Generating Data Visualizations of Longitudinal Cohort Ambient Air Pollution Exposure: Report Back Intervention Development in Participatory Action Research

Abstract

A civic engagement and data science design was used to develop a report-back intervention to address stakeholder concerns related to air emissions surrounding a coke oven factory near Buffalo, NY, United States. This factory had historically emitted high levels of benzene pollution and ceased operation in October 2018 due to violations of the United States Clean Air Act and United States Resource Conservation and Recovery Act. Using publicly-available air pollution and weather data, descriptive time series and windrose data visualizations were developed using open-source software as part of a 2-page report-back brief. Data from two air toxics monitoring sites in this direction suggest that industrial sources were likely the major contributor to the benzene in the air at these locations prior to May 2018, after which traffic emissions became the likely major contributor. Windrose visualizations demonstrated that the wind typically blew towards the northeast, which was qualitatively consistent with locations of stakeholder concerns. With the factory closed, collective efforts subsequently shifted to address traffic emission air pollution sources, factory site clean-up, ground and water pollution mitigation. Because this intervention utilized open-source software and publicly-available data, it can serve as a blueprint for future data-driven nursing interventions and community-led environmental justice efforts.

Background

Report-back¹ and teach-back² of objective research or laboratory results are essential nursing interventions to support self-management at the individual, family, organizational, community, and population levels.^3,4 The report back process is defined as the active and intentional sharing of research findings with the community and individuals participating in research.¹ In the teach-back process, the participant becomes the expert and actively explains the information back to the researcher in a manner meaningful to participant actions and behaviors.² Report-back and teach-back are effective components of the health education role in nursing practice. As we enter the advanced digital age, where health-related personal and environmental data can be quantified in real time, it is essential that nurse-led and designed report-back and teach-back interventions are responsive to those who may have low literacy, who may experience disparities, or whose limited technology access and use place them at risk in the growing digital divide.^5,6 The report-back of environmental exposure findings to community stakeholders is a crucial step towards the meaningful interpretation of findings and the determination of appropriate next steps, including proposed health behavior change and policy implications.^1,7 While there has been a historical disciplinary divide in clinical science and environmental exposure intervention, environmental hazards that are harmful to health are an emerging priority topic for nursing intervention development.^4,8,9 The active sharing of results in report-back and teach-back requires iterative communication between the research team and community stakeholders to provide context (local and cultural) and clarification of concepts to account for varying levels of science literacy.^3,10 Here, we relay our process of report-back intervention development that was used in subsequent informal teach-back participatory action research. We provide code snippets for nurse informatics experts to replicate our data visualization process using open access, freely downloadable software as shown in supplemental content 2 as programming code.

The 2021 National Academies of Sciences, Engineering, and Medicine report entitled “The Future of Nursing 2020–2030: Charting a Path to Achieve Health Equity” emphasizes that nursing and nursing education needs to be substantially strengthened to identify and intervene on complex environmental factors that affect human health.¹¹ Outdoor air quality assessments and regulatory standard setting are intricately linked with nursing informatics. Data derived from the clinical record and health care encounter have been vital to scientific evaluation of the health effects of outdoor air pollutants. For example, the United States Environmental Protection Agency (EPA) conducts an extensive Integrated Science Assessment (ISA) approximately every 5 years for each criteria air pollutant to inform the regulatory standards for levels of outdoor air pollutants.¹² These comprehensive and systematic reviews of the literature are used to develop potential biological pathways for health effects following pollutant exposure. At the population level, emergency department visits and hospitalizations are a significant epidemiology outcome measures associated with increases in outdoor air pollutants, strengthening the multi-level evidence about the health-harming effects of these exposures.¹³ Nurse informaticists and nurse content experts have unique and in-depth understanding of the data storage, aggregation, retrieval, meaning, validity, plausible alternate explanations, and unmeasured confounding related to these variables and associations derived from the clinical encounter information.

Members of our research team had completed a previous civic engagement project to address community inquiries regarding county-level changes in criteria air pollutants and asthma emergency visits coinciding with changes in the polluting activities at the criminally-convicted factory.¹⁴ While we found seasonal variations and associations, strongest in the month of June, we concluded that influential seasonal triggers such as infections or allergies likely masked the associations of air pollutants and asthma emergency department visits. In collaboration with a graphic designer, an infographic was developed and disseminated as a report-back intervention.¹⁴ In this present project, researchers were asked to generate a report-back of benzene levels using existing, publicly available data for local community stakeholders, in order to better understand the data’s relevance to differences between benzene exposure from a coke oven factory site and traffic emission exposures. Since 2008, the New York State Department of Environmental Conservation and Department of Health had provided routine updates to the community on air pollution levels, but had not scheduled a community meeting in over a year when our research team’s engagement was requested.¹⁵

Benzene is a volatile organic compound (VOC) and a common outdoor air pollutant in many urban areas.^16,17 A known carcinogen, the primary route of benzene exposure is inhalation.¹⁸ Benzene has also been associated with a number of other human health outcomes, including respiratory irritation, hematological disease (aplastic anemia or leukemia), immunological alterations to both humoral and cellular immunity, and neurological effects (drowsiness, dizziness, and delirium).¹⁹ High levels of acute exposure (20,000 ppm over 5–10 min) can be fatal. Health disparities and environmental justice issues have been identified in community benzene exposures, as elevated pollution levels have been associated with low income regions.²⁰ Benzene is often co-emitted with toluene, another VOC, and most outdoor benzene and toluene pollution has been associated with traffic-related emissions from vehicle exhaust.^21–24 However, there is growing health concerns and evidence linked to both industrial activity and oil or natural gas production on these emissions.^25–27 Tan and colleagues²⁸ observed a background benzene-to-toluene ratio of approximately 0.67 in an urban environment. However, in the vicinity of two coke oven factories, this ratio was increased by roughly an order of magnitude (~10), indicating a relative increase in benzene in the proximity to these factory point sources.²⁷ Outdoor benzene is a dominant source of subsequent indoor benzene levels, particularly for non-smoking households.²⁹ Due to a short half-life and rapid metabolism, biomonitoring for ambient, long-term benzene exposure rarely produces clinically meaningful or actionable results.³⁰

The local site for this present study was the community next to a coke oven factory. The coke oven factory was found guilty on criminal charges of violating the United States Clean Air Act and Resource Conservation and Recovery Act. The company was placed on criminal probation. Due to ongoing emissions, the factory was found guilty of violating probationary terms of the Clean Air Act conviction on September 17, 2018 and ceased operation on October 14, 2018. The aim of this study was to create descriptive data visualizations as a part of report-back intervention using publicly-available air pollution data in response to community and multi-sectoral stakeholder concerns regarding factory emissions from a specific point source. This report clarifies the methods and programming code used to generate the data visualization and report-back intervention for other nurse informaticists to tailor and replicate.

Methods

This study utilized a civic engagement and data science design. Following community-based participatory research principles, all elements of the study were co-designed with community members and civic leaders. The stakeholders were comprised of interested community members or leaders attending a grassroots, non-profit environmental organization meeting; local elected officials in the three surrounding municipalities; and Department of Environmental Conservation experts.

Data sources

This data science study was designed to produce data visualizations from longitudinal, observational pollutant data that is freely available through the Environmental Protection Agency Air Quality System. Wind speed and direction data was downloaded from the National Climatic Data Center (NCDC), National Centers for Environmental Information (NOAA).

Ethical Considerations

All data analyzed was obtained from public sources and did not meet the definition of human subjects research (45 CFR 46).

Variables

Benzene and toluene data were collected every sixth day at two ambient air monitoring stations using standard Environmental Protection Agency Air Method for Toxic Organics (TO) from January 1, 2018 to August 12, 2018. The rationale for this time period was the public reporting of a waste heat tunnel collapse during or before spring of 2018, which was associated with substantial changes observed in the visual opacity of smokestack emissions. At the time of the initial community meeting on October 8, 2018, data were only available and certified through the quality assurance processes through the first and second quarters of 2018. The two ambient air monitoring stations are located at latitude, longitude of 42.98844,-78.91859 and 42.99813,-78.89926. Benzene and toluene are collected as part of the Environmental Protection Agency’s Integrated Urban Air Toxics regulatory program since 1999, implemented to reduce hazardous air pollutants in densely populated areas. The gaseous pollutants are collected in a 6 Liter pressurized Summa (passivated stainless steel) canister for 24 hours and analyzed using gas chromatography-mass spectrometry by Environmental Protection Agency standard operating procedure.

Analysis

The data processing was initially coded in STATA, Version 14.2 (College Station, Texas) using ambient monitor data from the New York State Department of Environmental Conservation, which was provided to researchers by email from the government agency. For future reproducibility, the research team began developing the programming code to automate downloading data from the application program interface (API) at the publicly available Environmental Protection Agency’s (EPA) Air Quality System (AQS) Data Mart. Here, we provide the process in Python programming code snippets for other teams to freely replicate the data visualizations as Python is open source and freely downloadable.

The pollutant data was analyzed by descriptive statistics and descriptive visualizations, placing the results into time series line graphs and overlays. The benzene to toluene ratio was calculated and depicted in a time-series as a standard indicator of local traffic versus point-source factory emissions. Benchmark and thresholds were added for meaningful health effect risk levels.

Report-back

The visualizations were contextualized into a 2-page report back hand-out. This written report-back used a simple template of who? what? where? why? and when? the data, along with what else? Drafts of the handout were circulated to all identified stakeholders for feedback and coordination. After the first report-back draft, the units for the benzene to toluene ratios were unchanged, but the units for the benzene data from Environmental Protection Agency were converted from parts per billion of carbon (ppbC) to parts per billion volume (ppbV) using the number of carbon atoms in benzene (#C = 6) in order to align with units of cancer risk benchmarks (supplemental Figure 1). A windrose plot was generated to visualize wind direction and speed during the study period.

Informal Teach-back

Teach-back opportunities were informally offered to stakeholders. For this collaboration, we opted not to formalize teach-back as the method can been perceived as condescending, judgmental, or reinforcing power differentials in past research.²

Results

A total of 38 pollutant and 224 daily weather measurements were obtained. The average benzene level at the monitor closest to the factory was 0.4 parts per billion volume (ppbV). The average benzene levels for the distal neighborhood monitor was 0.16 parts per billion volume (ppbV). The resulting report-back with the benzene-toluene ratios and time series overlay with results from the two sites are depicted in the report-back hand out (Figures 2 & 3). Benzene levels from each monitor site were plotted with common lifetime cancer risk benchmarks (Supplemental Figure 1). The report back example is in supplemental content 1.

Figure 2.

Benzene to Toluene Ratio at the GIBLVD Monitor

Figure Note: GIBLVD is an ambient air pollution monitor at latitude, longitude of 42.98844,-78.91859

Figure 3.

Benzene to Toluene Ration at the Brookside Terrace Monitor

Figure Note: Brookside Terrace is an ambient air pollution monitor at latitude, longitude of 42.99813,-78.89926

The windrose plot (Figure 4) is a circular shape that provides the direction that the wind is coming from with the length of each spoke showing the percentage of time the wind blew from that direction. For example, wind blew from west-southwest 26.3% of the time. The different colors on the spokes represent the different wind speeds and proportion of days this wind speed occurred. Using the same west-southwest spoke, the light blue represents 13% which means that 13% of the time west-southwest winds were at speeds between 7.8–12.3 mph.

Figure 4.

Windrose Plot

No formal data was collected on teach-back. Informally, elected officials and a grassroots community organizer requested individual appointments with the researcher to discuss the information, clarify, and briefly practice potentially confusing portions of their own presentation of the information to their constituents and members.

Discussion

Environmental exposures are crucial determinants of health and developing nursing interventions to prevent, mitigate, and treat the health impacts of environmental exposures are an emerging priority for the nursing discipline.^4,9 Here, we developed a written report-back intervention. Report-back of exposure research findings must be ethical and effective,³¹ considering potential legal ramifications.³² The general principles necessary to guide best-practice report-back include considerations for privacy, culture, health literacy and the use of standard templates to facilitate the process.^7,10,31,33 The report-back process is more effective and satisfactory to community participants when it is tailored to the learning styles and cultural needs of individuals and communities. In pursuit of an effective report-back process, the integration of narrative text, data visualizations and community discussions can enhance report-back by accounting for different learning styles and community culture. In addition, research team members must cultivate the ability to convey basic environmental health concepts to answer patient or community questions regarding environmental exposure and risk reduction. Efforts to tailor report-back to each unique community have the potential to make report-back as efficacious as possible. The research community has begun to develop best practices for report-back including guidelines for how to report personal exposure results to participants using tailored materials, evaluation guides, and interview questions.³¹ Further research is needed regarding best practices for report back by format (written, verbal, virtual) and level (community, organization, family, individual). Representing the discipline that constitutes the largest proportion of the health care workforce, nurses are in an ideal position to develop, refine, and disseminate report-back community and clinical interventions. Nurse informaticists and data scientists provide crucial expertise for developing and tailoring accessible, meaningful and actionable data visualizations as part of these report back interventions.³⁴ Given the recommendations in the 2021 National Academies of Sciences, Engineering, and Medicine report entitled “The Future of Nursing 2020–2030: Charting a Path to Achieve Health Equity,” we anticipate nurses, nurse informaticists, and nurse scientists will engage in an increasingly prevalent role in preventing and mitigating environmental justice community environmental exposures to improve health.

In this study, we created descriptive data visualizations and a sample written report back intervention that was tailored to respond to grassroots community member concerns about exposure to ambient pollution from a factory criminally convicted of violating the Clean Air Act. While this nursing work was notably part of cross-sector coalition building, multidisciplinary and community engaged, nursing expertise presented a trusted catalyst among stakeholders to direct the relevance and meaningful reach of an environmental health project.³⁵ Similar narratives about the often overlooked, but essential role of nurses in environmental health and environmental justice include Charlotte Brody and Liz Boornizian. Nurse Brody, committed to healthy lifestyle habits, became an evidence-based advocate after her own comprehensive testing of the body burden of chemicals.³⁶ Oncology nurse Boornazian shared her astute observations and experiences about a potential childhood cancer cluster with a contact in the Environmental Protection Agency, subsequently elevating the credibility of community reports and concerns in Toms River, New Jersey.³⁷

Here, the descriptive statistic visualizations and 2-page report back form provided sufficient detail to inform participants about the likely pollution source (industrial vs. traffic) over time and common wind direction and speed. The findings were coherent with informal triangulation in our community participatory dialogue experiences. Namely, engaged members had discussed neighborhoods (mostly northeast, east-northeast, and west-southwest of the factory) and timing when industrial odors were most unpleasant and noxious (during calm winds and weather inversions in the middle of the night to early morning) while those to the northwest reported few to no problems with crowdsourced or anecdotal perceptions of industrial soot or odors. This report contributes uniquely to the literature by providing nurses and other health experts with example code snippets and processes written in open-source software (Python) to generate descriptive data visualizations and tailored report-back intervention using publicly available pollution and weather data to address community data-to-action needs. While our intervention included important informal elements of teach-back techniques, we did not formalize this teach-back process to maintain the value of equal participatory co-development and avoid perceptions of power differentials, judgment, or condescension that can be inherent to the method.² Our process may have implications for future initial outdoor community exposure assessments, report-back, and teach-back in communities near substantial wildland fire, factory, agricultural, or traffic pollution.³⁸ New open access software packages have been developed since our project was completed that may aid informaticists in advancing this work, including the RAQSAPI R stats package designed to import air quality data from the USA EPA’s Air Quality System datamart at https://cran.r-project.org/web/packages/RAQSAPI/index.html.

Our initial results corroborated previous evidence on benzene and toluene ratios near coke oven emission sources.²⁷ From January to May of 2018, our results became more closely aligned with the general mean benzene and toluene ratios from a previous study across 28 US cities of 1 to 3 (± 1.5).³⁹ Calculating benzene and toluene ratios is a crude and preliminary method for ascertaining potential pollution sources.²⁷ There are numerous, sophisticated and advanced source attribution techniques that are available in the scientific community.^40,41 However, these computational methods may require resources that are not rapidly available to most environmental justice communities. Further, the communication of advanced source apportionment methods and results may fail to match the numeracy and literacy needs of community members and elected officials in order to meaningfully inform their personal and collected decisions.

While the descriptive analyses we conducted may be commonly conducted and reported by local, state, or federal environmental agency officials, many communities and grassroots organizations may still experience a need for more frequent data-to-action translation instruments, or for report-back information more tailored to health concerns.¹⁵ As one of the most commonly encountered health experts in underserved communities, nurses are in a crucial position to fill this gap. Here, we tailored commonly reported information about benzene and toluene separately into a ratio that was meaningful for ongoing aggregate community action. Nurse informaticists can replicate our data visualization program using Python to develop data visualizations and report-back for their own communities where publicly available pollution and weather data are available.

The New York State Department of Environmental Conservation had been closely engaged with the community about the polluting factory site for over a decade. At the time the present work was requested by community stakeholders of our team, it had been slightly over one year since the agency had provided publicly available and updated reports. However, agency officials were rapidly responsive to phone and email requests and provided the initial dataset for our informatics program development.

Because we intended to run the program as an automated process, our preliminary results to key stakeholders were developed on the pollution units of ppbC, but using data from the federal Environmental Protection Agency’s Air Quality System listed in ppbV. Thus, the initial line charts appeared much higher relative to the programmed benchmarks than anticipated. The differences in the units for each pollutant was not easily determined by comparing the two websites and publicly available data information files at the time we developed our program, nor was there a sound rationale for the difference in the units used between the two agencies. This was an important lesson learned to screen for, identify, and correct errors and inconsistencies. This discrepancy in units is an important weakness we identified in the public usability and utility of the available data. We will use this example in the future to inform policymakers, agency scientists, government agency administrators about the need for usability and utility improvements.

Additional nursing interventions relevant to increased ambient air pollutants include community environmental management,⁴² environmental management for worker safety, patient education,⁴³ health policy monitoring,^44,45 risk identification,^46,47 and public health nursing surveillance.⁴⁸ Examples of community health management nursing interventions may include biomarker or physiologic screening for health risks from environmental exposures; participating in multidisciplinary initiatives, programs, or coalitions to reduce environmental health hazards; promote health protective government policy; community health education; and care coordination for at-risk groups and communities. Nurse-led community partnerships focused on environmental health have implemented programs to reduce home environmental toxin exposures (such as pesticide, chemical, and mold exposures), environmentally friendly transportation initiatives, and healthy food access.³⁵

Engaged stakeholders from the grassroots meetings in this project had expressed a great deal of interest in a subsequent health study to ascertain short and long-term health effects from the air pollution at the site, which had now decreased substantially.¹⁵ Our data visualizations and study revealed achieving adequate statistical power is highly unlikely to study excess cancer risks associated with ambient and community, non-occupational, benzene exposures. Noting the limitations of biomonitoring with excess cancer risk of 10 in a million, decreasing to 1 in a million in the more residential areas, the research team did not recommend a single site research study for the most common health outcomes from past ambient benzene exposure that did not consider indoor-outdoor measurements and modeling (Supplemental Figure 1). Rather, participants were encouraged to partner with national studies, including the National Institutes of Health All of Us⁴⁹ cohort for potential retrospective case-control grouping and comparisons with members of other communities experiencing high levels of community pollution at coke oven factories or petrochemical emissions sites.^17,50,51 Retrospective asthma development and exacerbation, acute cardiorespiratory effects, and pregnancy outcomes tied to the period of high pollution exposures require further attention.^51,52 Additional research is warranted on health outcomes linked to retrospective air pollution mixture exposures, indoor penetration of outdoor mixtures at the sites, total VOC exposures, and industrial vs. neighborhood ozone formation from VOC and nitrogen oxide emission sources.^51,53

Stakeholders received additional and timely information that the contribution of factory source benzene to overall emissions in the region were likely decreasing.^14,15 These efforts coincided with informing additional grassroots advocacy with the goal of decreasing idling vehicle exhaust emissions by removing a traffic toll booths in exchange for cashless tolling systems. This project also coincided with successful advocacy to secure additional air sampling resources from a Department of Environmental Conservation Community Air Screen program where state resources are utilized for additional air sample collections around the factory site, analyzed in the state laboratory facilities. Our participatory action research project achieved the intended result of informing stakeholder action with meaningful data.

Figure 1.

Study Process Diagram