Worldwide socioeconomic status and stroke mortality: an ecological study

Stroke causes 5.5 million deaths and the loss of 49 million disability-adjusted life years worldwide each year [1]. It has been estimated that by the year 2020 stroke will remain the second leading cause of death, and in terms of disability it will be among the five most important causes of disability worldwide [2].

It has been suggested that stroke burden and time trend of stroke varied among populations [3, 4], but trends in common risk factors may not fully explain time or inter-population variations in stroke occurrence [5‐10]. According to Global Health Statistics in 1996, the higher stroke mortality rates for men aged 45–59 years were observed in Formerly Socialist Economies of Europe (FSE, 135.5/100,000), China (CHN, 134.3), Latin America and the Caribbean (LAC, 104.1). The lower rates were reported from countries with Established Market Economies (EME, 36.9, such as USA, Canada, Japan and Australia); Middle Eastern Crescent (MEC, 73.2) and Other Asia and Islands (OAI, 85.7). In women, the higher rates were observed in CHN (110.5), LAC (83.9) and FSE (83.1). The lower rates were in EME (22.7), MEC (65.5) and OAI (67.2) [3]. Sarti and Rastenyte presented trends of stroke mortality during 1968–1994 [4]. The most developed countries experienced remarkably declining trends about 20 years, but less developed countries like eastern European countries experienced increasing trends for stroke mortality during the same period. Recent studies on life course exposures demonstrated that children exposed to low socioeconomic status (SES) were at high risk of stroke at adulthood [11‐13]. The findings from adult population studies examining the relationship between SES and the risk of stroke are conflicting, with some reporting positive and some negative associations [14‐17]. The effect of population SES on stroke mortality at country level has not been addressed. We carried out an ecological study to explore the association between SES in childhood and in adulthood with stroke mortality among countries/regions, to help define populations at a higher risk and predict the trends of stroke mortality for health policy.

Countries/regions with the HDI and some confounding factors used in the analysis lists in Table 1. The status of socioeconomic development among the 39 countries/regions ranged from HDI as low as 0.70 in Slovakia to HDI as high as 0.939 in Norway in 1999. Many countries/regions had no nationwide survey data of prevalence of hypertension or obesity. Stroke mortality rate was high in Mauritius (98.5 in men and 56.1in women) and low in Switzerland (7.7 per 100,000 in men and 7.5 in women) in recent years.

Pearson correlation analysis among SES indicators in childhood and adulthood shows that the indicators were correlated each other (Table 2). The choice of SES indicators was determined by comparing sum of ranks of correlation coefficients between each indicator and stroke mortality (Table 3). The correlation coefficients and the rank of the correlations were very close for GDP/capita (slightly higher in childhood) and HDI (slightly higher in adulthood) for both 1960 and 1999. HDI was chosen because it was a more comprehensive indicator of socioeconomic development status. HDI consisted of not only economic productivity but also life-expectancy, health care services and education levels. Those factors are largely related to the susceptibility to development of stroke and efficiency in primary and secondary prevention of stroke in the populations.

Table 4 and Figure 1 show that both HDI in 1960 and HDI in 1999 were inversely associated with stroke mortality in men and women. HDI in 1960 explained 37% of variance of stroke mortality among countries/regions in men and women (P < 0.001). HDI in 1999 explained 35% in men and 53% in women (P < 0.001), respectively. Table 4 shows that HDI in 1960 and HDI in 1999 was still significantly and negatively correlated with stroke mortality in men and women, respectively after adjusting for each confounding variable.

Stroke mortality in the countries/regions at the bottom tertile of HDI both in 1960 and 1999 was 2.52 and 2.61 times higher than those at the top tertile of HDI in both periods in men (38.5 per 100,000 per year, 95% CI: 15.5-61.6 vs. 14.8, 10.9-18.7), and women (28.8, 14.2-43.4 vs. 11.4, 8.5-14.4) separately.

Figure 2 illustrates that the peak of age-standardized stroke mortality was exhibited at about HDI = 0.77-0.81 for men and 0.75-0.79 for women (P < 0.01). Stroke mortality increased with improvement of HDI if HDI < 0.81 for men and HDI < 0.79 for women, while stroke mortality decreased with increasing HDI if HDI > 0.77 for men and HDI > 0.75 for women during 1974–2001 when each country/region at each of six time periods was considered simultaneously. The strength of the association was stronger as the year of the SES indicatorbecame closer to the year of mortality data used (see Additional file 1: Table S1).

This ecological study quantitatively estimated the effect of SES on stroke mortality at population level. HDI was used as a SES indicator in this study due to its significant association with stroke mortality and more comprehensively represented the social and economic development status at country level. In general, both childhood SES (HDI in 1960) and adulthood SES (HDI in 1999) were inversely associated with stroke mortality in recent years for men and women in middle- or high-income countries/regions. The direction of the association between SES and stroke mortality was opposite for low and high HDI levels: the association was positive for the countries/regions at low HDI level, but it was negative for the countries/regions at high HDI level (p < 0.01 for both directions).

Our finding that the risk of stroke is associated with low SES in childhood at population level suggests that exposure to socioeconomic deprivation in early life might increase the risk of stroke in adulthood. Possible pathophysiological mechanisms may include underlying lasting changes in vascular structure in response to under-nutrition in early life, activation of the renin-angiotensin system predisposing to the development of hypertension [28, 29], impact on hormonal and metabolic programming resulting in insulin resistance and diabetes [30, 31], adverse lipid and thrombogenic profiles [32, 33] predisposing to atherogenesis [11, 34].

A quadratic relationship revealed that the risk of stroke increased with improvement of SES in regions at a lower stage of socioeconomic development, while it decreased with increasing SES in regions at a higher stage of socioeconomic development. This finding likely reflected the changing pattern lifestyle, health literacy and access to health services through different stages of socioeconomic development. High blood pressure, obesity and hyperlipidemia are risk factors for stroke that are related to an unhealthy lifestyle pattern consisting of energy-dense diet, physical inactivity, smoking, and alcohol use. Such high risk profiles have been documented to be negatively associated with SES in middle- or high-income countries/regions [35‐40], but positively associated with SES in low-income countries [41‐44]. SES may affect health status through lifestyle choices [45], and healthy lifestyle was associated with lower risk of stroke [46, 47]. SES also related to education and health literacy in populations, which in turn related to a lower risk of stroke [48, 49]. Finally, SES is related to the accessibility and quality of health care services which are directly related to the prevention and treatment of stroke [24, 49]. The association between stroke mortality and the HDI in the same time period was stronger than the association of stroke mortality with the HDI in an earlier time period for the same cohort. This may suggest that health care services play a more important role for the prognosis of stroke patients.

Our study examined the association between the risk of stroke and SES among countries/regions. In agreement with our results, some studies reported inverse associations between SES and stroke mortality in middle- or high-income countries/regions [14, 15]. Studies from USA and northern European countries in the 1980s reported high stroke mortality in people with manual occupations and low mortality in the non-manual class [50]. An analysis from the 1990s assessed stroke mortality by educational level in ten European countries based on longitudinal data [51]. For all countries in all age groups, higher mortality rates were reported in groups with an education below upper secondary or equivalent. This was consistent for men (relative risk 1.27, 95% CI 1.24-1.30) and women (1.29, 1.27-1.32). In contrast, observations from low-income countries showed that populations with high socioeconomic status were at a higher risk of stroke compared with those with low socioeconomic status [16, 17]. For instance, stroke mortality was higher in urban compared with rural Tanzania, there being a graded response depending on wealth. The yearly age-adjusted rates per 100,000 in the group aged 15–64 year for the three project areas (urban, fairly prosperous rural, and poor rural, respectively) were 65 (95% CI: 39–90), 44 (31–56), and 35 (22–48) for men, and 88 (48–128), 33 (22–43), and 27 (16–38) for women, respectively [17]. Our study indicated the high correlation between childhood and adulthood SES and therefore we could not determine if childhood/adulthood SES related with stroke risk independently of adulthood/childhood SES due to their colinearity. However, we found that stroke mortality in the countries/regions at the bottom tertile of HDI both in 1960 and 1999 was 2.52 and 2.61 times higher than those at the top tertile of HDI in both periods.

There are limitations in our study. Given the nature of the ecological study design, our results could be biased by the ecological fallacy. We could not completely rule out the possibility of residual confounding due to unmeasured or inadequately measured covariates. The results from the population level could not be directly applied to individual patients. HDI was a comprehensive country level index estimated with comparable information; therefore, the bias in SES measurement was limited. Data on stroke mortality is not available for very low-income countries/regions, so that the results only applied to low to middle or high income countries/regions. Stroke mortality at different time points could be affected by the changing of ICD codes system over time. In the present study, the data available comprise deaths for stroke registered in national vital registration systems, with underlying cause of death as coded by the relevant national authority. Moreover, our analyses are restricted to overall category of stroke which is essentially similar in the 8th and 9th, and 9th and 10th [4, 52, 53], and for which death certificate data are more accurate than for specific diseases. The accuracy and consistency over time and between regions in the diagnosis may affect the comparison of stroke mortality at different time periods among countries/regions. However, the concerns about diagnostic accuracy in these statistics are minimized by inclusion of all cerebrovascular disease for analysis, thus improving comparability across countries and time. We did not analyze subtypes of stroke mortality due to the unavailability of the data, and the study is not able to examine regional differences and time trends in mortality from different stroke subtypes. Different SES indicators can establish groups with differential exposures and identify specific as well as generic mechanisms relating SES to health [54]. There is no best single indicator. Heath is captured in HDI, so it could be a limitation to study HDI and health outcomes. However, HDI is a composite measure of the three factors key in allowing people to lead more fulfilling lives and strong association with stroke mortality. Therefore, it was used as the SES indicator in this study. It should be mentioned that there was no log-linear or quadratic linear association between HDI and mortality from other cardiovascular/vascular diseases were observed (data not shown here).

In conclusion, SES is an important population indicator of stroke mortality risk. From low to low-middle income countries/regions, stroke mortality increases with improvement of SES (measured by HDI). From middle to high income countries/regions, stroke mortality decreases with advancing SES. In low SES environment, advantaged populations are at high risk of stroke. In high SES environment, disadvantaged populations are at high risk of stroke. For the countries/regions experiencing transition, the high and low risk populations are under epidemiological transition. This information allows health policy makers to mobilize appropriate resources and identify target populations for more efficient health protection and better health care services.