Is area-level socioeconomic deprivation associated with mortality due to circulatory system diseases in Poland?

Data on deaths during 2010–2014 was used to analyze mortality. According to the International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10), the causes for the deaths were encoded as CSD (codes: I00–I99), including IHD (I20–I25) and CeVD (I60–I69).

The age-standardized number of deaths was used for statistical analyses. A direct standardization method was utilized for the analyses, considering the demographic structure in 5-year age groups (from 15 to 19 years to 80–84 years, and above 84 years old) separately for each of the 66 sub-regions of Poland [35]. European Standard Population – 2013 edition was assumed as the standard population, and standardized mortality coefficients were determined accordingly [36]. The coefficients of standardized mortality were calculated as 95% confidence intervals (CIs).

The dynamics of changes in mortality rates in 2010–2014 were assessed by calculating the deaths prevented or postponed (DPP) index [37]. The index value was expressed in percentage and estimated as the difference between the expected number of deaths in 2014, assuming that the mortality rate in 2014 was the same as in 2010, and the actual number of deaths in 2014 relative to the expected number of deaths in 2014:

Five SED indicators were selected from the available variables in the system of routine statistical records from the Central Statistical Office. The selected indicators of SED were as follows: people with higher education, people employed in industry and construction, average monthly salary, unemployment rate, and people on social support due to poverty. These variables were chosen as they showed a strong correlation with all-cause mortality [38]. According to published literature data, the SED variables are potential predictors of the geographical distribution of CSD mortality [39‐44]. Education not only increases the awareness of beneficial health behaviors through medical and prophylactic care and treatment recommendations, but is also one of the essential determinants of employment [39, 40]. The professional class, characterized by a balance between effort and reward, is associated with health. Similarly, people employed in industry and construction are manual workers and exhibit worse health outcomes [41]. Income determines an individual’s access to better-quality goods, services and housing to stay healthy [42]. Thus, unemployment is linked with disease development due to reduced access to medical care and screening programs resulting from the lack of health insurance [43]. Poverty associated with a lack of social support leads to chronic stress and adverse health behaviors [18, 44]. To calculate the SED index, its component variables were standardized via linear transformation such that the expected value of the variables was equal to 0 and the standard deviation was equal to 1. In addition, the sign was reversed for destimulants (proportion of people with higher education and average monthly salary). Finally, the SED index was assumed to be the arithmetic mean of the transformed components and calculated from the averaged values of component variables separately for the years 2010–2014. Therefore, the proposed SED index might be considered the mean of Z-Scores of its components, taking into account inverted scales for destimulant variables. The calculated index values were used to explain the differences in mortality across sub-regions of Poland. In addition, based on the 5-year averaged values of the SED index, the 66 sub-regions were divided into three groups with equal number of areas and with lower and upper tercile of averaged SED index as cut-off points:

This classification was used to emphasize the inequalities in mortality for descriptive purposes.

Tobacco consumption and body mass index (BMI) were calculated for the 66 sub-regions based on the results of the Social Diagnosis Survey conducted in 2011. The survey involved 12,386 households selected based on the two-stage stratified sampling method—first at the voivodeship level, followed by sampling based on residential location (large cities, small cities, and villages) [34]. The study sample comprised 36,753 persons aged 16 years and over. The participation rate of the study was 72% (i.e., 26,453 persons were finally examined). To eliminate the socioeconomic differences in urbanization between the sub-regions in the model, population density was included as a standardizing factor [19, 45].

The relationship between mortality and the SED index was determined using the multivariate Poisson regression model for count data. As this model reflected the mortality coefficient per 100,000 residents rather than the number of deaths, the age-standardized number of deaths in individual sub-regions was taken as a dependent variable, with the number of people in sub-regions serving as the offset. The generalized linear models with Poisson distribution as probability distribution and logarithm as a link function were used for calculations.

Due to repeated measurements in the same statistical units (mortality in sub-regions in subsequent years during 2010–2014), generalized estimating equations (GEE) were used to obtain generalized linear models for the correlated data [46]. An exchangeable structure was assumed for the working correlation matrix in GEE models.

Variance Inflation Factors were calculated to check the presence of a potential multicollinearity problem in multivariable models. In no case did the VIF exceed the value of 4 taken as a threshold, and the maximum value observed was 3.84.

The regression models developed for the study included three causes of death (CSD, IHD, CeVD) for men and women. Furthermore, as independent variables, calendar year (for assessing the time effect on the dependent variable) and the aforementioned index - level of socioeconomic deprivation in the sub-regions - were included in the models. In addition, the models included lifestyle factors, such as mean BMI and the average percent of smokers in the sub-region (for the female and male population, respectively) and population density in the sub-region using logistic transformation at base 2.

Simple, univariate relationships between the SED index and mortality have been assessed with Spearman’s nonparametric correlation coefficients.

The study showed a considerable geographic variation in CSD mortality and SED in the 66 sub-regions of Poland. For both men and women, the largest mortality was found in the more deprived and middle group sub-regions, and the lowest in the less deprived sub-regions. However, the more deprived and middle group sub-regions benefited more from decreased CSD mortality in men and women. This indicates a reduction in health inequalities, which was also observed in other European countries [47, 48]. In women, the differences in the number of DPP between the SED groups were similar to that in men, except in the case of CeVD for which a nonsignificant trend was observed. The relationship between SED and CSD mortality was significant in men after adjusting for mean BMI, mean smoking rate, and population density. In women, a significant relationship was observed with mean BMI. However, the relationship between CSD mortality and SED was insignificant after adjusting for covariates, although the β coefficients were similar to that of men.

Our study indicates that the variation of CSD mortality in Poland’s sub-regions can be explained at least partially by the differences in SED. The results comply with the studies in which synthetic SED indexes were used to assess the relationship between deprivation and mortality due to CSD, IHD and CeVD [13‐15, 19‐22, 49]. It has been suggested that deprivation significantly influences the mortality due to CSD in men, but not women [13, 50]. We did not show a significant relationship in women, however the effect of SED is similar in size to that in model for men, which may indicate insufficient statistical power of the study.

While the impact of lifestyle on cardiovascular health is well known, it should be noted that there are gender differences in behavioral risk factors. For example, smoking prevalence traditionally affected more men than women [51]. In our study, CSD mortality was shown to be associated with tobacco smoking only in men. This could reflect the higher smoking rates in the group of more deprived sub-regions compared with the less deprived sub-regions. According to the MORGAM study, in Poland, smoking is less prevalent among women with low and secondary education than women with higher education, particularly in less urbanized areas [52]. A study indicated that in Eastern European countries, economic development and social and cultural processes associated with gender empowerment affect the differences in smoking between educated and uneducated women. It appears that the education related different patterns for smoking in Eastern European countries may interfere with the relationship between SED and mortality due to CSD in women, which may have influenced the lack of significance of the association [52, 53]. In our study, another important behavioral factor was BMI, which was associated with mortality due to CSD only in women. Although the prevalence of BMI varies by gender, obesity is more common in men below 45 years old, but at later ages above 45 years old, it predominates in women [54]. Some studies suggest that the risk of developing heart disease in women is increased by obesity coexisting with some classical CSD factors, i.e. diabetes, hypertension and hypercholesterolemia. Compared to men, these factors predispose women to heart disease to a greater extent [55].

It is worth noting the results obtained in assessing the association between SED and IHD mortality. A significant relationship was not found in univariate analysis, but the inclusion of additional explanatory variables nevertheless showed the existence of such a relationship. This may be related to the problem of omitted-variable bias: in the analysis ignoring the additional variables, the univariable model necessarily imputed their effect to the SED scale. By adding the omitted variables, the SED scale no longer assumes their partial effect, but reflects its “true” effect, which turns out to be statistically significant. Furthermore, this study confirmed the relationship between SED and mortality from IHD, with increased deprivation more strongly increasing mortality in women than in men (by 30.3% vs. 19.5% per SED scale unit). This significant result in both sexes may indicate similar risk factors for IHD, which may be explained by the fact that deprivation co-occurs with stress-related biological risk factors such as hypertension and diabetes [8, 18]. However, for the higher score in women, significant hormonal changes during menopause may play a role, which may also aggravate preexisting risk factors [56]. As in other studies [57, 58], we found the relationship between SED and mortality due to CeVD only in women. An explanation for this meaningful relationship may be psychosocial factors, which play a more significant role in women; for instance, depression may mediate the relationship between low SED and stroke [57]. However, the unclear result in the male population may be associated with a lower risk of stroke among unskilled manual workers compared with high-grade civil servants and executives [58].

Our results showed inequalities in mortality across sub-regions ranked by deprivation. The lowest CSD mortality was found in the less deprived sub-regions, which predominantly include large urban agglomerations and small cities. This observation may be associated with the fact that these sub-regions have access to different resources, including educational infrastructure, services, and job opportunities [59]. Furthermore, these sub-regions offer a more favorable environment, such as access to gyms and shops selling healthy foods, as well as health care services, which may contribute to better health outcomes in their residents [45, 60, 61]. The more deprived sub-regions are located in the eastern and northwest parts of Poland and are characterized by a low population density. These are considered less attractive to investors, which affects economic development. This confirms, among others, the fact that the areas of many sub-regions with more deprived sub-regions overlap with those areas in which national farm holdings had been liquidated and more harmful effects of the economic transformation are experienced. These areas lack support programs targeting social groups deprived of earlier forms of employment, and are consequently linked with poor health outcomes [26, 27].

Furthermore, the area encompassing sub-regions with a high deprivation in the western part of the country was associated with a more clear trend toward job migrations to work legally abroad. The migrations may exacerbate the optimal development of these areas due to the outflow of human capital and are also associated with a prevalence of social exclusion and poverty, leading altogether to health hazards [26, 62]. Some analyses indicate that the eastern and northwestern parts of the country have high total mortality due to CSD of the age group 25–64 years [63]. The issue of worse cardiovascular health in areas where national farm holdings were eliminated and areas with intensive labor migration is an interesting topic that needs further research.

For IHD, mortality was the highest in the less deprived sub-region group. In the case of CeVD, the highest mortality was found in middle sub-regions, which may suggest the occurrence of other specific factors that are not included in the analysis [64]. Indeed, for both IHD mortality and CeVD, positive relationship with the SED index has been confirmed using multivariable models that account for additional sources of mortality variation. It should also be emphasized that a rapid decrease in CSD mortality was noted in more deprived sub-regions with high SED index values. In the more deprived and middle group sub-regions, the percentage of DPP was higher compared to the less deprived sub-regions.

While the mortality due to CSD and CeVD was the lowest in the less deprived sub-regions, the reduction of mortality between 2010 and 2014 in these sub-regions was smaller compared to the sub-regions with the high SED index. This could be explained by the fact that in the best-developed—highly urbanized—sub-regions, mortality (due to better availability of cardiologic care services and invasive cardiology procedures) had been reduced in earlier years and the scope for improvement of cardiovascular health was narrower [65]. Simultaneously, modern prevention and therapeutic methods are increasingly becoming accessible in more deprived and middle sub-regions. For example, life-saving invasive cardiology procedures have been available after 2000 not only in academic centers but also in district hospitals, which might have resulted in reduced mortality even in smaller centers with a higher SED index. This is also supported by the results of other analyses [65], according to which the reduction of mortality due to CSD among people with higher education was particularly pronounced between 1991 and 1993 and 2001–2003, whereas during the period 2001–2003 and 2010–2012, the reduction was considerably lower. Noticeably, this mortality reduction was observed at a lower rate in people with low education compared to those with higher education.

Our results have shown that mortality from CSD, IHD and CeVD has decreased, and similar trends prevail in most European Union (EU) countries [1, 4, 5, 63]. It is also worth noting the progressive improvement in the health status of the Polish population, as evidenced by the systematically decreasing mortality rate of both younger and older people [63, 66]. At the same time, however, it must be said, against the background of the EU countries in the analyzed period 2010–2014, the mortality rates due to CSD in Poland were 60% higher compared to the average level for the EU, but already comparing to the old EU-15 countries were higher by 90%. For example, the level of mortality due to CSD in Poland is three times higher compared to France and the Netherlands. In contrast, against the background of the EU-13 countries that joined the EU after 2003, the level of mortality due to CSD in Poland is relatively low, i.e. the rates are 15% lower. Among Poland’s neighboring countries, the level of mortality due to CSD in Poland was 50% higher compared to Germany and similar to the level of mortality in the Czech Republic and Slovakia [67]. The existence of significant disparities in the level of mortality due to CSD between Poland and EU countries, especially the old EU-15, may indicate that there are opportunities for further mortality reduction. The stratification using terciles allowed to identify the residence-related deprivation, which may be necessary at intervention strategies and activities addressing improving cardiovascular health. In the policy process aimed to alleviate health inequalities, the activities should focus not only on poverty reduction throughout income redistribution. It is also critical ensuring equality of health opportunity in the entire population, through education, employment, improved working conditions and preventive care [68]. Decreasing health inequalities ought to be a political and social priority, given that they deteriorate economic productivity and the potential for sustainable and inclusive growth.

To our knowledge, this study is the first to use the SED index to assess the relationship with mortality due to CSD at the population level in Poland. The study was performed considering the whole population of the country. In the 66 sub-regions, a large variation was found in CSD mortality and SED. Furthermore, the sub-regions represented all the characteristics that are typical for Poland. The synthetic SED index enabled an approximate estimation of the singular variables (education, structure in employment, salary, unemployment, and poverty). A database concerning sub-regions defined based on NUTS-3 classification, which is used in the EU member states, was used for the first time in this study [32]. A unique strength of this study was the comparison of mortality due to CSD and the time-related changes using the DPP index in three different environments regarding deprivation level, thus, contrary to other studies reporting mortality trends in administrative areas of the country [65]. Our results showed a potent effect of health inequalities. Therefore, they may contribute to limited literature field dealing with the associations between area-related deprivation and mortality from CSD in Central and Eastern European countries [21, 22]. Similar and comparable socioeconomic levels in specific countries of this region may facilitate cross-comparisons and allow consistently conducting research across the populations.

The results of the study should be interpreted in light of certain limitations. The ecological design does not allow addressing the causality of the relationships. As sub-regions were considered as statistical units in this study, instead of individual persons, it was possible to investigate the inequalities between them, while inequalities within them remain unexplored. Epidemiological analyses for geographical areas lead to the best results when statistical units are populations of small size [69, 70]. It allows for better homogeneity and decreases the problem of averaging of different populations within a single geographical area. Such averaging results in attenuation of studied effects in statistical models. In this paper, population sizes in sub-regions were relatively large, so it was expectable that real effects might be significantly stronger and so their detection could be difficult, if not impossible. However, we were able to confirm statistically significant associations of the SED index with mortality from each of the three causes of death analyzed, although not always for both sexes. One explanation for the difficulty in demonstrating these relationships in each case may be the aforementioned high level data aggregation at the sub-regional level and the blurring of the relationships studied. In contrast to mortality and SED, information on covariates such as education, smoking, and BMI was based on one-point observation (Census 2011, Social Diagnosis Survey 2011), which was the only available for the studied sub-regions within the study period. It is unlikely that these characteristics changed much within the observation time. The relationship between SED and mortality due to CSD could be confounded by the sub-region differences in the exposure to the other uncontrolled factors; therefore, residual confounding should be considered. For example, we did not utilize stress, an important variable, because of the limited accessibility of such data. However, stress is regarded as a significant mediator linking the associations between deprivation and mortality due to cardiovascular diseases [71]. Noticeably, some components of the SED (e.g. percent of people on social support due to poverty) may have, at least partly, reflected chronic stress [18]. Thus, the role of stress should be investigated in future research. Another limitation of the study is the quality of the data on deaths associated with the differences in the reliability of death-cause coding by physicians who filled out death certificates. Our results showed that the highest mortality due to IHD was observed in the less deprived (most urbanized) sub-regions. This may be partially explained by the discrepancies in death certification encoding described earlier in Poland [72]. Such an issue was even reported in countries with highly advanced health information systems [73]. However, territorial differences in death-cause coding could rather contribute to the greater impact of random variability on our results, and it is less likely that the occurrence of a systematic error would explain the observed relationships. The latter is supported by the overall consistency of relationships found for IHD, CeVD, and CSD.