INTRODUCTION
Many previous studies have been conducted under the hypothesis that sociodemographic factors, such as educational attainment, may influence the development and prognosis of cardiovascular and cerebrovascular diseases. A number of these studies have demonstrated that lower educational levels are associated with both a higher risk of disease onset and worse prognosis following disease occurrence [1-4]. However, several studies have reported no significant association between education and these outcomes, likely due to differences in study populations, definitions of educational attainment, and adjustment for socioeconomic and healthcare-related factors [5,6]. Most evidences comes from western populations, underscoring the need for further research in non-Western populations.
In contrast, systematic review in Asian populations have demonstrated a clear association between lower education and higher cardiovascular mortality, particularly in younger populations, low- to middle-income economies, and lower income inequality (i.e., Gini coefficient below 0.3) [7]. Studies from China have shown that lower education levels are linked to higher recurrence and mortality rates in patients with cardiovascu-lar or cerebrovascular disease [3,8].
Considering discrepancies in the findings from Western and non-Western studies, we aimed to analyze the association between education level and the long-term prognosis of cardiovascular and cerebrovascular disease—specifically over a follow-up period of more than 7 years. Given Korea’s unique socioeconomic and healthcare environment, we hypothesized that the relationship between education level and prognosis may not be as clearly observed as in Western countries. Furthermore, we postulated that specific clinical factors may interact with this association, potentially moderating the impact of education level on long-term outcomes. Suicidal ideation (SI), a clinically important marker of acute psychological distress, is frequently observed in patients with cardiovascular and cerebrovascular illness [9-12]. SI represents a qualitatively distinct form of psychological distress whose pathophysiological underpinnings are partly independent of depression itself [13]. Beyond reflecting depression severity, SI has been linked to impaired coping, reduced treatment adherence, and increased healthcare utilization, suggesting a broader influence on disease course [14]. Therefore, we considered SI as a potential moderating factor, hypothesizing that the impact of educational attainment—a proxy for socioeconomic resources, health literacy, and coping capacity—might differ depending on the presence or absence of SI. This approach allows us to test whether the adverse prognostic significance of cardio-cerebrovascular events is uniform across all educational subgroups, or whether the association between educational attainment and outcomes is modified by acute psychological distress.
METHODS
Study design and participants
This study utilized data from two prospective cohorts conducted at Chonnam National University Hospital, Gwangju, South Korea [15,16]. The first cohort enrolled patients with recent onset acute coronary syndrome (ACS) (2006-2012), and the second enrolled patient with acute ischemic stroke (2006- 2009). Participants were enrolled after providing written informed consent, based on eligibility criteria detailed in the Supplementary Material. Baseline assessments, including all psychiatric and sociodemographic measures, were performed 2 weeks after the index event. The Chonnam National University Hospital Institutional Review Board approved both protocols (06-026, I-2008-02-028).
Baseline assessments
Education attainment
This was quantified as the total number of years of formal schooling. For analysis, it was dichotomized based on the study median into “lower education level” (≤10 years) and “higher education level” (>10 years) which ensures balanced group distribution for optimal statistical power and facilitates interpretability. In addition, sensitivity analyses were performed using a 12-year education cutoff (high school completion) to evaluate the robustness of findings across different educational thresholds.
Suicidal ideation
SI was evaluated at the baseline using item 10 (suicidal thoughts) of the Montgomery-Åsberg Depression Rating Scale (MADRS) [17]. The MADRS was administered as a faceto-face, semi-structured interview by trained research nurses blinded to psychiatric diagnoses, under psychiatrist supervision, within 2 weeks of hospitalization once patients were medically stable. Standardized training was provided to ensure inter-rater reliability. In stroke patients, those with severe aphasia or cognitive impairment (Mini-Mental State Examination <16) were excluded to ensure response validity. This suicidal thought item has been extensively used to assess SI in patients with severe physical illnesses including ACS and stroke [18,19]. and suicidal-thought items derived from depression rating scales have demonstrated acceptable validity for predicting subsequent suicidal behavior [20]. Furthermore, the pathophysiological characteristics of SI have been shown to be partly independent of underlying psychiatric conditions, including depression [13]. Consistent with established precedents [18,19], the presence of SI was defined as a score of ≥2, which indicates at least fleeting suicidal thoughts (e.g., feeling tired or disillusioned with living).
Covariates
A comprehensive set of covariates was collected. Sociodemographic data including age, sex, living status and employment. Psychiatric factors included a Diagnostic and Statistical Manual of Mental Disorders, Fourth Edition (DSM-IV) diagnosis of a depressive disorder [21] and prior psychiatric history. Vascular risk factors included hypertension, diabetes, hypercholesterolemia, obesity, smoking status, and family history of stroke or ACS. Cohort-specific clinical data were also recored: treatment group (escitalopram, placebo, or usual care), housing status, ACS diagnosis (myocardial infarction [MI] vs. unstable angina), Killip and Kimball [22], left ventricular ejection fraction (LVEF), and cardiac biomarkers (troponin I, CK-MB) in the ACS cohort while cardiac disease history, National Institutes of Health Stroke Scale (NIHSS) [23], and lesion location on neuroimaging in the stroke cohort.
Long-term outcomes
Long-term prognoses were systematically ascertained for all participants. For the ACS cohort, data were sourced from the Korean Acute Myocardial Infarction Registry (http://kamir5.kamir.or.kr) [24] over a 5-12 year follow-up. The primary endpoint was major adverse cardiac events (MACE), a composite outcome including all-cause mortality, recurrent MI, and percutaneous coronary intervention (PCI).
For the stroke cohort, data were gathered over an 8-14 year follow-up via electronic medical records and structured interviews. For deceased patients, caregiver interviews and death certificates were used. The primary endpoint was cerebrocardiovascular events (CCVE), defined as a composite of recurrent stroke, MI, or vascular death (VD).
The secondary endpoints consisted of the individual components of these primary composites. For the ACS cohort, these included all-cause mortality, cardiac death (defined as sudden death, death from arrhythmia, MI, heart failure, or cardiac procedures), recurrent MI (ST-elevation MI [STEMI], non-STEMI, or unstable angina requiring urgent catheterization), and recurrent PCI. For the stroke cohort, secondary outcomes included recurrent stroke (defined as a new neurologic deficit ≥24 hours or confirmed by neuroimaging evidence of a new lesion), MI, and VD (defined as death from ischemic/hemorrhagic stroke, MI, or other vascular causes). All potential endpoint events were formally adjudicated by independent clinical committees (cardiologists for ACS, neurologists for stroke) blinded to participants’ baseline psychiatric status.
Statistical analysis
Baseline characteristics were compared using independent t-tests or χ² tests as appropriate. To identify confounders, covariates were selected based on prior evidence [25] and bivariate associations (p<0.05) with the exposure or outcome, assessing for multicollinearity. Cox proportional hazards models were constructed to assess the association between education attainment and outcomes, presented as unadjusted and three hierarchical multivariable-adjusted models. Model 1 adjusted for age, sex, and vascular risk factors (hypertension, diabetes, hypercholesterolemia, obesity, previous and family history of ACS for ACS; smoking, cardiac disease, hypercholesterolemia, and previous stroke history for stroke). Model 2 adjusted for age, sex, and psychosocial, and disease severity factors (living alone, housing, employment, life stressors, depression comorbidity and treatment status, prior depression, ACS diagnosis, Killip class, LVEF, baseline troponin I and CK-MB for ACS; living alone, life stressors, NIHSS score, DSM-IV depression, prior depression for stroke). Model 3 combined all covariates from Models 1 and 2 for comprehensive adjustment. To test our hypothesis, an interaction term (educational attainment×SI) was introduced into the final multivariable model. A significant interaction was subsequently explored using stratified analyses to examine the effect of educational attainment within SI-present and SI-absent groups. We also tested for age-byeducation interactions in Model 3 to assess potential effect modification by age. To assess robustness, we repeated analyses using a 12-year education cutoff (high school completion), applying the same hierarchical models and stratification by SI status. All analyses were conducted separately for the ACS and stroke cohorts using SPSS (version 27.0, IBM Corp.), with two-tailed p-values <0.05 considered statistically significant.
RESULTS
Participants characteristics
The participant recruitment flow for both cohorts is detailed in Figure 1. From an initial 4,809 ACS patients and 567 stroke patients screened, the final analysis included 1,152 patients with ACS and 396 patients with stroke.
Baseline sociodemographic and clinical characteristics are detailed in Supplementary Tables 1-4. In the ACS cohort, patients with a lower educational level (≤10 years, n=578) were significantly older, more likely to be female, and had a higher prevalence of unemployment, hypertension, and diabetes compared to those with a higher education level (n=574). In the stroke cohort (n=396), those with lower educational level (n=247) were similarly older and more likely female, but conversely, were less likely to be current smokers.
Baseline SI was identified in 19.5% (n=225) of the ACS cohort and 15.2% (n=60) of the stoke cohort. The presence of SI was significantly associated with higher rates of comorbid depression, previous depression, presence of life stressors and greater disease severity (i.e., higher Troponin I levels or higher NIHSS scores).
Educational attainment and long-term prognosis
Lower educational attainment was associated with a significantly higher cumulative incidence of MACE (46.2% vs. 31.2%, p<0.001), all-cause mortality (23.9% vs. 12.7%, p<0.001), and cardiac death (12.1% vs. 7.1%, p=0.004) in the ACS cohort, while no such differences were observed in the stroke cohort (Table 1).
However, hierarchical Cox regression analyses (Table 2) revealed that these educational disparities in MACE (harard ratio [HR]=0.57, p<0.001) was substantially attenuated after adjusting for vascular risk factors in Model 1 (HR=0.85, p=0.145), with minimal further change in Model 2 (HR=0.91, p=0.424) and Model 3 (HR=0.90, p=0.327). Similar patterns were observed for all-cause mortality and cardiac death. In the stroke cohort, educational attainment showed no significant association with outcomes across any hierarchical models (Model 3 HR=1.06, p=0.805).
Moderating effect of SI on educational attainment
We next tested the hypothesis of a moderating effect of SI on the prognostic impact of educational attainment. The education ×SI interaction for MACE in the ACS cohort was statistically significant across all hierarchical models (Model 1 p=0.031; Model 2 p=0.042; Model 3 p=0.046). This interaction is illustrated in Figure 2A and detailed in Supplementary Table 5. Stratified analyses revealed distinct patterns by SI status. Among patients with SI (n=225), higher education was associated with a significant 46% reduction in MACE risk in the fully adjusted Model 3 (HR=0.54, 95% confidence interval [CI] 0.34-0.86, p=0.009), while no association existed among those without SI (n=927; HR=1.05, 95% CI 0.81-1.37, p=0.690). This significant interaction in the ACS cohort was specific to the composite MACE outcome and was not observed for secondary outcomes (all-cause mortality, cardiac death, recurrent MI, or recurrent PCI).
In the stroke cohort, the interaction patterns were similar, with statistical significance achieved in Models 1 and 3 (Model 1 p=0.043; Model 3 p=0.047) and marginal significance in Model 2 (p=0.063). This interaction is illustrated in Figure 2B and detailed in Supplementary Table 6. Similar to the ACS cohort, the interaction effect in the stroke cohort was observed for the CCVE, but not consistently across individual secondary outcomes (recurrent stroke, MI, or VD). We tested for ageby-education interactions in Model 3 to assess potential effect modification by age. No significant age modification was detected in either cohort (ACS: p=0.986; Stroke: p=0.361), indicating that the education-outcome association was consistent across age groups.
Sensitivity analyses
To evaluate robustness to different educational thresholds, we repeated analyses using a ≥12-year education cutoff (Supplementary Table 7). In the ACS cohort, the protective effect of higher education among patients with SI remained significant across all models (Model 3: HR=0.58, p=0.026), although the interaction term showed attenuated significance in the fully adjusted model (p=0.096). In the stroke cohort, interaction significance persisted in Models 1 and 3 (p=0.038 and p=0.040, respectively), but the stratified effect among SI patients did not reach statistical significance (Model 3: HR=0.68, p=0.529), likely reflecting limited statistical power in this smaller subgroup.
DISCUSSION
This prospective cohort study yielded three primary findings regarding the relationship between education, SI, and long-term prognosis after acute cardio-cerebrovascular events. First, educational attainment was not an independent predictor of long-term MACE or CCVE in patients with ACS or stroke. This finding was consistent across hierarchical adjustment models that sequentially controlled for vascular risk factors (Model 1), psychosocial/disease severity factors (Model 2), and their combination (Model 3), suggesting that the crude association was mediated by these covariates rather than representing an independent effect of education. Second, we identified a significant moderating effect of baseline SI on the relationship between education and outcomes, most robustly demonstrated in the ACS cohort. Third, in the ACS cohort, this interaction revealed a paradoxical protective effect: among patients with baseline SI, higher education was associated with a significantly reduced risk of MACE. A similar interaction pattern was observed in the stroke cohort, though with limited statistical power for stratified effects.
The attenuation of the association between education and outcomes after multivariable adjustment suggests that the significant unadjusted effect was mediated by the included covariates, such as disease severity, comorbidities, and demographic factors. Our findings are partly consistent with studies from Western countries reporting weak or nonsignificant associations between education and cardio-cerebrovascular outcomes after covariate adjustment, particularly in settings with broadly accessible or universal healthcare [5,6]. This finding, situated within Korea’s universal healthcare system, implies that equitable access to acute and long-term care may mitigate the direct prognostic disparities often attributed to education. In healthcare systems with greater access barriers, education often serves as a proxy for healthcare quality and access; our findings suggest this relationship may be less pronounced in a universal access environment.
The study’s most critical finding is the significant education ×SI interaction, which was consistent across all hierarchical models in the ACS cohort (p=0.031-0.046) and achieved significance in Models 1 and 3 of the stroke cohort (p=0.043, p=0.047). While acute SI is a known risk factor for poor outcomes and mortality following an cardio-cerebrovascular events, our results refine this association, suggesting the risk is not uniform across educational strata [26,27]. The paradoxical protective effect (HR=0.54) observed in highly educated patients with SI is notable. We hypothesize this may reflect differential health-seeking behaviors. Prior studies have shown that higher education is associated with healthier behaviors and reduced risk factors such as smoking, hypertriglyceridemia, depression and cardiovascular disease [28]. Highly educated individuals may possess greater health literacy, enabling them to recognize SI as a severe symptom and more effectively navigate the healthcare system for both mental and physical health support [29]. Conversely, acute SI in this group may represent a transient, acute distress response to the event, prompting enhanced adherence to secondary prevention measures and lifestyle modifications, thereby improving their long-term prognosis.
In the stroke cohort, the interaction pattern was directionally consistent but the stratified protective effect did not reach statistical significance (HR=0.68, p=0.529), likely reflecting the smaller sample size (n=396) and correspondingly lower statistical power. Furthermore, the unique pathophysiology of stroke [30] which can directly impact cognitive and functional capacity [31], may introduce different mechanisms linking education, psychosocial distress and health behaviors compared to ACS, potentially attenuating the buffering effect of education.
This study has several notable strengths, including its prospective design, long-term follow-up (5-14 years), and the inclusion of two distinct, well-characterized clinical cohorts. The use of comprehensive Cox models and formally adjudicated endpoints, blinded to psychiatric status, strengthens the internal validity of our findings. However, limitations must be acknowledged. First, the data originates from a single tertiary center, which may limit the generalizability of the findings. Second, the dichotomization of education (≤10 years vs. >10 years) is a simplification and may not capture the full doseresponse relationship of education. Sensitivity analyses using a ≥12-year cutoff demonstrated directional consistency, with the protective effect among ACS patients with SI remaining significant (HR=0.58, p=0.026), though the interaction showed some attenuation (p=0.096). Larger prospective studies are needed to establish precise educational thresholds and doseresponse patterns. Third, limited statistical power in the stroke cohort and among SI-positive subgroups precluded age-stratified analyses. However, tests for age-by-education interactions revealed no significant effect modification by age (ACS: p=0.986; stroke: p=0.361), suggesting consistent patterns across age groups. Fourth, SI was identified using a single MADRS item rather than a formal instrument; however, this approach has been applied in prior studies of patients with severe physical disorders [19] and depressive disorders [18], and has demonstrated good validity [20]. As SI was analyzed as a targeted construct rather than a comprehensive measure of psychological distress, future studies should examine whether other MADRS symptom dimensions or total scores provide additional prognostic value. Finally, despite our comprehensive adjustments, residual confounding from unmeasured variables cannot be entirely excluded.
In conclusion, educational attainment did not independently predict long-term outcomes after acute cardio-cerebrovascular events; instead, its prognostic impact was significantly moderated by acute-phase SI across hierarchical models in both cohorts. Higher education conferred substantial protection (46% risk reduction) specifically among ACS patients with SI, suggesting that health literacy and adaptive health behaviors buffer the adverse effects of psychological distress. Sensitivity analyses confirmed this interaction pattern. Clinically, these findings underscore the necessity of routine SI screening in all patients following acute cardio-cerebrovascular events. Lower-educated patients with SI represent a highrisk subgroup requiring targeted psychosocial support and enhanced care navigation. Integrating educational status into risk stratification protocols may identify vulnerable populations and guide tailored interventions to reduce outcome disparities.
PDF Links
PubReader
ePub Link
XML Download
Full text via DOI
Download Citation
Supplement1
Print
