Geographical variation and mediation analysis of incidence in stroke: Insights from CHARLS cohort study

Yuanyuan Anna Wang, Jiaying Liu, Huazheng LiangmedRxiv 2025 · DOI

This study, based on the China Health and Retirement Longitudinal Study (CHARLS), explores the geographical "North-to-South" gradient of stroke incidence in China. By analyzing over 11,000 participants over a decade, researchers found that the significantly higher stroke risk in Northern China is largely driven—or "mediated"—by regional prevalence of hypertension and dyslipidemia.

TL;DR

Researchers identified a stark geographical gradient in Chinese stroke incidence, with Northern regions facing the highest risks. Crucially, mediation analysis reveals that hypertension and dyslipidemia are the primary mechanisms through which geography influences stroke risk, suggesting that targeted blood pressure and lipid management in Northern provinces could significantly close the regional health gap.

What Problem Does This Paper Address?

Stroke is the leading cause of death in China, but the burden isn't shared equally across the country. Epidemiologists have long observed a "stroke belt" in the North, yet the why remains complex. Is it just the climate? Diet? Or is it because certain intermediate health conditions are more prevalent there?

While previous studies have mapped these disparities, this paper seeks to move beyond mere observation. The authors use mediation analysis to disentangle the direct effect of living in a specific region from the indirect effects caused by regional differences in chronic conditions like hypertension, diabetes, and dyslipidemia (unhealthy lipid levels). Understanding these pathways is vital for policymakers to decide whether to focus on broad environmental factors or specific clinical interventions.

Methods

The study utilized a longitudinal cohort design, following participants from 2011 through 2020.

  • Statistical Modeling: The team employed Cox proportional hazard regression models to estimate the Hazard Ratio (HR) of stroke across seven distinct regions (Northeast, North, Northwest, Central, East, Southwest, and South China).
  • Mediation Analysis: This is the core of the paper’s contribution. Using the "difference method," they calculated the "Proportion Mediated"—essentially asking: "If we controlled for hypertension, how much would the stroke risk difference between the North and South shrink?"
  • Adjustments: The models were stepwise-adjusted for demographics (age, gender), lifestyle (smoking status, residence), and comorbidities to ensure the regional findings weren't simply due to one region having an older population.

Datasets

The study relies on the China Health and Retirement Longitudinal Study (CHARLS), a high-quality, nationally representative survey of Chinese residents aged 45 and older.

  • Sample Size: 11,625 participants were included after rigorous exclusion criteria (e.g., removing those with missing health data or those lost to follow-up).
  • Timeline: Data spanned multiple waves from 2011 to 2020, providing a median follow-up of 9.8 years.

Key Insights and Ideas

  • The North-South Divide: The risk of stroke is significantly higher in the Northeast and North compared to South China. Even after adjusting for age and lifestyle, a resident of Northeast China is roughly 3 times more likely to suffer a stroke than someone in the South.
  • Hypertension as the "Bridge": Hypertension was the most powerful mediator. In North and Northwest China, hypertension explained approximately 8.5% to 9.5% of the regional disparity in stroke incidence.
  • The Role of Dyslipidemia: While hypertension was a universal mediator in the "stroke belt," dyslipidemia (high cholesterol/triglycerides) played a specific mediating role in North and Northwest China, but less so in the Northeast.
  • Vulnerable Subgroups: The "Northeast effect" was particularly pronounced among female urban residents with lower education levels and existing diabetes. Conversely, in the Southwest, rural residents faced higher risks than their urban counterparts.

Key Results

  • Incidence: Over the ~10-year period, 5.1% of the participants developed a stroke.
  • Hazard Ratios (Model 4 - Fully Adjusted):
    • Northeast China: HR 3.07 (95% CI: 2.53–3.74)
    • North China: HR 2.48 (95% CI: 2.07–2.97)
    • South China: Reference group (lowest risk).
  • Mediation Proportions:
    • In North China, the association between geography and stroke was mediated by hypertension (9.46%) and dyslipidemia (5.64%).
  • In the Northeast, hypertension mediated 1.55% of the risk, suggesting other factors (perhaps diet or extreme cold) play a larger direct role there than in the North.

Limitations and Future Work

The authors acknowledge several hurdles:

  1. Self-Reporting: Stroke diagnosis was self-reported by participants (based on prior physician diagnosis), which may lead to underestimation compared to clinical imaging data.
  2. Data Gaps: There was a significant loss of participants during the decade-long follow-up, which could introduce selection bias.
  3. The "COVID Gap": Because the CHARLS data currently only goes up to 2020, the long-term impact of the COVID-19 pandemic on stroke incidence and management habits remains unknown.
  4. Unmeasured Mediators: While the study looked at three major comorbidities, other factors like dietary salt intake, air pollution (PM2.5), and temperature fluctuations—all of which vary geographically in China—were not included in the mediation model. Future research integrating environmental sensors with cohort data could provide a fuller picture.

Summary updated

Abstract

Abstract Introduction Stroke is leading cause of mortality in China and it has regional disparities in the incidence among Chinse populations. The study aimed to investigate whether and how much of the association between regional disparities and stroke is explained by an mediator. Methods We applied Cox proportional hazard regression model and mediation analysis to analyse results from the follow-up of of the CHARLS database. Results Over the median follow-up of 9.8 years in CHARLS, 5.1 (%) of CHARLS participants developed a stroke. We found that North China had the highest incidence of stroke, followed by Central, Northeast, Northwest, Eastern, and Southwest China. South China had the lowest incidence of stroke, and was the reference group. Rural residents who were older and smokers had a higher risk of stroke. Subgroup analysis revealed that the incidence of stroke in Northeast China was significantly higher among female urban residents with lower education levels, diabetes, and hypertension. The mediation analysis showed that hypertension had a significant mediation effect in Northeast, North, and Northwest China. Dyslipidemia had similar mediating effect in the Northwest and North China, but not in other regions. Conclusion Geographical variation is associated with a different incidence of stroke. This association is partially mediated by hypertension in Northern, Northeastern, and Northwestern China, and by dyslipidemia in Northwestern and Northern China.
BibTeX
@misc{wang2025geographical,
  author = {Yuanyuan Anna Wang and Jiaying Liu and Huazheng Liang},
  title = {{Geographical variation and mediation analysis of incidence in stroke: Insights from CHARLS cohort study}},
  year = {2025},
  doi = {10.1101/2025.04.21.25326174},
  howpublished = {medRxiv},
}
Figure