Publicly available data reveals association between asthma hospitalizations and unconventional natural gas development in Pennsylvania.

Since the early 2000s, unconventional natural gas development (UNGD) has rapidly grown throughout Pennsylvania. UNGD extracts natural gas using a relatively new method known as hydraulic fracturing (HF). Here we addressed the association of HF with asthma Hospitalization Admission Rates (HAR) using publicly available data. Using public county-level data from the Pennsylvania Department of Health (PA-DOH) and the Pennsylvania Department of Environmental Protection for the years 2001-2014, we constructed regression models to study the previously observed association between asthma exacerbation and HF. After considering multicollinearity, county-level demographics and area-level covariables were included to account for known asthma risk factors. We found a significant positive association between the asthma HAR and annual well density for all the counties in the state (3% increase in HAR attributable to HF, p<0.001). For a sensitivity analysis, we excluded urban counties (urban counties have higher asthma exacerbations) and focused on rural counties for the years 2005-2014 and found a significant association (3.31% increase in HAR attributable to HF in rural counties, p<0.001). An even stronger association was found between asthma hospitalization admission rates (HAR) and PM2.5 levels (7.52% increase in HAR attributable to PM2.5, p<0.001). As expected, asthma HAR was significantly higher in urban compared to rural counties and showed a significant racial disparity. We conclude that publicly available data at the county-level supports an association between an increase in asthma HAR and UNGD in rural counties in Pennsylvania.

Introduction

Petroleum and natural gas production has steadily increased in the United States over the past two decades [1]. As of 2021, the United States is a world leader in the production and export of oil and gas [2]. The United States became a net total energy exporter in 2019, which had not occurred since 1952. The increase in natural gas production played a vital role in the transition of the US from net total energy importer to an exporter. The increase in gas production is attributable to technological advancements called unconventional natural gas development (UNGD). UNGD combines horizontal drilling with a well stimulation technique called hydraulic fracturing (HF), or fracking. UNGD horizontal drilling enables well operators to follow thin rock layers with deposits of natural gas or oil that reside 1 mile or further below the Earth’s surface, which were previously inaccessible using conventional methods and then stimulate production by HF [3]. Since 2008, UNGD has dramatically increased in Pennsylvania because it became economical for producers to extract gas from new regions of the Marcellus Shale, which lies underneath two-thirds of the state (Fig 1). Since the commercial introduction of UNGD in 2004, nearly 12,000 UNGD wells have been drilled in Pennsylvania [4]. Hydraulic fracturing rapidly changed Pennsylvania by bringing heavy industry to sparsely populated rural areas.

Fig 1

Line graphs of UNGD in Pennsylvania over time.

Panel A shows the number of active wells that have been drilled over time as determined through each well’s corresponding spud date (defined as the date well drilling commenced or date the largest pipe casing for the well was installed). Panel B shows the number of active wells that have been drilled each calendar year as determined through each well’s corresponding spud date.

There is broad public concern about the possible adverse health risks to people living near UNGD because, even though a majority of HF wells are in rural areas, more than 10% of Pennsylvanians live within one mile of an active HF well [5]. Prior epidemiology studies found associations between fracking and several adverse medical conditions including heart failure, low birth weight, preterm birth and asthma [6-9].

Asthma is a chronic airway disease that affects over 1 million Pennsylvanians making it a significant public health concern [10]. Severe asthma exacerbation is an appropriate health metric to consider when investigating the community impact of UNGD because it can be triggered by environmental factors, such as poor air quality and stress [11-13]. UNGD can release environmental toxicants that could exacerbate asthma, such as volatile organic compounds (VOCs), nitrogen dioxide (NOx), diesel exhaust, silica dust, and particulate matter [14]. Additionally, unlike other diseases such as cancer, there can be a short latency between exposure to an environmental trigger associated with UNGD and a severe asthma exacerbation [15].

Prior studies on the effects of HF on health relied on Protected Health Information (PHI) from private sources that are not publicly available, such as hospital or insurance records. For example, an asthma study relied on private patient health records that included residential addresses and individualized demographic information to geocode patients [7]. Publicly available data provides the opportunity to investigate the influence of UNGD on asthma exacerbation while circumnavigating concerns regarding individual patient reidentification. In public data bases, severe asthma exacerbations are recorded as hospital admissions and can be quantified as asthma hospital admission rates (asthma HAR).

Here, we investigated the association between UNGD and asthma hospitalization admission rates (HAR), using HAR as a proxy for severe asthma exacerbation. We developed two multiple linear regression models using publicly available data at the county-level for the state of Pennsylvania, while accounting for relevant environmental and socioeconomic covariables. Both models found an association between UNGD and asthma HAR.

Materials and methods

Approach and sourcing of data

Asthma is reported by the Pennsylvania Department of Health publicly as hospitalization admission rates. Asthma HAR are reported by 62 of the 67 counties each year. Other publicly available data for variables of interest were sourced from state and federal government agencies. All data sourced for statistical analysis was organized into a singular spreadsheet for importation and analysis in R studio. We used the years 2005 to 2014 because they cover the periods before and after fracking began in PA. Also, after 2014 changes in the medical coding system made comparisons of hospitalizations difficult with earlier years.

Asthma hospitalization admissions by county

The Pennsylvania Department of Health (PA-DOH) compiles and releases data on various health metrics, including age-adjusted asthma HAR sourced through the Enterprise Data Dissemination Informatics Exchange (EDDIE), an interactive health statistics dissemination web tool. These data were provided by the Division of Health Informatics, Pennsylvania Department of Health. The Department specifically disclaims responsibility for any analyses, interpretations, or conclusions [16].

Unconventional oil & gas wells by county

Data for active UNGD wells in the state were obtained from the publicly available operator well inventory maintained by Pennsylvania Department of Environmental Protection (PA-DEP) [4]. Active UNGD wells that did not have a specified spud date (defined as the date well drilling commenced or date the largest pipe casing for the well was installed) (771 wells) were excluded from analysis as this date is necessary to determine if drilling or construction had commenced for a specific well. Active UNGD wells from the years 2001–2014 (7489 wells) were then sorted by date and matched to their county to determine the number of wells spudded each year for every county. We calculated the well density for active UNGD wells spudded during a calendar year in each county using the area of each county provided by the US Census Bureau through TIGERweb [17].

Geographic data for counties

We used spatial data to generate county-level maps of Pennsylvania for visual analysis. Spatial data for county boundaries were sourced from the Pennsylvania Department of Transportation provided through Pennsylvania Spatial Data Access (PASDA) and joined by county name to visualize relevant data of interest, including asthma hospitalization rates, population density, and fine particulate (PM2.5) pollution [2]. Unconventional wells were mapped as point data utilizing each well’s latitude and longitude and layering over a county-level map of Pennsylvania. All mappings were done in QGIS 3.16.1-Hannover and R version 4.0.3, utilizing the coordinate reference system NAD83 and the ggmap, rgdal, sf, and sp packages [18-21]. The full reproducible code is available in GitHub.

Demographic covariates

Demographic covariables were considered during statistical model building to control for potential confounding effects when examining the relationship between the annual well density of unconventional natural gas wells and asthma hospitalization. All demographic data were publicly sourced at the county-level from the US Census Bureau through the Small Area Health Insurance Estimates (SAHIE), the American Community Survey (ACS) five-year estimates for 2005–2009 and 2010–2014, and the 2010 Census Urban and Rural Classifications. We extracted data on the percentage of people uninsured under the age of 65 from the SAHIE. The ACS provided estimates of county-level median household income, percentage of white population and percentage of people over 25 years of age with a high school diploma or equivalent. The latter two demographic variables were computed for each county in PA using the population estimates provided by each ACS five-year survey. The 2010 Census Urban and Rural Classifications provided the percent of people living in urban areas at the county-level. For our analyses, we also classified PA counties as rural or urban, based on population density being either below or above the population density for the whole state (Fig 2A). The assignment of counties as urban or rural is based on the classification presented in the PA-DOH 2012 Asthma Burden Report, which matches the classifications of The Center for Rural Pennsylvania (CRP) [2, 22]

Fig 2

Comparison of annual average asthma hospitalization admission rates (HAR) in rural and urban counties of Pennsylvania.

Panel A shows a map of Pennsylvania with counties designated as urban or rural as distinguished by the PA-DOH. The orange counties are those designated as urban, while the green represent counties that are rural. Panel B shows a line graph of the average asthma HAR per year from 2001–2014 for counties designated as rural and urban by the PA-DOH. The orange line represents the average asthma HAR per year for urban counties, while the green represents rural counties.

Area-level covariates

Relevant environmental covariables were also considered during statistical model building to control for confounding effects, which were PM2.5 pollution and smoking prevalence at the county-level. We extracted modeled and monitored data for annual average ambient concentration of PM2.5 in μg/m3 from the Centers of Disease Control and Prevention (CDC). The percentage of current smokers was sourced from the PA-DOH through the Behavioral Risk Factor Surveillance System (BRFSS).

Statistical analysis

Data analysis was performed, and graphs created using R version 4.0.3 and the ggplot2 and dplyr packages [23, 24]. Reproducible code is provided in the supplementary materials with additional packages [18–21, 25–27]. We used data for county-level asthma HAR as the response variable for this study after performing a natural log-transformation to help normalize its right-skewed distribution. However, an age-adjusted hospitalization admission rate for a county was not displayed by the PA-DOH if its raw count was below 20, regardless of county population size, so these missing data points were not included for analysis. After computing the annual well density on the county-level, we used this metric as our main explanatory variable for this study.

Multicollinearity between covariables was investigated to minimize redundancy in our final models. With some subjectivity in terms of what is appropriate for the upper limit of correlation between covariables, we choose a conservative threshold of approximately 50% correlation as the threshold for multicollinearity permitted. After performing correlation tests, we determined the following: (a) median household income could not be included in the same model as the percentage of people percentage of people over 25 with a high school diploma or equivalent, or percentage of current smokers, (b) percentage of people uninsured under 65 could not be included in the same model as percentage of people over 25 with a high school diploma or equivalent, and (c) percentage of population in urban areas could not be included in the same model as percentage of white population.

We utilized multiple linear regression models with fixed-year effects to investigate the relationship between the annual density of HF wells and asthma HAR. To meet the objects of our study, we built two separate regression models. The first regression model was built to investigate the relationship between asthma HAR and HF across all the counties, including both urban and rural counties. The second regression model was built to optimize the regression model for rural counties.

When building these models, we considered covariables with a socioeconomic or biological plausibility to exacerbate asthma (specifically, (a) annual average ambient air concentration pf PM2.5, (b) percentage of people uninsured under 65, (c) median household income, percentage of white population, (d) percentage of population over 25 with at least a high school diploma or equivalent, (e) percentage of population that are current smokers, (f) percentage of population living in urban areas) [7, 14, 28]. We utilized a stepwise regression tool that is AIC-based and uses backward elimination for initial investigation of the input variables’ influence on the model. To adequately consider issues of multicollinearity, models were then manually evaluated with consideration for the models’ residual standard error, adjusted R-squared, p-value, and covariables’ significance to select our model with annual well density as the main explanatory variable. Due to the response variable being natural log-transformed, the partial slopes for these regression models were appropriately back-transformed by exponentiating the slope, subtracting one, and multiplying by 100. This enabled the partial slopes to be interpreted as a percent increase in asthma HAR per unit increase for that explanatory variable. 95% confidence intervals were calculated using raw values for the partial slopes and standard error, backtransforming the lower and upper bounds in the same manner.

Results

We first conducted a statewide analysis including all counties in the state of Pennsylvania for an association between UNGD and asthma hospitalization admission rates (HAR). Because the statewide analysis showed strong urban and rural differences in asthma HAR, we additionally carried out a sensitivity analysis including only the rural counties. Both approaches showed an association between asthma and HAR.

Model 1: Multiple linear regression model for all counties

To investigate the relationship between asthma HAR and annual well density, we began by designing a multiple linear regression model for all counties in the state reporting asthma HAR while controlling for fixed-year effects. Our multiple linear regression model resulted in a significant relationship between asthma HAR and annual well density (Table 1). In this model, a 0.01 increase in annual well density is associated with a 3.0% increase in asthma HAR. Median income, percentage of population in urban areas, and PM2.5 were all included as significant covariables. For median income, a $1000 increase on a county-level is associated with a 1.56% decrease in asthma HAR. The percentage of population in urban areas is positively associated with a 1% increase associated with a 0.87% increase in asthma HAR. Lastly, a 1 μg/m3 increase in annual average concentration of PM2.5 pollution is associated with an 2.70% increase in asthma HAR. This regression model has a prediction error rate of 16.07%, an adjusted R-squared value of 0.3211, and significant p-value (<0.05).

Table 1

Results of the multiple linear regression model using the 62 of the 67 counties in PA that report asthma HAR to investigate the relationship between UNGD and asthma hospitalization.

Model 1	Range	Unit Increase	Associated % Change in Asthma HAR	95% Confidence Intervals (as % Change)	p-value (α = 0.05)
Annual well density	0.00–0.43 wells/sq mi	0.01 wells/sq mi	+3.0%	[0.95%, 7.21%]	0.000127
Median Income	$34,018 - $86,093	$1000	-1.46%	[-1.83%, -1.10%]	4.16 x 10−15
% Urban Population	0–100%	1%	+0.87%	[0.71%, 1.03%]	< 2.0 x 10−16
PM 2.5	7.8–21.5 μg/m3	1 μg/m3	+2.70%	[0.25%, 5.23%]	0.028029

Ranges for variables are based on counties included in the model. Since the response variable was natural log-transformed, the associated percent change in asthma HAR for each explanatory variable was computed from backtransforming the partial slopes of the regression model. The 95% confidence intervals were computed using the raw partial slope and standard error, and then backtransforming the lower and upper bounds. A percent change highlighted in green represents an associated percent increase; a percent change highlighted in red represents an associated percent decrease.

Urban v. rural asthma hospitalization admission rates

The risk of asthma hospitalization is well established for urban dwellers, so we compared asthma HAR in rural versus urban counties as a useful control for our study [29-32]. A line graph revealed the urban counties’ average asthma HAR were consistently higher than the average asthma HAR of rural counties in every year from 2001 to 2014 (Fig 2B). Notably, there was more variability in the asthma HAR in urban counties due to inclusion of counties with historically high asthma HAR, such as Philadelphia. Beginning in about 2010, there is a distinct downward trend for average asthma HAR for both rural and urban counties over time. Rather than a two-sample t-test, a Wilcoxon sum rank test was performed due to the small number of data points being compared and that our data otherwise met the test’s assumptions. This test resulted in a significant p-value of 1.31 x 10−4, indicating there is a significant difference between the distribution of average asthma HAR for urban & rural counties between 2001–2014.

Model 2: Multiple linear regression model for rural counties

To refine our analysis to consider only rural areas, we designed a multiple linear regression model to include only data for rural counties of Pennsylvania (Fig 3). After finalizing the model with fixed-year effects, the multiple linear regression analysis showed a significant relationship between asthma HAR and annual well density (Table 2). In this model, a 0.01 increase in annual well density is associated with a 3.31% increase in asthma HAR. Significant covariables in this model include median household income, percentage of white population, and PM2.5 pollution. For median household income, a $1000 increase on a county-level is associated with a 1.90% decrease in asthma HAR. We found that the percentage of white population is also negatively associated, with a 1.0% increase in the percentage of white population associated with a 2.54% decrease in asthma HAR. Lastly, a 1 μg/m3 increase in annual average concentration of PM2.5 pollution is associated with an 7.33% increase in asthma HAR. This regression model has a prediction error rate of 15.76%, an adjusted R-squared value of 0.2899, and significant p-value (<0.05).

Fig 3

Pennsylvania counties included for regression model optimization.

Active UNGD wells spudded through the end of 2014 shown by grey circles. The color designation of counties are as follows: green indicates rural counties that were included, grey indicates counties that lack or do not display asthma HAR for all years of interest, and white indicates urban counties that were excluded.

Table 2

Results of the multiple linear regression model using only rural counties to investigate the relationship between UNGD and asthma hospitalization.

Model 1	Range	Unit Increase	Associated % Change in Asthma HAR	95% Confidence Intervals (as % Change)	p-value (α = 0.05)
Annual well density	0.00–0.43 wells/sq mi	0.01 wells/sq mi	+3.0%	[0.95%, 7.21%]	0.000127
Median Income	$34,018 - $86,093	$1000	-1.46%	[-1.83%, -1.10%]	4.16 x 10−15
% Urban Population	0–100%	1%	+0.87%	[0.71%, 1.03%]	< 2.0 x 10−16
PM 2.5	7.8–21.5 μg/m3	1 μg/m3	+2.70%	[0.25%, 5.23%]	0.028029

Ranges for variables are based on counties included in the model. Since the response variable was natural log-transformed, the associated percent change in asthma HAR for each explanatory variable was computed from through backtransforming the partial slopes of the regression model. The 95% confidence intervals were computed using the raw partial slope and standard error, and then backtransforming the lower and upper bounds. A percent change highlighted in green represents an associated percent increase, while highlight in red represents an associated percent decrease.

Using outdoor PM2.5 as measurement of air quality

Particulate matter is one of the six criteria air pollutants identified by the EPA to harm human health and commonly used as a proxy for overall air quality [33]. In our study, we relied on outdoor PM2.5 as a general county-level metric for air pollution connected to combustion-based power generation, such as coal, and traffic emissions. Similar for asthma HAR, we observed a statewide downward trend for annual average ambient concentrations of PM2.5 as we did for asthma HAR over time in both urban and rural counties (S1 Fig). Although not statistically significant according to a Wilcoxon sum rank test, there was a consistent trend of urban counties having higher PM2.5 pollution on average than rural counties over time. In a simple regression between asthma HAR and PM2.5 pollution, there is a strong positive relationship, which was also observed in the statewide multiple linear regression analyses (S1 Table). Notably, this simple regression model was slightly improved with a 1-year lag on the annual average PM2.5 level (S2 Table).

Discussion

We tested two models in our analysis, Model 1: statewide for all counties. Because of the significant differences in air pollution and higher asthma HAR in urban counties, we developed Model 2: Rural alone, which compared only rural counties. Also, in Pennsylvania a majority of the wells are in rural counties. When we analyzed only rural counties across the state, we still observed a significant association between asthma HAR and annual well density after controlling for several confounding variables [34].

We found a significant association between asthma HAR and unconventional annual well density at the county-level in Pennsylvania when considering all counties, after controlling for several potentially confounding variables. The association was seen when we adjusted for the urban metric of counties and when we examined the rural counties alone. Our models were validated by finding several well-established associations with asthma. (1) We found a negative association between income and hospitalization as previously observed in studies on the influence of socioeconomic status on asthma [35]. (2) There was a strong positive relationship between the percent of individuals in a county living in urban areas and asthma hospitalization. This relationship may be representative of asthma triggers associated with urban living, such as pollution, violence, and housing quality [29-32]. (3) We also observed a positive association between annual PM2.5 concentrations and asthma HAR [36]. Air pollution, specifically PM2.5, is a well-known environmental trigger for asthma [11]. Although excluded from the model reported in this study due to issues of multicollinearity, we observed a strong racial disparity across the state, with the percentage of white population negatively associated with asthma hospitalization rates.

Previous studies have indicated that geographic proximity to UNGD is associated with adverse health outcomes, including increases in asthma [5, 7, 14, 37, 38]. Our general exposure metric of well density has a low spatial granularity compared to methods using precise residential geocoding from patient charts, such as zip codes or home addresses. By using well density, we are forced to assume a uniform well distribution over a county’s land area. Despite this assumption, a clear association can still be observed at the county-level, likely due to the high UNGD and supporting activity during our study period. The populations of the highest UNGD counties, Bradford, Washington, Susquehanna, and Greene are exposed extensively as these counties all had a cumulative well density close to or over 1 unconventional well per square mile by the end of 2014. Other studies found that patients within about a mile were significantly affected [7, 39].

An unexpected finding was a significant reduction in average asthma HAR since 2010. The decrease was seen throughout the state when counties were separated in rural and urban groups. This decrease parallels the downward trend of asthma hospitalization rates throughout the nation for the same period [40]. Several changes in health care management could impact asthma HAR during our study period. The first was the passage of the Affordable Care Act in 2010 (Obamacare), and the subsequent expansion of Medicaid enrollment under this law in Pennsylvania in 2015. These changes provided health insurance to 1.1 million people in the state [41]. The second was the increase in corticosteroid prescriptions [42]. Asthma has a chronic inflammatory component that can be managed by regular medication with inhaled corticosteroids, thus reducing the severe exacerbations that cause hospitalizations. The large number of newly insured patients likely improved the management of their asthma. One consequence of better management might be increased corticosteroid usage.

Hydraulic fracturing could also have contributed to steady declines in asthma HAR as the boom in natural gas extraction led to a reduction of the use of coal for electricity generation from 31% to 11% between 2006 and 2016 [43]. Coal powered electrical generation plants are a significant source of PM2.5, which exacerbates asthma. Since 2011, Pennsylvania retired six coal powered generators in the following counties: Beaver (2), Chester, Snyder, Washington, and Greene. The latter three counties are designated as rural according to our definition based on population density. We note that the drop in asthma HAR was seen statewide and not confined to these counties, and that from the list of counties with retired coal generators, only Washington county is in the highest quartile of HF counties. Thus, while asthmatics might benefit from HF by replacing coal with cleaner natural gas as an energy source, the benefits may be reduced for asthmatics living near HF wells.

We also observed a strong racial disparity in asthma HAR (Table 2). The percentage of white population was negatively associated with asthma hospitalization. Therefore, the percentage of minority populations in rural counties (determined by non-white racial categories reported by the ACS, including Black, Native American, Asian, Pacific Islander, and biracial identities) is positively associated with asthma HAR, a health disparity for minorities that has been observed in past studies [44, 45]. Surprisingly, the disparity was seen even when the analysis was restricted to rural counties alone.

The last significant covariable for this model was annual PM2.5 concentrations, which were positively associated with asthma HAR. For rural counties, PM2.5 was associated with a percent increase in asthma HAR 3 times greater than when considering both urban and rural counties in the other statewide model. Outdoor PM2.5 may be a more influential environmental trigger in rural areas relative to urban areas as rural patterns of activity and occupations may favor spending more time outdoors [46]. Although we did not observe a strong association between UNGD and PM2.5 levels at the county-level in Pennsylvania, we have found that both UNGD and PM2.5 were associated with asthma HAR separately (Table 1, Model 1). Other researchers have associated local PM2.5 pollution with UNGD in Pennsylvania and other states [37, 47, 48]. However, we found that both UNGD and PM2.5 were associated with asthma HAR but not each other indicating that they are each contributing to asthma HAR. We believe that association between the increased number of HF wells being spudded and the increase of asthmatics with severe exacerbations could be due to ineffective set-back distances.

This study illustrates the potential to address associative trends between UNGD and asthma exacerbation from publicly available data. The public data are less granular than data derived from protected health information, but typically have longer longitudinal frames as well as a wider geographical range. For example, this is the first statewide analysis of asthma and UNGD. Exploration of asthma exacerbation in other states with high UNGD, such as Colorado and Texas, using public data may also be informative, particularly if the other states have different asthma metrics publicly available that Pennsylvania lack, such as emergency department visitation.

Limitations of the study

The major limitation of this study is that by solely relying on publicly sourced data, the data are significantly less granular than primary hospital records. The county level asthma HAR permit only imprecise estimates of proximity to wells. Another limitation was inconsistent data availability for asthma HAR and relevant covariables. As of 2021, statewide public health data collection and management can be challenging as there are only six Pennsylvania counties and four municipalities with departments of public health. No publicly available data for asthma HAR was available for Forest, Cameron, Potter, Sullivan, and Fulton counties (S3 Table). Data was also sparse for 10 rural counties between 2005–2014, especially for Juniata, Montour, Snyder, and Union (S3 Table). Despite having asthma HAR dating to 2001, we were unable to consider data prior to 2005 due to a lack of reliable data for the majority of our covariables.

Conclusion

Our study confirms and extends the growing literature associating UNGD with asthma HAR. Our models were robust enough to confirm an expected association of asthma HAR with urban areas, PM2.5, income, and racial disparities (in model 2) using public data only. A major contribution of our work is that we utilized publicly available data without any protected health information. Both publicly available data and protected health information can contribute to our understanding of the health risks of environmental pollutants. Performing analyses using publicly available data enables the results to be shared widely and helps support research conducted on protected health information that is not publicly accessible. Future research investigating asthma exacerbation and UNGD should focus on elucidating casual mechanisms.

Line graphs comparing trends of asthma HAR and average annual PM 2.5 concentrations in rural and urban counties.

Panel A shows a line graph for rural counties with average asthma HAR (per 10,000 cases) on the left axis in purple and average annual PM 2.5 concentrations (μg/m3) on the right axis in gray over time. Panel B shows a line graph for urban counties of average asthma HAR (per 10,000 cases) on the left axis in purple and average annual PM 2.5 concentrations (μg/m3) on the right axis in gray over time.

(PDF)

Click here for additional data file.

Results from a model using rural PA counties: relationship between asthma HAR and average annual PM 2.5.

Ranges for PM 2.5 are based on counties included in the model. Since the response variable was natural log-transformed, the associated percent change in asthma HAR for the explanatory variable, PM 2.5, was computed from through backtransforming the partial slope. A percent change highlighted in green represents an associated percent increase.

(PDF)

Click here for additional data file.

Results from a model using rural PA counties: relationship between asthma HAR and average annual PM 2.5 with a 1-year temporal lag.

Ranges for PM 2.5 are based on counties included in the model. Since the response variable was natural log-transformed, the associated percent change in asthma HAR for the explanatory variable, PM 2.5, was computed from through backtransforming the partial slope. A percent change highlighted in green represents an associated percent increase.

(PDF)

Click here for additional data file.

List of the Pennsylvanian counties included in each model with information on age-adjusted asthma HAR data available and prevalence of HF wells.

Cells highlighted in blue indicate counties that did not have data available for age-adjusted asthma HAR for the years of study. Cells highlighted in yellow indicate counties with incomplete data available for age-adjusted asthma HAR for the years of study. Cells with * indicates a county in the third quartile for HF well count based on the cumulative well count in 2014. Cells with ** indicates a county in the fourth quartile for HF well count based on the cumulative well count in 2014.

(PDF)

Click here for additional data file.

19 Jan 2022

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Kjell Torén, MD, PhD

Academic Editor

PLOS ONE

Journal Requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf  and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Thank you for stating the following in the Funding Section of your manuscript:

“This study was supported in part by the National Institute of Health R25ES021649 and P30ES013508.”

We note that you have provided additional information within the Funding Section that is not currently declared in your Funding Statement. Please note that funding information should not appear in other areas of your manuscript. We will only publish funding information present in the Funding Statement section of the online submission form.

Please remove any funding-related text from the manuscript and let us know how you would like to update your Funding Statement. Currently, your Funding Statement reads as follows:

“This study was supported in part by the National Institute of Health R25ES021649 and P30ES013508 to JF. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.”

Please include your amended statements within your cover letter; we will change the online submission form on your behalf.

3. We note that you have included the phrase “data not shown” in your manuscript. Unfortunately, this does not meet our data sharing requirements. PLOS does not permit references to inaccessible data. We require that authors provide all relevant data within the paper, Supporting Information files, or in an acceptable, public repository. Please add a citation to support this phrase or upload the data that corresponds with these findings to a stable repository (such as Figshare or Dryad) and provide and URLs, DOIs, or accession numbers that may be used to access these data. Or, if the data are not a core part of the research being presented in your study, we ask that you remove the phrase that refers to these data.

4.  Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

5.  We note that Figures 4 and 6  in your submission contain [map/satellite] images which may be copyrighted. All PLOS content is published under the Creative Commons Attribution License (CC BY 4.0), which means that the manuscript, images, and Supporting Information files will be freely available online, and any third party is permitted to access, download, copy, distribute, and use these materials in any way, even commercially, with proper attribution. For these reasons, we cannot publish previously copyrighted maps or satellite images created using proprietary data, such as Google software (Google Maps, Street View, and Earth). For more information, see our copyright guidelines: http://journals.plos.org/plosone/s/licenses-and-copyright.

We require you to either (1) present written permission from the copyright holder to publish these figures specifically under the CC BY 4.0 license, or (2) remove the figures from your submission:

a. You may seek permission from the original copyright holder of Figures 4 and 6 to publish the content specifically under the CC BY 4.0 license.

We recommend that you contact the original copyright holder with the Content Permission Form (http://journals.plos.org/plosone/s/file?id=7c09/content-permission-form.pdf) and the following text:

“I request permission for the open-access journal PLOS ONE to publish XXX under the Creative Commons Attribution License (CCAL) CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). Please be aware that this license allows unrestricted use and distribution, even commercially, by third parties. Please reply and provide explicit written permission to publish XXX under a CC BY license and complete the attached form.”

Please upload the completed Content Permission Form or other proof of granted permissions as an "Other" file with your submission.

In the figure caption of the copyrighted figure, please include the following text: “Reprinted from [ref] under a CC BY license, with permission from [name of publisher], original copyright [original copyright year].”

b. If you are unable to obtain permission from the original copyright holder to publish these figures under the CC BY 4.0 license or if the copyright holder’s requirements are incompatible with the CC BY 4.0 license, please either i) remove the figure or ii) supply a replacement figure that complies with the CC BY 4.0 license. Please check copyright information on all replacement figures and update the figure caption with source information. If applicable, please specify in the figure caption text when a figure is similar but not identical to the original image and is therefore for illustrative purposes only.

The following resources for replacing copyrighted map figures may be helpful:

USGS National Map Viewer (public domain): http://viewer.nationalmap.gov/viewer/

The Gateway to Astronaut Photography of Earth (public domain): http://eol.jsc.nasa.gov/sseop/clickmap/

Maps at the CIA (public domain): https://www.cia.gov/library/publications/the-world-factbook/index.html and https://www.cia.gov/library/publications/cia-maps-publications/index.html

NASA Earth Observatory (public domain): http://earthobservatory.nasa.gov/

Landsat: http://landsat.visibleearth.nasa.gov/

USGS EROS (Earth Resources Observatory and Science (EROS) Center) (public domain): http://eros.usgs.gov/#

Natural Earth (public domain): http://www.naturalearthdata.com/

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: No

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: The article reports on using data routinely collected in the US state of Pennsylvania to investigate the association of hospital admission rates for asthma with unconventional natural gas production. Several changes are need before the manuscript would be ready for publication.

-Introduction

The Introduction is unnecessarily long at approximately 1,000 words. Several paragraphs address the EPA’s failed attempt to impose the “secret science rule” and the advantage of using publicly available data. The text addressing these issues is excessively long, and the same could be accomplished in about half the number of words. Since the EPA never implemented this rule, it is unclear this is a particularly strong motivation, especially since it is likely other studies that rely on personal health information provide more convincing evidence.

Most of the last paragraph of the Introduction (except the final sentence) belongs in the Methods or the Results. This paragraph should clearly state the goals and objectives of the study.

-Methods

Minor comment: Explain the meaning of “spud” when first used in the text, which is in the second paragraph of the Methods.

The model building part of the Methods is unclear. Please clarify which variables were candidates for inclusion in the different full models, whether the models were built using forward addition or backward elimination or some combination of the two, and the criteria for inclusion or exclusion of candidate variables. The criteria for including variables that are potential confounders should not be an alpha of 0.05. Statistical significance alone is insufficient to judge potential confounding because variables with p values greater than 0.05 can still confound the effect of the exposure of interest. Use a higher p value (e.g., 0.10) as the criterion for which variables are included/retained as potential confounders. Also, the point of the models is to estimate the effect of Annual Well Density, so this variable should never be deleted from the model, as the authors did in the 3rd model.

-Results

The first paragraph of the Results belongs in the Methods.

In the second paragraph of the Results, the authors describe results from Model 1 in the following way: “In this model, a 0.01 increase in annual well density correlates to a 3.0% increase in asthma HAR.” The words “correlates to” are not optimal in this setting, partly because the results are not from a test of correlation. I expected to see something like the following, in which results are presented as the association of the health outcome with the exposure of interest: “In this model, a 3.0% increase in asthma HAR was associated with a 0.01 increase in annual well density.” The “correlates to” and similar expressions should be changed throughout the manuscript.

Figure 3 includes results from Model 1 presented in both a table and a graph. In the table, delete “Correlated” from the column heading “Correlated % Change in Asthma HAR.” In this column, put a negative sign before any result that is a percentage decrease to improve clarity of communication. Coloring the table cell is insufficient to indicate the direction of the effect. Follow this column with a new column for 95% confidence intervals for the effect estimates. The table alone communicates all relevant information, so the graph is redundant and should be deleted. These comments apply to Figures 5 and 7 as well. After deleting the graphs, re-label these three figures as tables.

Section on Model 2: The first sentence of this section is an example of using more words than needed. The original sentence is: “To refine our analysis to consider only rural areas, we designed a multiple linear regression model intended to optimize the study design for the publicly available data to include all available data for rural counties of Pennsylvania (Figure 4).” I suggest the following instead: “To refine our analysis to consider only rural areas, we designed a multiple linear regression model to include _only_ data for _the_ rural counties of Pennsylvania (Figure 4).” Sentences like this appear throughout the manuscript and are candidates for similar modification.

-Discussion

The authors did a good job with the Discussion, providing context for the findings and identifying limitations.

I recommend commenting on the potential for using the publicly available data to conduct ongoing surveillance. It appears they could be used to monitor longitudinal trends, which might exceed what most studies based on personal health information can accomplish.

Reviewer #2: GENERAL COMMENTS

This well-written manuscript reports the results of an ecological study of the association of area-density of unconventional oil and gas extraction (hydraulic “fracking”) and asthma hospitalizations in the U.S. state of Pennsylvania. The study methods and data analysis are appropriate if a bit conventional. The results of the study are not novel, but the contribution of the study lies in the confirmatory support this state-wide analysis brings to the fracking-asthma association observed in previous smaller studies. The county-wide frame used by the authors in this ecological analysis is characterized by exposure misclassification due to lack of individual patient residential address, and the lack of individual covariate data may mean that there is residual confounding of the observed association. That said, the exposure misclassification likely biases towards the null and the covariate adjustment is reasonable given the county-level analysis frame.

The authors state that the primary rationale for their study was to use publicly available data so the results could be used by the U.S. EPA for regulatory purposes given the Trump Administration’s efforts to promulgate a “Secret Science” rule. While this is a laudable rationale, the rule was blocked by court action and subsequently rescinded by the Biden EPA in early 2021. To this reviewer, the manuscript could be both shortened and better focused if the text on the Trump EPA “Secret Science” rule was eliminated from the manuscript, including the Abstract.

The authors attempt to replicate the results of a previous Johns Hopkins study using Geisinger Clinic patient data – Model 3. I also suggest dropping this component of the manuscript. It is unnecessary and distracting. The results of the state-wide analysis stand on their own.

SPECIFIC COMMENTS

The lack of page numbers makes the review unnecessarily difficult.

Introduction, 2nd page, 2nd full paragraph The statement “This rule was revoked when a minor procedural flaw in its final implementation was challenged in court” Is both factually incorrect – not a minor procedural flaw (see https://eelp.law.harvard.edu/2021/03/final-secret-science-rule/ -- and out of date. As noted above, I suggest eliminating all mention of this now rescinded rule in the manuscript. It is unnecessary and distracting.

Methods, Unconventional Oil & Gas Wells by County section “Spud” is a technical jargon term that should be explained at first use in the manuscript.

Discussion, 2nd page, 2nd full paragraph “Cortical steroid” should be replaced by “corticosteroid.”

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: Yes: John R. Balmes, MD

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

26 Feb 2022

Response to Reviewers:

Date: Jan 19 2022 10:21AM

To: "Jeffrey M. Field" jfield@mail.med.upenn.edu

From: "PLOS ONE" plosone@plos.org

Subject: PLOS ONE Decision: Revision required [PONE-D-21-35951]

PONE-D-21-35951

Publicly available data reveals association between asthma hospitalizations and unconventional natural gas development in Pennsylvania

PLOS ONE

Dear Dr. Bushong,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

The manuscript has been evaluated by two very experienced referees. They raise a number of significant issues, that must be handled before the manuscript is suitable for decision.

Please, read the comments by the referees and respond to them, point-by-point.

However, I will especially stress two of them.

1. The part dealing with the domestic politics has to be deleted from the manuscript. This is a journal with a global readership.

Author’s Response: We removed discussion of politics.

2. I also agree with one of the reviewers that Model 3 could be deleted.

Author’s Response: We removed Model 3.

Please submit your revised manuscript by Mar 05 2022 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

• A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.

• A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.

• An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Kjell Torén, MD, PhD

Academic Editor

PLOS ONE

Journal Requirements:

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

Author’s Response: Thank you to the academic editor for this clarification. We reformatted the manuscript to fit PLOS ONE’s style requirements. Minor edits and edits for style were not tracked in the “track changes draft” so the reviewers can focus on the responses to their points.

2. Thank you for stating the following in the Funding Section of your manuscript:

“This study was supported in part by the National Institute of Health R25ES021649 and P30ES013508.”

We note that you have provided additional information within the Funding Section that is not currently declared in your Funding Statement. Please note that funding information should not appear in other areas of your manuscript. We will only publish funding information present in the Funding Statement section of the online submission form.

Please remove any funding-related text from the manuscript and let us know how you would like to update your Funding Statement. Currently, your Funding Statement reads as follows:

“This study was supported in part by the National Institute of Health R25ES021649 and P30ES013508 to JF. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.”

Please include your amended statements within your cover letter; we will change the online submission form on your behalf.

Author’s Response: This is the appropriate funding statement, which we also stated in the cover letter for the revisions. “This study was supported in part by the National Institute of Health R25ES021649 and P30ES013508 to JF. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.”

3. We note that you have included the phrase “data not shown” in your manuscript. Unfortunately, this does not meet our data sharing requirements. PLOS does not permit references to inaccessible data. We require that authors provide all relevant data within the paper, Supporting Information files, or in an acceptable, public repository. Please add a citation to support this phrase or upload the data that corresponds with these findings to a stable repository (such as Figshare or Dryad) and provide and URLs, DOIs, or accession numbers that may be used to access these data. Or, if the data are not a core part of the research being presented in your study, we ask that you remove the phrase that refers to these data.

Author’s Response: We included a citation for this statement in the revised manuscript.

4. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

Author’s Response: Thank you to the editor for clarifying this requirement for the Supporting Information. Captions for the Supporting Information files have been added at the end of the manuscript and in-text citations changed to match this formatting.

5. We note that Figures 4 and 6 in your submission contain [map/satellite] images which may be copyrighted. All PLOS content is published under the Creative Commons Attribution License (CC BY 4.0), which means that the manuscript, images, and Supporting Information files will be freely available online, and any third party is permitted to access, download, copy, distribute, and use these materials in any way, even commercially, with proper attribution. For these reasons, we cannot publish previously copyrighted maps or satellite images created using proprietary data, such as Google software (Google Maps, Street View, and Earth). For more information, see our copyright guidelines: http://journals.plos.org/plosone/s/licenses-and-copyright.

We require you to either (1) present written permission from the copyright holder to publish these figures specifically under the CC BY 4.0 license, or (2) remove the figures from your submission:

a. You may seek permission from the original copyright holder of Figures 4 and 6 to publish the content specifically under the CC BY 4.0 license.

We recommend that you contact the original copyright holder with the Content Permission Form (http://journals.plos.org/plosone/s/file?id=7c09/content-permission-form.pdf) and the following text:

“I request permission for the open-access journal PLOS ONE to publish XXX under the Creative Commons Attribution License (CCAL) CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). Please be aware that this license allows unrestricted use and distribution, even commercially, by third parties. Please reply and provide explicit written permission to publish XXX under a CC BY license and complete the attached form.”

Please upload the completed Content Permission Form or other proof of granted permissions as an "Other" file with your submission.

In the figure caption of the copyrighted figure, please include the following text: “Reprinted from [ref] under a CC BY license, with permission from [name of publisher], original copyright [original copyright year].”

b. If you are unable to obtain permission from the original copyright holder to publish these figures under the CC BY 4.0 license or if the copyright holder’s requirements are incompatible with the CC BY 4.0 license, please either i) remove the figure or ii) supply a replacement figure that complies with the CC BY 4.0 license. Please check copyright information on all replacement figures and update the figure caption with source information. If applicable, please specify in the figure caption text when a figure is similar but not identical to the original image and is therefore for illustrative purposes only.

The following resources for replacing copyrighted map figures may be helpful:

USGS National Map Viewer (public domain): http://viewer.nationalmap.gov/viewer/

The Gateway to Astronaut Photography of Earth (public domain): http://eol.jsc.nasa.gov/sseop/clickmap/

Maps at the CIA (public domain): https://www.cia.gov/library/publications/the-world-factbook/index.html and https://www.cia.gov/library/publications/cia-maps-publications/index.html

NASA Earth Observatory (public domain): http://earthobservatory.nasa.gov/

Landsat: http://landsat.visibleearth.nasa.gov/

USGS EROS (Earth Resources Observatory and Science (EROS) Center) (public domain): http://eros.usgs.gov/#

Natural Earth (public domain): http://www.naturalearthdata.com/

Author’s Response: Both figures were created using publicly available data sourced directly from the Pennsylvania Department of Environmental Protection for the well locations and the spatial data for county boundaries from the Pennsylvania Department of Transportation, which was accessed through an open geospatial data portal known as PASDA. We cited these data sources initially but pending acceptance we will upload the primary data and reproducible code to GitHub. The files will be for Figure 4 (now revised in-text as Figure 3). Figure 6 was removed during the process of removing Model 3 from the manuscript.

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: Yes

Reviewer #2: Yes

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: Yes

Reviewer #2: Yes

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: Yes

Reviewer #2: Yes

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: No

Reviewer #2: Yes

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1

Reviewer #1: The article reports on using data routinely collected in the US state of Pennsylvania to investigate the association of hospital admission rates for asthma with unconventional natural gas production. Several changes are need before the manuscript would be ready for publication.

-Introduction

The Introduction is unnecessarily long at approximately 1,000 words. Several paragraphs address the EPA’s failed attempt to impose the “secret science rule” and the advantage of using publicly available data. The text addressing these issues is excessively long, and the same could be accomplished in about half the number of words. Since the EPA never implemented this rule, it is unclear this is a particularly strong motivation, especially since it is likely other studies that rely on personal health information provide more convincing evidence.

Author’s Response: Thank you for this comment. Our motivation to initiate the study was initially sparked by the “secret science rule”, but defer to the judgment of the reviewer and editor that discussion of this rule detracts from the manuscript and the core findings. All mention of the EPA’s rescinded rule has been removed from the introduction and discussion, though we continue to emphasize the public sourcing of the data, without the use of protected health information, as well as discuss the limitations of public datasets.

Most of the last paragraph of the Introduction (except the final sentence) belongs in the Methods or the Results. This paragraph should clearly state the goals and objectives of the study.

Author’s Response: Thank you for this helpful comment. We have moved this portion of the introduction to the beginning of the Results and refocused the last paragraph of the introduction to clearly define the main goal and objective of the study.

-Methods

Minor comment: Explain the meaning of “spud” when first used in the text, which is in the second paragraph of the Methods.

Author’s Response: Thank you for this helpful comment. In response, we added the definition of “spud” according to the Pennsylvania Department of Environmental Protection when the term is first used in the legend of Figure 1 and the first time it was used in the body of the manuscript.

The model building part of the Methods is unclear. Please clarify which variables were candidates for inclusion in the different full models, whether the models were built using forward addition or backward elimination or some combination of the two, and the criteria for inclusion or exclusion of candidate variables. The criteria for including variables that are potential confounders should not be an alpha of 0.05. Statistical significance alone is insufficient to judge potential confounding because variables with p values greater than 0.05 can still confound the effect of the exposure of interest. Use a higher p value (e.g., 0.10) as the criterion for which variables are included/retained as potential confounders. Also, the point of the models is to estimate the effect of Annual Well Density, so this variable should never be deleted from the model, as the authors did in the 3rd model.

Author’s Response: We thank the reviewer for this comment. We have clarified this section in the methods to help clarify our system of selecting variable and assessment of each variables' inclusion in the model along with collinearity assessment to create more parsimonious models. Furthermore, one of the models - was designed to most closely fit the prior published literature (Model 3). Though model 3 is no longer in the manuscript, it is a useful comparator point for variables accepted in the literature, so we included all terms that they included and cite the study as well as two other studies (refs 7, 14 and 28).

-Results

The first paragraph of the Results belongs in the Methods.

Author’s Response: The authors agree this paragraph is more appropriate under the Methods since it explains reasoning and sourcing of publicly available data. It has been moved to be the first paragraph of the methods section.

In the second paragraph of the Results, the authors describe results from Model 1 in the following way: “In this model, a 0.01 increase in annual well density correlates to a 3.0% increase in asthma HAR.” The words “correlates to” are not optimal in this setting, partly because the results are not from a test of correlation. I expected to see something like the following, in which results are presented as the association of the health outcome with the exposure of interest: “In this model, a 3.0% increase in asthma HAR was associated with a 0.01 increase in annual well density.” The “correlates to” and similar expressions should be changed throughout the manuscript.

Author’s Response: Thank you to the reviewer for this important comment. You are correct with the suboptimal usage of “correlated”, rather than the appropriate term “associated”. This has been changed through the body of manuscript and supplementary materials, including the relevant figures and tables.

Figure 3 includes results from Model 1 presented in both a table and a graph. In the table, delete “Correlated” from the column heading “Correlated % Change in Asthma HAR.” In this column, put a negative sign before any result that is a percentage decrease to improve clarity of communication. Coloring the table cell is insufficient to indicate the direction of the effect. Follow this column with a new column for 95% confidence intervals for the effect estimates. The table alone communicates all relevant information, so the graph is redundant and should be deleted. These comments apply to Figures 5 and 7 as well. After deleting the graphs, re-label these three figures as tables.

Author’s Response: Thank you for this comment. We have made the suggested improvements to these tables and agree these changes have improved the presentation of the results. The manuscript has been edited to reflect the re-labeling of these figures as tables. A new column for 95% confidence intervals for the effect estimates was also created, reporting these intervals as percent change. The graphs were also removed as suggested.

Section on Model 2: The first sentence of this section is an example of using more words than needed. The original sentence is: “To refine our analysis to consider only rural areas, we designed a multiple linear regression model intended to optimize the study design for the publicly available data to include all available data for rural counties of Pennsylvania (Figure 4).” I suggest the following instead: “To refine our analysis to consider only rural areas, we designed a multiple linear regression model to include _only_ data for _the_ rural counties of Pennsylvania (Figure 4).” Sentences like this appear throughout the manuscript and are candidates for similar modification.

Author’s Response: Thank you for this helpful comment. We corrected this sentence as suggested and have carefully reviewed the manuscript making appropriate changes to sentences when found. Note that not all of the rephrasing was saved in the track changes draft for clarity of reading.

-Discussion

The authors did a good job with the Discussion, providing context for the findings and identifying limitations.

I recommend commenting on the potential for using the publicly available data to conduct ongoing surveillance. It appears they could be used to monitor longitudinal trends, which might exceed what most studies based on personal health information can accomplish.

Author’s Response: Thank you to the reviewer for this comment. We agree explicitly commenting on this benefit of publicly available data improves the discussion. Therefore, we incorporated a statement regarding the potential for ongoing surveillance in the last paragraph of the Discussion.

Reviewer #2

This well-written manuscript reports the results of an ecological study of the association of area-density of unconventional oil and gas extraction (hydraulic “fracking”) and asthma hospitalizations in the U.S. state of Pennsylvania. The study methods and data analysis are appropriate if a bit conventional. The results of the study are not novel, but the contribution of the study lies in the confirmatory support this state-wide analysis brings to the fracking-asthma association observed in previous smaller studies. The county-wide frame used by the authors in this ecological analysis is characterized by exposure misclassification due to lack of individual patient residential address, and the lack of individual covariate data may mean that there is residual confounding of the observed association. That said, the exposure misclassification likely biases towards the null and the covariate adjustment is reasonable given the county-level analysis frame.

The authors state that the primary rationale for their study was to use publicly available data so the results could be used by the U.S. EPA for regulatory purposes given the Trump Administration’s efforts to promulgate a “Secret Science” rule. While this is a laudable rationale, the rule was blocked by court action and subsequently rescinded by the Biden EPA in early 2021. To this reviewer, the manuscript could be both shortened and better focused if the text on the Trump EPA “Secret Science” rule was eliminated from the manuscript, including the Abstract.

Author’s Response: We are encouraged by the reviewer’s positive comments. We defer to the reviewer’s comments that focus on the “Secret Science” rule distracts from the focus of the manuscript as mentioned in response to the first reviewer and have eliminated it from the manuscript.

The authors attempt to replicate the results of a previous Johns Hopkins study using Geisinger Clinic patient data – Model 3. I also suggest dropping this component of the manuscript. It is unnecessary and distracting. The results of the state-wide analysis stand on their own.

Author’s Response: Thank you for this straightforward, helpful comment. The authors agreed the manuscript would benefit from streamlining and have removed model 3. The remaining portion of the manuscript has been edited to reflect this change.

SPECIFIC COMMENTS

The lack of page numbers makes the review unnecessarily difficult.

Introduction, 2nd page, 2nd full paragraph The statement “This rule was revoked when a minor procedural flaw in its final implementation was challenged in court” Is both factually incorrect – not a minor procedural flaw (see https://eelp.law.harvard.edu/2021/03/final-secret-science-rule/ -- and out of date. As noted above, I suggest eliminating all mention of this now rescinded rule in the manuscript. It is unnecessary and distracting.

Author’s Response: We added page numbers. As discussed above, we eliminated all mention of the rescinded rule from the manuscript.

Methods, Unconventional Oil & Gas Wells by County section “Spud” is a technical jargon term that should be explained at first use in the manuscript.

Author’s Response: Thank you to the reviewer for this comment. “Spud” is now defined when first used in the legend of “Figure 1” and in the body of the manuscript to clarify the meaning of this technical term.

Discussion, 2nd page, 2nd full paragraph “Cortical steroid” should be replaced by “corticosteroid.”

Author’s Response: Thank you for this comment. “Cortical steroid” has been replaced by “corticosteroid” throughout the manuscript.

Submitted filename: response to reviewers.docx

Click here for additional data file.

3 Mar 2022

Publicly available data reveals association between asthma hospitalizations and unconventional natural gas development in Pennsylvania

PONE-D-21-35951R1

Dear Dr. Field,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Kjell Torén, MD, PhD

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

The manuscript has been considerably improved, and I consider that it is ready for acceptance, i.a. the manuscript is accepted.

Reviewers' comments:

22 Mar 2022

PONE-D-21-35951R1

Publicly available data reveals association between asthma hospitalizations and unconventional natural gas development in Pennsylvania

Dear Dr. Field:

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now with our production department.

If your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information please contact onepress@plos.org.

If we can help with anything else, please email us at plosone@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Kjell Torén

Academic Editor

PLOS ONE