Cognitive impairment profiles in patients with cerebrovascular diseases: a latent profile analysis.

Cognitive impairment profiles in patients with cerebrovascular diseases: a latent profile analysis

Background

Cognitive impairment is an increasingly recognized condition predominantly attributable to cerebrovascular diseases (CVDs). It substantially diminishes patients’ quality of life and imposes considerable burdens on caregivers and healthcare systems. However, how cognitive impairment manifest across multiple functional domains in individuals with CVDs remains poorly understood, often resulting in generalized care plans that fail to meet patients’ specific rehabilitation needs. This study aimed to identify latent profiles of cognitive impairment and their associated factors in patients with CVDs using latent profile analysis (LPA).

Methods

This cross-sectional study was conducted using convenience sampling in a tertiary hospital in southwestern China from November 2023 to September 2025. LPA was employed to identify distinct cognitive impairment profiles among patients with CVDs. Multinomial logistic regression was subsequently performed to examine factors associated with profile categories.

Results

The final three-profile model classified 474 participants into three groups: a mild impairment group (82.5%), a moderate impairment group (11.4%), and a severe impairment group (6.1%). The three profiles reflected increasing severity of cognitive impairment among patients with CVDs. Compared with the mild impairment group, higher NIHSS scores (OR = 1.222, 95 %CI = 1.126–1.326, p < 0.001) were associated with higher odds of membership in the moderate impairment profile. Male gender and lower NIHSS scores were associated with lower odds of membership in the severe impairment profile (all p < 0.05).

Conclusion

In this predominantly hemorrhagic and neurosurgical cohort, the retained three-profile model represented different levels of MMSE-based global cognitive severity among patients with CVDs. NIHSS score and gender showed exploratory associations with profile membership. The applicability of these findings to broader CVD populations requires further validation.

1Introduction

Cerebrovascular diseases (CVDs) remain a leading cause of death and loss of healthy life among Chinese residents, presenting a critical challenge to public health systems (1). Despite advances in acute management, the burden of CVDs continues to rise, with a reported crude mortality rate of 149.6 per 100,000 individuals (2). Beyond physical impairments, cognitive impairment has emerged as a pervasive and debilitating complication, encompassing a spectrum from mild deficits to dementia (3, 4). The prognostic implications of cognitive impairment are profound, evidence suggests that even among individuals adhering to optimal lifestyle behaviors, the presence of cognitive impairment significantly elevates the risk of all-cause and cerebrovascular mortality (5). High prevalence rates have been reported across various CVD subtypes, including stroke (66%) and moyamoya disease (54.59%), severely compromising patients’ quality of life and increasing the burden on healthcare resources (6, 7).

Traditionally, clinical and nursing research on cognitive impairment has predominantly followed a “variable-centered” research path. This methodology relies on global cognitive screening instruments, such as the Mini-Mental State Examination (MMSE) or the Montreal Cognitive Assessment (MoCA), to categorize patients via dichotomous cut-off scores (8, 9). While effective for identifying universal risk factors (e.g., age, comorbidities, or lesion sites), this approach assumes population homogeneity. It fails to account for latent heterogeneity or distinct patient subgroups, as it cannot simultaneously analyze individual performance across multiple cognitive domains (e.g., memory, orientation, and language ability). Consequently, such generalized metrics may lead to “one-size-fits-all” care plans that overlook the specific rehabilitation needs of diverse patient profiles.

To address these limitations, an “individual-centered” analytical approach is warranted. Latent Profile Analysis (LPA) offers a sophisticated statistical alternative by identifying heterogeneity among individuals (latent profiles) within a population based on response patterns across multiple continuous indicators (28). Recent studies have employed LPA to identify cognitive impairment profile categories in older adults, stroke survivors, and individuals with cardiometabolic diseases (10–12). These investigations consistently identified three cognition profiles based on MMSE performance: a high-performing group, a severely impaired group, and an intermediate group characterized by domain-specific deficits. Notably, the distinguishing features of the intermediate profile varied across disease populations, suggesting disease-specific patterns of cognitive compromise. However, research applying LPA to characterize cognitive impairment categories targeting the broader CVDs population remains scarce.

Therefore, this study aimed to identify the latent profiles of cognitive impairment in patients with CVDs and explore their associated factors. By characterizing different levels in global cognitive impairment severity among patients with CVDs, this study may support the management of cognitive impairment among patients with CVDs and the development of effective interventions.

2Materials and methods

2.1Study design, setting and participants

The study was prepared and reported according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) checklists to enhance transparency and mitigate reporting bias.

This cross-sectional study was conducted using convenience sampling in a tertiary hospital in southwestern China from November 2023 to September 2025. This hospital serves as a leading tertiary referral center for the management of diverse CVDs, including AVMs, intracranial aneurysms, stroke, et al. All research participants were recruited from the hospital and had a primary diagnosis of CVDs. The inclusion criteria were: (1) age 18 years or older; (2) diagnosis of CVDs via ICD-10 codes (I60-I68), and confirmed without misdiagnosis by cranial CT or MRI; (3) volunteer to participate and signed informed consent by themselves or their families. Exclusion criteria included: (1) had a previously established diagnosis of cognitive impairment or dementia attributable to other medical conditions; (2) a history of severe auditory impairment, psychiatric illness or communication disorders; (3) acute critical medical conditions (e.g., shock, respiratory failure, acute heart failure, acute myocardial infarction, or stroke); (4) were concurrently enrolled in other clinical trials.

2.2Sample size estimation

Although there is currently no sample formula or calculator to estimate required sample size in the LPA, some simulation studies have suggested samples of 300 to 500 would qualify as a minimum sample (13). And LPA was considered a ‘large sample’ method. Models and fit statistics have been shown to consistently perform with low reliability, when the sample size was less than 300 (14). Ultimately, the sample size of this study was 474 participants.

2.3Data collection

Data were collected by research assistants in the outpatient departments from patients who were treated at the study hospital with a primary diagnosis of CVDs. These assistants were employed under the supervision of specialized physicians in CVDs at the hospital. Prior to data collection, all research assistants underwent comprehensive training on the study design, objectives, procedures, and measurement protocols. Informed consent was obtained from each participant before data collection began. Following this, the data were carefully reviewed for accuracy.

2.4Measures

2.4.1Demographic and disease-related characteristics

The demographic and disease-related characteristics included in the questionnaire were determined by the research team following a comprehensive review of the relevant literature and subsequent discussions. Demographic variables encompassed age, body mass index (BMI), gender, ethnicity, educational background, marital status, residence, employment status, and a history of smoking and alcohol consumption. Disease-related variables included disease subtypes, comorbidities, follow-up interval (discharge-to-follow-up interval), surgery, use of analgesics, sedation, and mechanical ventilation, as well as the Nutritional Risk Screening 2002 (NRS-2002) score, and the length of hospital stay.

2.4.2The Chinese Mini-Mental State Examination (MMSE)

The MMSE has been widely used for screening for cognitive function for patients, which was developed by Folstein et al. (15). The tool consists of 30 items, assessing five dimensions of orientation, memory, attention and calculation, recall and language ability. The total score ranges from 0 to 30. Higher scores indicate less impairment of cognitive function. According to educational level, cognitive impairment was defined as MMSE scores < 24 with education ≥ 7 years or < 20 with education < 7 years. The Chinese version of the MMSE translated and adopted by Zhang et al. [17]is widely utilized in clinical screening for patients with cerebrovascular disease-related cognitive impairment in China. The Cronbach’s α was 0.83 (16).

2.4.3National Institutes of Health Stroke Scale (NIHSS)

It is primarily used to assess the severity and neurological function of stroke but has also been extended to evaluate other cerebrovascular conditions, such as AVMs and MMD, as recommended by Chinese clinical practice guidelines (17, 18). Healthcare providers use the NIHSS to measure neurological function and deficits through a series of 11 questions and both physical and mental assessments. The score ranges from 0, indicating no symptoms, to 42, indicating severe stroke, with higher scores correlating with increased disease severity.

2.5Statistical analysis

Data were analyzed using IBM SPSS version 29.0. Descriptive statistics were used to characterize patient demographics, with continuous variables reported as means ± standard deviations (SD) or medians (interquartile range), and categorical variables as frequencies and proportions. Kruskal-Wallis and chi-squared tests were applied to compare demographic and disease-related variables across subgroups. The statistically significant variables were further analyzed using multinomial logistic regression. A significance level of p < 0.05 was used for all analyses.

Demographic adjustment and LPA were conducted using R version 4.4.0 (R Foundation for Statistical Computing, Vienna, Austria). The five MMSE domain scores were separately adjusted for age, gender, and years of education using ordinary least-squares linear regression, implemented with the lm() function in the stats package. These demographically adjusted residual z-scores were subsequently used as indicators in the LPA performed using the tidyLPA package (19). The complete annotated analysis code has been deposited in Zenodo and is publicly available at [DOI: https://doi.org/10.5281/zenodo.22998558]. The number of classes was incrementally increased until model fit indices no longer improved. Model selection was based on the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), and Adjusted BIC, with lower values indicating better fit. Entropy values closer to 1 indicated more precise classification. The Lo–Mendell–Rubin likelihood ratio (LMR) and Bootstrapped Likelihood Ratio Tests (BLRT) were used to compare model fits, with a p-value < 0.05 indicating a better k-class model than a (k-1)-class model. Selection of the retained solution was informed by statistical fit, parsimony, and substantive interpretability (20). These considerations were evaluated in relation to the objective of describing broad levels of cognitive impairment severity.

2.6Ethical considerations

After receiving a comprehensive description of the study’s objectives and procedures, each participant gave signed informed consent. Throughout the investigation, no personally identifiable information about the patients was revealed. The Declaration of Helsinki and Good Clinical Practice principles were followed in the conduct of the study. This study was approved by the West China Hospital of the Sichuan University Biomedical Research Ethics Committee (no.888[2022]).

3Results

3.1Demographic and disease-related characteristics

As shown in Table 1, in this cohort of 474 patients with cerebrovascular disease, the median age was 52 years (ranging 18–84), with a near-equal gender distribution (49.6% male) and a majority of Han ethnicity (91.8%). Clinically, intracerebral hemorrhage was the most prevalent diagnosis (52.3%), and hypertension was a common comorbidity (45.0%). A key characteristic of this population was the high acuity of care required, as evidenced by the fact that the majority underwent surgical intervention (79.7%) and received intensive supportive management, including analgesia (80.1%), sedation (73.2%), and mechanical ventilation (67.0%). With this high level of intervention, the median NIHSS score was 0 (range 0–26), indicating minimal neurological deficits at assessment for a substantial portion of the cohort, and the median hospital length of stay was 9 days (IQR 6–12).

Table 1Demographic and disease-related characteristics of participants(n = 474).

Variables Category Median (IQR)/N (%)

Age, y 52 (41, 59)

BMI 24 (22, 26)

Gender Male 235 (49.6)

Female 239 (50.4)

Ethnicity Han 434 (91.8)

Minority 39 (8.2)

Educational background, y 9 (9, 15)

Marital status Married 399 (84.2)

Single/Divorced/Widowed 75 (15.8)

Residence Urban 327 (69.1)

Rural 146 (30.9)

Employment status Employment 137 (28.9)

Unemployment/Unknown 337 (71.1)

Smoking Yes 78 (16.5)

Alcohol consumption Yes 69 (14.6)

Subtypes of disease Intracerebral hemorrhage 248 (52.3)

Ischemic stroke 22 (4.6)

AVMs 28 (5.9)

Intracranial aneurysms 71 (15.0)

MMD 89 (18.8)

Other 16 (3.4)

Comorbidities Hypertension 213 (45.0)

Diabetes 48 (10.1)

Coronary heart disease 10 (2.1)

Hyperlipidemia 35 (7.4)

Diseases duration, d 14 (7, 108)

Underwent surgery Yes 378 (79.7)

History of analgesic use Yes 379 (80.1)

History of sedative use Yes 346 (73.2)

History of receiving mechanical ventilation Yes 317 (67.0)

NRS-2002 1 (0, 3)

NIHSS score 0 (0, 3)

Length of hospital stay, d 9 (6, 12)

IQR, Interquartile range; BMI, Body mass index; AVMs, Arteriovenous malformations; MMD, Moyamoya disease; NRS-2002, Nutritional risk screen-2002; NIHSS, National Institutes of Health Stroke Scale.

3.2Scores of MMSE dimensions among patients with CVDs

The cognitive profile of the cohort, as assessed by the MMSE, revealed a mean total score of 22.92 ± 7.25, indicating overall mild cognitive impairment. Analysis of the specific domains showed notable variations. The most significant deficits were observed in recall (mean 1.66 ± 1.20 out of 3) and orientation (mean 7.72 ± 2.79 out of 10). In contrast, memory (registration) and language ability were relatively better preserved, with mean scores of 2.75 (±0.75) out of 3 and 7.79 (±2.06) out of 9, respectively. Performance in attention and calculation was moderately affected (3.01 ± 1.93 out of 5). Detailed information was shown in Table 2.

Table 2Scores of MMSE dimensions among patients with CVDs (n = 474).

Dimensions Number of items Score range Scores (Means±SD)

Orientation 10 0–10 7.72±2.79

Recall 3 0–3 1.66±1.20

Attention and calculation 5 0–5 3.01±1.93

Memory 3 0–3 2.75±0.75

Language ability 9 0–9 7.79±2.06

Total score 30 0–30 22.92±7.25

SD, standard deviation.

3.3Latent profile analysis of cognitive impairment in patients with CVDs

The five demographically adjusted residual z scores were entered as continuous indicators in an LPA to identify profiles representing different levels of global cognitive impairment severity. One- through five-profile solutions were estimated (Table 3). AIC, BIC, and aBIC decreased as profiles were added, and the LMR and BLRT comparisons were significant for the two- through five-profile solutions (all p < 0.001). Thus, the four- and five-profile solutions showed better statistical fit than the three-profile solution according to these criteria. However, the three-profile model showed greater classification certainty (entropy = 0.955) than the four- and five-profile models (0.859 and 0.815, respectively). We retained the three-profile model to provide a parsimonious and clinically interpretable description of broad differences in cognitive impairment severity within the cohort.

Table 3Fit indices for latent profile models of cognitive impairment in patients with CVDs.

Model LL FP AIC BIC aBIC Entropy LMR (P) BLRT (P) Probability of class (%)

1 −3360.382 10 6740.763 6782.375 6750.637 - - - 100.0

2 −2686.641 16 5405.281 5471.861 5421.079 0.997 <0.001 <0.001 92.2/7.8

3 † −2507.467 22 5058.935 5150.481 5080.657 0.955 <0.001 <0.001 82.5/11.4/6.1

4 −2477.727 28 5011.454 5127.968 5039.100 0.859 <0.001 <0.001 57.4/9.7/26.8/6.1

5 −2443.929 34 4955.858 5097.339 4989.428 0.815 <0.001 <0.001 44.5/9.5/25.3/14.6/6.1

†The three-profile model was retained for subsequent analyses as a pragmatic choice based on parsimony and clinical interpretability. Profile-specific indicator means were freely estimated, indicator-specific variances were constrained to be equal across profiles, and within-profile covariances were fixed at zero. Candidate models were first fitted using the default mclust initialization. Local-solution sensitivity was examined using 1,000 randomized hierarchical initializations for each two- through five-profile model. The BLRT was based on 1,000 bootstrap draws.

LL, log-likelihood; FP, free parameters; AIC, Akaike information criterion; BIC, Bayesian information criterion; aBIC, sample-size adjusted BIC; LMR, Lo–Mendell–Rubin likelihood ratio; BLRT, bootstrap likelihood ratio test.

3.4Cognitive impairment severity profiles

Following LPA of the five demographically adjusted MMSE domain z-scores, we retained a three-profile model to describe differences in global cognitive impairment severity within the cohort (Figure 1). The largest profile was the “mild impairment” group (Class 1; 82.5%, n = 391). Participants in this group performed relatively well across all five cognitive domains. The “moderate impairment” group (Class 2; 11.4%, n = 54) showed lower performance across all domains, with the greatest deficits in orientation and attention/calculation, while memory was relatively preserved. The smallest profile was the “severe impairment” group (Class 3; 6.1%, n = 29), which was characterized by pervasive deficits across all domains. Overall, the three profiles reflected increasing severity of cognitive impairment among patients with CVDs.

3.5Comparison of baseline information and univariate analysis across cognitive impairment profiles

Patient demographic and disease-related characteristics were used as the independent variable and three-class cognitive impairment profiles were used as the dependent variable. Univariate analyses revealed significant associations between the derived cognitive profiles and key demographic and clinical characteristics.

The three classes were demonstrated statistically significant differences in gender, educational background, alcohol consumption, subtypes of disease, follow-up interval, nutritional risk (NRS-2002), disease severity (NIHSS score) and length of hospital stay (all p < 0.05; refer to Table 4 for detailed results).

Table 4Univariate analysis to explore inter-profile differences.

Variables Category Class 1(n = 391) Class 2(n = 54) Class 3(n = 29) (χ²/H) P

Age, y 52 (40,59) 51 (42.61) 56 (48.62) 3.816 0.145

BMI 24 (22, 27) 24 (22.26) 23 (21,25) 2.675 0.262

Gender Male 189 (48.3) 37 (68.5) 9 (31.0) 11.979 0.003

Female 202 (51.7) 17 (31.5) 20 (69.0)

Ethnicity Han 355 (91.0) 51 (94.4) 28 (96.6) - 0.589*

Minority 35 (9.0) 3 (5.6) 1 (3.4)

Educational background, y 12 (9,15) 9 (9.15) 9 (6.12) 7.412 0.025

Marital status Married 334 (85.4) 40 (74.1) 25 (86.2) 4.683 0.096

Single/divorced/widowed 57 (14.6) 14 (25.9) 4 (13.8)

Residence Urban 275 (70.5) 36 (66.7) 16 (55.2) 3.151 0.207

Rural 115 (29.5) 18 (33.3) 13 (44.8)

Employment status Employment 117 (29.9) 16 (29.6) 4 (13.8) 3.434 0.180

Unemployment/unknown 274 (70.1) 38 (70.4) 25 (86.2)

Smoking Yes 61 (15.6) 12 (22.2) 5 (17.2) 1.504 0.471

Alcohol consumption Yes 51 (13.1) 14 (25.9) 4 (13.8) 6.301 0.043

Subtypes of disease Intracerebral hemorrhage 187 (47.8) 38 (70.4) 23 (79.3) - 0.013*

Ischemic stroke 18 (4.6) 3 (5.6) 1 (3.4)

AVMs 24 (6.1) 3 (5.6) 1 (3.4)

Intracranial aneurysms 66 (16.9) 4 (7.4) 1 (3.4)

MMD 82 (21.0) 4 (7.4) 3 (10.3)

Other 14 (3.6) 2 (3.7) 0 (0.0)

Comorbidities Hypertension 172 (44.1) 27 (50.0) 14 (48.3) 0.798 0.671

Diabetes 37 (9.5) 8 (14.8) 3 (10.3) 1.478 0.478

Coronary heart disease 8 (2.1) 0 (0.0) 2 (6.9) - 0.122*

Hyperlipidemia 31 (7.9) 4 (7.4) 0 (0.0) - 0.353*

Follow-up interval, d 1 (0,103) 0 (0.1) 0 (0.1) 21.542 <0.001

Underwent surgery Yes 315 (80.6) 44 (81.5) 19 (65.5) 3.897 0.142

History administration of analgesia Yes 311 (79.7) 44 (81.5) 24 (82.8) 0.224 0.894

History administration of sedation Yes 282 (72.3) 41 (75.9) 23 (79.3) 0.913 0.633

History of receiving mechanical ventilation Yes 259 (66.4) 37 (68.5) 21 (72.4) 0.502 0.778

NRS-2002 0 (0.2) 2 (2.3) 2 (2.5) 27.366 <0.001

NIHSS score 0 (0.2) 3 (1.7) 9 (4.15) 104.586 <0.001

Length of hospital stay, d 9 (6.11) 9 (6.5, 12) 12 (7.17) 6.213 0.045

MMSE score 27 (23.29) 15 (12.16) 3 (0.5) 195.720 <0.001

*Fisher’s exact test.

IQR, Interquartile range; BMI, Body mass index; AVMs, Arteriovenous malformations; MMD, Moyamoya disease; NRS-2002, Nutritional risk screen-2002; NIHSS, National Institutes of Health Stroke Scale; MMSE, the Chinese Mini-Mental State Examination.

3.6Multivariate analysis of cognitive impairment profiles in patients with CVDs

To identify factors associated with different classes of cognitive impairment profile, a multinomial logistic regression analysis was performed. The model included variables that were statistically significant in univariate analyses or deemed clinically important, with the “mild impairment group” group designated as the reference category.

The multinomial logistic regression model, using the “mild impairment group” as the reference, identified several factors independently associated with membership in the “moderate impairment group “and “severe impairment group” (Table 5). Higher NIHSS scores (OR = 1.222, 95 %CI = 1.126–1.326) were associated with higher odds of membership in the moderate impairment group. Male gender (OR = 0.218, 95 %CI = 0.067–0.704) was associated with lower odds of membership in the severe-impairment profile, whereas higher NIHSS scores (OR = 1.397, 95 %CI = 1.254–1.557) were associated with higher odds of membership in this profile (see Table 5).

Table 5Multinomial logistic regression analysis of latent cognitive impairment profiles in patients with CVDs.

Independent variables Class 2 Class 3

P OR 95%CI P OR 95%CI

Age, y 0.761 0.996 0.970–1.022 0.675 1.008 0.970–1.047

Gender (ref: female)

Male 0.420 1.351 0.650–2.809 0.011 0.218 0.067–0.704

Educational background, y 0.233 0.951 0.875–1.033 0.112 0.902 0.794–1.024

Alcohol consumption (ref: yes) 0.097 0.511 0.231–1.129 0.240 0.422 0.100–1.780

Subtypes of disease (ref: intracerebral hemorrhage)

Ischemic stroke 0.954 1.042 0.254–4.271 0.959 0.942 0.095–9.331

AVMs 0.823 1.172 0.292–4.710 0.637 0.583 0.062–5.513

Intracranial aneurysms 0.371 1.760 0.510–6.079 0.228 4.111 0.412–40.988

MMD 0.318 1.863 0.550–6.313 0.539 0.604 0.121–3.009

Other - - - - - -

Follow-up interval, d 0.062 0.996 0.992–1.000 0.359 0.997 0.992–1.003

NRS-2002 0.934 1.009 0.819–1.243 0.535 1.103 0.810–1.502

NIHSS score <0.001 1.222 1.126–1.326 <0.001 1.397 1.254–1.557

Length of hospital stay, d 0.452 0.983 0.942–1.027 0.980 0.999 0.943–1.059

OR, Odds ratio; CI, Confidence interval; AVMs, Arteriovenous malformations; NRS-2002, MMD, Moyamoya disease; Nutritional risk screen-2002; NIHSS, National Institutes of Health Stroke Scale.

Multinomial logistic regression analysis of factors associated with moderate (Class 2) and severe impairment (Class 3) profiles (reference = mild impairment group).

4Discussion

This study used LPA to characterize heterogeneity in cognitive impairment among patients with CVDs. We retained a three-profile model to describe broad differences in global cognitive impairment severity and examined factors associated with profile membership. The largest profile in the final three-profile model was the “mild impairment” group (Class 1; 82.5%, n = 391). Participants in this group performed relatively well across all five cognitive domains. The “moderate impairment” group (Class 2; 11.4%, n = 54) showed lower performance across all domains, with the greatest deficits in orientation and attention/calculation, while memory was relatively preserved. The smallest profile was the “severe impairment” group (Class 3; 6.1%, n = 29), which was characterized by pervasive deficits across all domains. Overall, the three profiles reflected increasing severity of cognitive impairment among patients with CVDs. It is important to note that these class proportions reflect the distribution within the present sample and are not intended to represent population prevalence. The emergence of these distinct profiles underscores the multidimensional nature of cognitive impairment captured by the MMSE, challenging the notion of a unitary, linear trajectory of cognitive decline and instead indicating the existence of specific-subgroup variations with potential implications for targeted nursing assessment and intervention.

Male participants exhibited statistically significant lower odds of severe cognitive impairment compared to females (OR = 0.218, p = 0.011). These findings align with previous research demonstrating sex differences in cognitive trajectories following cerebrovascular disease (21, 22), particularly the increased vulnerability of older women to rapid cognitive decline (10). This disparity may be mediated by physiological mechanisms involving sex steroid hormones. Females typically demonstrate higher baseline cerebral blood flow (CBF) than males (23), potentially rendering them more sensitive to hemodynamic fluctuations following cerebrovascular events. Furthermore, while sex hormones confer neuroprotection in younger physiology, preclinical evidence suggests these effects may become ineffective or even deleterious with advancing age (24). Notably, 64.0% (153/239) of the female participants in the present study were aged over 50 years. Additionally, the observed gender disparity may be influenced by sociodemographic factors, as women had significantly lower levels of educational attainment compared to men (9.85 ± 4.41 vs. 11.62 ± 4.11, p < 0.001). Further studies could assess biological or social mechanisms that might underlie this association.

Greater neurological deficit severity, as reflected by higher NIHSS scores, was associated with significantly higher odds of being classified into the moderate and severe impairment groups (OR = 1.222 and 1.397, respectively; both p < 0.001, reference group = the mild impairment group). This strong association aligns with the well-established prognostic utility of the NIHSS score in predicting cognitive outcomes following some CVDs (11). The effect of stroke severity and burden on the risk of dementia was evident in all age groups, with severe stroke (NIHSS score >10) bringing forward dementia by about 25 years (25). Mechanistically, a higher NIHSS score typically correlates with a larger infarct or hematoma volume and a lesion in a more critical location (e.g., lesions affecting the dominant hemisphere or key cognitive networks) (1). This greater initial tissue damage depletes the brain’s cognitive reserve, leaving the patient less capable of compensating for subsequent neuronal loss. However, these possible mechanisms could not be examined in our study because information on lesion volume, white matter hyperintensities, brain atrophy, and blood or cerebrospinal fluid biomarkers was not available. These unmeasured factors may have contributed to the observed differences in cognitive severity and may have resulted in residual confounding. The American Heart Association/American Stroke Association scientific statement also notes that post-stroke cognition may be affected by noncerebral factors, including infection, frailty, and social conditions, whose independent roles remain unclear (26). Future studies should combine assessments with neuroimaging, fluid biomarkers, and measures of infection, frailty, rehabilitation, and social conditions to identify independent predictors of cognitive impairment severity levels.

The nonsignificant estimates for other covariates should not be taken as evidence that these factors have no clinical relevance. The small number of participants in the severe-impairment profile (n = 29), corresponding to a category-specific events-per-parameter ratio of 2.4, limited the statistical precision of this comparison and resulted in wide confidence intervals (27). In addition, we adjusted for the potential influence of some of these variables in our analyses by using demographically (age, gender, and years of education) adjusted z-scores. The nonsignificant estimates for these demographic variables in the multinomial logistic regression should not be interpreted as evidence that they are unrelated to cognitive impairment.

5Limitations

This study provides valuable insights into the heterogeneity of cognitive impairment in patients with CVDs; however, several limitations should be acknowledged. Participants were recruited from the neurosurgery department of a tertiary referral hospital. The cohort consisted mainly of patients with intracerebral hemorrhage (52.3%), moyamoya disease (18.8%), and intracranial aneurysms (15.0%), whereas ischemic stroke accounted for only 4.6%. This may introduce a selection bias, potentially limiting the generalizability of the identified latent cognitive profiles to different healthcare settings or patient populations. The cross-sectional design of the study inherently precludes causal inferences and does not allow for the characterization of individual cognitive trajectories over time. While the overall sample size was robust, the subgroup of patients exhibiting severe cognitive impairment had a relatively small sample size. This limitation may affect the stability and precision of the latent profile estimates for this specific group and could lead to reduced statistical power in subsequent analyses comparing this profile to others. Finally, the MMSE is a brief screening tool with limited coverage of executive function, processing speed, visuospatial ability, and episodic memory. Although the five MMSE domain scores were separately adjusted for age, gender, and years of education before the LPA, this adjustment cannot recover information lost because of these measurement limitations. Future multicenter, longitudinal studies with larger, more diverse samples are warranted to validate the current latent cognitive profiles and assess their prognostic utility.

6Conclusion

The retained three-profile model provided a pragmatic summary of cognitive impairment severity among patients with CVDs: a mild impairment group (82.5%), a moderate impairment group (11.4%), and a severe impairment group (6.1%). The three profiles reflected increasing severity of cognitive impairment among patients with CVDs. Compared with the mild impairment group, male gender was associated with lower odds of membership in the severe-impairment profile. Higher NIHSS scores were associated with higher odds of membership in the moderate and severe impairment profile. Given the predominance of hemorrhagic and neurosurgical cases in this cohort, the generalizability of these findings to broader CVDs populations requires further validation.

References

1.Hu SS . Epidemiology and current management of cerebrovascular disease in China . J Geriatr Cardiol . (2024 ) 21 :465 –74 . doi: 10.26599/1671-5411.2024.05.009 , 38948893 PMC11211905

2.Lv B Song G Jing F Li M Zhou H Li W . Mortality from cerebrovascular diseases in China: exploration of recent and future trends . Chin Med J . (2024 ) 137 :588 –95 . doi: 10.1097/CM9.0000000000002760 , 37415525 PMC10932538

3.Chen C . Cognitive and neuropsychiatric consequences of cerebrovascular disease? J Neurol Sci . (2023 ) 455 :120894 . doi: 10.1016/j.jns.2023.120894

4.Rundek T Tolea M Ariko T Fagerli EA Camargo CJ . Vascular cognitive impairment (VCI) . Neurotherapeutics . (2022 ) 19 :68 –88 . doi: 10.1007/s13311-021-01170-y , 34939171 PMC9130444

5.Li C Bai C Wang L Zhang M Zhou M Chen J . Association between healthy lifestyle and cognitive decline, all-cause mortality, and mortality from cardiovascular and cerebrovascular diseases: a 10-year population-based prospective cohort study . Alzheimers Dement . (2025 ) 21 :e70021 . doi: 10.1002/alz.70021 , 40110644 PMC11923566

6.Melkas S Jokinen H Hietanen M Erkinjuntti T . Poststroke cognitive impairment and dementia: prevalence, diagnosis, and treatment . Degener Neurol Neuromuscul Dis . (2014 ) 4 :21 . doi: 10.2147/DNND.S37353 PMC7337160 32669898

7.Toh KZX Koh MY Loh EDW Sia CH Chong Y Yeo LLL . Prevalence and associations of cognitive impairment in adult patients with Moyamoya disease: a systematic review and Meta-analysis . J Alzheimer's Dis . (2024 ) 97 :541 –52 . doi: 10.3233/JAD-230979 , 38108354

8.Wen B Zhang P Niu X Cheng H Li Z Zhang M . Factors associated with depression and cognitive impairment after cerebral hemorrhage surgery . Sci Rep . (2025 ) 15 :40075 . doi: 10.1038/s41598-024-53437-x , 41249161 PMC12623810

9.He M Wang J Liu N Xiao X Geng S Meng P . Effects of blood pressure in the early phase of ischemic stroke and stroke subtype on poststroke cognitive impairment . Stroke . (2018 ) 49 :1610 –7 . doi: 10.1161/STROKEAHA.118.020827 , 29895539

10.Xiang J Xiong Y Liang H . Latent profiles and correlates factors of cognitive function in older adults: a cross-sectional study . Front Aging Neurosci . (2025 ) 17 :2804 . doi: 10.3389/fnagi.2025.1622804 PMC12504479 41069966

11.Ma Y He X Yang T Qin J Nan J Wang X . Cognitive function in older adults with stroke: a latent profile analysis . Geriatr Nurs (Minneap) . (2025 ) 62 :269 –75 . doi: 10.1016/j.gerinurse.2024.10.053 , 39532601

12.Wang Y Wan B Xu X Lv Q He Y Liu J . Latent transition and predictors of cognitive function among community-dwelling Chinese older adults with cardiovascular metabolic diseases . Int J Geriatr Psychiatry . (2025 ) 40 :e70144 . doi: 10.1002/GPS.70144 , 40827006

13.Sinha P Calfee CS Delucchi KL . Practitioner’s guide to latent class analysis: methodological considerations and common pitfalls . Crit Care Med . (2021 ) 49 :e63 –e79 . doi: 10.1097/CCM.0000000000004710 , 33165028 PMC7746621

14.Bauer J . "A primer to latent profile and latent class analysis". In: Goller M Kyndt E Paloniemi S , editors. Methods for Researching Professional Learning and Development: Challenges, Applications and Empirical Illustrations . Cham: Springer International Publishing. p. 243 –68 .

15.Folstein MF Folstein SE McHugh PR . “Mini-mental state”: a practical method for grading the cognitive state of patients for the clinician . J Psychiatr Res . (1975 ) 12 :189 –98 . doi: 10.1016/0022-3956(75)90026-6 , 1202204

16.Zhang M Katzman R Salmon D Jin H Cai G Wang Z . The prevalence of dementia and Alzheimer’s disease in Shanghai, China: impact of age, gender, and education . Ann Neurol . (1990 ) 27 :428 –37 . doi: 10.1002/ana.410270412 , 2353798

17.Wang M Jiao Y Zeng C Zhang C He Q Yang Y . Chinese cerebrovascular neurosurgery society and Chinese Interventional & Hybrid Operation Society, of Chinese Stroke Association clinical practice guidelines for management of brain arteriovenous malformations in eloquent areas . Front Neurol . (2021 ) 12 :651663 . doi: 10.3389/fneur.2021.651663 PMC8219979 34177760

18.Chinese Neurosurgical Society. Guidelines for the clinical management of patients with moyamoya disease and moyamoya syndrome (2024 edition) . Chin J Neurosurg . 40 :220 –9 . doi: 10.3760/cma.j.cn112050-20231119-00177

19.Rosenberg JM Beymer PN Anderson DJ van Lissa CJ Schmidt JA . tidyLPA: An R Package to Easily Carry Out Latent Profile Analysis (LPA) Using Open-Source or Commercial Software A tidy user-interface (2018 ) 3 :978 . doi: 10.21105/joss.00978

20.Spurk D Hirschi A Wang M Valero D Kauffeld S . Latent profile analysis: a review and “how to” guide of its application within vocational behavior research . J Vocat Behav . (2020 ) 120 :103445 . doi: 10.1016/j.jvb.2020.103445

21.Bako AT Potter T Tannous J Pan AP Johnson C Baig E . Sex differences in post-stroke cognitive decline: a population-based longitudinal study of nationally representative data . PLoS One . (2022 ) 17 :e0268249 . doi: 10.1371/journal.pone.0268249 , 35522611 PMC9075630

22.Guo J Sun F Tao L Luo Y Liu L Cao K . Influence of cognitive function on cerebrovascular disease among the elderly . Acta Neurol Scand . (2017 ) 135 :308 –15 . doi: 10.1111/ane.12599 , 27145203

23.Robison LS Gannon OJ Salinero AE Zuloaga KL . Contributions of sex to cerebrovascular function and pathology . Brain Res . (2019 ) 1710 :43 –60 . doi: 10.1016/j.brainres.2018.12.030 , 30580011

24.Cai M Ma YL Qin P Li Y Zhang LX Nie H . The loss of estrogen efficacy against cerebral ischemia in aged postmenopausal female mice . Neurosci Lett . (2014 ) 558 :115 –9 . doi: 10.1016/j.neulet.2013.11.007 , 24246902

25.Pendlebury ST Rothwell PM Oxford Vascular Study. Incidence and prevalence of dementia associated with transient ischaemic attack and stroke: analysis of the population-based Oxford vascular study . Lancet Neurol . (2019 ) 18 :248 –58 . doi: 10.1016/S1474-4422(18)30442-3 , 30784556 PMC6390174

26.El Husseini N Katzan IL Rost NS Blake ML Byun E Pendlebury ST . Cognitive impairment after ischemic and hemorrhagic stroke: a scientific statement from the American Heart Association/American Stroke Association . Stroke . (2023 ) 54 :e272 –91 . doi: 10.1161/STR.0000000000000430 , 37125534 PMC12723706

27.de Jong VMT Eijkemans MJC van Calster B Timmerman D Moons KGM Steyerberg EW . Sample size considerations and predictive performance of multinomial logistic prediction models . Stat Med . (2019 ) 38 :1601 –19 . doi: 10.1002/sim.8063 , 30614028 PMC6590172

28.Tan Y . Application of LPA in mental health research . Adv Soc Sci . (2024 ) 13 :383 –9 . doi: 10.12677/ASS.2024.131054