in even mild stroke survivors, can act as predictors of PSD.

The flow network model examining the association between PSD, PSCI and QoL showed that ‘Sleep disturbances’ (CESD10) was the symptom most strongly correlated with a decline in QoL, which is in partial agreement with a prior CLHLS study using CESD10, suggesting that ‘Sleep disturbances’ was the strongest depressive symptom associated with QoL.³³ In this study, ‘Sleep disturbances’ was negatively associated with QoL and occurred often following a stroke, manifesting as sleep apnoea, insomnia or daytime sleepiness among stroke survivors. Previous studies showed that QoL was highly influenced by sleep behaviour, and stroke survivors with poor sleep quality often experienced prolonged stroke rehabilitation, which could negatively affect stroke outcomes and increase recurrence rates, thereby contributing to lower QoL.⁴⁰

In clinical practice, conventional treatments for PSD and PSCI may lack targeted interventions for specific symptoms. Our network analysis identified key symptoms, such as ‘Feeling blue/depressed’, ‘Attention and calculation’ and ‘Hopelessness’. Such key symptoms can guide more personalised treatment strategies, potentially improving outcomes for stroke survivors. Combining pharmacological, psychological and adjuvant therapies tailored to these pivotal symptoms can likely improve treatment efficacy and QoL.

The strengths of this study included the representative sample drawn from across the nation with a large cohort of older stroke survivors, along with the innovative network approach to explore the central and bridge symptoms within the PSD-PSCI network model and the interactions among PSD, PSCI and QoL. However, the study had several limitations. First, as the data on stroke phase (eg, acute or chronic), lesion localisation, severity, subtypes (eg, ischaemic or haemorrhagic) and the duration since stroke onset were not documented in the CLHLS, their potential impact on the findings could not be evaluated. Additionally, the absence of available data on neuroimaging and biomarkers (eg, cortisol, inflammatory markers) would limit any mechanistic exploration. Second, the use of the MMSE, which inadequately assesses executive function, and the CESD-10, which omits important criteria for depression as outlined in the Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition-5, such as psychomotor retardation, might limit the accuracy of assessment. Third, the network modelling was conducted using cross-sectional data, thereby preventing any causal inference between PSD and PSCI symptoms or between PSD, PSCI and QoL. Furthermore, we could not examine any dynamic changes between PSD and PSCI symptoms over time. Fourth, the assessment of PSD and PSCI symptoms was derived from self-reporting, potentially leading to recall bias. Fifth, certain potential confounding factors related to PSD, PSCI and QoL, such as comorbidities, medication use and social support, were not recorded in the CLHLS. Finally, the study sample of older stroke survivors (mean age of 80.17 years) might limit the generalisability of the findings.

Conclusion

In conclusion, our study revealed that PSD and PSCI were prevalent among older stroke survivors. According to the network analysis, ‘Feeling blue/depressed’ (CESD3) and ‘Attention and calculation’ (At_C) were determined to be the most influential symptoms, while ‘Naming’ (Nam) was the most significant bridge symptom in the PSD-PSCI network. Additionally, ‘Sleep disturbances’ (CESD10) was the symptom most closely associated with QoL. To improve outcomes for older stroke survivors, treatments focusing on the central, bridge and QoL-related symptoms are crucial.

Future research could consider the following directions. First, integrating functional MRI might help investigate the neural correlates of ‘attention/calculation deficits’, particularly in relation to the dorsal attention network and default mode network dysconnectivity. Additionally, conducting randomised controlled trials for targeted interventions, such as SSRIs combined with computerised attention training, could provide insights into how symptom network reconfiguration impacts cognitive recovery. Finally, the development of artificial intelligence-based predictive models using symptom networks could dynamically identify high-risk individuals, enabling early intervention and tailored treatments for post-stroke cognitive deficits.

Data availability statement

Data are available upon reasonable request.