The Minneapolis–St Paul (Twin Cities) metropolitan area has cleaner air than many cities and currently meets federal air quality standards. Nonetheless, fine particulate (PM2.5) levels in this 7-county area are elevated compared with other parts of the state, and community members and local officials are concerned about significant elevations in child asthma exacerbations rates in parts of the core cities. Moreover, research has shown that a relationship exists between fine particulate pollution and premature mortality, even at the relatively low concentrations in the Twin Cities metropolitan area.1,2 Infants, children, seniors, and people who have chronic lung and heart conditions are especially vulnerable, and prior work suggests that exposure disparities in urban neighborhoods place disproportionate health risks on populations of color and low-income communities.3,4
Over the past 10 years, national, state, and local pollution reduction policies and initiatives have required sizable ongoing investments in cleaner fuels, new vehicles, and technology upgrades. And while air quality has improved substantially, there is a need to demonstrate how these improvements are impacting public health. State agencies now have the analytical tools to apply population health surveillance and environmental monitoring data and indicators to track the progress and efficacy of public health actions, to predict changes in the magnitude and distribution of air pollution–related health impacts, and to weigh the benefits against the cost of continuing investment in air quality improvements. Officials and communities alike want assurance that the health benefits of pollution reduction policies are shared equitably across all communities.
In 2014, the Minnesota State Legislature funded the Urban Air Quality and Respiratory Health Initiative to address important questions about health impacts, to use the findings for stakeholder engagement, and to help inform interventions and policy decisions. Under this initiative, the Minnesota Department of Health (MDH) Environmental Public Health Tracking Program (MN Tracking) and the Minnesota Pollution Control Agency (MPCA) collaborated on an environmental health impact analysis. Patterned after similar analyses conducted by the New York City Department of Health and Mental Hygiene,5 we used local air pollution data and public health surveillance data to address 3 key questions: (1) What are the health impacts of fine particulate matter and ozone on the metro area population? 2) Are the impacts distributed equally across the area or are there patterns of geographic and demographic disparities? and (3) What benefits can be achieved through further reduction in air pollutant concentrations?
The MPCA works with the US Environmental Protection Agency (EPA) to monitor, identify, and regulate sources of air pollution to comply with state and federal regulations. The MN Tracking program at MDH, with support from the Centers for Disease Control and Prevention, conducts routine analysis and dissemination of mortality and morbidity data for environmentally related chronic diseases. Under the Healthy Minnesota 2020 framework, both agencies share a responsibility for ensuring equal opportunity for health for all Minnesotans.6
We applied health impact analysis methods that use research findings about the causal relationship between pollutant exposures and health outcomes to estimate the fraction of cases in a population that can be attributed to air pollution. In these analyses, we quantified the counts and rates of premature deaths, hospital admissions, and emergency department (ED) visits for respiratory and cardiovascular outcomes attributed to exposure to PM2.5 and ozone at 2008 concentrations for populations living in 165 zip codes that comprise the Twin Cities metropolitan area.
The analyses were conducted using EPA's Environmental Benefits Mapping and Analysis Program Community Addition (BenMAP-CE), version 1.0.8, an open-source software tool used by states and the EPA for estimating the health benefits, and economic values of those benefits, associated with improvements in ambient air pollution.7 Inputs into the BenMAP-CE model and data sources are described later and include a real or potential change in the air quality, one or more concentration-response (C-R) coefficients for each of the selected health outcomes, the estimated count and rate of each disease endpoint in the population, and the total population at risk. Air pollution–attributable counts, rates, and 95% confidence intervals (CIs) were calculated for each of the 165 zip codes that lie entirely or partly within the 7-county metro area. Zip code–level impacts were summed together to estimate the impacts for the entire metro area in a single year, 2008. Disparate impacts by poverty and race were estimated by summing impacts among zip codes stratified according to the percentage of the population in poverty and percentage of population that are residents of color using American Community Survey (ACS) data.
Input: Change in air quality data
Annual average PM2.5 and summertime ozone concentrations were derived for the year 2008 from the US EPA Downscaler model.8 The model merges measured daily PM2.5 and 8-hour maximum ozone concentrations from regulatory air monitors with a corresponding modeled concentration surface from the Community Multiscale Air Quality model using spatial regression methods.
Using geographic information system techniques, modeled daily census tract-level PM2.5 and ozone concentrations were averaged to annual and summertime concentrations, respectively, and then aggregated to each of the 165 zip codes in the study area to correspond with health outcome data resolution (see Figure, Supplemental Digital Content 1, available at http://links.lww.com/JPHMP/A345, which shows the 2008 air quality surface for both ozone and PM2.5). Annual concentrations of PM2.5 are relatively uniform throughout the metropolitan area and ranged from 9.7 to 11.6 μg/m3 across all zip codes, with higher concentrations found in the more densely populated Minneapolis–St Paul zip codes.
The 2008 averages of daily 8-hour maximum ozone concentrations by zip code range from 40.3 to 43.9 ppb. Because of the atmospheric dynamics of ozone formation and dispersion, higher concentrations occur in the southern suburban zip codes of the metro area. Overall, average ozone concentrations are considered to be relatively uniform throughout the metro area.
Input: Health and mortality outcome data
Morbidity and mortality data are annual average counts of each outcome for each zip using data from 2006 to 2010. The 5-year period was selected to derive a stable measure of annual average counts centered on the year 2008 and to provide sufficient population density and privacy protections for a zip code–level analysis. Outcomes included are described in Table 1.
Inpatient hospitalization and outpatient ED hospital discharge data are provided to MDH by the Minnesota Hospital Association and include hospital billing data from all reporting Minnesota and neighboring state hospitals but not from federal and sovereign hospitals. Hospital discharge data are de-identified and provide the patient's date of birth, sex, diagnostic code (International Classification of Diseases, Ninth Revision), and the zip code of the patient's home address at the time of hospitalization. Mortality data are from MDH's Center for Health Statistics and provide the codes for the underlying cause of death and zip code of residence at the time of death. For each health outcome, age and diagnostic codes or underlying causes of death codes were matched to case definitions from the studies used as the source of the C-R coefficients.
To protect patient privacy, small hospitalization and ED visit counts were suppressed, which resulted in a sizable number of zip codes with suppressed data. Therefore, for estimates of PM2.5 pediatric asthma and ozone-attributable asthma hospitalization, a Geographical Aggregation Tool developed by New York State,9 a CDC National Tracking Network partner, was used to join neighboring zip codes until a minimum number of cases was reached to support data privacy protection.
Input: CR coefficients
Although hundreds of peer-reviewed studies have examined air pollution–related morbidity and mortality, we selected C-R coefficients from several large US multicity studies as well as smaller studies deemed most relevant to our Upper Midwest urban population based on study date (most current) and study population characteristics.1,2,10–15 For PM2.5 impacts, outcomes included all-cause premature mortality (ages 25 years and older, 30 years and older), asthma hospital admissions (age younger than 18 years), respiratory hospital admissions (ages 18-64 years, 65 years and older), cardiovascular hospital admissions (age 65 years and older), and asthma ED visits (all ages). To test the sensitivity of our impacts analysis to choice of C-R coefficient, we used two C-R coefficients from separate all-cause mortality studies (and age thresholds).1,2 For ozone impacts, outcomes included cardiopulmonary mortality (all ages), asthma hospital admissions (all ages), and asthma ED visits (all ages) (see Table, Supplemental Digital Content 2, available at http://links.lww.com/JPHMP/A346, which lists the studies chosen, the specific health outcomes, age groups, and the corresponding C-R coefficients).
Input: Population and demographic data
Baseline health and death rates by zip code and age group were drawn from 5-year population estimates from ACS.16 Population estimates from 2007 to 2011 were the earliest time period for which ACS provides data by zip code.
In addition, ACS zip code–level poverty status and racial composition were used to explore potential differential exposure and impacts. Zip codes were stratified according to the percentage of the population in poverty and the percentage of residents of color. To ensure comparability of our results with other local studies and initiatives, we used poverty and racial composition categories developed by the Twin Cities Metropolitan Council for defining Racially Concentrated Areas of Poverty.17 Zip codes were apportioned into “low,” “medium,” and “high” poverty strata following percentage ranges of 0%-19%, 20%-39%, or 40% or more residents with incomes 185% or less of the federal poverty line, the threshold that many federal assistance programs set for low-income families to be eligible. Zip codes were apportioned into “low,” “medium,” and “high” strata by percentages of residents of color as 0%-24%, 25%-49%, and 50% or more residents in a zip code that do not identify as white non-Hispanic.
Health impact analysis approach
To determine local pollution concentration changes to be evaluated, we first compared the 2008 Downscaler modeled concentrations with estimated naturally occurring “background” concentrations that would exist without sources of air pollution from human activity and that therefore cannot be affected by pollution control efforts. This background concentration in the Twin Cities metropolitan area was identified as 0.84 μg/m3 for PM2.518 and 27.7 ppb for ozone.19 Next, we evaluated the potential health benefits associated with a modest 10% improvement in air quality. This modest improvement level corresponds to air quality targets recommended by Environmental Initiative's Minnesota Clear Air Dialogues, a consortium of leaders from Minnesota's business, government, and nonprofit sectors.20
The BenMAP-CE model calculates the predicted change in the number of health impacts that are associated with a user-defined change in air pollutant concentration. The general form of the equation used to estimate health impacts is as follows:
where ΔY = change in the number of health events associated with the change in the level of air pollution, ΔAQ; Y0 = baseline morbidity or mortality rate; β = C-R coefficient as determined by an epidemiological study; ΔAQ = change in the air pollutant concentration (PM2.5 or ozone); and Pop = population size of the relevant age group.
In BenMAP-CE, we estimated 95% CIs for each of the impact estimates at the zip code level, based on the standard error of the effect coefficient from the epidemiology study (β). As such, CI does not account for other sources of uncertainty, such as uncertainties in air pollution concentration estimates or in the population age distribution.
Metro-wide pollution-attributable health impacts
In the Twin Cities metropolitan area, we estimate that, in 2008, fine particle pollution (using 2008 baseline concentration minus the background) contributed to about 2.1% of respiratory hospitalizations among residents aged 65 years and older (95% CI, 1.2-3.0) and 2.1% of child asthma hospitalizations (95% CI, 0-10.6); 0.7% of cardiovascular hospitalizations among residents aged 65 years and older (95% CI, 0.3-1.1); and 2.9% of all-ages asthma ED visits (95% CI, 0.8-4.9) (Table 1).
The premature mortality estimate analysis showed that this method for estimating the attributable fraction can be highly sensitive to the choice of C-R coefficient. Using the C-R coefficient from the latest update to the Harvard 6-Cities (H6C) cohort study2 yields a result of 12.6% (95% CI, 6.5-18.3) of all deaths (age 25 years and older) annually attributed to PM2.5 whereas the C-R coefficient from the latest update to the American Cancer Society cohort study1 yields a result of 5.8% (95% CI, 4.0-7.6) of all deaths (age ≥30 years and older).
We considered these two results as “low” and “high” scenarios to estimate a range of potential impacts. Specifically, a 10% reduction in PM2.5 concentrations from 2008 baseline levels could annually prevent approximately 105 to 247 premature deaths, 35 hospital admissions for respiratory and cardiovascular outcomes, and 44 asthma ED visits across the metro area.
In the Twin Cities metropolitan area, we estimated that in 2008 ozone pollution (2008 Downscaler baseline May-September levels minus background) contributed to 1.1% of all-ages premature cardiopulmonary deaths (95% CI, 0.4-1.8), 4.9% of all-ages asthma hospitalizations (95% CI, 3.0-6.7), and 3.2% of all-ages asthma ED visits (95% CI, 0-67.0) (Table 1). We estimated that a 10% reduction in ozone concentrations from 2008 levels could annually prevent approximately 7 premature deaths, 14 hospital admissions, and 57 ED visits.
In a separate analysis of impacts by age group (data not shown), we found that PM2.5-attributable respiratory hospitalization rates and death rates are highest among the elderly, aged 65 years and older, whereas PM2.5-attributable asthma ED visit rates were highest among children (age younger than 18 years). Both the elderly and children also experienced the highest ozone-attributable respiratory hospitalization rates. These age-related disparities are reflective of the large underlying disease rate disparities in these age groups. With the exception of PM2.5-attributable respiratory hospitalizations, which use different effect estimates by age, the proportion of impacts attributable to air pollution was consistent across age groups.
Geographic analysis of Twin Cities pollution-attributable health impacts
Although pollutant-level differences across the area and by zip code were relatively small, a consistent finding in this analysis was that geographical differences in air pollution–attributable health impacts ranged widely due to underlying health disparities. For example, using the H6C cohort C-R coefficient for all-cause mortality, the estimated zip code–level rate of PM2.5-attributable premature death ranged from 27 to 636 premature deaths per 100 000 people aged 25 years and older, with higher rates in the core cities of Minneapolis and St Paul. The underlying death rates in these zip codes ranged from 213 to 5337 deaths per 100 000 people and followed the same geographic pattern. Similarly, the estimated zip code–level rates of PM2.5-attributable asthma hospitalizations among children younger than 18 years ranged from 0.5 to 27.5 cases per 100 000 individuals, with the highest impacts occurring in the central urban zip codes where the underlying rates are significantly elevated. We observed that for all health outcomes examined in this study, the spatial distribution of PM2.5-attributable impacts closely corresponds to the spatial distribution of underlying rates. The spatial patterns of ozone-attributable premature deaths and morbidity impacts also mirrored the underlying outcome rates and varied widely across zip codes.
Premature mortality by area-level poverty status and racial composition
Results of the analysis of zip codes stratified by the proportion of the population in poverty and populations of color revealed that air pollution–attributable rates of premature mortality (age 65 years and older) were higher in populations with greater poverty and populations of color and reflect differences in the underlying rates (Table 2). However, the differences between strata were not statistically significant. We found that the proportion of premature mortality (age ≥65 years and older) attributable to PM2.5 from all causes of death (12.1%) and to ozone from cardiopulmonary causes (1.1%) was generally consistent across all strata of poverty and populations of color.
We applied a previously published method of health impact analysis using environmental public health surveillance data to quantify the population health impacts from air pollution across the Twin Cities metropolitan area toward informing the work of local organizations to improve air quality and public health. In summary, we estimated that air pollution contributed to around 2% to 5% of respiratory and cardiovascular hospitalizations and ED visits and between 6% and 13% of premature deaths in a single year (2008). These findings are consistent with similar studies in other urban areas of the United States using this method. For instance, studies in Los Angeles, Philadelphia, and Chicago found that approximately 10% of deaths were attributable to air pollution.19 With nearly 3 million people living in the Twin Cities metropolitan area, these results suggest that substantial numbers of preventable health impacts are occurring.
Importantly, the geographical differences in air pollution–attributable health impacts that we observed largely reflected underlying disparities in occurrence rates and population vulnerability, given that variation in pollution levels was relatively small.
Implications for Policy & Practice
* To support application by a broad range of partner organizations, the conclusions of this study were distilled into 3 key messages: (1) air pollution causes significant health impacts even with the relatively clean Twin Cities air; (2) everyone is susceptible to the health impacts of air pollution, but some groups are more vulnerable than others; and (3) reducing air pollution and decreasing underlying health disparities are important for reducing the health impacts. These messages were shared with the media, diverse community groups, legislators, and the public through multiple venues. We produced a Web site for businesses, communities, and individuals to access information about air quality and public health and to promote actions that reduce these impacts.23 Data from this project served to build new zip code–level data displays of respiratory disease disparities for the MN Tracking online data access portal.24
* Reducing the public health impacts of air pollution requires a 2-pronged approach: environmental partners working to improve air quality, and public health partners working to improve overall health equity. In a report to the Minnesota Legislature, MDH laid out a plan for advancing health equity statewide, noting that inequities in social and economic factors remain as major contributors to health disparities.25 Practices described included applying a “Health in All Policies” approach through collaborations, strengthening community relationships, and expanding current data capacity.
* Clean Air Minnesota, a coalition of air quality leaders from the private and nonprofit sectors and state and local governments, is currently implementing 24 initiatives to reduce emissions associated with fine particles and ozone formation.20 These include the MPCA Diesel Emission Reduction Act support for vehicle fleet improvements and Project Stove Swap replacements of wood-burning appliances.
* This study and engagement with stakeholders represent a new level of collaboration in Minnesota. It has informed a data-driven, integrated approach to addressing environmental public health equity problems. Building on prior research for developing indicators that track progress over time, MN Tracking is exploring analyses for calculating a local C-R coefficient using case-crossover and time-series methods.26 Although it is not addressed in this article, the MPCA has estimated the economic values of these health impacts and presented the findings to the Minnesota Legislature.27 The analysis and results described in this article serve as a baseline and a benchmark for future year assessments of public health in relation to air quality improvements.
This is consistent with other studies of urban air pollution health impacts and disparities, notably a similar analysis conducted in New York City.5,21 Pollution concentration differences across zip codes and demographic categories in the Twin Cities metropolitan area appear to be slight, and the attributable fractions of air pollution health impacts were fairly consistent across geographic and demographic categories.
There are significant limitations to this analysis due to uncertainties in each of the parameter estimates and model inputs. This analysis is based on single annual average air pollutant concentration for each zip code to estimate population exposure and does not use measurements of individual exposures. Temporal averaging smooths out daily changes in air quality, and spatial averaging smooths out the heterogeneity of air quality within each zip code. Population estimates from ACS are published with a margin of error and so this analysis aggregated racial/ethnic groups to reduce error in the demographic analysis due to small numbers.
The choice of a C-R coefficient introduces additional uncertainty into the results.22 For estimating the effect of PM2.5 pollution on premature death, we primarily used the effect coefficient from the H6C cohort study for estimating disparities because it was the study that followed a racially and economically diverse population most resembling the Twin Cities metropolitan area population. As the results show, the American Cancer Society study C-R coefficient produces a significantly smaller impact estimate.
Given the limitations of this method, we caution that these estimates may be used to gauge the magnitude and variability of health impacts and population vulnerability, and to show the benefits of broad strategies to address them, but to be clear that this method does not provide a precise measure.
1. Krewski D, Jerrett M, Burnett RT, et al Extended Follow-up and Spatial Analysis of the American Cancer Society Study Linking Particulate Air Pollution and Mortality. Boston, MA: Health Effects Institute; 2009. HEI Research Report No. 140.
2. LePeule J, Laden F, Dockery D, Schwartz J. Chronic exposure to fine particles and mortality: an extended follow-up of the Harvard Six Cities Study from 1974-2009. Environ Health Perspect. 2012;120:965–970.
3. Clark LP, Millet DB, Marshall JD. National patterns in environmental injustice and inequality: outdoor NO2
air pollution in the United States. PLoS One. 2014;9(4):e94431.
4. Pratt GC, Vadali ML, Kvale DL, Ellickson KM. Traffic, air pollution, minority and socio-economic status: addressing inequities in exposure and risk. Int J Environ Res Public Health. 2015;12:5355–5372.
5. Kheirbek I, Wheeler K, Walters S, Kass D, Matte T. PM2.5
and ozone health impacts and disparities in New York City: sensitivity to spatial and temporal resolution. Air Qual Atmos Health. 2013;6:473–486.
6. Minnesota Department of Health. Healthy Minnesota 2020, statewide health improvement framework. http://http://www.health.state.mn.us
/healthymnpartnership/hm2020/1212healthymn2020fw.pdf. Published 2012. Accessed January 10, 2017.
7. US Environmental Protection Agency. Environmental Benefits Mapping and Analysis Program (BenMAP). http://http://www.epa.gov
/benmap. Published 2014. Accessed January 10, 2017.
8. US Environmental Protection Agency. Fused air quality surfaces using downscaling. http://www.epa.gov
/hesc/rsig-related-downloadable-data-files. Published 2014. Accessed January 10, 2017.
9. New York State Department of Health. Geographic Aggregation Tool, SAS Beta Version 4. Troy, NY: New York State Department of Health; 2009.
10. Moolgavkar SH. Air pollution and hospital admissions for chronic obstructive pulmonary disease in three metropolitan areas in the United States. Inhal Toxicol. 2000;12(suppl 4):75–90.
11. Zanobetti A, Franklin M, Koutrakis P, Schwartz J. Fine particle air pollution and its components in association with cause-specific emergency admissions. Environ Health. 2009;8:58.
12. Peng RD, Bell ML, Gay AS, et al Emergency admissions for cardiovascular and respiratory diseases and the chemical composition of fine particulate air pollution. Environ Health Perspect. 2009;117(6):957–963.
13. Babin SM, Burkom HS, Holtry RS, et al Pediatric patient asthma-related emergency department visits and admissions in Washington DC, from 2001-2004, and associations with air quality, socio-economic status and age group. Environ Health. 2007;6:9.
14. Winquist A, Klein M, Tolbert P, Flanders W, Hess J, Sarnat S. Comparison of emergency department and hospital admissions data for air pollution time-series studies. Environ Health. 2012;11:70.
15. Huang Y, Dominici F, Bell ML. Bayesian hierarchical distributed lag models for summer ozone exposure and cardio-respiratory mortality. Envirometrics. 2005;16:547–562.
17. Metropolitan Council. Choice, place and opportunity: an equity assessment of the Twin Cities region. http://http://www.metrocouncil.org
/Planning/Projects/Thrive-2040/Choice-Place-and-Opportunity.aspx. Published March 2014. Accessed January 10, 2017.
18. US Environmental Protection Agency. Integrated science assessment for particulate matter. http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=216546. Published March 2012. Accessed January 10, 2017.
19. Fann N, Lamson AD, Anenberg SC, Wesson K, Risley D, Hubbell BJ. Estimating the national public health burden associated with exposure to ambient PM2.5
and ozone. Risk Anal. 2012;32(1):81–95.
20. Environmental Initiative. Minnesota's Clean Air Dialogue: a collaborative plan to reduce emissions. http://http://www.environmental-initiative.org
/images/files/MnCAD/*MnCADFinalReport24Apr13.pdf. Published 2013. Accessed December 15, 2017.
21. New York City Department of Health and Mental Hygiene. Air pollution and the health of New Yorkers: the impact of fine particles and ozone. http://www1.nyc.gov/assets/doh/downloads/pdf/eode/eode-air-quality-impact.pdf. Published March 2011. Accessed June 5, 2017.
22. Hubbell BJ, Fann N, Levy JI. Methodological considerations in developing local-scale health impact assessments: balancing national, regional, and local data. Air Qual Atmos Health. 2009;2(2):99–110.
24. Minnesota Department of Health. MN public health data access. https://apps.health.state.mn.us/mndata. Published 2016. Accessed January 15, 2017.
25. Minnesota Department of Health. Advancing health equity in Minnesota, report to the legislature. http://http://www.health.state.mn.us
/divs/chs/healthequity/ahe_leg_report_020414.pdf. Published 2014. Accessed December 1, 2017.
26. Johnson J, Pratt G, Yawn B. Measuring the impact of particulate matter reductions by environmental health outcome indicators. http://cfpub.epa.gov/ncer_abstracts/index.cfm/fuseaction/display.abstractDetail/abstract/8644/report/F. Published 2012. Accessed January 15, 2017.
27. Minnesota Pollution Control Agency. Air quality in Minnesota, 2015 report to the legislature. https://http://www.pca.state.mn.us
/sites/default/files/lraq-1sy15.pdf. Published 2015. Accessed January 10, 2017.
air pollution; attributable fraction; environmental public health tracking; health disparities; health impact analysis