Impact of socio-economic factors on stroke prevalence among urban and rural residents in Mainland China

Previous studies using different study designs and various definitions of socioeconomic status (SES) have shown that people with lower SES are at increased risk of stroke [1–6]. Such socioeconomic gradient for the risk of stroke tends to be flattened, though still significant, after taking into account the conventional risk factors of stroke, including cigarette smoking, alcohol drinking, body mass index (BMI), physical activity, diabetes, and hypertension [6–12].

SES is a complex construct containing either a single measure or a combination of several components. Using different SES measures may reflect different specific exposures and can provide various information on peoples' relative socioeconomic positions within their community [13]. It has been suggested that family average income (FAI) could be considered as a single measure of SES as it is a more realistic reflection of material measures than other indicators such as educational attainment or occupation [14–16].

In our previous studies, we found that overweight and diabetes, two conventional risk factors of stroke, [17–19] were positively associated with FAI in China [15, 16]. We thus hypothesized that Chinese people with higher FAI might also be at an increased risk of stroke. The purpose of this study was to examine the relationship between stroke and SES, indicated separately by FAI, education and occupation, and to see if the SES gradient, if present, could be explained by differences in conventional risk factors of stroke, using representative data obtained through a large population-based survey conducted in both urban and rural areas of a region in Mainland China between October 2000 and March 2001.

Among a total of 15,563 families and 32,568 eligible subjects approached in this survey, 15,300 households and 29,340 subjects agreed to participate in the interview, giving a response rate of 98.3% (households) and 90.1% (individual participants) respectively. The average family size was 3.3 members (SD = 1.3, Min = 1 and Max = 9).

No significant differences were detected between the respondents and the non-respondents in terms of age, gender and residence area. No family had more than one stroke patients at the time of interview. There were 67.7% of participants from urban areas and 32.3% from rural areas; 49.8% of subjects were male and 50.2% were female.

Our study consistently revealed that the risk of stroke was associated with increasing levels of all SES indices, including FAI, education, and occupation. A positive and significant gradient of FAI with stroke prevalence remained after the adjustment of the basic demographic variables (age, gender and area) and a group of defined conventional risk factors. The findings of our study show similar patterns to those of previous decades for Western populations but were in contrast to the results in more recent developed areas in which SES is inversely associated with the risk of stroke [1–12]. We hypothesize that this reflects the earlier stage of epidemiological transition within China at this time, in which more affluent areas are adopting more Westernized lifestyles including diet.

Socioeconomic inequalities in health have been attributed to a number of different mechanisms, including unhealthy behaviors, inadequate accesses to health care service, nutritional inadequacies, other inequalities in material circumstances, and psychosocial stress [22–24]. The mechanism of how the specific SES influences stroke is not entirely understood. Potential explanations for the patterns of inequalities may be related to the differences in major risk factors of stroke, in psychosocial factors, or in access to and use of health care services [1, 5], since in China ability to pay is linked to access to health care.

However, psychosocial factors and inequalities of health care did not adequately explain the SES gradient in stroke, because the social gradient in stroke has been shown to be largely driven by conventional stroke risk factors [12]. Previous studies suggested that the conventional risk factors (components were similar to our study) might have accounted for more than half of stroke disparities [6, 25]. It is well known that, in developing countries, people in lower SES categories are generally less likely to be obese and more likely to be physically active in both their workplace and leisure time than their counterparts in developed countries [26–32]. People with lower socioeconomic status may also have a lower total food (calorie) intake, especially of animal proteins; they are more likely to consume vegetables [29–32]. As seen in many areas of Africa, people of low SES status were less likely to be overweight and develop diabetes due to the nature of their work and also access to transportation or car ownership [15, 16]. Our study found that the proportion of subjects who consumed red meat at least once a day was significantly higher in the high FAI group than the low FAI category (87.8% vs. 46.0%, p < 0.001), while the rate of participants who consumed vegetables at least 100 gram each day in the higher FAI group was significantly lower than those in the lower FAI group (59.6% vs. 72.4%, p < 0.001). These disparities in conventional risk factors might act as intermediates for the impact of SES on stroke development. However, the proportion of cigarette smokers among the lower FAI group was significantly higher than that in the higher FAI group (37.4% vs. 29.1%, p < 0.01). Our finding on the negative association between the prevalence of cigarette smoking and SES level was inconsistent with the evidence from previous studies that smoking was an important factor explaining stroke disparities in the different socioeconomic groups [6–9]. However, no firm conclusion could be drawn based on the current cross-sectional study design. It would be of great interest to analyse the trend of stroke prevalence and its association with SES in the context of China to see whether and when the relationship between stroke and SES would reverse and become similar to that observed in developed countries.

We believe that per capita income of each family unit was probably a more sensitive reflection of the influences of SES on participants' lifestyles than per capita GDP of the surveyed area. Compared to other SES indicators (such as educational attainment, occupation, total household income or deprivation score), FAI in China was recognized as a new and a single measure of SES based on our findings since it was considered to be a more realistic reflection of material measures. In China, most families have only one child, but meanwhile a lot of families have their parents and/or grandparents living together, especially in the rural areas. The family size is mostly affected by the number of supported elderly. Most supported elderly may have lower retirement salaries or even no incomes (rural elders) at all. Total household incomes can thus not reflect per capita incomes in the families. With the same household incomes, a larger family with more elderly living together may have poorer life quality than that of a smaller family due to increased costs of supporting a larger family. Although education and occupation may partially determine an adult's SES in Mainland China where the economy is undergoing a rapid transition, they were considered to be less valid and sensitive than FAI and could only act as an indirect index of SES in the context of China.

An important question is whether competing causes of death might play an important role in the explanation of the inverse association between FAI and stroke prevalence. It was possible that subjects in the low income groups could have died from other causes such as infectious diseases in earlier life that prevented them from living long enough to develop stroke later in life. Vital statistics showed that the ranking of mortality from specific diseases among urban population was similar to that of the rural areas. In addition, the life expectancy was over 73 years in the rural areas of Nanjing since 1995. Thus, we believe that the lower prevalence of stroke among the lower SES groups could not be adequately explained by competing causes of deaths in earlier lives.

Elevated OR of stroke was significantly associated with higher SES. However, this gradient pattern remained significant only for the age group of 65 years or above after stratification by age. This might be due to the low prevalence of stroke among the younger (0.12%) and middle age groups (1.40%). The prevalence tended to be unstable after the stratification by FAI (three categories). Another possible explanation was that the relationship between SES and stroke has been transiting to that observed in Western countries, because younger people adopted more westernised lifestyles (e.g. high fat intakes) relative to their old counterparts.

The strengths of the current study is that the analysis include a large representative sample of both urban and rural populations, a very high response rate of more than 90% for household interviews, a sensitive and realistic index of SES (the FAI), and a full consideration of important stroke risk factors. However, when interpreting these findings, several limitations of this study should be considered. First, although strong agreement was demonstrated between self-reported stroke cases and those confirmed by medical record review or CT scan [19, 34, 35], the possibility of underestimating the prevalence of diagnosed stroke could not be completely ruled out in our study, especially in rural areas. However, in the current context of healthcare in China, all outpatients can keep their personal medical records and/or hospital discharge notes from which we were able to check and extract the information regarding the dates of hospital visits, diagnoses of diseases, prescriptions, treatment, and recommendations. We believe most people knew well their disease status, especially for strokes requiring hospitalization with certain recognized disabilities. Based on this scenario, we believe that the possibility of underreporting diagnosed strokes by patients themselves, if present, would be minimal. The second limitation was related to self-reported family incomes. People living in rural areas might have underestimated their earnings and inaccurately estimated the value of their home-grown foods and informal barter arrangements. We consider such misclassification was likely to be random and biased the associations towards null, as subjects with stroke and those without stroke should not have estimated their incomes differently. Third, no data was obtained on participants' accesses to health care, which makes it difficult to understand the role of inequitable access to health care on stroke prevalence. The prevalence of stroke was affected by both the incidence and duration of survival. We found that people living in urban areas had a higher prevalence of stroke regardless of the SES levels (Table 4). More accessible and affordable health care for urban people might result in a better post-stroke survival and could partly explain the disparities of stroke prevalence between urban and rural populations. Similarly, people with higher FAI might have better access to medical facilities and better post stroke health care, which could contribute to extending their life expectancies beyond that of those with lower FAI status. The positive gradient between stroke prevalence and FAI might be influenced by the survivor effect, as prevalent cases instead of incident cases were studied with the current cross-sectional design. Fourth, our study did not allow us to infer causality for the stroke-FAI relationships. As the study was cross-sectional, the temporal relationship between SES and stroke could not be accurately defined. Part of their association might reflect the impact of stroke on the FAI due to the loss of work time of both the stroke patients and the family members. A prospective study design to confirm the potential causal relationship between family average income and stroke is recommended. Finally, this study was mainly designed to investigate the prevalence of selected chronic diseases in Nanjing municipality, thus only each responder's identification number was originally required to be recorded on each questionnaire without consideration of family identification number. At this time, several years after household interview, it was impossible for us to check up each responder's family identification by manual counting the local residence registration information. This inherent limitation makes us unable to control for possible clustering effect in the analyses.

We recommend conducting further in-depth studies to elucidate the relationships between stroke (including its conventional risk factors) and SES under different circumstances in countries undergoing demographic and epidemiological transitions. In a developing society like China with rapid social and economic transition to industrialization, more and more people are being employed as office workers and their lifestyles become westernized (e.g., consuming foods with high energy and high fat but taking less physical activity and having less Chinese traditional vegetable and rice foods) [36]. As a consequence of 'unhealthy' lifestyles, more and more people are tending to become overweight, especially in urban areas [36]. Health promotion strategies and lifestyle interventions targeting people with higher family average income in the current context of China could be helpful in the campaigns against stroke. Meanwhile, to prevent stroke world-wide, it is also important to spread the knowledge to all populations experiencing economic and lifestyle transitions in the developing communities so that they can avoid taking up the 'unhealthy' lifestyle and behaviours associated with modernization and westernization.

This study shows a positive association between stroke prevalence and better SES using different indicators. However, compared with occupation and education, FAI may be a more sensitive and realistic index of socio-economic status for public health studies, and may inform the targeting of campaigns or other initiatives, particularly in populations where material prosperity is high in some social groups. People with higher family average income should be the priority target in the fight against stroke in China, and more attention on monitoring stroke risk factors shall be paid by themselves and health care providers.