Hydraulic fracturing and hospitalization: a tentative link

Debates about the merits and pitfalls of hydraulic fracturing have only intensified over the past year, with New York banning the natural gas extraction method altogether even as North Carolina lifted its moratorium earlier this year.  Hard evidence about the environmental and health effects of fracking are scarce – for example, scientists still do not understand some of the basics about how well-drilling affects surrounding wildlife – which makes it difficult for lawmakers to bring substantive claims for more regulation (even though anecdotes abound!)

This lack of knowledge is beginning to change, however, and a new article in PLOS One provides the first glimpse into how building hydraulic wells affects hospitalization rates in nearby communities.  The results don’t bode well for those claiming that fracking is not a health risk, although the study relies on correlative associations that will require more data to confirm.

To determine whether a relationship exists between fracking and medical care, the study authors collected geographical data from the Environmental Protection Agency about new shale gas wells drilled in several Pennsylvanian counties from 2007-2011.  For example, the green and red dots in the maps below identify new wells drilled in 2009 and 2011 across counties in northeast Pennsylvania:

Figure courtesy of [1]

Figure courtesy of [1]

As the figures show, a dramatic increase in wells occurred over these years, which can also be seen from this bar graph:

Figure courtesy of [1]

Figure courtesy of [1]

The increase is quite surprising: by 2011, about 20 zip codes have almost 1 well per square kilometer!  This type of change across years provides a robust set of data to look at how this flux of wells affects hospitalization rates.

Using this data, the authors created two main predictors: wells per zip code and well density (wells per square kilometer), calculated for each year.  These variables were then entered into a model that tested their relationship to the prevalence of inpatient visits at each of the same zip codes for a variety of medical conditions: cardiology, neonatology, oncology, dermatology, neurology, etc.  The model also accounted for any other unobserved variables that  correlated with the number of wells drilled or inpatient rates to prevent confounding relationships.

What relationship between well-drilling and health care did the authors find?  Out of 25 different medical conditions examined, inpatient rates related to cardiology, neonatology, neurology, dermatology, endocrinology, oncology, and urology all increased with well density (wells per square kilometer).  Only cardiology inpatient rates were significantly associated with an increased number of wells per zip code.

So what does this mean?  This a a great first look at some much-needed data about how fracking affects health care.  The significant association with cardiology – the most robust finding given its association with both number of wells per zip code and well density – makes sense to me.  Toxic chemicals from fracking sites could likely lead to heart problems, as well as neonatalogical issues that have been previously documented.  Some of the other connections seem more mysterious – I would think that oncological issues would take a long time to develop (although I may be wrong about this), but this study only spans four years.

One of my first reactions to reading this article was that socioeconomic status should be accounted for.  I imagine that many of these drill sites occur in rural regions of Pennsylvania where access to health care may be limited and average income may be lower, so I thought this may be a confounding relationship affecting inpatient rates.  But the authors had an interesting way of indirectly addressing this: one of the counties studied (Wayne County) did not have any wells drilled at all but has a similar demographic to the other counties with greater well densities.  Including this data from Wayne County serves as a type of control  – if a spurious relationship such as SES was related to inpatient rates, then this would have been seen in Wayne County as well, where no wells were drilled.  The significant association between drilled wells and inpatient rates persisted even with the inclusion of Wayne County data, suggesting there is more going on than SES (the fixed effets model they used also helps control for other variables like SES).  Despite these findings, I do believe that SES and other control variables should be included in future models to provide a more holistic picture of all the factors at play.

I would call these findings a tentative link, but an extremely important one!  Articles such as this are crucial to begin to wrap our heads around what long-term effects fracking has on the general population.  The study used very strict statistical tests when testing the relationship between well and inpatient visits, and these findings provide one of the only quantitative looks at the health risks moving beyond scary anecdotes.  This is the type of work that will help push policy forward, and we need more of it to confirm, deny, or understand some more of the complexities about how fracking affects our society.

References

1)

Jemielita, T., Gerton, G., Neidell, M., Chillrud, S., Yan, B., Stute, M., Howarth, M., Saberi, P., Fausti, N., Penning, T., Roy, J., Propert, K., & Panettieri, R. (2015). Unconventional Gas and Oil Drilling Is Associated with Increased Hospital Utilization Rates PLOS ONE, 10 (7) DOI: 10.1371/journal.pone.0131093

Advertisements
This entry was posted in Article Reviews, Energy, Energy and Society and tagged , , , , . Bookmark the permalink.

2 Responses to Hydraulic fracturing and hospitalization: a tentative link

  1. andyextance says:

    To be cautious about the statistics here, as we must be increasingly, how did they determine the link? Was it the classic p=0.05 (ie happens one time in 20 at random)? If so, and they looked at 25 different conditions, wouldn’t they have expected to get one in that range just at random? My stats isn’t strong, but my sceptical instinct is.

    • jptrinastic says:

      I completely agree – this is something I also considered as I read the article. The authors use a Bonferroni correction to account for multiple tests, in which they divide the standard p-value (0.05) by the number of tests done (52 – 25 medical conditions for each outcome, wells per km2 and wells per zip code, and two more tests for overall inpatient rates). This gives a p < 0.00096 significance test. The Bonferroni correction is somewhat adhoc but accepted in the statistical community. with this stricter level of significance, cardiology inpatient rates were still associated with wells per zip code and neurology inpatient rates with well density.

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