International Journal of Environmental Health Research, 2015 Vol. 25, No. 3, 265–276, http://dx.doi.org/10.1080/09603123.2014.938027

An examination of the effects of mountaintop removal coal mining on respiratory symptoms and COPD using propensity scores Michael Hendryxa* and Juhua Luob a Applied Health Science, Indiana University, Bloomington, IN, USA; bEpidemiology & Biostatistics, Indiana University, Bloomington, IN, USA

(Received 13 December 2013; final version received 12 May 2014) Previous research on public health consequences of mountaintop removal (MTR) coal mining has been limited by the observational nature of the data. The current study used propensity scores, a method designed to overcome this limitation, to draw more confident causal inferences about mining effects on respiratory health using non-experimental data. These data come from a health survey of 682 adults residing in two rural areas of Virginia, USA characterized by the presence or absence of MTR mining. Persons with a history of occupational exposure as coal miners were excluded. Nine covariates including age, sex, current and former smoking, overweight, obesity, high school education, college education, and exposure to coal as a home-heating source were selected to estimate propensity scores. Propensity scores were tested for balance and then used as weights to create quasi-experimental exposed and unexposed groups. Results indicated that persons in the mountaintop mining group had significantly (p < 0.0001) elevated prevalence of respiratory symptoms and chronic obstructive pulmonary disease. The results suggest that impaired respiratory health results from exposure to MTR environments and not from other risks. Keywords: coal mining; Appalachia; COPD; respiratory health; propensity score

Introduction Mountaintop removal (MTR) coal mining is a form of large-scale surface coal mining practiced in the steep mountain terrain of Central Appalachia (i.e. areas of Kentucky, Tennessee, Virginia, and West Virginia.) The practice involves clear-cutting forests and using explosives and heavy machinery to remove up to hundreds of feet of rock and soil above and between coal seams. The resulting spoil, or overburden, is dumped into adjacent valleys, permanently burying headwater streams. Using diesel equipment, the coal is extracted and transported to local processing facilities and then shipped out by truck or train. MTR mining impacts an area 12 million acres in size, roughly equivalent to the combined areas of Vermont and New Hampshire (EPA 2010). Environmental studies indicate that MTR mining creates local air and water pollution (Palmer et al. 2010; Knuckles et al. 2013). Air sampling indicates elevated levels of ultrafine particulate matter in mining residential communities compared to control communities (Kurth et al. 2014); exposure to ultrafines increases risk for respiratory disease (Oberdorster 2001; Sioutas et al. 2005). Particulate matter collected from residential sites in mining *Corresponding author. Email: [email protected] © 2014 Taylor & Francis

266

M. Hendryx and J. Luo

communities consists of elevated levels of silica, aluminum, and other inorganic lithogenic constituents, and organic compounds from coal and diesel combustion (Knuckles et al. 2013). There is strong evidence that silica and coal dust lead to various forms of lung disease including chronic obstructive pulmonary disease (COPD) and pneumoconiosis in coal miners (Beeckman et al. 2001; Cohen et al. 2008; Cullinan 2012). Thus, there is reason to believe that MTR coal mining operations impair air quality through blasting and excavation, and through the transport and processing of coal in and around local mining communities. Impaired air quality, in turn, may reasonably be expected to adversely impact the respiratory health of local residents. Previous studies have found that persons who live in areas of Central Appalachia where MTR coal mining is practiced experience elevated risks for a set of poor health outcomes. These outcomes include higher total mortality rates (Hendryx 2011) as well as higher mortality for lung cancer (Hendryx et al. 2008), cardiovascular disease (Esch & Hendryx 2011), and kidney disease (Hendryx 2009). Elevated morbidity rates have been found for cancer (Christian et al. 2011), self-reported health (Zullig & Hendryx 2011), and depression (Hendryx & Innes-Wimsatt 2013). A survey of Kentucky residents reported higher levels of respiratory symptoms and elevated risk of COPD (Hendryx 2013). A common concern of these studies is that they are based on non-experimental observational data and that residents in mining communities and non-mining communities may not be comparable in term of lifestyle and socioeconomic factors, despite controlling for these confounds using traditional linear or logistic regression techniques. Propensity scoring offers a rigorous approach to improve the capacity to make causal inference in the absence of experimental data (Rosenbaum & Rubin 1983; D’Agostino 1998). It is a statistical technique that can create quasi-experimental groups that are in effect equivalent on measured confounds. With a few exceptions (Siddiqui et al. 2008; Juhn et al. 2010; Boutwell et al. 2011), it appears that propensity scoring has rarely been used to investigate possible effects of environmental exposures in observational studies. Furthermore, propensity scoring has not been previously applied to the study of possible impacts of MTR mining on public health outcomes. In this paper, we follow the procedures recommended by Lanza et al. (2013) to conduct a propensity score analysis to determine whether or not there is evidence that MTR mining contributes to poor respiratory health among community residents. The research question driving this study is, “What is the effect of living in a MTR environment vs. a non-mining environment on occurrence of respiratory symptoms and risk of COPD?” Methods Design Data for this study were collected from a cross-sectional face-to-face survey conducted in Spring 2013, in rural communities in western Virginia. This part of Virginia contains areas heavily impacted by MTR mining operations as well as areas where mining does not take place due to the absence of economically minable coal. The mining communities were drawn from Lee and Wise counties in Virginia, and the non-mining communities from Smyth County. Participants in the survey were drawn from communities located three miles or less from MTR mining sites or from non-mining control communities that were at least 20 miles from mountaintop mining activity. The areas that have been most heavily mined are west and north of the mining communities, which are even

International Journal of Environmental Health Research

267

farther, often more than 40 miles, from the non-mining communities. A map of the survey areas is shown in Figure 1. Previous community surveys have been conducted and results reported for West Virginia (Hendryx et al. 2012) and Kentucky (Hendryx 2013); the current paper offers the first presentation of results from a more recent survey in Virginia, includes additional covariates not previously available, and is the first study on this topic to employ propensity scoring. Sampling Maps of the study area that contained all roads and structures (households and other buildings) were examined in consultation with local residents to plan the sampling strategy. Households are generally clustered in unincorporated villages or hamlets in hollows, which are narrow valleys containing rivers or streams. Over the course of the sampling period, every hollow was canvassed at least once in an effort to reach every household. The sampling strategy was not otherwise stratified or selected. Business establishments and institutional settings such as nursing homes were not included. Contact attempts took

Figure 1.

(Color online) Approximate location of mining and non-mining survey areas.

268

M. Hendryx and J. Luo

place primarily during daylight hours over Monday–Friday of the sampling weeks; however, surveying also occurred in the early evening hours and on Saturdays. Interviewers received training led by persons with previous experience conducting similar surveys and by local residents. Training consisted of background information on local culture, maintaining personal safety, conflict avoidance, and survey and data recording procedures including practicing mock surveys. The survey took approximately 15–20 min on average per participant to complete. Eligible participants were self-identified household residents at least 18 years old who spoke English and gave verbal assent to participate; to preserve participant anonymity, written consent was not required. Variables The survey included measures on self-rated health and disease conditions as primary outcome measures of interest, and demographic and lifestyle questions for use as possible covariates. The current study used measures of self-reported COPD and current respiratory symptoms as dependent measures. The value for the current respiratory symptoms variable was determined by asking participants to report whether they experienced each of five symptoms: shortness of breath, persistent cough, wheezing, coughing up mucus or phlegm on a daily basis, and daily activity limitations due to breathing problems. The measure of symptoms was thus a count that varied from zero to five. Self-reported COPD was measured dichotomously based on a positive response to the question, “Have you ever been told by a doctor, nurse, or other health professional that you had emphysema, chronic bronchitis, or COPD?” Most covariate measures were collected using items from the CDC’s Behavioral Risk Factor Surveillance System (BRFSS) (CDC 2012). Potential covariates included participant age, sex, and race/ethnicity (coded into non-Hispanic White, African-American, Hispanic/Latino, or Other.) Race/ethnicity was simplified into White or non-White for analysis because of the small numbers of minority persons in the sample. Marital status was coded as married or not married. Years of education completed was scored on a sixpoint scale ranging from never attended school, grades 1 through 8, grades 9 through 11, grade 12 or GED, 1–3 years of college, to 4 or more years of college; for analysis, persons were categorized into three dichotomous measures including less than high school, high school graduate, or 1 or more years of college. Less than high school was used as the referent in regression models. Height and weight were recorded and converted to body mass index (BMI); persons were classified as obese (BMI ≥ 30), overweight (25–25.9), or normal (< 25). Normal weight was used as the referent. Smoking was categorized into three dichotomous measures using BRFSS items including current smoker, former smoker, or lifetime never smoker (the latter used as the referent). Current use of smokeless tobacco was measured using a BRFSS item. Participants were asked how they heated their homes, and we recorded up to three responses including natural gas, electricity, oil, propane, wood, coal, or other; the variable used for analysis categorized persons who used coal as a sole or supplementary heat source vs. those who did not. Persons who reported current or past occupational experience working as a coal miner were excluded from the analysis. Analysis The propensity score analysis was conducted following the four-step process described by Lanza et al. (2013). The four steps are:

International Journal of Environmental Health Research    

269

Estimating propensity scores. Using propensity scores to adjust for confounding. Assessing balance. Estimating the propensity-score adjusted effect of MTR residence compared to non-mining residence on respiratory health.

Estimating propensity scores Propensity scores were obtained by logistic regression analysis of the exposure (i.e. MTR mining) on the potential confounders. Potential confounders were selected based on two criteria, one empirical and one theoretical, as recommended by Lanza et al. (2013). The empirical method involved the calculation of standardized mean differences in covariate values between the exposed (e) and unexposed (u) groups divided by the standard deviation (SD) of the exposed group: Xe  Xu =SDe A standardized mean difference > 0.20 was used as the criterion to include the covariate in the calculation of propensity scores, as per the recommendation of Lanza et al. (2013). In addition to the empirical criterion, covariates were also included in calculation of the propensity scores if there was a strong theoretical rationale for doing so, regardless of the standardized mean difference. As shown in the Results section, covariates selected for the propensity score calculation on theoretical grounds included age, obesity, and current smoking. These three were selected because of evidence from other research that they are related to respiratory health in expected ways: respiratory health declines with advancing age (Raherison & Girodet 2009), obesity (Salome et al. 1985), and smoking (Raherison & Girodet 2009). Once the set of covariates was selected, they were used as independent variables in a logistic regression model where the dependent variable was residence in a MTR mining community or a non-mining community. The propensity score for each individual was the predicted probability of being in the MTR group. The propensity scores were then tested for overlap between the two groups: the score distributions should overlap, indicating that individuals in the exposed group overlap individuals in the unexposed group on the potential confounders. Using propensity scores to adjust for confounding The second step of the analysis was to adjust the exposure variable for confounding. In this paper, we used inverse probability of treatment weighting (Hirano & Imbens 2001). More specifically, we used the Average Treatment Effect (ATE) weight. Persons in the exposed group received a weight of 1=p: Persons in the unexposed group received a weight of 1=ð1  pÞ where π = the propensity score. In effect, this approach mimics what would be expected if individuals had been randomly assigned to mountaintop mining or non-mining conditions and were equivalent on the confounds.

270

M. Hendryx and J. Luo

Assessing balance Third, balance on the potential confounds was assessed to determine whether differences between exposed and unexposed groups remained on the confounders after the data were adjusted using propensity scores. Balance was tested by conducting the same standardized mean difference analysis as was undertaken in the first step; in this case, standardized mean differences on all potential confounds between the exposed and unexposed groups should be < 0.20. Estimating the propensity-adjusted effect Fourth, once balance was demonstrated, the effect of exposure was assessed by using the ATE values as weights in models where respiratory health variables were predicted from the exposure variable (i.e. MTR). We tested two respiratory health-dependent measures, including the count of respiratory symptoms and self-reported COPD diagnosis. For this analysis, we used SAS software version 9.4 Proc Genmod, specifying a log link function and a Poisson distribution; resulting coefficients were exponentiated to find prevalence ratios (PRs). Results The total sample included 821 persons. After excluding 139 persons who reported occupational exposure as coal miners (mostly men in the mining communities), 682 cases remained for analysis, including 401 from mining communities and 281 from nonmining communities. Table 1 presents a descriptive summary of study variables, plus the results of the first standardized mean difference test prior to implementation of propensity scores. Table 1 shows that 27 % of the sample in the mountaintop mining area reported a COPD diagnosis, compared to 19 % in the non-mining area. Persons in the mining group on average reported experiencing 1.03 current respiratory symptoms, compared to an average of 0.79 current symptoms in the non-mining group. All five of the individual respiratory symptoms were reported at greater frequency in the mining vs. the non-mining group (results not in Table). However, based on chisquare tests, the differences in individual respiratory symptoms were most apparent for persistent cough (21 % vs. 14 %, p < 0.02) and wheezing (27 % vs. 20 %, p < 0.03). Mucus or phlegm discharges were reported by 14 % and 9 % (marginally significant at p < 0.09). The other two symptoms were not significantly different between groups: shortness of breath was reported by 30 % vs. 29 % (p < 0.66), and activity limitations were reported by 10 and 7 % (p < 0.17). Based on the standardized mean differences shown in the last column of the table, the mining and non-mining groups were comparable on most covariates. Four potential confounders were selected from Table 1 for use in calculation of propensity scores based on standardized mean differences > 0.20: high school education, female sex, former smoking, and use of coal as a home-heating source (in almost all cases coal was used as a supplemental, and not a sole, heating source). In addition, we included three potential confounders in the propensity score calculation based on theoretical grounds even though the standardized mean differences on these variables were less than 0.20: age, obesity, and current smoking. To keep the full range of categorical measures for the subsequent regression models, when we retained high school education we also retained college education, and when we retained obesity we also retained overweight.

International Journal of Environmental Health Research

271

Table 1. Descriptive statistics for variables in propensity score analysis for MTR and non-mining groups. Columns labeled “Mean or %” show either the mean value or the % value depending on the variable. Mountaintop removal Variable Dependent measures Mean respiratory symptoms % COPD Potential confounds Mean age % Non-white % Female % College education % High school education % Married % Overweight % Obese % Use of smokeless tobacco % Current smoker % Former smoker % Use of coal for home heating

Non-mining

SD

N (Total N = 281)

Mean or %

SD

Effect size Std. mean diff.

1.03

1.39

278

0.79

1.28

0.33

394

0.266

0.443

278

0.187

0.391

0.29

397 401 397 397 397

51.8 0.045 0.710 0.395 0.322

18.7 0.207 0.454 0.490 0.468

280 281 278 280 280

55.0 0.042 0.619 0.332 0.432

18.8 0.203 0.487 0.472 0.496

0.17 0.01 0.20 0.13 0.24

401 384 384 397

0.531 0.292 0.422 0.078

0.500 0.455 0.495 0.269

281 273 273 280

0.477 0.286 0.374 0.061

0.500 0.453 0.485 0.239

0.11 0.01 0.10 0.06

394 394 401

0.378 0.192 0.092

0.486 0.394 0.290

279 279 281

0.323 0.289 0.004

0.468 0.454 0.060

0.11 0.25 0.30

N (Total N = 401)

Mean or %

397

The only potential confounds that were not retained were race, marital status, and smokeless tobacco use. The nine selected potential confounds were used to find the propensity scores. The results are shown in Figure 2. The distributions of the predicted probabilities in the exposed and unexposed groups largely overlap. We, therefore, proceeded with the next step in the analysis and used the propensity scores to weight each observation as described under Methods. Table 2 presents the results of the propensity score-weighted means, percentages, and standardized mean differences. Compared to original data, all of the potential confounds had comparable values between groups and reduced standardized mean differences after propensity score weighting, and all standardized mean differences were < 0.20. Results of the final propensity score analysis are summarized in Figure 3. We observed significant effects (p < 0.0001) of MTR coal mining on both dependent measures of respiratory health. The PRs indicate that MTR mining increased risk of COPD by 63 %, and risk of each increment in respiratory symptoms by 44 %. Discussion Persons in this study who resided in communities near MTR coal mining experienced significantly greater respiratory disease symptoms and increased prevalence of selfreported COPD. These differences were observed after conducting a propensity score analysis that created exposed and unexposed groups that were equivalent on measured

272

Figure 2.

M. Hendryx and J. Luo

Boxplots for logit propensity: MTR mining and non-mining groups.

Table 2. Balance results: group means or percentages, standard deviations (SD), and standardized mean differences after weighting by propensity scores. Mountaintop removal Covariate Mean age % Female % High school education % College education % Overweight % Obese % Current smoker % Former smoker % Use of coal for home heating

Non-mining

Mean or %

SD

Mean or %

SD

Standardized mean difference after propensity score weighting

53.1 0.670 0.366

25.2 0.616 0.631

53.2 0.639 0.349

28.7 0.746 0.741

0.002 0.05 0.03

0.364 0.291 0.395 0.355 0.229 0.058

0.630 0.595 0.640 0.626 0.550 0.307

0.407 0.325 0.379 0.372 0.222 0.056

0.763 0.728 0.754 0.751 0.646 0.357

0.07 0.06 0.03 0.03 0.01 0.01

confounds including age, smoking, use of coal for home heating, and others. Persons who were present or former coal miners were not included in the analysis, so the findings are not due to differences in occupational exposure. The results of this and other studies on cardiovascular, respiratory, and cancer outcomes in Appalachian mining communities (Hendryx et al. 2008; Hendryx 2009; Christian et al. 2011; Esch & Hendryx 2011; Hendryx 2013), coupled with environmental assessments that document water and air pollution from this form of mining (Pond et al. 2008; MSHA 2010; Palmer et al. 2010; Orem et al. 2012) strongly suggest that MTR coal mining in Appalachia is harmful to population health. Findings from the current study using propensity scoring serves as a validation of the previous observational studies documenting health disparities in mountaintop mining populations. These findings may be combined with additional studies conducted in Europe, Australia, and Asia that also find poor public health outcomes such as respiratory illness or cancer from exposure to surface coal mining activity (Temple & Sykes 1992; Yapici et al. 2006; Onder & Yigit 2009;

International Journal of Environmental Health Research

273

Figure 3. Propensity score PRs and 95 % confidence intervals (CI): effect of exposure to MTR mining on respiratory health. * Respiratory symptoms: PR = 1.44, 95 % CI = 1.28 − 1.63, p < 0.0001. **COPD: PR = 1.63, 95 % CI = 1.28 − 2.09, p < 0.0001.

Higginbotham et al. 2010; Fernández-Navarro et al. 2012; Morrice & Colagiuri 2013). The precautionary principle in environmental science calls for efforts to reduce exposure risk even when exact causes or mechanisms of disease are unclear (Kriebel et al. 2001). Efforts to reduce risk in the present case range from establishing and enforcing air quality standards for communities proximate to coal extraction, processing, and transportation activities, to abolishing MTR mining altogether. Calls for the latter have been made in other scientific forums (Palmer et al. 2010). Mechanisms by which surface coal mining impair respiratory health for community residents require additional study. However, it is known that surface mining operations generate dust in respirable ranges in both fine and ultrafine scale (Onder & Yigit 2009; Higginbotham et al. 2010; Kurth et al. 2014). Dust from surface mining operations contains both overburden material such as crystalline silica, and organics and other chemicals from the coal itself, and from chemicals used in blasting, extraction, transportation, and processing (Ghose & Majee 2007; MSHA 2010; Knuckles et al. 2013). Coal dust and silica have been linked to increased risk of COPD and other lung diseases in other research (Cohen et al. 2008; Cullinan 2012; Graber et al. 2014). Inhalation of silica and other chemicals, especially in the ultrafine range, likely contribute to observed respiratory health disparities. The study is limited by a number of considerations. First, propensity scoring corrects for measured confounds but does not adjust for possible bias from unmeasured confounds. In terms of respiratory health as a consequence of exposure to mining environments, available covariates such as age, smoking, and occupational exposure would appear to be the major confounds of interest (Raherison & Girodet 2009), and they were captured in this study. Second, the study design is cross-sectional and was not able to represent the temporal relationships between exposure and subsequent development of respiratory symptoms or disease. However, residency status in these areas is quite stable (the average age of participants was 54.5 and the mean time they reported living in their

274

M. Hendryx and J. Luo

current community was 39.2 years, including on average 39.8 years in the mining area and 38.4 years in the non-mining area.) Furthermore, it seems unlikely that people in poor respiratory health will be more likely to move into mining communities. Third, exposure was operationally defined as living in a MTR mining community; data on person-level dose from exposures to air pollution resulting from MTR mining activity are currently unavailable. Fourth, most sampling occurred during traditional workday hours such that the samples available may not fully represent working populations and may over-represent older or non-working persons, although this limitation will affect both mining and non-mining groups. Finally, self-report respiratory symptoms and COPD may result in misclassification for both measurements relative to confirmed medical diagnosis; however, there is no evidence to believe that misclassification is differential with respect to exposure. Propensity scoring offers certain advantages relative to traditional covariance adjustment. First, if the variance among covariates for the two groups is unequal, covariance adjustment can either increase bias or over-adjust for bias (Rosenbaum & Rubin 1983). Second, propensity scoring allows one to test whether there is balance, and to proceed with analysis only when this test is satisfied; in traditional covariate analysis, exposed and unexposed groups may be compared despite being highly dissimilar. A third advantage of a propensity score used as a weight is that it preserves degrees of freedom relative to using multiple covariates. This study provides the first application of propensity scoring to research on the public health effects of living in areas where large-scale surface coal mining operations occur. It is also one of few studies to use this technique to address possible environmental exposures of any type. The study of environmental exposures must necessarily contend with observational data where persons cannot be randomized to exposure conditions, and where exposed and unexposed persons often vary on other disease-risk variables. The use of propensity scoring offers a means to address this problem by providing a rigorous test of possible causal relationships between exposure and disease in non-experimental designs. In conclusion, this study confirms previous findings (Hendryx et al. 2008; Hendryx 2009; Hendryx 2013) of respiratory health disparities among persons living in communities where mountaintop mining is practiced, using a technique designed to draw stronger causal inferences from non-experimental data. The evidence across studies clearly indicates that people living in areas where MTR is practiced experience health problems that cannot be explained by traditional risks such as age, smoking, obesity, occupational exposure, or socioeconomic status. In addition, other research shows that air and water quality in these areas is impaired by mining in ways consistent with observed health disparities (McAuley & Kozar 2006; Palmer et al. 2010; Orem et al. 2012; Knuckles et al. 2013; Kurth et al. 2014). Based on the combined environmental (Pond et al. 2008; Palmer et al. 2010; Lindberg et al. 2011; Kurth et al. 2014) and public health evidence, and consistent with previous evidence-based recommendations (Palmer et al. 2010), MTR mining should be discontinued. References Beeckman LA, Wang ML, Petsonk EL, Wagner GR. 2001. Rapid declines in FEV1 and subsequent respiratory symptoms, illnesses, and mortality in coal miners in the United States. Am J Respir Crit Care Med. 163:633–639.

International Journal of Environmental Health Research

275

Boutwell BB, Beaver KM, Gibson CL, Ward JT. 2011. Prenatal exposure to cigarette smoke and childhood externalizing behavioral problems: a propensity score matching approach. Int J Environ Health Res. 21:248–259. CDC. 2012. Behavioral risk factor surveillance system. National Center for Chronic Disease Prevention and Health Promotion, Centers for Disease Control; [cited 2012 June 29]. Available from: http://www.cdc.gov/brfss/index.htm. Christian WJ, Huang B, Rinehart J, Hopenhayn C. 2011. Exploring geographic variation in lung cancer incidence in Kentucky using a spatial scan statistic: elevated risk in the Appalachian coal-mining region. Pub Health Rep. 126:789–796. Cohen RA, Patel A, Green FH. 2008. Lung disease caused by exposure to coal mine and silica dust. Semin Respir Crit Care Med. 29:651–661. Cullinan P. 2012. Occupation and chronic obstructive pulmonary disease (COPD). Brit Med Bull. 104:143–161. D’Agostino RB. 1998. Propensity score methods for bias reduction in the comparison of a treatment to a non-randomized control group. Stat Med. 17:2265–2281. EPA. 2010. EPA issues comprehensive guidance to protect Appalachian communities from harmful environmental impacts of mountaintop mining; [cited 2010 Sep 26]. Available from: http:// yosemite.epa.gov/opa/admpress.nsf/e77fdd4f5afd88a3852576b3005a604f/4145c96189a1723985 2576f8005867bd!OpenDocument. Esch L, Hendryx M. 2011. Chronic cardiovascular disease mortality in mountaintop mining areas of central Appalachian states. J Rural Health. 27:350–357. Fernández-Navarro P, García-Pérez J, Ramis R, Boldo E, López-Abente G. 2012. Proximity to mining industry and cancer mortality. Sci Total Environ. 435–436:66–73. Ghose MK, Majee SR. 2007. Characteristics of hazardous airborne dust around an Indian surface coal mining area. Environ Monit Assess. 130:17–25. Graber JM, Stayner LT, Cohen RA, Conroy LM, Attfield MD. 2014. Respiratory disease mortality among US coal miners: results after 37 years of follow-up. Occup Environ Med. 71:30–39. Hendryx M. 2009. Mortality from heart, respiratory, and kidney disease in coal mining areas of Appalachia. Int Arch Occup Environ Health. 82:243–249. Hendryx M. 2011. Poverty and mortality disparities in central Appalachia: mountaintop mining and environmental justice. J Health Dispar Res Pract. 4:44–53. Hendryx M. 2013. Personal and family health in rural areas of Kentucky with and without mountaintop coal mining. J Rural Health. 29:S79–S88. Hendryx M, Innes-Wimsatt KA. 2013. Increased risk of depression for people living in coal mining areas of central Appalachia. Ecopsychology. 5:179–187. Hendryx M, O’Donnell K, Horn K. 2008. Lung cancer mortality is elevated in coal mining areas of Appalachia. Lung Cancer. 62:1–7. Hendryx M, Wolfe L, Luo J, Webb B. 2012. Self-reported cancer rates in two rural areas of West Virginia with and without mountaintop coal mining. J Com Health. 37:320–327. Higginbotham N, Freeman S, Connor L, Albrecht G. 2010. Environmental injustice and air pollution in coal affected communities, Hunter Valley, Australia. Health Place. 16:259–266. Hirano K, Imbens GW. 2001. Estimation of causal effects using propensity score weighting: an application to data on right heart catheterization. Health Serv Outcomes Res Methodol. 2:259–278. Juhn YJ, Qin R, Urm S, Katusic S, Vargas-Chanes D. 2010. The influence of neighborhood environment on the incidence of childhood asthma: a propensity score approach. J Allergy Clin Immunol. 125:838–843. Knuckles T, Stapleton P, Minarchick V, Esch L, McCawley M, Hendryx M, Nurkiewicz T. 2013. Air pollution particulate matter collected from an Appalachian mountaintop mining site induces microvascular dysfunction. Microcirculation. 20:158–169. Kriebel D, Tickner J, Epstein P, Lemons J, Levins R, Loechler EL, Quinn M, Rudel R, Schettler T, Stoto M. 2001. The precautionary principle in environmental science. Environ Health Perspect. 109:871–876. Kurth LM, McCawley MA, Hendryx M, Lusk S. 2014. Atmospheric particulate matter size distribution and concentration in West Virginia coal mining and non-mining areas. J Expo Sci Environ Epidemiol. 24:405–411. Lanza ST, Moore JE, Butera NM. 2013. Drawing causal inferences using propensity scores: a practical guide for community psychologists. Am J Com Psychol. 52:380–392.

276

M. Hendryx and J. Luo

Lindberg TT, Bernhardt ES, Bier R, Helton AM, Merola RB, Vengosh A, Di Giulio RT. 2011. Cumulative impacts of mountaintop mining on an Appalachian watershed. Proc Nat Acad Sci USA. 108:20929–20934. McAuley SD, Kozar MD. 2006. Ground-water quality in unmined areas and near reclaimed surface coal mines in the northern and central Appalachian coal regions, Pennsylvania and West Virginia. Scientific Investigations Report 2006-5059. Reston (VA): US Department of the Interior US Geological Survey. Morrice E, Colagiuri R. 2013. Coal mining, social injustice and health: a universal conflict of power and priorities. Health Place. 19:74–79. MSHA. 2010. Effects of blasting on air quality. Arlington (VA): Mining Safety and Health Administration, United States Department of Labor. Oberdorster G. 2001. Pulmonary effects of inhaled ultrafine particles. Int Arch Occup Environ Health. 74:1–8. Onder M, Yigit E. 2009. Assessment of respirable dust exposures in an opencast coal mine. Environ Monit Assess. 152:393–401. Orem W, Tatu C, Crosby L, Varonka MS, Bates A, Engle M, Geboy NJ, Hendryx M. 2012. Water chemistry in areas with surface mining of coal. Paper presented at: The Geological Society of America Annual Meeting and Exposition; 2012 Nov; Charlotte, NC. Palmer MA, Bernhardt ES, Schlesinger WH, Eshleman KN, Foufoula-Georgiou E, Hendryx M, Lemly AD, Likens GE, Loucks OL, Power ME, et al. 2010. Mountaintop mining consequences. Science. 327:148–149. Pond GJ, Passmore ME, Borsuk FA, Reynolds L, Rose CJ. 2008. Downstream effects of mountaintop coal mining: comparing biological conditions using family- and genus-level macroinvertebrate bioassessment tools. J N Am Benthol Soc. 27:717–737. Raherison C, Girodet PO. 2009. Epidemiology of COPD. Eur Respir Rev. 18:213–221. Rosenbaum PR, Rubin DB. 1983. The central role of the propensity score in observational studies for causal effects. Biometrika. 70:41–55. Salome CM, King GG, Berend N. 1985. Physiology of obesity and effects on lung function. J Appl Phys. 108:206–211. Siddiqui AR, Gold EB, Yang X, Lee K, Brown KH, Bhutta ZA. 2008. Prenatal exposure to wood fuel smoke and low birth weight. Environ Health Perspect. 116:543–549. Sioutas C, Delfino RJ, Singh M. 2005. Exposure assessment for atmospheric ultrafine particles (UFPs) and implications in epidemiological research. Environ Health Perspect. 113:947–955. Temple JMF, Sykes AM. 1992. Asthma and open cast mining. BMJ. 305:396–397. Yapici G, Can G, Kiziler AR, Aydemir B, Timur IH, Kaypmaz A. 2006. Lead and cadmium exposure in children living around a coal-mining area in Yatagan, Turkey. Toxicol Ind Health. 22:357–362. Zullig K, Hendryx M. 2011. Health-related quality of life among central Appalachian residents in mountaintop mining counties. Am J Pub Health. 101:848–853.

Copyright of International Journal of Environmental Health Research is the property of Taylor & Francis Ltd and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.

An examination of the effects of mountaintop removal coal mining on respiratory symptoms and COPD using propensity scores.

Previous research on public health consequences of mountaintop removal (MTR) coal mining has been limited by the observational nature of the data. The...
356KB Sizes 0 Downloads 3 Views