GeographicalDifference, Rural-urban Transition and Trend in Stroke Prevalence in China: Findings from a National Epidemiological Survey of Stroke in China

Accurate and up-to-date provincial and regional-level stroke prevalence estimates are important for research planning and targeted strategies for stroke prevention and management. However, recent and comprehensive evaluation is lacking over the past 30 years in China. This study aimed to examine the geographical variations in stroke prevalence based on data from the National Epidemiological Survey of Stroke in China (NESS-China) and demonstrate urban-rural transition and trend over three decades. The stroke prevalence (prevalence day, August 31, 2013) was estimated using the world standard population. The stroke prevalence was 873.4 per 100,000 population, and varied from 218.0 in Sichuan to 1768.9 in Heilongjiang. Stroke prevalence exhibited a noticeable north-south gradient (1097.1, 917.7, and 619.4 in the north, middle, and the south, respectively; P < 0.001) and showed a 2.0-fold, 1.5-fold, and 1.2-fold increase in rural areas in the north, the middle, and the south, respectively, from 1985 to 2013. Overall, stroke prevalence was higher in the rural regions than in the urban (945.4 versus 797.5, P < 0.001) regions. However, the converse was depicted in 12 provinces. A noticeable geographical variation in stroke prevalence was observed and was evolving overtime in China. It is imperative that effective public health policies and interventions be implemented, especially in those regions with higher prevalence.


Provincial-level variations.
shows the crude and age-standardized stroke prevalence. All 7,679 participants were diagnosed with stroke. The overall age-standardized prevalence of stroke was 873.4 per 100,000 population. There were remarkable provincial-level differences in the distribution of stroke prevalence. The age-standardized prevalence of total (all types) stroke (per 100,000 population) was the highest in Helongjiang  9) and Tibet (508.2, 95%CI: 490.1-526.2). A more than 8 times difference in stroke prevalence occurred between the highest and the lowest provincial regions (Fig. 1).
The hemorrhagic stroke age-standardized prevalence (per 100,000 population) was 178.3. The age-standardized hemorrhagic stroke prevalence was highest in Henan (311.7, 95%CI: 291.7-331. 6 The prevalence of ischemic stroke (676.7) was significantly higher than that of hemorrhagic stroke (178. 3). The ratio of the prevalence for the two types of stroke ranged from 2.4 to 4.9, and averaged 3.8. However, there was an exception; Tibet had a higher prevalence of hemorrhagic stroke (249.2) than ischemic stroke (93.5) ( Table 2). www.nature.com/scientificreports www.nature.com/scientificreports/ Regional-level variations. Withthe Yellow River and Yangtze river acting as the boundary, the total stroke age-standardized prevalence (per 100,000 population) was 1097.1(95%CI: 1059. 8-1134.3) in the north of the Yellow River, 917.7(95%CI: 883.6-951.8) in the middle, and 619.4(95%CI: 591.3-647.4) in the south of the Yangtze River; exhibiting a north-south gradient. There were significant differences in the distribution of total stroke prevalence between different regions (P < 0.001) ( Table 3). Compared to the rural study in 22 provincial regions in 1985 10 , there was a significant increase of stroke prevalence (age-standardized to the 1960 United States population) in rural areas, including a 2.0-fold increase in the north (from 480.0 to 1433.1), a 1.5-fold increase in the middle (from 420.0 to 1051.4) and a 1.2-fold increase in the south (from 280.0 to 623.1) (Supplemental Fig. 1).
Compared tothe rural study in 22 provincial-regions in 1985 7 , in the past 3 decades, rural areas in these regions have experienced a greater increase in stroke prevalence (from 393.7 to 1014.7). The prevalence of stroke has increased most significantly by 5.3, 3.9, 3.3, and 3.1 times in Inner Mongolia,Guangxi, Tianjin, and Hubei, respectively. The minimum increase was found in Sichuan (from 210.4 to 219.1), Hunan (from 432.5 to 652.6), and Guizhou (from 267.1 to 456.7) (Supplemental Fig. 2).
Compared to the urban study in 6 cities in 1983 8 , the prevalence of stroke in urban areas has increased slightly over the past 3 decades (from 719.0 to 816.    www.nature.com/scientificreports www.nature.com/scientificreports/ The stroke age-standardized prevalence was higher in rural areas than in urban areas in the various regions surveyed, except for the southern regions (rural, 580.9; urban, 657.6) (Supplemental Table 1).

Discussion
The present study, which differs from the stroke registries, cohort studies, or reviews, is the first to report stroke prevalence across the 31 provincial regions based on a large nationwide representative survey for all ages. Although the estimates from previous studies 7, 10 showed that there were significant geographical differences in stroke prevalence, our analysis provided a more detailed insight into the prevalence of stroke, with rural-urban transition, and the trend in the past three decades. Most studies 9,11 in China have shown geographical variations in stroke incidence and mortality. However, provincial and regional level stroke prevalence estimates for all ages are not available. In the earlier 22 rural population study 7 , the age-standardized stroke prevalence (age-standardized to the 1960 total US population)was 393.7, and the highest stroke prevalence was 869.6 in Heilongjiang and lowest (95.4)in Guangxi. In the 6-city study 8 , the age-standardized stroke prevalence (age-standardized to the 1960 total US population) was 719.0, with the highest (909.0) in Harbin of Heilongjiang and the lowest in Shanghai.
In Germany, a population-based survey of 28,090 participants (aged ≥50 years) reported an increased stroke prevalence to 7.6% (men 8.4%; women 7.2%). Factors associated with the higher prevalence were increasing age, sex (male), non-German nationality, lower education, positive family history of stroke, and solitary living 12 . In Urban Sri Lanka, a population-based, cross-sectional study conducted among 2,313 adults aged ≥18 years found a stroke prevalence of 10.4 per 1,000 (95%CI: 6.3-14.5) 13 . In the present analysis, stroke prevalence was 873.4 per 100,000 population for all ages, and varied substantially among different Chinese populations in different regions. The provincial difference in stroke prevalence was also striking and varied from 218.0 in Sichuan province to 1768.9 in Heilongjiang. It was in general, higher in the north provincial region than in the south. However, for some provincial regions with high-income, such as Beijing in north China and the eastern and southern coastal  Table 4. The prevalence of stroke in rural and urban areas of the 31 provincial regions. Age-standardized prevalence of all stroke was standardized to the WHO world standard population.
www.nature.com/scientificreports www.nature.com/scientificreports/ provinces, the prevalence of stroke was relatively low. Adequate health services and strategies for stroke prevention and care (including blood pressure control, smoking cessation, and acute stroke units) are the most likely explanations for the lower stroke prevalence shown. Our study found that the high provincial prevalence of stroke was associated with a high provincial incidence of stroke (r = 0.905, p < 0.001), but was not related to the low provincial mortality (r = 0.185, p = 0.320).
Stroke prevalence exhibited a noticeable north-south gradient, which is consistent with the results of a previous study 10 . High stroke prevalence regions were mainly located in the north of the Yellow River, while low stroke prevalence regions are mainly in the south of the Yangtze River. The striking variation in stroke prevalence could be explained by the difference in the prevalence of hypertension between the northern and southern provincial regions. The China hypertension survey found geographical gradients in the prevalence pattern of hypertension, which is highest in northeast, north, and southwest China 14,15 . He et al. suggested that the geographic difference in stroke and hypertension in China is consistent with the "salt hypothesis, " which proposes that high salt intake increases both the risk of hypertension as well as stroke 11 . Since hypertension remains the most important risk factor for all types of stroke, we therefore retrieved the hypertension prevalence for 31 provincial regions from a review 9 to analyze the correlation with stroke prevalence. We observed that the prevalence of hypertension correlated significantly with all types of stroke prevalence (r = 0.415, p = 0.020) and hemorrhagic stroke (r = 0.370, p = 0.041). The recent hypertension survey in China 16 found that 23.2% (estimated 244.5 million) of the Chinese adult population aged ≥18 years had hypertension, while another 41.3% (estimated 435.3 million) had prehypertension. Among individuals with hypertension, 46.9% were aware of their condition, 40.7% were taking prescribed antihypertensive medications, and 15.3% had controlled hypertension. Therefore, hypertension, as one of the major risk factors for stroke, should be better managed. Studies 2,17 have suggested that regions and country income levels are related to the burden of stroke, and the age-specific prevalence of stroke is significantly greater in high-income countries. Our study found that the central (middle-income) and eastern (high-income) regions showed higher prevalence of stroke compared to the western (low-income).
Stroke types also vary by country, regional income level, and ethnic differences 18,19 . A recent review of stroke epidemiology in the South, East, and South-east Asia reported that ischemic stroke occurs more commonly than hemorrhagic stroke, except in India and Vietnam, where the converse was observed 5 . According to previous studies in China 7,8,20,21 , ischemic stroke is the most common type of stroke. The proportion of stroke types varied from 43.7-78.9% for cerebral infarction to 18.8-47.6% for intracerebral hemorrhage 22 . The results of this study showed significantly higher ischemic stroke prevalence than that of hemorrhagic stroke. In general, it is more meaningful to use stroke incidence for classification. This is due to the survival of stroke patients, mostly mild or ischemic stroke patients. Patients with hemorrhagic stroke are generally more serious and have a higher proportion of deaths in the acute phase, thus, the higher prevalence of ischemic stroke compared to that of hemorrhagic stroke. However, there was an exception in Tibet, where the prevalence of hemorrhagic stroke was 2.7-fold higher than the prevalence of ischemic stroke. This finding may be explained as follows: (1) consideringthe fact that hypertension is the strongest contributor to hemorrhagic stroke, the highest hypertension prevalence (19.54%) occurred in Tibet 9 and (2) there are only two survey sites included in Tibet, which might have influenced the findings.
With dramatic transformations in the social, economic, and environmental conditions over the past 30 years, the lifestyle in China has changed rapidly, especially in rural areas. As a result of an aging population, urbanization, and westernization, the main risk factors for stroke such as hypertension, diabetes mellitus, hypercholesterolemia and obesity have increased substantially, and resulted in a greater increase in stroke prevalence in rural areas than in urban areas for three decades. But the 12 provincial-level regions showed lower stroke prevalence in rural areas and 8 of them (Chongqing, Guangxi, Guizhou, Ningxia, Sichuan, Tibet, Yunnan, and Xinjiang) were in the western regions with underdeveloped economy.
The early stroke epidemiological surveys were mainly based on the clinical level and less dependent on the objective diagnostic methods. Along with the development of medical diagnosis, the use of brain CT or MRI in this study reached 89.4%, which guaranteed the reliability and accuracy of the results. The findings from our study should well reflect the current actual picture in China. The reasons for the difference in stroke prevalence in different regions are complex. These geographical differences may be attributed to the differences in risk factors, lifestyle, environmental, and sociocultural factors between Chinese populations. The specific reasons require further study.

Limitations
There were some potential limitations to this study. First, the cross-sectional retrospective design is susceptible to recall bias, which resulted in the missing of some important information. However, repeated verification measures by the CDC preliminary screening and neurologists review minimized the recall bias. Secondly, the distribution of sampling sites was uneven. There were fewer sampling sites in the more populated regions or western China. Considering the operability, only two survey points were selected from Beijing and Shanghai with large populations, which may not have completely reflected the level of stroke prevalence. Compared to the study conducted 30 years ago, there was indeed a marked increase in stroke prevalence in the rural areas of China. The increase could be attributable to the markedly changed health behavior and lifestyles of people in rural areas with social and economic development. However, we should be cautious in interpreting the comparison results because of bias from different sampling methods in the two studies, as well as the recall bias. Thirdly, the data collected did not include risk factors, environmental, and sociocultural factors; thus, the relationships between stroke prevalence and these factors could not be established.

Conclusions
The prevalence of stroke in China showed a noticeable geographical difference and was evolving overtime. The burden of stroke was increasingly severe and it is urgently needed to grasp provincial and regional level epidemiological data to better understand the burden of stroke. The comparison of stroke prevalence in different regions is helpful to understand the mechanism of stroke that is specific to risk factors, lifestyle, environmental, and sociocultural factors and also helpful to implement targeted secondary prevention measures. Further and better epidemiological studies should be needed to clarify the real reasons for the geographic differences. A better understanding of the reasons for these differences could guide public health prevention programs. The policies and interventions should target, especially those regions with higher stroke prevalence.

Methods
Study design and participants. We conducted an analysis on data derived from the NESS-China. The NESS-China was a nationally representative cross-sectional survey conducted in 157 sites of the 31 provincial regions in mainland China from September 1, 2013 to December 31, 2013. The 31 provincial regions included 22 provinces, 5 autonomous regions (Xinjiang, Ningxia, Tibet, Inner Mongolia, and Guangxi), and 4 municipalities (Beijing, Shanghai, Tianjin, and Chongqing). A multistage, stratified complex sampling method was used to select a nationally representative sample of all populations. The survey was conducted face-to-face by households. All the family members in a household were interviewed. For children younger than 14 years, older adults who had no articulate answers to the questions, and individuals with language barriers from disease or other reasons, the questions could be answered by other family members on their behalf. If a participant was absent (e.g., away on a business trip, study, or work), he/she could answer the questions by telephone with the help of a family member. Specific details on sample size calculation, sampling process, and participant selection for the NESS-China study have been published previously 3,23,24 . A total of 596,536 participants of all ages were included in our analysis.
Data collection. The NESS-China was the first national survey to use the door-to-door and face-to-face data collectionmethod. The trained interviewers of the Center for Disease Control and Prevention (CDC) administered the preliminary screening tool to identify participants with positive stroke symptoms. A comprehensive questionnaire, which included family information on demographic characteristics, medical history, stroke-related symptoms and death information of family members, who died from stroke between September 1, 2012 and August 31, 2013. After the preliminary screening, participants with the symptoms or history suggestive of stroke were invited to see a neurologist in the town/village clinic of their choice. Their medical records (with information on e.g. cardiovascular disease risk factors, computed head tomography [CT], magnetic resonance imaging [MRI] scans, and autopsy protocols) were carefully reviewed and relevant data were recorded. When appropriate, some study participants were requested to undergo a brain-neuroimaging examination (e.g. to exclude brain disorders mimicking stroke) and another neurological examination.
www.nature.com/scientificreports www.nature.com/scientificreports/ Next, neurologists arranged face-to-face interviews with all confirmed and suspected cases. Based on the information obtained (e.g., history, neurological symptoms/signs, and imaging data such as CT and/or MRI), the neurologist would make a final diagnosis according to the stroke diagnostic criteria. Any participant who was not interviewed was contacted again to arrange for another interview. For complicated cases that could not be confirmed by the neurologist, the diagnosis confirmed on the basis of the discussion between two or more neurologists. If a diagnosis could still not be confirmed, relevant case data was documented and submitted to the provincial technical consultants for further consultation. Finally, a total of 28,800 participants with stroke or positive stroke symptoms were reviewed and 7,679 of these were confirmed as stroke by August 31, 2013. Two of the 157 sites failed to complete the investigation as required and were therefore excluded from the analysis (Fig. 5). Additional details on stroke confirmation can be found in the literature 3 .
The epidemiological data 7,8,10 from two comparable studies (the 6-city stroke study and the 22 rural population study are similar to the present study in the methodology used for ascertaining stroke and the survey used) were used to confirm the urban-rural transition and trend over the three decades.
The study protocol was approved by the ethical review committees of Beijing Tiantan Hospital. All methods were performed in accordance with relevant guidelines and regulations. All participants provided written informed consent before data collection.
Definitions. Based on the World Health Organization (WHO) criteria, stroke was defined as "rapidly developing clinical syndrome of focal (or global) abnormalities of cerebral function, lasting more than 24 hours or leading to death, with no apparent cause other than that of vascular origin 25 ". Any nervous system abnormalities induced by trauma, metabolic disorder, tumor, or central nervous system infections were excluded. All cases identified as definite stroke were further classified into ischemic, hemorrhagic, and undetermined stroke, and coded using the tenth revision of the international Classification of Diseases. The classification of ischemic and hemorrhagic stroke was performed on the basis of the clinical presentation, and confirmation by CT or MRI was required, with reference to the Atherosclerosis Risk in Communities criteria 26 . Undetermined stroke included all cases that did not have CT or MRI data or could not be classified by CT or MRI. The prevalence of stroke in this study was expressed as a point prevalence ratio (prevalence day, August 31, 2013).
With the Yellow River and Yangtze River acting as the boundary, the 157 survey sites were divided into three categories as follows: the north, the middle, and the south. The north of the Yellow River included 57 survey sites, the middle area between the Yellow River and the Yangtze River included 41 survey sites, and the south of the Yangtze River included 59 survey sites.
According to the level of economic development, the 31 provincial-regions were divided into three categories as eastern (11 provincial-regions including Beijing, Fujian, Guangdong, Hainan, Hebei, Jiangsu, Liaoning, Shandong, Shanghai, Tianjin, and Zhejiang), central(8 provincial-regions including Anhui, Henan, Heilongjiang, Hubei, Hunan, Jilin, Jiangxi, and Shanxi), and western (12 provincial-regions including Gansu, Guizhou, Ningxia, Qinghai, Shaanxi, Sichuan, Tibet, Xinjiang, Yunnan, Chongqing, Guangxi, and Inner Mongolia). www.nature.com/scientificreports www.nature.com/scientificreports/ Statistical analyses. With regard to sociodemographic factors, continuous variables were described as mean with standard deviations or as median (interquartile range [IQR]) according to their distribution. Categorical variables were reported as frequency (percent). All reported prevalence of stroke were estimated and age-standardized to the WHO world standard population 27 . For prevalence data, 95% confidence intervals (CIs) were calculated. When compared with the 6-city stroke study in 1983 and the 22 rural population study in 1985, stroke prevalence was standardized to the 1960 U.S. population. The comparison of prevalence or rates between different groups was performed by Chi-square test. Correlation analysis was used to identify the relationship between provincial stroke prevalence and provincial hypertension prevalence. All p values that were 2-tailed and <0.05 were considered statistically significant. All data analyses were conducted using SPSS for Windows version 13.0 (SPSS Inc., Chicago, IL, USA).

Data availability
All data generated or analysed during this study are included in this published article (and its Supplementary  Information Files).