OUP user menu

Association of proximity to polluting industries, deprivation and mortality in small areas of the Basque Country (Spain)

Koldo Cambra , Teresa Martínez-Rueda , Eva Alonso-Fustel , Francisco B. Cirarda , Covadonga Audicana , Santiago Esnaola , Berta Ibáñez
DOI: http://dx.doi.org/10.1093/eurpub/ckr213 171-176 First published online: 7 February 2012

Abstract

Background: The study is aimed at assessing social inequities in the location of polluting industries in the Basque Country, and at exploring if the effect on mortality of living near air polluting industries is modified by economic deprivation. Methods: This is a cross-sectional ecological study that uses the census sections as analysis units. Mortality from all causes, lung cancer, respiratory diseases and ischaemic heart disease were studied. Ordinal logistic regression models were fitted to assess if proximity of census sections to polluting industries is associated with deprivation. Bayesian Poisson regression models were used to explore if the association between proximity to polluting industries and mortality is modified by socio-economic deprivation. Results: Proximity to a polluting industry and deprivation are positively associated, showing a clear gradient across deprivation quintiles. In women, the risk associated with proximity to metal-processing industries grows as the deprivation of the area increases in the case of total and lung cancer mortality. In men, the interaction terms between proximity and deprivation are positive for total, ischaemic heart disease mortality, with a credibility level approaching 90%. High levels of deprivation are associated with greater risk of mortality, excepting lung cancer in women. Conclusion: There is a higher proportion of more deprived census sections around polluting industries in the Basque Country. Risks of mortality associated with proximity to polluting industries tend to be higher in more deprived areas.

Introduction

Association between deprivation and increased mortality and morbidity has been found across a wide range of diseases and areas.13 In a small area study conducted in 11 Spanish cities, Borrell et al.4 reported that deprivation is positively associated in men with mortality from lung cancer, ischaemic heart diseases, respiratory diseases and cirrhosis, and in women that it is positively associated with diabetes and cirrhosis mortality but negatively with lung cancer mortality.

Exposures to environmental pollution have been proved to be important risk factors for many diseases. Short- and long-term increases in mortality rates in relation to air pollution, particularly airborne concentrations of particulate matter, have been well established on the basis of large multicity studies conducted in the USA5,6 and in Europe.7,8 Industrial activities are a recognized source of environmental pollutants. Cancer, mortality and congenital malformations have been the health problems most widely addressed in international literature in relation to residential proximity to industrial sites and landfills.911

Recent attention has been focused on the question of whether environmental hazards and their impacts on health are equitably distributed in the population or, on the contrary, most disadvantaged communities end up bearing a disproportionate part of them. In the USA, larger and more chemical intensive facilities tend to be located in counties with larger African-American populations.12 In Europe, the studies on this question have been relatively scarce and mainly focused on air pollution and traffic. Higher levels of air pollution have been reported in lower socio-economic status (SES) neighbourhoods,13,14 but no association or even a negative association with traffic and traffic-related pollutants have also been reported in some European cities.15,16 Briggs et al.17 confirmed the existence of environmental inequities related to a series of environmental risk factors in England, and they observed that the association of deprivation tends to be stronger with measures of air pollution than with other type of hazards, and with environmental concentrations rather than with proximity to emission sources. In a review of literature published in Europe and the USA, Martuzzi et al.18 claim that available data provide consistent indications that waste facilities are more often located in areas with more deprived residents or with residents from ethnic minorities. The spatial correlation between different environmental hazards also implies that exposures will rarely occur singly, and that those exposed to an environmental hazard are likely to be subject to complex mixtures. Whether there is an interaction between environmental and socio-economic factors is still relatively unexamined.

The Basque Country is a highly industrialized region situated in the north of Spain, on the coast of the Bay of Biscay. It has an area of 7234 km2, a population slightly over two million, and a population density of 295 inhabitants/km2. In a small area study conducted in the region, some slight excesses in mortality from lung cancer in men, and from ischaemic heart and respiratory diseases in women were found, respectively, in the vicinity of energy producing plants and metal-processing industries. Those excesses remained after adjusting for deprivation and spatial correlation. No association was found with other types of industries and/or other mortality causes.19 The objectives of this study have been (i) to assess the extent in which location of polluting industries is related to the socio-economic characteristics of the nearby neighbourhood and (ii) to explore if the association between proximity to polluting industries and mortality is modified by area-based socio-economic deprivation.

Methods

This is a cross-sectional, ecological study, with census sections (CSs) as units of analysis. The study population consisted of all residents in the Basque Country (2 082 587 according to the 2001 census), and all deaths (146 195) occurring during the 1996–2003 period were included. CS with less than 500 inhabitants were integrated into adjacent sections, resulting in a total of 1645 units for analysis, with a mean size of 1257 inhabitants ranging from 500 to 3500. Population data were obtained from the 2001 Population and Housing Census and death data from the Basque Country Death Statistics conducted by the Basque Statistics Institute (EUSTAT). We studied the mortality from tracheal, bronchial and lung cancer (IDC-10: C33), from ischaemic heart disease (IDC-10: I20–I25), from chronic lower respiratory tract diseases (ICD-10: J40–J44, J47). These specific mortality causes were selected for being the causes for which we had previously found some association with proximity to energy or metal-processing industries.19 Additionally, as a more general indicator, mortality from all causes was also included. All analyses were done separately for women and men.

Data of type of industry and location were obtained from the European Pollutant Emission Register (EPER) (Decision 2000/479/EC). The total number of EPER industries in the Basque Country in 2001 is 66; of those energy and metal-processing industries make up 4 and 28, respectively. The great majority of them were established in their actual locations before the 1970s. Two variables of proximity were used as proxy variables of exposure to pollutants: a dichotomous variable taking value 1 for CS whose centroid lies within a 2 km buffer from the industries and 0 otherwise, and an ordinal variable taking four different values according to the distance (d) of the centroid of the CS to the nearest EPER industry: ‘0’ if d >2 km, ‘1’ if 1 km < d < 2 km, ‘2’ if 0.5 km < d < 1 km and ‘3’ if d < 0.5 km.

We used the continuous score deprivation index (DI) proposed in the MEDEA project,20 which combines the indicators from 2001 related to unemployment (percentage of people aged 16 years seeking a job), low educational level (percentages of people with <5 years of schooling or/and not completing the basic compulsory education, in population over 16 and between 16 and 29 years), manual workers and temporary workers (percentages over the total employed population). The index was normalized with a mean 0 and standard deviation of 1.The DI reflects the material deprivation of CS: the higher the index, the higher its deprivation.

Statistical analysis

To assess the relationship between deprivation and proximity to EPER industries, we used two approaches. First, a trend test for proportions to check if the proportion of CS near EPER industries (d < 2 km) changes along the quintiles of DI. Secondly, in order to quantify the magnitude of the association, we adjusted ordinal logistic regression models, with proximity taking values from 0 to 3 as the dependent variable and the quintiles of DI as the only explanatory variable.

To assess if there is an interaction on mortality between proximity to polluting industries and deprivation, we derived smoothed estimates using Poisson generalized linear-mixed models with two random effects, one corresponding to spatial dependence (ui) and the other to unstructured heterogeneity (vi), as proposed by Besag, York and Mollié (BYM).21 For each mortality cause and type of industry, we adjusted two models, with and without an interaction term, as follows: Embedded Image Embedded Image where ei are the expected cases for each census tract obtained by indirect standardization with the age-specific mortality rates of the Basque Country in the study period (1996–2003), Xid is the dichotomous variable of proximity to an energy or metal-processing industry and Xie is the DI.

The relative risk estimates were obtained through a measure of central tendency of the posterior distribution, the median and its 90% credibility interval (90% CI). This distribution was obtained using Monte–Carlo methods based on three Markov chains (MCMC) with Gibbs sampling algorithm, with 148 500 iterations, a burn-in period of 12 000 iterations and a thin of 75. Model convergence was controlled using the Brooks–Gelman–Rubin statistic and the effective sample size of the chains. Criteria for convergence were for the former to be less than 1.1 and for the latter greater than 100. The hyperprior distributions assigned to the variance of the random effects was the uniform distribution U(0,5). More detailed information about models, estimation and computing procedures can be found elsewhere.22 WinBUGS 1.4.1 and R 2.9.2 software were used.

Results

In the Basque Country, the number of CS near EPER industries is relatively large. Of the total number of 1465 CS, 61 (3.7%) are closer than 0.5 km to an EPER industry, 223 (13.6%) between 0.5 and 1 km, 495 (30.1%) between 1 and 2 km and 866 (52.6%) further than 2 km.

Proximity to an EPER industry and deprivation are positively associated. The proportion of CS within a 2 km buffer of EPER industries is higher in more deprived CS. This tendency is observed in relation to all EPER industries and more noticeably in relation to energy and metal-processing plants. The number of CS in the lowest quintile of deprivation (most affluent) near EPER industries is low. Trend tests are statistically significant both for all industries together and by activity (table 1).

View this table:
Table 1

Association between proximity of CSs to an EPER industry and DI (Basque Country 2001)

Type of industryDI quintilesaNumber (%) of CSOrdinal logistic regression
d > 2 kmbd  < 2 kmbPOR (95% CI)cP
Energy1326 (22)3 (2)<0.001Ref.
2311 (21)18 (12)6.32 (1.84–21.67)0.003
3295 (20)34 (23)12.56 (3.82–41.34)<0.001
4294 (20)35 (24)12.92 (3.93–42.52)<0.001
5273 (18)56 (38)22.26 (6.89–71.95)<0.001
Metal1299 (25)30 (7)<0.001Ref.
2251 (21)78 (18)3.06 (1.95–4.80)<0.001
3245 (20)84 (19)3.39 (2.17–5.30)<0.001
4218 (18)111 (26)5.03 (3.25–7.78)<0.001
5197 (16)132 (30)6.40 (4.15–9.85)<0.001
All EPER1249 (29)80 (10)<0.001Ref.
2199 (23)130 (17)2.02 (1.46–2.82)<0.001
3166 (19)163 (21)2.93 (2.12–4.05)<0.001
4149 (17)180 (23)3.75 (2.72–5.18)<0.001
5103 (12)226 (29)6.28 (4.55–8.66)<0.001
  • a: quintiles: one corresponds to the least deprived areas and five to the most deprived ones

  • b: d > 2 km and d < 2 km: distance of CS centroid to an EPER industry greater/smaller than 2 km

  • c: OR calculated taking the first quintile of deprivation (highest SES) as the category of reference

The odds ratios (ORs) of being closer to EPER industries show a clear gradient through deprivation quintiles and they increase as deprivation of CS increases. People living in CS with DI in the fifth quintile of deprivation have an odds of living nearer to EPER industries six times higher than those in CS of the first quintile, and people living in CS of the second quintile have an odds twice as large. ORs for metal-processing industries are slightly higher than those derived for all EPER industries; in the case of energy plants, however, they appear to be much larger, yet precision of the estimates are rather low (table 1).

In models without interaction term (left part of tables 2 and 3), living within a 2 km buffer from metal-processing industries is associated with higher mortality from ischaemic heart disease and respiratory diseases in women. Results related to deprivation differ between sexes. While higher levels of deprivation are constantly associated with greater risks of mortality in males, deprivation is not significantly associated with mortality in women, excepting lung cancer, for which the association between mortality and deprivation is negative i.e. higher risks for lower deprivation.

View this table:
Table 2

Association of mortality in men with proximity to energy/metal-processing industries and deprivation (DI) of CSs

Sex; IndustryMortality causeProximitya, β1 (90% CI)DI, β2 (90% CI)Proximitya, β1 (90% CI)DI, β2 (90% CI)Proximity × DI, β3 (90% CI)
Men; EnergyLung cancer0.060 (−0.042–0.158)0.076 (0.053–0.103)0.041 (−0.073–0.160)0.117 (0.092–0.142)0.059 (−0.034–0.151)
Ischaemic heart disease0.011 (−0.093–0.110)0.033 (0.010–0.057)0.013 (−0.102–0.130)0.063 (0.036–0.089)0.014 (−0.081–0.110)
Respiratory diseases−0.006 (−0.132–0.140)0.086 (0.053–0.118)−0.105 (−0.243–0.034)0.177 (0.145–0.208)0.065 (−0.056–0.188)
Total mortality0.000 (−0.050–0.048)0.044 (0.033–0.056)−0.016 (−0.066–0.035)0.083 (0.072–0.095)0.024 (−0.016–0.069)
Men; MetalLung cancer0.024 (−0.038–0.088)0.077 (0.054–0.102)0.012 (−0.052–0.077)0.125 (0.095–0.153)−0.005 (−0.054–0.052)
Ischaemic heart disease0.010 (−0.048–0.063)0.032 (0.009–0.056)−0.050 (−0.114–0.014)0.051 (0.024–0.081)0.049 (−0.006–0.102)
Respiratory diseases0.040 (−0.037–0.120)0.084 (0.053–0.116)−0.036 (−0.111–0.044)0.160 (0.124–0.194)0.065 (0.000–0.133)
Total mortality0.008 (−0.021–0.036)0.044 (0.033–0.055)−0.023 (−0.052–0.006)0.081 (0.069–0.093)0.015 (−0.008–0.039)
  • Results of BYM regression models with (right part) and without (left part) an interaction term between proximity and DI

  • a: Proximity: dichotomous variable taking the value 1 when the centroid of CS is <2 km from an EPER industry and 0 otherwise

View this table:
Table 3

Association of mortality in women with proximity to energy/metal-processing industries and DI of CS

Sex; IndustryMortality causeProximitya, β1 (90% CI)DI, β2 (90% CI)Proximitya, β1 (90% CI)DI, β2 (90% CI)Proximity × DI, β3 (90% CI)
Women; EnergyLung cancer−0.003 (−0.235–0.249)0.101 (−0.161 to 0.044)−0.034 (−0.320–0.225)−0.135 (−0.200 to −0.075)0.063 (−0.212–0.346)
Ischaemic heart disease0.015 (−0.110–0.146)0.000 (−0.029 to 0.031)−0.072 (−0.212–0.081)0.024 (−0.008 to 0.056)0.098 (−0.025–0.233)
Respiratory diseases−0.199 (−0.391–0.006)0.036 (−0.007 to 0.077)−0.183 (−0.414–0.015)0.060 (0.017 to 0.104)−0.024 (−0.231–0.172)
Total mortality0.021 (−0.029–0.066)0.008 (−0.003 to 0.020)0.011 (−0.038–0.065)0.031 (0.018 to 0.043)−0.011 (−0.062–0.036)
Women; MetalLung cancer0.093 (−0.044–0.235)−0.111 (−0.173 to −0.052)0.020 (−0.125–0.166)−0.176 (−0.248 to −0.105)0.151 (0.020–0.277)
Ischaemic heart disease0.088 (0.014–0.155)−0.005 (−0.034 to 0.024)0.018 (0.062–0.099)0.023 (−0.013 to 0.060)0.016 (−0.052–0.085)
Respiratory diseases0.114 (0.002–0.212)0.020 (−0.021 to 0.065)0.048 (−0.073–0.157)0.042 (−0.009 to 0.093)0.025 (−0.075–0.124)
Total mortality0.011 (−0.018–0.040)0.009 (−0.003 to 0.019)−0.004 (−0.035–0.025)0.022 (0.009 to 0.036)0.029 (0.003–0.055)
  • Results of BYM regression models with (right part) and without (left part) an interaction term between proximity and DI

  • a: proximity: dichotomous variable taking the value 1 when the centroid of CS is <2 km from an EPER industry and 0 otherwise

Nevertheless, in women, there appears to be a significant effect modification between deprivation and proximity to metal processing for total and lung cancer mortality (right part of tables 2 and 3). In men, positive interactions almost reaching 90% credibility are found for respiratory, ischaemic and total mortality. The magnitude of the estimated effect modification is the difference in slopes between the straight lines of those CS within the 2 km buffer and those out of it (figure 1).

Figure 1

Regression plots showing the interaction on mortality between DI and location of the CS within a 2 km buffer from a metal-processing EPER industry. Results from BYM models

No significance was found for the interaction between deprivation and proximity to energy producing plants, neither for women nor for men.

The fraction of the variance with spatial structure differs depending on the mortality cause, sex and type of industry, ranging from 0.66 to 0.85 for lung cancer, from 0.20 to 0.9 for respiratory diseases, from 0.14 to 0.39 for ischaemic heart diseases and from 0.08 to 0.30 for total mortality.

Discussion

There is a higher proportion of more deprived CSs around polluting industries in the Basque Country. Risks of mortality associated with proximity to polluting industries tend to be higher in more deprived areas, though the results are not conclusive.

The ecological design, the use of proximity to industries as a proxy for exposure to environmental pollutants and the limited statistical power due to the low number of CS of low deprivation near energy producing plants are weaknesses of this study. Also, workplace hazards also raise the possibility of confounding, as persons exposed to high risk at work could more likely live in deprived neighbourhoods situated near polluting areas, and differences among areas in social patterns and mobility of the residents would affect the accuracy of the use of proximity to industries as a proxy of exposure to pollutants.

The most noteworthy strong points of the study are the use of CS as the analysis unit and the statistical models employed. The use of such small analysis units as CSs reduces heterogeneity within areas in relation to exposure to risk factors. BYM models have low sensitivity and high specificity for detecting excess risks, reducing the probability of false positives23 and they provide risk estimators controlled for location, i.e. controlled by a set of unknown risk factors varying smoothly in space.24

Deprivation in the Basque Country is positively associated with mortality from lung cancer, ischaemic heart disease, respiratory diseases and total mortality in men. On the contrary, in women, no statistically significant associations have been found, excepting mortality from lung cancer, for which greater mortality rates are observed in women of higher SES. These results are quite in line with what has been reported in several cities of Spain.4 The fact that in Spain, tobacco consumption has been higher among men of lower SES and among women of higher SES has been suggested as the main explanation for the different pattern in lung cancer mortality between men and women.25

Urban and industrial density in the Basque Country is high and not evenly distributed. The likelihood of Basque people to live near polluting industries depends on the SES of their neighbourhood, showing a clear positive gradient along the whole range of deprivation: the higher the DI the higher the odds of living near a polluting industry. The lack of similar data in Spain makes it difficult to assess the relative magnitude of this environmental inequity. In UK, Elliot et al.26 reported that 34% of the population living in the area within 2 km of landfill vs. 23% of the reference population were in the most deprived tertile of Castairs score; these proportions do not seem to be different from the ones we have got for EPER industries in the Basque Country.

The reasons why lower SES sections are nearer to polluting industries are possibly complex and historically rooted. During the economic booms of the 20th century, mainly in the 1900s and 1960s, tens of thousands of manual workers from other regions arrived, and the Basque Country experienced unprecedented social and urban changes. Some towns became residential areas for wealthy classes. But in many others, urban areas actually developed in a messy and disorganized way, devoting the best and most accessible places for factories and leaving residential land trapped between the mountainous orography and the factories themselves. Often, out of necessity and following the paternalistic policies of the time, local companies would promote or directly build houses for their workers.27,28 Despite the subsequent closure of mines and industries, and the gentrification that industrial surroundings have undergone in the last decades, we think that this disorganized urban development is to a great extent responsible for the social gradient observed nowadays. However, new forces may be still generating environmental inequalities related to the location of industrial polluting facilities. Among them, silent factors ruling the housing market and land planning are likely to be influential. The presence of polluting industries depresses the housing market in their vicinity, making it more affordable for lower income groups, and also attracting new industrial polluting facilities, as the price of the land is also lower and the probability of a social response against is smaller than in places with no industries and higher environmental preservation.

The social gradient in the odds of proximity is more noticeable for both metal-processing industries and energy industries than for all EPER industries together. This can be meaningful because these types of industries are the main source of atmospheric emissions of heavy metals, polycyclic aromatic hydrocarbons (PAHs), benzene and particles in the Basque Country. In 2003, in fact, the steel industry and the power stations were responsible for 32% of the total amount of particles <10 µm in diameter (PM10) emitted into the air.29

Additionally to exposure differentials related to socio-economic conditions, there is some evidence that the effect of air pollutants may be higher in lower SES populations and that more disadvantaged people may have a higher susceptibility to disease.16,30 Some authors have hypothesized that some of the observed social disparities in health outcomes may be explained by air pollution exposure.31,32 The routes proposed by O'Neill et al.32 are that people with lower SES would be more exposed to air pollution, and also that their vulnerability to air pollution-related health consequences would be increased. Deguen and Zimirou-Navier,33 in a review of European studies, found that the general pattern in terms of health consequences is that deprived populations experience greater harmful effects of air pollution, suggesting as explanatory factors differential exposure beyond ambient pollutant levels—due to differences in activity patterns and housing conditions across social classes—and differential susceptibility—due to higher prevalence of chronic conditions and nutritional deficiencies among the most deprived. Other authors have referred this issue as the triple jeopardy: if more deprived people, with poorer health, are more likely to be exposed to environmental hazards, and they are more susceptible to their effects, then their health will be more compromised.17 Our results on the effect modification on mortality between deprivation and proximity to EPER industries cannot be conclusive but they constantly show the same tendency. All the interaction terms between proximity and deprivation that are at least marginally significant have been positive, which suggests that the effect on mortality of the proximity to polluting industries may increase as the deprivation of the area increases or, conversely, that deprivation may have a stronger effect on mortality in areas near EPER industries than in areas located further away. For lung cancer mortality in females, the negative association with deprivation almost disappears in CS located within 2 km buffers from metal-processing industries. Furthermore, models including the interaction term between deprivation and proximity to industries suggest that living near metal-processing industries may be associated in more deprived groups with increased mortality from other causes than the ones identified for the whole population in the log-additive models. Thus, the combined effect of environmental exposures to risk factors and deprivation is likely to be more complex than additive, which makes necessary to consider socio-economic factors when evaluating environmental hazards.

Further research is needed to better understand what is beyond socio-economic constructs such as the DI, namely, epidemiological research, including individual-based studies, should be led to identify the main risk factors that are responsible for the social health inequalities observed. Knowing the role that environmental pollution may play among them would enable to better define and implement health-targeted environmental programmes and policies.

Funding

This work has been partially funded by the Instituto de Salud Carlos III (PI04/0489 and PI04/0388), Spain.

Conflict of interest: None declared.

Key points

  • Association between deprivation and increased mortality and morbidity has been found across a wide range of diseases and areas.

  • Environmental pollutants have been proved to be important risk factors for many diseases but its relationship with socio-economic factors is relatively unknown.

  • We found greater deprivation in neighbourhoods near polluting industries.

  • Risks of mortality associated with proximity to polluting industries tend to be higher in more deprived areas.

  • It seems necessary to consider socio-economic factors in evaluating environmental health hazards.

Acknowledgements

This study was conducted within the project Inequalities in Health and Environmental Pollution in the Basque Country Autonomous Community (MEDEA). Information about Medea Project is available at: http://www.proyectomedea.org/eng/medea.html.

References

View Abstract