Abstract
Background
Stroke disproportionately affects socioeconomically deprived populations, which leads to poor outcomes in disadvantaged groups. This study aimed to assess associations of socioeconomic status (SES) and long‐ and short‐term functional outcome after stroke in Mongolia.
Methods
A prospective, population‐based incidence stroke study was performed in Ulaanbaatar, Mongolia, between 2019 and 2020. SES indicators were average household family income and highest level of education attainment. The primary outcome was an unfavorable change in functional outcome on the binary modified Rankin Scale (scores 3–6 versus 0–2) measured at 28, 90, and 365 days. Logistic regression was used to analyze the association of SES and outcomes. Generalized linear mixed‐effects models were used to assess associations according to repeated measures of modified Rankin Scale over 12 months, and a mediation analysis aimed to determine the mechanism by which SES affects outcome.
Results
A total of 2205 patients (mean age, 58.0±12.8 years; 1330 [60.3%] men) were included. Patients with high‐level family income were less likely to have unfavorable short‐term (adjusted odds ratio, 0.36 [95% CI, 0.26–0.49]) and long‐term (adjusted odds ratio, 0.41 [95% CI, 0.30–0.58]) outcomes. Patients with higher SES experience faster recovery from disability within 1 year, and patients with intracerebral hemorrhage have the slowest recovery rate among the stroke subtypes. Mediation analysis revealed that the impact of SES on functional outcome was mostly direct rather than through other mediators.
Conclusions
Patients with higher SES exhibit a better recovery pattern up to 1 year after stroke. SES directly and significantly influenced stroke outcome.
Keywords: developing countries, outcome, socioeconomic status, stroke
Subject Categories: Intracranial Hemorrhage, Ischemic Stroke
Nonstandard Abbreviations and Acronyms
- HRQoL
health‐related quality of life
- ICH
intracerebral hemorrhage
- LMIC
low‐ and middle‐income country
- mRS
modified Rankin Scale
- NIHSS
National Institutes of Health Stroke Scale
Research Perspective.
What Is New?
In a prospective, population‐based study conducted in Mongolia, we observed that the association between socioeconomic status and stroke outcomes was consistent in both direction and magnitude at 28, 90, and 365 days after stroke, individuals in the medium socioeconomic status group exhibited more pronounced recovery after 90 days.
The association between family income and long‐term stroke outcomes was not mediated by covariates, stroke severity (measured by the modified Rankin Scale and National Institutes of Health Stroke Scale) at admission partially mediated the relationship between education and stroke outcomes.
What Question Should Be Addressed Next?
Future research should focus on investigating the underlying mechanisms linking socioeconomic status to stroke outcomes, including the potential influence of health care access and rehabilitation services on changes in outcomes over time.
Stroke is a major public health concern, with the global burden disproportionately affecting low‐ and middle‐income countries (LMICs), and is expected to increase with the increased aging population. 1 , 2 Stroke disproportionately affects socioeconomically deprived populations, leading to an increase in severity, 3 higher likelihood of unfavorable functional outcomes, 4 and higher case fatality. 5 , 6 Most evidence pertaining to the impact of stroke in LMICs comes from selected series and has not been specifically designed to investigate the influence of socioeconomic status (SES). Studies often used a single‐measure SES, and only a few covariables were included in an adjusted analysis, limiting both quality and generalizability of the data. 4 , 7 , 8 , 9 , 10 Additionally, few studies have specifically focused their attention on intracerebral hemorrhage (ICH) as compared with ischemic stroke, which has higher rates in LMIC settings. 4 , 7 , 8 , 11 , 12 To address these gaps, we conducted a population‐based study of adult residents (aged ≥16 years) of 6 urban districts of Ulaanbaatar, Mongolia, between 2019 and 2021. 13 Mongolia has one of the highest rates of stroke in the world, particularly for ICH, where the age‐standardized incidence (153.3 per 100 000 people), case fatality (98.1 per 100 000 people), and disability‐adjusted life years (2174.5 per 100 000 people) are the highest worldwide, according to the Global Burden of ICH Study 2021. 14 The high incidence of ICH has been attributed to the high prevalence of hypertension, 2 which is linked to excessive salt and alcohol consumption, 15 , 16 air pollution, 17 , 18 and inadequate fruit and vegetable intake. 16
In this population‐based study, we aimed to determine associations of individual‐level SES measures and functional outcomes after stroke overall and by subtypes. We hypothesized that patients with lower SES would have a higher risk of experiencing unfavorable functional outcomes and death in the short and long term. We also examined whether the patterns of functional recovery differ by SES up to 1 year after stroke. Mediation analysis was conducted to evaluate multiple mediators and determine whether the impact of SES on long‐term functional outcomes is primarily direct or mediated through an indirect pathway of vascular risk factors.
Methods
Data Availability Statement
Individual deidentified participant data may be shared upon request with qualified investigators, provided that the request complies with the Mongolian Law on Personal Data Protection and is approved through an appropriate research protocol and a signed data access agreement via the Research Office of the George Institute for Global Health, Sydney, New South Wales, Australia.
Study Population
The Mongolia stroke study was a prospective, population‐based incidence study with 3803 stroke cases, conducted in adult (aged ≥16 years) residents of 6 urban districts of Ulaanbaatar, Mongolia (population person‐years, N=1 896 965), between January 1, 2019, and December 31, 2020. All cases of stroke were identified through surveillance of multiple overlapping sources of hospitalized, ambulatory, and deceased individuals, using standardized diagnostic criteria. Further details of the study can be found elsewhere. 13 The study was approved by the ethics committee of the Ministry of Health, Mongolia (97/20190318), and written consent was obtained from all participants or appropriate surrogates.
Data Collection Procedures
Details on sociodemographic, medical history, clinical features, transport to hospital, management, and in‐hospital outcomes were obtained for all registered participants. This information was obtained as soon as possible after patients were notified by the study office and was entered directly by hospital physicians into the study database, according to standard procedures and with subsequent checks made against source documents by research staff. Sociodemographic data were marital status, highest attained level of education, employment status, household family size, and self‐reported monthly household family income. Details of the participant’s medical history (hypertension, diabetes, hypercholesterolemia, atrial fibrillation) and corresponding drug therapy were determined by self‐report or extracted from medical records. Information on smoking status (ie, lifelong nonsmoker, current, or past [stopped within the previous 6 months]) and regular alcohol consumption were also determined in a similar way. Body mass index is an estimate of body fat and was calculated by dividing a person’s weight (in kilograms) by the square of their height (in meters). Stroke cases were diagnosed primarily using computed tomography or magnetic resonance brain imaging. Trained neurologists used structured questionnaires to assess the participants’ disability (modified Rankin Scale [mRS]) and neurological impairment (the National Institutes of Health Stroke Scale) of participants at admission. Patients were followed‐up 3 times through telephone interviews conducted by research staff at 28, 90, and 365 days after the onset of symptoms. 13
Outcomes
The primary outcome was functional status, defined by the distribution of scores on the mRS. For analyses involving the primary outcome of functional status (ordinal shift in the distribution of scores on the mRS), we first evaluated the proportional odds assumption, which was violated in all cases; we thus used a binary outcome of the mRS (scores 0–2 [nondisabled] versus 3–6 [death or disabled]). Other secondary outcomes were: mRS scores 0 to 2 (nondisabled) versus 3–5 (major disability); health‐related quality of life (HRQoL) evaluated on the EuroQol‐5 Dimensions‐3 Levels scale (above and below the median), a standardized instrument that assesses HRQoL on the basis of 5 dimensions (mobility, self‐care, usual activities, pain/discomfort, and anxiety/depression), with each related to 5 levels of severity 19 and death. The outcome of the analysis was the adjusted odds ratio (aOR) and 95% CI for an unfavorable outcome comparing levels of mRS and HRQoL.
SES Indicators
In this analysis, the indicators of SES were household family income and level of education. Monthly household income was divided into low (≤700 000 Mongolian Tughrik), medium (>700 000 to ≤1 000 000 Mongolian Tughrik), and high (>1 000 000 Mongolian Tughrik) (100 000 Mongolian Tughrik=US$34.9), to balance the number of participants in each category. The level of education was categorized from low to high: elementary school or no education, middle school or vocational training, and high school or above; participants with unknown family income and education were excluded from the analysis.
Statistical Analysis
Analyses were conducted on participants with data available on stroke history, SES, and stroke subtypes; 841 participants with previous stroke were excluded from the study. Additionally, 755 participants missing SES indicators and 2 participants with known stroke subtypes were excluded. Finally, a total of 2205 participants were included in this study. The remaining covariates had missing values <10%; thus, it was not deemed necessary to perform multiple imputation. 20 Patient characteristics were compared using the Kruskal–Wallis test or χ2 test. Age, sex, and statistically significant variables from univariate analysis were included as covariates in multivariable logistic regression models to estimate the association between SES and outcome across all stroke subtypes; family size was added as an extra covariate to the family income model. Effect modification by subgroup was tested by fitting models with main effects for income and education, the subgroup variable all the covariates used in the primary analyses, as well as all the interactions between the subgroup variable. When the P value for the interaction between the subgroup variable and income was <0.05, we took this to be evidence of heterogeneity. A generalized linear mixed‐effects model was used to investigate the impact of SES on functional recovery trajectories, 21 with adjustment for age, sex, and time (28, 90, and 365 days). The recovery trajectory is plotted as a line chart, and the analysis was performed using the “lme4” R package (R Foundation for Statistical Computing, Vienna, Austria).
To separate the direct and indirect effects of SES on long‐term mRS (score 0–2 versus 3–6), a parallel mediation analysis was performed using structural equation modeling to test the effect of multiple mediators simultaneously. 22 The direct effect reflects the part of the association that is directly attributable to the effect of SES on the outcome, while the indirect effect reflects the part of the association that can be entirely explained by the risk factors, comorbidities, and neurological severity upon admission to the hospital. The total effect is defined as the sum of direct and indirect effects. Path coefficients (β), defined as the response of dependent variables to 1 unit change in independent variables, were estimated using the maximum likelihood method; SD and P values were estimated via nonparametric bootstrapping (1000 replications) (Data S1). The mediation analysis was performed using the “lavaan” package in R. 23 Statistical significance was set at a P value of 0.05. Given the exploratory design of this study, we did not apply formal corrections for multiple comparisons. Our goal was to identify potential signals for future investigation, and we caution readers to interpret P values accordingly. All analyses were performed using R version 4.3.3.
Results
Baseline Characteristics
Figure S1 is a flowchart for the study design. Overall, 2205 patients (mean age, 58.0±12.8 years; 1330 [60.3%] men) were included in this analysis (Table 1). Compared with patients with a low income, patients with higher family income were younger (low: 60.0±12.2 years; medium: 57.50±12.7 years; high: 55.0±13.2 years), more likely to be married, had higher body mass index, were less likely to have had a severe stroke on admission, and to have had hypertension and hypercholesterolemia. Patients with higher education exhibited a similar trend. Most patients were diagnosed with ischemic stroke (54.6%), followed by ICH (34.4%) and subarachnoid hemorrhage (11.1%). At hospital admission, 16.7% of participants had a premorbid mRS score of 0 to 1, and the median National Institutes of Health Stroke Scale was 6 (interquartile range, 3–13). A summary of follow‐up outcomes on 28, 90, and 365 days can be found in Table S1.
Table 1.
Baseline Characteristics of Patients According to Household Family Income and Education
| Variable | Total (N=2205) | Family income | Education | ||||
|---|---|---|---|---|---|---|---|
| Low (n=798) | Medium (n=772) | High (n=635) | Low (n=118) | Medium (n=1119) | High (n=968) | ||
| Sociodemographic | |||||||
| Age, y, mean±SD | 58.0±12.8 | 60.0±12.2 | 57.50±12.7 | 55.0±13.2 | 73.5±13.9 | 57.0 (12.0) | 57.0±12.8 |
| Male sex, n/N (%) | 1330/2205 (60.3) | 479/798 (60.0) | 459/772 (59.5) | 392/635 (61.7) | 61/118 (51.7) | 654/1119 (58.4) | 615/968 (63.5) |
| Marital status, n/N (%) | |||||||
| Divorced/widowed | 394/2202 (17.9) | 242/796 (30.4) | 94/771 (12.2) | 58/635 (9.1) | 49/118 (41.5) | 192/1117 (17.2) | 153/967 (15.8) |
| Married | 1553/2,202 (70.5) | 449/796 (56.4) | 606/771 (78.6) | 498/635 (78.4) | 47/118 (39.8) | 793/1117 (71.0) | 713/967 (73.7) |
| Single | 255/2202 (11.6) | 105/796 (13.2) | 71/771 (9.2) | 79/635 (12.4) | 22/118 (18.6) | 132/1117 (11.8) | 101/967 (10.4) |
| Vascular risk factors | |||||||
| Body mass index, mean±SD | 26.4±4.2 | 25.9±4.5 | 26.6±3.9 | 26.7±4.2 | 25.5±4.5 | 26.1±4.3 | 26.6±4.1 |
| Hypertension, n/N (%) | 1350/2088 (64.7) | 552/740 (74.6) | 468/738 (63.4) | 330/610 (54.1) | 91/106 (85.8) | 683/1059 (64.5) | 576/923 (62.4) |
| Hypercholesterolemia, n/N (%) | 103/2200 (4.7) | 27/796 (3.4) | 48/769 (6.2) | 28/635 (4.4) | 7/118 (5.9) | 27/1116 (2.4) | 69/966 (7.1) |
| Diabetes, n/N (%) | 278/2200 (12.6) | 102/796 (12.8) | 105/769 (13.7) | 71/635 (11.2) | 12/118 (10.2) | 128/1116 (11.5) | 138/966 (14.3) |
| Atrial fibrillation, n/N (%) | 96/2200 (4.4) | 27/796 (3.4) | 33/769 (4.3) | 36/635 (5.7) | 7/118 (5.9) | 44/1116 (3.9) | 45/966 (4.7) |
| Smoking, n/N (%) | 589/2200 (26.8) | 217/796 (27.3) | 209/769 (27.2) | 163/635 (25.7) | 25/118 (21.2) | 314/1,116 (28.1) | 250/966 (25.9) |
| Regular alcohol consumption, n/N (%) | 527/2200 (24.0) | 185/796 (23.2) | 186/769 (24.2) | 156/635 (24.6) | 24/118 (20.3) | 288/1116 (25.8) | 215/966 (22.3) |
| NIHSS at admission, median (IQR) | 6.0 (3.0–13.0) | 6.0 (3.0–11.0) | 6.0 (3.0–13.0) | 5.0 (2.0–13.0) | 7.0 (4.0–13.0) | 6.0 (3.0–15.0) | 5.0 (3.0–11.0) |
| Estimate premorbid function, n/N (%)* | |||||||
| 1 | 368/2205 (16.7) | 98/798 (12.3) | 121/772 (15.7) | 149/635 (23.5) | 12/118 (10.2) | 148/1119 (13.2) | 208/968 (21.5) |
| 2 | 557/2205 (25.3) | 205/798 (25.7) | 211/772 (27.3) | 141/635 (22.2) | 24/118 (20.3) | 302/1119 (27.0) | 231/968 (23.9) |
| 3 | 405/2205 (18.4) | 185/798 (23.2) | 123/772 (15.9) | 97/635 (15.3) | 27/118 (22.9) | 206/1119 (18.4) | 172/968 (17.8) |
| 4 | 399/2205 (18.1) | 149/798 (18.7) | 136/772 (17.6) | 114/635 (18.0) | 25/118 (21.2) | 187/1119 (16.7) | 187/968 (19.3) |
| 5 | 476/2205 (21.6) | 161/798 (20.2) | 181/772 (23.4) | 134/635 (21.1) | 30/118 (25.4) | 276/1119 (24.7) | 170/968 (17.6) |
| Blood pressure–lowering medicine, n/N (%) | 1557/2180 (71.4) | 590/786 (75.1) | 543/762 (71.3) | 424/632 (67.1) | 96/117 (82.1) | 778/1107 (70.3) | 683/956 (71.4) |
| Stroke subtype, n/N (%) | |||||||
| Ischemic stroke | 1203/2205 (54.6) | 452/798 (56.6) | 430/772 (55.7) | 321/635 (50.6) | 68/118 (57.6) | 600/1119 (53.6) | 535/968 (55.3) |
| ICH | 758/2205 (34.4) | 283/798 (35.5) | 256/772 (33.2) | 219/635 (34.5) | 47/118 (39.8) | 405/1119 (36.2) | 306/968 (31.6) |
| Subarachnoid hemorrhage | 244/2205 (11.1) | 63/798 (7.9) | 86/772 (11.1) | 95/635 (15.0) | 3/118 (2.5) | 114/1119 (10.2) | 127/968 (13.1) |
ICH indicates intracerebral hemorrhage; IQR, interquartile range; mRS, modified Rankin Scale; and NIHSS, National Institutes of Health Stroke Scale.
Estimated on the mRS.
Association of SES and Functional Outcome
Patients with higher family income were less likely to have an unfavorable functional outcome for any stroke, compared with those with low income on day 28 (medium income: aOR, 0.70 [95% CI, 0.52–0.95]; high income: aOR, 0.41 [95% CI, 0.30–0.58]; missing values in Table S2), this association remained consistent at 90 and 365 days. Similar patterns were also observed for major disability (mRS score 3–5) at 28 days (medium: aOR, 0.68 [95% CI, 0.50–0.91]; high: aOR, 0.41 [95% CI, 0.29–0.57]), EuroQol‐5 Dimensions‐3 Levels at 28 days (medium: aOR, 0.56 [95% CI, 0.42–0.75]; high: aOR, 0.32 [95% CI, 0.23–0.45]) and death at 28 days (medium: aOR, 1.02 [95% CI, 0.64–1.63]; high: aOR, 0.53 [95% CI, 0.30–0.92]; Figure 1). Similar patterns were observed for education. We further examined the association for each stroke subtypes concerning family income and found that, while the overall pattern remained consistent, the odds ratios increased over time for all outcomes (Table 2). This suggests that the protective effect of income advantage diminishes as time progresses for patients with ICH. However, such a difference was not observed for education (Table S3). No significant heterogeneity was found for any subgroup analysis (Figure 2).
Figure 1. Multivariable logistic regression analysis of the association between socioeconomic status and unfavorable outcomes at 28, 90 and 365 days after the onset of stroke.
mRS scores were split to 0 to 2 vs 3 to 6; 0 to 2 vs 3 to 5; EQ‐5D‐3L was split on the basis of median; living status was based on the survival of participants throughout the follow‐ups. Unfavorable outcomes are defined as mRS score of 3 to 6, mRS score of 3 to 5, EQ‐5D‐3L less than median, and death. Smaller aOR indicates participants are less likely to experience unfavorable outcome. Models for mRS were adjusted for age, sex, marital status, hypertension, hypercholesterolemia, smoking, NIHSS at admission, mRS at admission, and stroke type. For EQ‐5D‐3L, adjustments were made for age, sex, marital status, hypertension, hypercholesterolemia, NIHSS at admission, mRS at admission, administer blood pressure–lowering drug, and stroke type. For living status, adjustments were made for age, sex, body mass index, hypertension, hypercholesterolemia, diabetes, NIHSS at admission, mRS at admission, administration of blood pressure–lowering drug, and stroke type. Major disability defined as mRS score of 3 to 5. aOR indicates adjusted odds ratio; EQ‐5D‐3L, EuroQol‐5 Dimensions‐3 Levels; NIHSS, National Institutes of Health Stroke Scale; mRS, modified Rankin Scale.
Table 2.
Analysis of the Association Between Household Family Income and Functional Outcome by Stroke Subtypes
| Family income | Ischemic stroke, aOR (95% CI) | Intracerebral hemorrhage, aOR (95% CI) | Subarachnoid hemorrhage, aOR (95% CI) | |
|---|---|---|---|---|
| mRS (3–6) | ||||
| 28 d | Medium vs low | 0.73 (0.50–1.05) | 0.47 (0.24–0.90) | 1.28 (0.46–3.58) |
| High vs low | 0.41 (0.27–0.62) | 0.24 (0.12–0.49) | 0.87 (0.31–2.49) | |
| 90 d | Medium vs low | 0.47 (0.33–0.67) | 0.55 (0.31–0.96) | 0.24 (0.08–0.68) |
| High vs low | 0.36 (0.23–0.54) | 0.27 (0.14–0.51) | 0.22 (0.07–0.67) | |
| 365 d | Medium vs low | 0.46 (0.32–0.64) | 0.62 (0.37–1.02) | 0.29 (0.10–0.80) |
| High vs low | 0.33 (0.22–0.50) | 0.37 (0.20–0.65) | 0.39 (0.13–1.15) | |
| mRS (major disability) | ||||
| 28 d | Medium vs Low | 0.74 (0.52–1.06) | 0.45 (0.23–0.86) | 1.22 (0.43–3.54) |
| High vs low | 0.45 (0.30–0.67) | 0.24 (0.11–0.48) | 0.86 (0.29–2.59) | |
| 90 d | Medium vs Low | 0.48 (0.34–0.68) | 0.50 (0.28–0.87) | 0.18 (0.05–0.59) |
| High vs Low | 0.41 (0.27–0.61) | 0.25 (0.13–0.47) | 0.17 (0.05–0.61) | |
| 365 d | Medium vs low | 0.44 (0.31–0.62) | 0.53 (0.31–0.91) | 0.11 (0.02–0.49) |
| High vs low | 0.31 (0.21–0.48) | 0.34 (0.18–0.61) | 0.25 (0.05–1.05) | |
| EQ‐5D‐3L (less than median) | ||||
| 28 d | Medium vs Low | 0.64 (0.45–0.90) | 0.50 (0.28–0.89) | 0.21 (0.07–0.59) |
| High vs low | 0.32 (0.21–0.49) | 0.28 (0.15–0.52) | 0.27 (0.09–0.76) | |
| 90 d | Medium vs low | 0.52 (0.37–0.74) | 0.60 (0.35–1.04) | 0.15 (0.04–0.49) |
| High vs low | 0.33 (0.22–0.49) | 0.38 (0.20–0.68) | 0.28 (0.08–0.89) | |
| 365 d | Medium vs low | 0.53 (0.38–0.75) | 0.60 (0.35–1.02) | 0.17 (0.05–0.50) |
| High vs low | 0.29 (0.19–0.42) | 0.42 (0.23–0.74) | 0.25 (0.08–0.76) | |
| Living status (death) | ||||
| 28 d | Medium vs low | 1.08 (0.52–2.25) | 0.88 (0.42–1.82) | 1.78 (0.49–6.88) |
| High vs low | 0.48 (0.19–1.16) | 0.49 (0.21–1.12) | 1.04 (0.23–4.73) | |
| 90 d | Medium vs low | 0.92 (0.49–1.72) | 0.93 (0.47–1.83) | 1.12 (0.33–3.93) |
| High vs low | 0.35 (0.15–0.79) | 0.63 (0.29–1.38) | 0.68 (0.15–2.82) | |
| 365 d | Medium vs low | 0.91 (0.55–1.51) | 0.99 (0.54–1.80) | 0.77 (0.24–2.47) |
| High vs low | 0.61 (0.33–1.12) | 0.67 (0.33–1.34) | 0.61 (0.16–2.29) | |
Models for mRS were adjusted for age, sex, marital status, hypertension, hypercholesterolemia, smoking, NIHSS at admission, mRS at admission, and stroke type. For EQ‐5D‐3L, adjustments were made for age, sex, marital status, hypertension, hypercholesterolemia, NIHSS at admission, mRS at admission, administration of blood pressure–lowering drug, and stroke type. For living status, adjustments were made for age, sex, body mass index, hypertension, hypercholesterolemia, diabetes, NIHSS at admission, mRS at admission, administration of blood pressure–lowering drug, and stroke type. Major disability defined as mRS score 3 to 5. aOR indicates adjusted odds ratio; EQ‐5D‐3L, EuroQol‐5 Dimensions‐3 Levels; mRS, modified Rankin Scale; and NIHSS, National Institutes of Health Stroke Scale.
Figure 2. Subgroup analysis of the association between household family income and functional outcome (mRS scores of 3–6) at 365 days.
For age, younger is defined as age median or less, while older is age greater than median. For NIHSS, minor is defined as NIHSS score ≤4, moderate is NIHSS score of 4 to ≤15, and severe is NIHSS score >15. For mRS, minor is defined as an mRS score of 0 to 2, and severe is a score of 3 to 6. mRS scores were split to 0 to 2 vs 3 to 6. Unfavorable outcomes are defined as an mRS score of 3 to 6. Smaller aOR indicates participants are less likely to experience unfavorable outcome. aOR indicates adjusted odds ratio; ICH, intracerebral hemorrhage; IS, ischemic stroke; mRS, modified Rankin Scale; NIHSS, National Institutes of Health Stroke Scale; and SAH, subarachnoid hemorrhage.
Impact of SES on Recovery Patterns
Patients in higher SES groups had a lower probability of an unfavorable outcome over time compared with those in lower SES groups, indicating faster recovery for those with higher SES. For family income (Figure 3A), both low‐ and high‐income groups followed a consistent recovery trajectory throughout the follow‐up period, whereas patients in the medium‐income group showed a more pronounced recovery between 90 and 365 days compared with the earlier period of 28 to 90 days. A similar trend was observed for education (Figure 3B), although its overall effect on recovery was less impactful than family income.
Figure 3. Generalized linear mixed‐effect models comparing different recovery trajectories according to socioeconomic status indicators and stroke subtypes at 28, 90 and 365 days after stroke.
mRS scores were split to 0 to 2 vs 3 to 6. Unfavorable outcome is defined as an mRS score of 3 to 6. Steeper trend means faster recovery. A, Recovery trajectory for patients with different levels of family income. B, Recovery trajectory for patients with different levels of education attainment. C, Recovery trajectory for each stroke subtype according to family income. D, Recovery trajectory for each stroke subtype for according to education. ICH indicates intracerebral hemorrhage; IS, ischemic stroke; mRS, modified Rankin Scale; and SAH, subarachnoid hemorrhage.
For stroke subtypes, patients with ischemic stroke and subarachnoid hemorrhage experienced a similar recovery trend when adjusted with family income and education (Figure 3C and 3D). However, for ICH, both family income and education had little impact on the recovery throughout follow‐up. Further analysis showed that the level of family income slightly affects the recovery rate (Figure S2). An analysis was also conducted to investigate the joint effect of family income and education, which suggested a gradient effect of SES (Figure S3). Participants with lower SES had a consistently higher risk of stroke, and the disparity between socioeconomic groups became more pronounced over the follow‐up period. However, these results should be interpreted with caution, due to the small sample sizes in some joint family income and education categories.
Mediation Analysis
An illustration of parallel mediation pathway is shown in Figure S4, and the summary of mediating effects is presented in Table 3. The total effect of family income was statistically significant (β=−0.215 [95% CI, −0.284 to −0.134]). Most of the effect was directly attributable to family income (β=−0.219 [95% CI, −0.285 to −0.147]), while the indirect effects were small and not statistically significant (β=0.004 [95% CI, −0.042 to 0.052]); none of the mediators had a statistically significant effect. In contrast, the indirect effect (β=−0.099 [95% CI, −0.156 to −0.045]) of education contributed significantly to the total effect (β=−0.252 [95% CI, −0.358 to −0.151]), while National Institutes of Health Stroke Scale admission (β=−0.051 [95% CI, −0.097 to −0.015]) and hypercholesterolemia (β=−0.016 [95% CI, −0.003 to −0.005]) were statistically significant mediators. This suggests that the association between family income and stroke is not mediated by covariates such as comorbidities and neurological severity on admission, while the covariates, particularly mRS and National Institutes of Health Stroke Scale at admission, partially mediate education and stroke outcome.
Table 3.
Summary of Mediation Analysis at 365 Days
| Family income | P value | Education | P value | |
|---|---|---|---|---|
| Regression coefficient (95% CI) | Regression coefficient (95% CI) | |||
| Total effect | −0.215 (−0.284 to −0.134) | <0.001 | −0.252 (−0.358 to −0.151) | <0.001 |
| Direct effect | −0.219 (−0.285 to −0.147) | <0.001 | −0.154 (−0.249 to −0.058) | 0.002 |
| Indirect effect | 0.004 (−0.042 to 0.052) | 0.861 | −0.099 (−0.156 to −0.045) | <0.001 |
| mRS admission | −0.004 (−0.02 to 0.012) | 0.636 | −0.032 (−0.065 to −0.006) | 0.056 |
| NIHSS admission | 0.010 (−0.018 to 0.038) | 0.494 | −0.051 (−0.097 to −0.015) | 0.022 |
| Hypertension | 0.004 (−0.006 to 0.016) | 0.451 | 0.001 (−0.005 to 0.007) | 0.837 |
| Hypercholesterolemia | −0.003 (−0.009 to 0.002) | 0.203 | −0.016 (−0.03 to −0.005) | 0.015 |
| Diabetes | 0.001 (−0.002 to 0.004) | 0.664 | 0.000 (−0.005 to 0.004) | 0.912 |
| Atrial fibrillation | 0.003 (−0.001 to 0.009) | 0.198 | 0.001 (−0.002 to 0.007) | 0.654 |
| Body mass index | −0.008 (−0.021 to 0.003) | 0.152 | −0.005 (−0.018 to 0.003) | 0.361 |
| Smoking | 0.001 (−0.001 to 0.006) | 0.473 | 0.003 (−0.002 to 0.01) | 0.377 |
| Alcohol | 0.000 (−0.002 to 0.002) | 0.998 | 0.001 (−0.006 to 0.008) | 0.781 |
Ordinal mRS (score 0–6) and NIHSS (score 0–42) were used at admission. The model fit indices were CFI=0.997, RMSEA=0.039, SRMR=0.075 for family income; and CFI=0.998, RMSEA=0.036, SRMR=0.074 for education. CFI indicates comparative fit index; mRS, modified Rankin Scale; NIHSS, National Institutes of Health Stroke Scale; RMSEA, root mean square error of approximation; and SRMR, standardized root mean square residual.
Discussion
Our population‐based stroke study shows that patients with high SES have a better functional outcome and HRQoL after stroke. The association was confirmed in an analysis with repeated measures of mRS, where patients with high SES were observed to have a better pattern of functional recovery over 1 year. Mediation analysis further revealed that SES had a direct influence on the outcome after stroke.
Our results offer comprehensive insight into the impact of SES on stroke outcomes in one of the highest‐stroke‐risk countries (Mongolia). In comparison with previous research, we minimized the potential for selection bias and included patients with a diverse range of SES indicators. Our findings align with earlier studies from LMICs that identified SES as a key predictor of unfavorable functional outcomes, albeit not within a population‐based research framework. 10 , 24 , 25 Our study provides a more comprehensive analysis of SES factors through the use of level of education and household income as key measures. Furthermore, we examined the association between SES and functional outcomes after stroke in a population with a high proportion of ICH cases, whereas previous studies have often included mixed types of stroke populations with a limited number of ICH (Table S4). Additionally, our results demonstrate that both family income and education are significantly associated with poststroke HRQoL, in both the short and long term. Family income was found to have a greater impact on stroke outcomes than education, which could be explained by participants from wealthier families being more often able to afford rehabilitation. Since Mongolia is an LMIC with limited medical resources, variation in income may contribute to greater disparities in stroke outcomes than would be the case in the West.
Moreover, whereas previous studies on SES have used a single point of measurement and seldom tracked changes in disability through the recovery process, 4 our study incorporated multiple assessments to investigate the impact of SES on stroke recovery during a follow‐up period of 12 months, as well as differences in recovery across stroke subtypes on the basis of SES.
Our mediation analysis indicates that the association between SES and functional outcome is primarily driven by SES factors themselves. This may be attributed to many patients in LMICs facing socioeconomic deprivation after stroke and other critical conditions due to the influence of comorbidities, neurological impairment, poor health literacy, and limited access to services, including rehabilitation. 5 , 26 , 27 Disparities in care reemphasize the need for health care providers to adopt inclusive strategies that address socioeconomic barriers. Specifically, stroke prevention programs should be tailored to target high‐risk and low‐SES groups by improving access to preventive care and health education, promoting healthier lifestyles, and ensuring equitable access to care and rehabilitation services.
Key strengths of our study include the large population‐based design and comprehensive assessment of functional outcomes that ensure reliable and valid results that are generalizable to similar to other LMIC populations. However, several limitations should be noted. First, some data, such as those pertaining to vascular risk factors, were self‐reported, which may introduce responder bias. This could help explain why the association between SES indicators and stroke outcome was not mediated by vascular risk factors. Second, the follow‐up on the functional outcome at 28, 90, and 365 days was conducted through telephone interview, and the lack of physical examination of the outcome, due to the COVID‐19 pandemic, may give an inaccurate estimation. Third, multiple testing of associations gives a high chance of false positives, and thus, our results must be interpreted with caution.
In summary, our population‐based study of stroke in Mongolia demonstrated that patients with higher SES exhibit a better pattern of functional recovery over the subsequent year for both ischemic stroke and ICH. Additionally, SES directly and significantly influenced stroke outcome.
Sources of Funding
This work was supported by National Health and Medical Research Council investigator grant APP1195237.
Disclosures
None.
Supporting information
Data S1: Supplemental Methods
Tables S1–S4
Figures S1–S4
Acknowledgments
This manuscript was prepared using the Mongolia stroke incidence funded by the Science and Technology Foundation of the Ministry of Education, Culture, and Science of Mongolia and the George Institute for Global Health. The authors thank the participants and the cooperative research group for their invaluable contributions to this study. C.S.A. and X.W. receive fellowship funding from the National Health and Medical Research Council of Australia.
This manuscript was sent to Ajay K. Gupta, MD, MSc, PhD, FRCP, FESC, Senior Associate Editor, for review by expert referees, editorial decision, and final disposition.
Supplemental Material is available at https://www.ahajournals.org/doi/suppl/10.1161/JAHA.124.040282
For Sources of Funding and Disclosures, see page 10.
References
- 1. Feigin VL, Forouzanfar MH, Krishnamurthi R, Mensah GA, Connor M, Bennett DA, Moran AE, Sacco RL, Anderson L, Truelsen T. Global and regional burden of stroke during 1990–2010: findings from the global burden of disease study 2010. Lancet. 2014;383:245–255. doi: 10.1016/S0140-6736(13)61953-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Feigin VL, Stark BA, Johnson CO, Roth GA, Bisignano C, Abady GG, Abbasifard M, Abbasi‐Kangevari M, Abd‐Allah F, Abedi V. Global, regional, and national burden of stroke and its risk factors, 1990–2019: a systematic analysis for the global burden of disease study 2019. Lancet Neurol. 2021;20:795–820. doi: 10.1016/S1474-4422(21)00252-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Yamanie N, Chalik Sjaaf A, Felistia Y, Harry Susanto N, Diana A, Lamuri A, Miftahussurur M. High socioeconomic status is associated with stroke severity among stroke patients in the National Brain Centre Hospital, Jakarta, Indonesia. Prev Med Rep. 2023;32:102170. doi: 10.1016/j.pmedr.2023.102170 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Nguyen MTH, Sakamoto Y, Maeda T, Woodward M, Anderson CS, Catiwa J, Yazidjoglou A, Carcel C, Yang M, Wang X. Influence of socioeconomic status on functional outcomes after stroke: a systematic review and meta‐analysis. J Am Heart Assoc. 2024;13:e033078. doi: 10.1161/JAHA.123.033078 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5. Addo J, Ayerbe L, Mohan KM, Crichton S, Sheldenkar A, Chen R, Wolfe CD, McKevitt C. Socioeconomic status and stroke: an updated review. Stroke. 2012;43:1186–1191. doi: 10.1161/strokeaha.111.639732 [DOI] [PubMed] [Google Scholar]
- 6. Bray BD, Paley L, Hoffman A, James M, Gompertz P, Wolfe CDA, Hemingway H, Rudd AG. Socioeconomic disparities in first stroke incidence, quality of care, and survival: a nationwide registry‐based cohort study of 44 million adults in England. Lancet Public Health. 2018;3:e185–e193. doi: 10.1016/s2468-2667(18)30030-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Farzadfard MT, Sheikh Andalibi MS, Thrift AG, Morovatdar N, Stranges S, Amiri A, Kapral MK, Behrouz R, Juibary AG, Mokhber N. Long‐term disability after stroke in Iran: evidence from the Mashhad stroke incidence study. Int J Stroke. 2019;14:44–47. doi: 10.1177/1747493018789839 [DOI] [PubMed] [Google Scholar]
- 8. Saberi A, Hosseininezhad M, Rezaei S, Salehil S, Neshin SAS, Hasanzadeh B. The effect of cognitive impairment on functional outcome after stroke: a cross sectional study. Act Nerv Super Rediviva. 2020;62:41–46. [Google Scholar]
- 9. Liu X, Lv Y, Wang B, Zhao G, Yan Y, Xu D. Prediction of functional outcome of ischemic stroke patients in northwest China. Clin Neurol Neurosurg. 2007;109:571–577. doi: 10.1016/j.clineuro.2007.05.008 [DOI] [PubMed] [Google Scholar]
- 10. Ouyang F, Wang Y, Huang W, Chen Y, Zhao Y, Dang G, Zhang C, Lin Y, Zeng J. Association between socioeconomic status and post‐stroke functional outcome in deprived rural southern China: a population‐based study. BMC Neurol. 2018;18:1–9. doi: 10.1186/s12883-018-1017-4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Abanto C, Ton TG, Tirschwell DL, Montano S, Quispe Y, Gonzales I, Valencia A, Calle P, Garate A, Zunt J. Predictors of functional outcome among stroke patients in Lima, Peru. J Stroke Cerebrovasc Dis. 2013;22:1156–1162. doi: 10.1016/j.jstrokecerebrovasdis.2012.11.021 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Marshall IJ, Wang Y, Crichton S, McKevitt C, Rudd AG, Wolfe CD. The effects of socioeconomic status on stroke risk and outcomes. Lancet Neurol. 2015;14:1206–1218. doi: 10.1016/s1474-4422(15)00200-8 [DOI] [PubMed] [Google Scholar]
- 13. Banzrai C, Bosookhuu O, Yadamsuren E, Dambasuren B, Turbat S, Erdenedalai T, Myadagsuren M, Munkhtur U, Baatar K, Boldbayar P, et al. Incidence and outcomes for stroke in Ulaanbaatar, Mongolia, during 2019–21: a prospective population‐based study. Lancet Glob Health. 2023;11:e942–e952. doi: 10.1016/s2214-109x(23)00130-4 [DOI] [PubMed] [Google Scholar]
- 14. Xu L, Wang Z, Wu W, Li M, Li Q. Global, regional, and national burden of intracerebral hemorrhage and its attributable risk factors from 1990 to 2021: results from the 2021 global burden of disease study. BMC Public Health. 2024;24:2426. doi: 10.1186/s12889-024-19923-7 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15. Enkhtungalag B, Batjargal J, Chimedsuren O, Tsogzolmaa B, Anderson CS, Webster J. Developing a national salt reduction strategy for Mongolia. Cardiovasc Diagn Ther. 2015;5:229. doi: 10.3978/j.issn.2223-3652.2015.04.11 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Pengpid S, Peltzer K. Trends in sedentary behaviour and associated factors among adults in Mongolia: results from national cross‐sectional surveys in 2009, 2013 and 2019. BMJ Open. 2022;12:e058330. doi: 10.1136/bmjopen-2021-058330 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Cousins S. Air pollution in Mongolia. Bull World Health Organ. 2019;97:79–80. doi: 10.2471/blt.19.020219 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Shah AS, Lee KK, McAllister DA, Hunter A, Nair H, Whiteley W, Langrish JP, Newby DE, Mills NL. Short term exposure to air pollution and stroke: systematic review and meta‐analysis. BMJ. 2015;350:h1295. doi: 10.1136/bmj.h1295 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19. Group TE . EuroQol—a new facility for the measurement of health‐related quality of life. Health Policy. 1990;16:199–208. doi: 10.1016/0168-8510(90)90421-9 [DOI] [PubMed] [Google Scholar]
- 20. Woodward M. Epidemiology: Study Design and Data Analysis. CRC Press; 2013. [Google Scholar]
- 21. Breslow NE, Clayton DG. Approximate inference in generalized linear mixed models. J Am Stat Assoc. 1993;88:9–25. doi: 10.1080/01621459.1993.10594284 [DOI] [Google Scholar]
- 22. Gunzler D, Chen T, Wu P, Zhang H. Introduction to mediation analysis with structural equation modeling. Shanghai Arch Psychiatry. 2013;25:390–394. doi: 10.3969/j.issn.1002-0829.2013.06.009 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Rosseel Y. Lavaan: AnRPackage for structural equation modeling. J Stat Softw. 2012;48:1–36. doi: 10.18637/jss.v048.i02 [DOI] [Google Scholar]
- 24. Song T, Pan Y, Chen R, Li H, Zhao X, Liu L, Wang C, Wang Y, Wang Y. Is there a correlation between socioeconomic disparity and functional outcome after acute ischemic stroke? PLoS One. 2017;12:e0181196. doi: 10.1371/journal.pone.0181196 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Wang S, Shen B, Wei L, Wu M, Wang J. Association between socioeconomic status and prognosis after ischemic stroke in South China. Neurol Res. 2019;41:916–922. doi: 10.1080/01616412.2019.1630165 [DOI] [PubMed] [Google Scholar]
- 26. Prvu Bettger J, Liu C, Gandhi DBC, Sylaja PN, Jayaram N, Pandian JD. Emerging areas of stroke rehabilitation research in low‐ and middle‐income countries: a scoping review. Stroke. 2019;50:3307–3313. doi: 10.1161/strokeaha.119.023565 [DOI] [PubMed] [Google Scholar]
- 27. Khatib R, Arevalo YA, Berendsen MA, Prabhakaran S, Huffman MD. Presentation, evaluation, management, and outcomes of acute stroke in low‐ and middle‐income countries: a systematic review and meta‐analysis. Neuroepidemiology. 2018;51:104–112. doi: 10.1159/000491442 [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data S1: Supplemental Methods
Tables S1–S4
Figures S1–S4
Data Availability Statement
Individual deidentified participant data may be shared upon request with qualified investigators, provided that the request complies with the Mongolian Law on Personal Data Protection and is approved through an appropriate research protocol and a signed data access agreement via the Research Office of the George Institute for Global Health, Sydney, New South Wales, Australia.