Proximity to mining industry and respiratory diseases in children in a community in Northern Chile: A cross-sectional study

The mining industry has been one of the most important contributors to the Chilean economy with a share of 13 % of the gross domestic product [1]. Gold and copper are extracted in open-pit mining facilities through excavation—a process that uses explosives and heavy machinery. Most of the air pollutants emerging form open pit mining are total suspended particulate (TSP) matter and particles, including aerodynamic diameter smaller than 10 μm (PM₁₀) [2, 3]. The gold and copper mining activities generating these particles are mainly drilling, blasting, sediments loading and unloading, road transport over unpaved roads and losses from exposed over-burden dumps, residuals handling plants and exposed pit faces [4]. Dust after the explosion and transportation of materials and have been identified as the main source of TSP and PM₁₀ pollution [5–7] in open-pit mining regions. High concentrations of particulate material (TSP and PM₁₀) have been associated with asthma and allergies among others diseases [8–11]. Asthma is a common disease that affects children and adults of all ages. According to findings of the International Study of Asthma and Allergies in Childhood (ISAAC), the prevalence of asthma in childhood in Latin America is on average around 13 % and ranging between 13 and 16 % in Chile [12, 13].

Using measurements of air pollutants, associations between elevated levels of ambient air pollution and frequency of asthma attacks (emergency room visits/hospital admissions) have been shown [14–17]. Using proximity to the air pollution source as a proxy of exposure, studies suggested associations between living in areas close to industrial zones and the risk of asthma in children based on spatial analyses [18, 19]. In these studies, mainly logistic regression models and generalized additive models (GAM) [20] have been used. One limitation of the first approach is the use of arbitrarily chosen cut-off points for quartiles of exposure. For the latter, the interpretation of the results might be challenging, particularly when the association between the exposure and the outcome is non linear.

Countries in South America are experiencing high prevalence of asthma with rates ranging from 18 to 27 % [13, 21]. In Chile, higher prevalence of asthma and allergies was found in urban than in rural areas [22]. To the best of our knowledge associations between point sources of air pollution and respiratory health in children have not been studied in Chile to date. Exposure to open-pit mining is of particular interest as the exposure is widespread in Chile and other Latin American countries. In some communities, residents live less than one kilometer away from these open-pit mines.

Therefore, we examined associations between average distance of child’s place of residence to the two open-pit mines in a rural mining community and the prevalence of respiratory disease in children. The community is a mining dependent town located in Region IV in Chile with less than 10.000 inhabitants located at 1050 m above the sea level. Using data from a cross-sectional study, we assessed associations between the distance to the mines and respiratory diseases applying Bayesian semiparametric and parametric models including a spatial effect.

The prevalence of asthma was 24 %, and the prevalence of rhinoconjunctivitis was 34 %. The mean (± sd) distance to the gold mine was 2.1 (± 0.26) km (median: 2.1 km; range 1.3–2.8 km) and 1.9 (± 0.37) km (median: 2.0 km; range 0.9–3.1 km) for the copper mine. For the average distance to both mines the mean was 2.1 km (± 0.26) km (median: 2.0; range 1.3–2.7). The prevalence of the diseases among the 25 % of children living closest to the mines, was between 30 and 45 %, and was slightly higher than for the remaining sample (Table 1). A geographical map of children’s place of residence is presented in the Additional file 1: Figure SA.1.

The adjusted STAR models confirmed the results of the logistic regressions for all outcomes (Figs. 1 and 2) with those living closer to the mines having higher risks for respiratory diseases. A slight “U” shape in the estimations of the associations was seen for the copper mine, and for average distance in both outcomes (Figs. 1 and 2). For the association between distances to the copper mine and asthma no decrease towards zero was estimated. The findings for the two outcomes were consistent, i.e., with closer distance to the gold mine the risk of all respiratory diseases increased about exp(1) with confidence interval reaching from exp(0) to exp(2). Within a distance of 2 km to the gold mine we estimated a positive effect suggesting an increased risk for the respiratory outcomes. Around the copper mine, the risk was increased by exp(0.5) with BCIs between exp(1) and exp(2). The odds ratios were positive up to a distance of around 1.7 km. Combining the proximity to both mines by using the average distance resulted in similar estimates as the analyses of the distance to the gold mine only. The unadjusted models did not differ appreciably from the adjusted models (Additional file 1: Figures SD.1 to SD.2).

Using the STAR models (Figs. 1 to 2), we found a maximum OR of exp(1) at the minimum distance to the mines, with the lower and upper limits of the confidence intervals around exp(1) and exp(2), respectively. The construction of the prior distributions is presented in the Additional file 1: (Section F.1).

The parametric Bayesian models 0 to 3 indicated that model 3, including the proximity-risk function together with the spatial terms resulted in a better fit for all outcomes and exposures (Additional file 1: Table SD.1); the DIC was lower than for other models.

Table 3 shows the parameter estimates and the 95 % Bayesian confidence intervals for each mine. Generally, the point estimates of \( \widehat{\alpha_k} \) lay within 1.6–2.4, for \( \widehat{\phi_k} \) had values of 1.05–1.13 and \( \widehat{\delta_k} \) between 0.72 and 0.95. Figure 3 shows the posterior \( \widehat{f_k} \) estimates for distance functions. These graphs present similar values at the source of around 2, and it maintains the risk constant until around 1 km of distance from the mines; at further distance, the risk declines to 1 at around 2.5 km of distance to the mine. The 95 % confidence intervals for \( \widehat{f_k} \) are above 1 up to a distance of 1.8 km. Plots of the random terms Ŝ _k (s) using T = 40 spatial nodes are presented in the Additional file 1: Figures SE.1 to SE.2.

Respiratory disease	Mine	\( \widehat{\boldsymbol{\alpha}} \) 95%BCI	\( \widehat{\boldsymbol{\Phi}} \) 95 % BCI	\( \widehat{\boldsymbol{\updelta}} \) 95 % BCI
Asthma	Gold	2.48 (0.51–6.80)	1.08 (0.80–1.46)	0.80 (0.31–1.88)
	Copper	1.92 (0.29–5.81)	1.13 (0.82–1.53)	0.95 (0.31–2.13)
	Average distance	2.26 (0.42–6.62)	1.10 (0.80–1.50)	0.86 (0.31–1.95)
Rhinoc.	Gold	1.85 (0.46–4.87)	1.08 (0.78–1.48)	0.78 (0.28–1.97)
	Copper	1.61 (0.52–3.27)	1.07 (0.81–1.47)	0.78 (0.29–1.67)
	Average distance	1.65 (0.61–4.28)	1.07 (0.80–1.46)	0.77 (0.30–1.59)

Abbreviations: BCI Bayesian confidence interval, Rhinoc. rhinoconjunctivitis

The findings of our community based study in Chile suggest that proximity of children’s residence to open-pit mines, one of the main income sources in many low and middle income countries, is associated with allergic rhinoconjunctivitis and asthma with markedly increased risks in children living closest to the mines. We used different measures of proximity as proxy for ambient air pollution exposure which resulted in comparable results.

The results of our study corroborate previous study from Latin America suggesting increased risks for asthma within a distance of two kilometers to different air pollution sources including power plants, quarries or petroleum refineries [18]. Using categorical distance measures, investigators also found an increased prevalence of respiratory diseases and allergies in children who lived near wood industries in Alaska [19]. Indeed, the use of cut-off points is arbitrary, making the use of STAR models an alternative allowing a broader perspective on the exposure-outcome association in this kind of analysis.

We also estimated the spatial effect to capture the residual variation not accounted for by the population at risk, and to estimate risks across the study area. Models including an effect of distance from the mine on the respiratory diseases resulted in a better fit; however, the improvements were small.

Variables included as potential confounders for asthma were parental history of atopic disease, child’s mother working and living with both parents. For rhinitis only parental history of atopic disease played a role and was included in the adjusted models. Exposure to indoor sources of pollutants such as type of heating and smoking in presence of the child was not relevant for the the respiratory diseases in our data. Parental occupation was not included, because the work environment did not involve exposure to contaminants that could be associated with asthma attacks or any other respiratory disease in the children. We considered residential area rather than school location, as children reported to spend most leisure time at home. We assumed children spent only the remaining time at school. Since our study population only consisted of children from two different schools, the variation of school-based exposure will be small.

Our study has several limitations. Asthma and rhinoconjunctivitis diagnosis was reported by parents or children themselves even though the questionnaire was intended to be answered by the children’s parents; this could have led to an inaccurate report of the prevalence of the diseases, because most parents or children may not have been aware of the diagnosis. However, the inclusion of asthma medication data should have improved the accuracy of the outcome assessment. Likewise, the symptoms of rhinoconjunctivitis were self-reported. However, both questions were previously validated within the ISAAC study [24].

There were no emission measurements available of the mining related exposures. Unfortunately, we do not have any information on the particle concentration or composition in the mining community. In general, the particles include minerals, chemicals, heavy metals, aluminum, mercury, and arsenic; several of these components are known to be associated with impaired respiratory health [42]. Moreover, it is also known that particulate material may act as allergenic compound and thus might not only contribute to non allergic asthma and rhinitis, but also to allergic and atopic asthma [43]. Additionally, we did not incorporate information on wind direction or other geographical information (i.e. the angle of the mines to each residence) [44]. Since these data were not available in our study region, we used spatial models to estimate risks. However, misclassification of exposure is likely non-differential which may bias our findings likely towards the null. However, this approach can be used for risk estimation in other communities in Latin America with similar mining procedures where measurement data are often not available.

We found wide Bayesian confidence intervals which may be explained by the “flat” likelihood of these kinds of models, and the high standard errors in the estimates that usually result in posterior estimates close to the mean of the prior distribution [38, 45], especially for the α _k . Therefore, we used different types of prior distribution to evaluate this choice of the α parameter (see Section F.2 and Additional file 1: Figures SF.1 to SF.3) as a sensitivity analysis. Our findings of two previous studies [45, 46] that suggested prior distributions for parametric form in the distance function should be carefully chosen.

This study has several strengths. We assessed the models specifications by permitting distance-risk relationships to be nonlinear and free of a pre-specified shape. We compared traditional statistical methods with most recent approaches proposed in the literature as easy to interpret. Semiparametric methods suffer from a number of limitations for incorporating information about a realistic shape of the proximity-risk [26, 35]. Therefore, we combined Bayesian semiparametric models to specify the prior distributions for parametric form, since no or little epidemiological data on the association was available.

We implemented complex models in standard statistical software. These implementations could be extended to consider a model for multiple diseases and multiple sources, avoiding the specification of a single model for each disease, potentially taking into account scenarios in which the outcomes are correlated.

Our findings suggest that proximity to mines using open-pit procedures increases the risk for asthma and rhinoconjunctivitis in children. In communities with high concentrations of environmental pollution from mining facilities, the adverse effect on children’s health highlights the importance of stricter emissions regulating policies. Most important in terms of public health would be emission control and reduction at the emission source. In addition, education campaigns regarding the identification and control of respiratory disease in order to improve the diagnosis and treatment of respiratory diseases in children are suggested. Finally, the proposed threshold distance might be used as an initial indicator to define zones which should be restricted for residential use. Such a policy might implicate the relocation of parts of the population. Adequate prediction models would allow estimations of possible health impacts of such a policy in terms children’s health in the affected areas. Impacts of such a policy should also be examined from different societal perspectives.

BCI, Bayesian confidence interval; CI, confidence interval; GAM, Generalized additive models; GPS, Global positioning system; ISAAC, International study of asthma and allergies in childhood; Rhinoc., Rhinoconjunctivitis