Cancer Mortality and West Virginia Coal Miners

Anesetti-Rothermel, A et al. (2010) used geographic information systems to  test the hypotheses that age-adjusted county cancer mortality rates were positively associated with distance-weighted population exposure to coal extraction and processing activities, and distance-weighted exposure measures are more strongly correlated to cancer mortality than exposure based on tons of coal mined in the county. Their results supported their hypothesis in that they found a relationship between population proximity to coal mine features and cancers. All three global spatial autocorrelation tests ran resulted in positive spatial autocorrelation. Other variables, like the tonnage exposure measure, race/ethnicity percentages, education, poverty, primary health care access, etc. contributed significant additional variance and for total cancer and three cancer subgroups, the exposure measure was correlated to higher mortality after controlling for smoking rates. The previous exposure measure, based on tonnage, was not related as strongly to cancer mortality.Rosemary Kulp
Anesetti-Rothermel, A., Fedorko, E. Hendryx, M. 2010 A geographical information system-based analysis of cancer mortality and population exposure to coal mining activities in West Virginia, United States of America. Geospatial Health 4(2), 2010, pp. 243-256

Approximately 50% of West Virginia’s population resides within just 11counties. West Virginia’s counties have an average population density of 94.9 persons, and a median of 51.1 persons, per square land mile. Persons who live in coal mining counties of Appalachia, like the counties of West Virginia have elevated all-cause cancer as well as lung cancer mortality, compared to non-mining counties or the nation, even after controlling for socio-economic, health services and behavioral variables.
The higher cancer mortality in the region has been attributed to behavioral risks such as smoking, poor socio-economic conditions and problematic access to medical care.  This experiment conducted an exploratory spatial data analysis to determine if there was a spatial relationship between the existing data. In this study a distance-based index describing, per county, the proximity of that county’s population to coal mine features was developed and then used in a regression analysis.
The experiment was conducted with the expectation that  distance-weighted population exposure is more highly correlated to cancer mortality rates than the previous measure of tons of county-level coal mining. If cancer mortality is not related to exposure to mining activity, and is only a reflection of socio-economic status or behavior, there would be no improvement in the capacity of the exposure measure to account for mortality rates.
Cancer mortality rates were taken from the Centers for Disease Control and Prevention. The rates were age-adjusted using the 2000 US standard population, and were found for West Virginia counties as the rate per 100,000 person-years for 1979-2004. The person-year approach allowed for aggregation across years to estimate cancer mortality in rural, less populated counties that typify most coal mining locations.
The time period, represented by covariates, was sometimes based on the 2000 Census, and sometimes on more recent estimates when available. These covariates include average poverty rate for 2000-2002, high school and college education rates in 2000, supply of primary care physicians per 1,000 population in 2001, and smoking rate in 2003. Geographic data on activities of the coal mining industry included mining permit boundaries for mining sites, the point locations of surface slurry impoundment dams, the point locations of permitted underground injection sites, and the point locations of coal processing facilities. Most of the data compilation and manipulation was performed using ArcView GIS software,versions 9.2 and 9.3
Within each of the study area’s census block groups the mean distance was calculated in km from the nearest mine boundary, the impoundment dam, the injection site, and the preparation plant. The inverse distance for each mine infrastructure type for each block group was then calculated. The mean inverse distances were multiplied by the population of the block group. This resulted in a value per block group/infrastructure type where closer distances and bigger populations have larger values, and farther distances and smaller populations have smaller values.
The tonnage measured, rather than the distance weighted measure, was used for the spatial analysis because tonnage measures from border counties outside West Virginia could be included. However, the mapping of mining activities was available onlywithin the state.
Anesetti-Rothermel, A et al. (2010) identified two sources of spatial information for coal processing facilities the US environmental Protection Agency and the the West Virginia Department of Environmental Protection (WVDEP) data sets. With the data a basic photo alignment of the point was relocated to a more accurate location. Of 76 entities, 46 were realigned.
The last coal mining data set used in the study was slurry injection sites. These are areas where waste water from mining, drilling or processing has been injected into underground areas for the purpose of storage. Injection sites are permitted by the EPA as National Pollution Discharge Elimination System (NPDES) points. The content of the slurry is monitored at the discharge points as part of the NPDES regulatory process. Mining operations often involve the capture of used water in artificial impoundments, held in place by earthen dams, for the purpose of removing contaminants and non-combustibles. Acid mine drainage (AMD) is also held in surface impoundments. These coal impoundment dams are regulated by WVDEP. Further information involving potential health effects of coal slurry and toxic water will be addressed later in the chapter.
Global measure of spatial autocorrelation was used to measure the level and direction (positive or negative) of association for the entire sample used. The resulting test is similar to a correlation coefficient as it varies between -1.0 and +1.0. This test of global spatial autocorrelation was computed using GeoDa 0.9.5-i software. The variables of interest, county-level age-adjusted combined cancer mortality rates (1979-2004) and tonnage of coal production per county (1986-2005), were joined to a georeferenced spatial county layer file of West Virginia and its neighboring counties from surrounding states.
All three global spatial autocorrelation tests yielded a value of P <0.001, showing that the data was not spatially random. The correlations between cancer mortality and the total distance-weighted exposure were higher than the corresponding correlations between cancer mortality and the tonnage exposure measure for all cancer sites. Respiratory cancer was found to be correlated to the distance-weighted measure but not to the tonnage measured.
The superior performance of the distance-weighted exposure measure is consistent with the possibility of environmental contamination from the mining industry as a causal factor in the etiology of cancer for populations residing in West Virginia. The strong association between respiratory cancer and mining boundaries, controlling for smoking, may reflect air quality problems around the mines, especially at mountaintops and other surface mining operations.
This experiment was done under the assumption that current mining reflects past mining and based off of  the long history of coal mining in the region but there is the possibility of error if there has been significant change in coal mining technology in the area. Another interesting fact is that the population in West Virginia decreased from 1.94 million to 1.79 million before becoming stable from 1990 to 2000 (1.79 to 1.81 million). The population loss to emigration affected coal-mining counties significantly more than non-mining counties: between 1980 and 1990, the average coal mining county lost 5,233 people to migration compared to a loss of 1,175 people for non-mining counties This could make observed mining effects more conservative than they are, because people  exposed in mining areas would develop cancer later in another area later on.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s