Taking advantage of the “natural experiment” scenario of the Shenango, Inc., coal coking plant closure and utilizing ITS analysis methods, we confirmed multiple short and longer term respiratory health benefits in the neighboring resident community after the plant shutdown, as compared with before its closure and compared with control sites and control outcomes (as summarized in Figure 6). Notably, the pediatric asthma ED visit risk precipitously dropped by 41% the month after closure and continued to decrease in the following years, significantly differing from the preclosure time trend. Overall respiratory ED visits also abruptly dropped by 21% after the shutdown in the near-plant exposure population but then continued to time-trend at a rate not different from that for the preclosure period. Overall hospitalizations for respiratory symptoms, as well as for people primarily diagnosed with asthma, decreased immediately in the quarter of the plant closure (by 56% and 16%, respectively) but followed the preclosure trend afterward, whereas COPD hospitalizations decreased in the longer term versus the prior trend. These near-plant community reductions in pulmonary medical visits after the closure were not found in the control populations during the same postclosure period, consistent with the interpretation of the near-plant reductions as due to the plant closure.

Known air pollution health effect mechanisms are consistent with the health benefits noted after the plant closure. The immediate respiratory health benefits observed are consistent with the significant large reduction of airway irritants, and longer term reductions may represent lung repair after the closure. These lung health benefits from reduced fossil-fuel air pollution appear remarkably consistent with the lung function and airway inflammation benefits experienced by smokers with asthma after they quit smoking (43). Coke ovens emit SO₂ and hydrogen sulfide (which spontaneously oxidizes to SO₂), which can induce airway irritation and bronchoconstriction and thus result in asthma exacerbation (20, 44). The immediate drop in ambient SO₂ concentrations at the exposure site is consistent with the significant immediate decreases observed for both asthma hospitalizations and pediatric ED visits. The immediate declines in respiratory ED visits and hospitalizations of patients admitted for respiratory symptoms are also consistent with the instant elimination of environmental SO₂.

Longer term downward trends were observed for COPD hospitalizations and for pediatric asthma ED visits, suggesting a gradual reduction of chronic systemic inflammation from improved air quality over time, consistent with the elimination of chronic oxidative stress inducing coking-related PM particles. Although the closure-associated PM_2.5 mass concentration reduction was just a few micrograms per cubic meter, the physical and chemical characteristics of particles emitted in the coking process can make them especially potent in eliciting inflammatory responses and systemic effects, as compared with other types of PM (e.g., windblown soil particles) that lack those especially toxic characteristics. In particular, coking-related particles are usually higher in number concentration and smaller in size than other more usual ambient PM particles (9), exhibiting a higher ratio of surface area to volume and, thus, a higher capacity for biochemical reactions and biological interactions, as well as a higher amount of toxic components adhered (45, 46). Commonly adhered gaseous compounds include SO₂, which not only trigger irritation themselves but also increase the solubility and bioavailability of transition metals present in fossil fuel– associated PM (20, 47). Such submicron particles can also penetrate deeper into the respiratory tract, enter the bloodstream, and translocate into the systemic circulation (48, 49). Once inhaled, metal oxides—widely used as catalysts in coke production—can be generally easily dissolved in lung fluids (50, 51). Moreover, high concentrations of heavy metals and transition metals have been found to be enriched in coking-related PM (52, 53), which is consistent with an enhanced adverse effect per micrograms per cubic meter than reported for PM_2.5 particles, in both the short and long term.

The presence of these fossil fuel– related PM particles can individually or synergistically induce oxidative stress and excess reactive oxygen species production, inflammation, and chronic systemic health effects (47, 54). The inflammatory responses in the airway epithelium and pulmonary circulation also involve the secretion of proinflammatory cytokines (e.g., tumor necrosis factor α, IL-1β, IL-6, IL-8, etc.) (47, 55, 56). The elevation of these proinflammatory cytokines are known to be associated with, and involved in, the pathogenesis of COPD; for example, recent scientific evidence suggests that IL-6 can interfere with the cell signaling pathway, induce apoptosis, and play a role in airway remodeling (57, 58). The shutdown of the coke plant, therefore, would mitigate prior exposure-induced systemic inflammation and, in the longer term, alleviate the exacerbation of chronic respiratory diseases such as COPD and childhood asthma. The hypothesized biological mechanism of alleviated systemic inflammation and oxidative stress is further supported by our previously reported findings of immediate and longer term cardiovascular health benefits in nearby residents (17).

The findings of this study must be considered in the context of its limitations, despite its strengths in design. First, the spatial-temporal resolution of health data acquired for the analyses was limited by the local agency data release rules. Notably, in contrast to the complete unredacted data before closure, nearly half of the numbers of monthly pediatric ED visits had to be redacted because of low (1 ⩽ N < 5) counts after the shutdown at the near-Shenango exposure site. Such significantly higher percentages of redacted data after versus before the closure can also be interpreted as further evidence of the reduction in incidences after the closure. Also, the findings remained robust in sensitivity analyses with only complete case data and with alternate imputation methods (Table E4). In addition, the transition from ICD-9 to ICD-10 codes in the last quarter of 2015 in hospitalization data may have contributed to a systematic reporting bias and result in less accurate intervention effect size estimates, but such count reductions were not seen at other sites, and multiple studies have reported no significant changes in coding validity or sensitivity in other similar databases (28, 59).

Second, our analytical areas were defined using zip codes, as the data were supplied or aggregated by residential zip code by the providing agencies. However, because of the complexity of terrain and wind patterns, and their associated heterogeneous dispersion of air pollutants emitted from point sources such as the Shenango coke plant, these zip code areas’ spatial heterogeneity and geographical complexity may have led to exposure misclassification and thus may have affected the estimated health effect size (60). In this study, we restricted our study area to zip codes in the immediate vicinity of the Shenango plant to examine the health impacts of the most directly affected population—those who were exposed to the fugitive emissions and unintended air pollution releases in its air quality violations (24). Higher spatiotemporal resolution air quality and health data, more accurate dispersion models considering plant characteristics and wind pattern, and source-specific air pollution data, if they were available, could potentially enable further analyses to assess the benefits from the intervention more accurately (60, 61).

Third, despite the extensive efforts to address potential confounding with the inclusion of internal and external control series, the control outcomes may still not exhibit the exact same temporal patterns to completely address the concern for unmeasured/residual confounding. By using the same population as its own control over time, the ITS study design inherently minimized potential confounding effects from individual participant level health condition and lifestyle factors; however, a long-term nonlinear trend and/or the population/area level confounding effects may still exist. Nevertheless, we were not aware of any other events or policy changes at the study sites that could have alternatively led to the robust patterns of decreased health impact trends observed.

In terms of the generalizability of our results, there were a few modest demographic differences between the study sites and the county and state population: The exposure population had a slightly higher poverty rate and percentage of non-White people, as well as a smaller population growth rate, limiting the generalization of the results (e.g., effect coefficients) to dissimilar (e.g., healthier and wealthier) populations. Although the type of health benefits found from this plant closure should apply to other areas with such plant closures, caution should be given in the direct quantitative application of effect size findings to another population with very different population characteristics. Moreover, air pollution from different emissions sources can vary drastically in its composition and particle properties. Therefore, interventions targeting different air pollution components can have different health implications; with varying susceptibility of population exposed, the responses would very likely differ. Therefore, future environmental and health benefit assessments of interventions would help generate further scientific evidence to inform policymaking.

Another possible complication lies in the potential demographic and socioeconomic changes after implementations of energy transition and industrial transformation policies (62). In this case, the closure of the Shenango coke plant affected 173 workers in the local neighborhood (63), representing some 0.2% of the population at the exposure site, which is small compared with the percent change found in respiratory impacts. Despite their potentially higher occupational exposures, it is not plausible that they could account for all the reductions in respiratory hospitalizations and ED visits observed. Also, in terms of the local and spillover socioeconomic impacts, we did not observe significant immediate changes in the demographic and socioeconomic factors at our study sites, and adjusting for poverty rate did not change our effect estimates. However, it should be noted that environmental interventions that target industrial emissions, especially those with larger scale interventions and on a longer term basis, can exert complex systemic implications on the health and well-being of the local communities (62).

Overall, the aforementioned potential limitation considerations are deemed unlikely to be able to explain away our robust findings, and prior epidemiologic studies lend strong support to our findings. Those studies have found that the respiratory health of young children improved after the closure of coal-fired powerplants (64) and steel mills (65). Although past literature has suggested that living in the vicinity of coking plants can potentially increase health risks in resident populations (13, 20, 66), a previous investigation on the Shenango coke plant closure, which was based on annual counts but not the immediate and long-term trend changes, was reluctant to conclude a causal relationship between the shutdown and the health benefits observed (32). Our analysis, considering more immediate time periods before and after the closure, are more definitive as to the cause-and-effect relation, as it revealed the strong time coincidence of the closure and multiple subsequent health benefits, especially in comparison with control health outcomes and other locales at the same time, meeting the temporality, strength, consistency, specificity, and biological plausibility criteria for causal inference (67). Our findings on respiratory benefits, together with the previously reported cardiovascular benefits, are consistent with the conclusion of a causal inference of significant cardiopulmonary health benefits from the closure of the Shenango coke plant being caused by the noted air pollution reductions (17, 67, 68). Our findings add to the growing body of evidence that regulating and reducing fossil fuel–related air pollution emissions produce significant public health benefits, in both the short term and the longer term (69–71).