Introduction

The mission of Centers for Disease Control and Prevention (CDC)’s Environmental Public Health Tracking (EPHT) program is “…to provide information from a nationwide network of integrated health and environmental data that drives actions to improve the health of communities,…” (CDC 2006). Consistent with this mission, a key objective of the EPHT program is tracking the public health burden of ambient air pollution with analyses that are timely and relevant to local stakeholders. To address methodological issues relevant to this objective, in January 2008 a 2-day workshop in Baltimore, Maryland, jointly organized and sponsored by Health Effects Institute, CDC, and US EPA, brought together representatives of state and national public health and environmental agencies and academic researchers from the USA, Canada, and Europe. Among the workshop participants were persons with expertise and experience in the development and application of statistical and epidemiologic tools for air pollution health impact assessment in both academic and policy settings in North America and in Europe (see Appendix 1).

Workshop participants were charged with producing recommendations for use at the state and local levels for communicating the results of the analyses to stakeholders. These recommendations would suggest ways to analyze linked air quality and health data and to estimate and track over time health impact indicators for two pollutants: fine particulate matter (PM2.5) and ozone.

Specifically, the participants were asked to recommend:

  1. 1.

    Approaches for using analyses of state data to generate state and sub-state impact estimates for acute effects of air pollution;

  2. 2.

    Approaches for using current quantitative estimates—and to develop further estimates—of the relationship between air pollution exposure and health outcomes from the scientific literature, as well as ways to generate estimates of acute and chronic health impacts in local areas; and

  3. 3.

    Approaches for communicating to stakeholders the estimates and the limitations of those estimates.

In working papers commissioned before the workshop (Bachmann 2008; Fuentes 2009; Hubbell et al. 2009; Medina et al. 2009; Shin et al. 2009; Talbot et al. 2009; Wartenberg 2009; White 2009) and in workshop presentations, relevant work of the EPHT program was reviewed, previous health impact assessments were described, and methodological topics were discussed. The workshop participants then discussed key methodological issues and made recommendations regarding further development and application of indicators of the health effects of air pollution—specifically, indicators that would be suitable for public health tracking of air pollution impacts at the state and local level.

This paper summarizes the workshop discussions and recommendations for developing ongoing and consistent implementation of air pollution health impact assessments by state and local agencies.

The need for timely and locally relevant information on the health effects related to air pollution

Over the past 50 years, the USA has made considerable progress in reducing levels of health-damaging ambient air pollution, progress that has resulted in substantial public health benefits (US EPA 1997, 1999). Nevertheless, in some areas of the country, PM2.5 and ozone levels currently exceed National Ambient Air Quality Standards (US EPA 2008). In terms of increased mortality and morbidity, especially cardiovascular and respiratory disease, exposure to air pollution continues to affect the health of the US population (US EPA 2004a, b, 2005a, b, 2006, 2007).

While air quality improvements are projected to continue, increasing urban sprawl, traffic volume, and congestion may be slowing progress and threatening future gains, especially in some regions (Frumkin et al. 2004). At the same time, growing evidence points to the importance of health impacts associated with intra-urban gradients in ambient air pollution, particularly those related to traffic (Miller et al. 2007; Jerrett et al. 2005). Thus, future advances in public health protection will likely require improved understanding of those local pollution sources and potential control measures that impinge on state and local land use and transportation policy. The more such measures affect local stakeholders, the greater the need for local engagement in the process. Providing local stakeholders timely, understandable and locally relevant data on air pollution health impacts will be a necessary part of such engagement.

Environmental public health tracking has been defined as the “ongoing collection, integration, analysis and dissemination of data from environmental hazard monitoring, human exposure tracking, and health effects surveillance” (Meyer et al. 2006; Environmental Health Tracking Project Team 2000). Consistent with CDC’s model of public health surveillance, the communication of findings to those with a “need to know” (Thacker and Berkelman 1988) is a key component. The tracking model broadly defines stakeholders who are targets of dissemination efforts to include policy makers, government and nongovernmental agencies, businesses, researchers, the media, and members of the public. Because the health effects of PM2.5 and ozone are well established, the EPHT program identifies ambient air pollution as a priority for tracking indicators of exposure and for health impact assessment. As a result of a collaboration between CDC, US EPA, and state EPHT programs, an infrastructure within the Environmental Public Health Tracking Network (EPHTN) is under development (Talbot et al. 2009; Boothe et al. 2005). This infrastructure will enable ongoing, periodic, and timely analyses of the health impacts of air pollution at the state and local levels and will address such questions as:

  • What is the public health burden attributable to ambient PM2.5 and ozone levels?

  • How does the burden vary within and between states?

  • Is the burden changing over time, for example, in response to efforts to reduce air pollution levels and population exposure?

Methodological challenges

Estimating the health impacts of air pollutants is more complex than estimating some health impacts of certain environmental exposures. For example, acute carbon monoxide poisonings (which generally result from indoor, not ambient, exposures) can be diagnosed with specificity when compatible signs and symptoms occur in the presence of an elevated level of carboxyhemoglobin. The number and rate of such events can be tracked using hospital inpatient and emergency department administrative data. For the EPA’s criteria air pollutants, generally, there is, unfortunately, no similarly straightforward way to track public health impacts. The most serious known health effects of exposure to air pollutants such as PM2.5 and ozone, both indicators of complex mixtures with no well-accepted or easily measured specific biomarkers, involve morbidity and mortality from diseases of the cardiovascular and respiratory systems that have multiple other complex causes. Therefore, although the contribution of the diseases themselves to the overall burden of disease in a population can be estimated categorically by identifying individuals who meet pre-specified clinical criteria, the burden attributable to PM2.5 or ozone cannot be unambiguously identified and can only be estimated from statistical models.

While an extensive body of research has established causal links between exposure to these pollutants and human health, concentration–response (CR) relationships quantifying these links have, for a variety of reasons, been shown to vary among geographic areas and over time (Samet et al. 2000; Samet 2008; Katsouyanni et al. 2001; Hubbell et al. 2009). PM2.5 and ozone may serve in part as indicators of complex pollutant mixtures (White 2009), and variation in composition by space and time may alter the concentration/health impact relationship. Additional modifying factors include population susceptibility and local health care utilization, services, and recording practices. In addition to these relatively stable local differences, exceptional local weather events or emission sources, including forest or structural fires or construction demolition, may introduce within a local area different pollutant species or extreme pollutant levels. In addition, local interventions, including enhanced air quality alerts, land use and transportation changes, and control of local point sources may alter local air quality and affect human exposure and its relationship to health.

Although estimates from purely local analyses should in theory best reflect local modifiers of a CR function, “true” CR relationships at prevalent ambient air pollution levels are small relative to random error and potential bias affecting a single local study. Thus, estimates based on local data only may not accurately reflect the underlying CR relationship and may even give indications of either anomalous “protective” effects or implausibly large effect risk estimates. The methodological challenges stemming from use of local data are even greater for estimating time trends at the state and sub-state levels for air pollution’s impact on health.

Communicating with stakeholders in a meaningful and complete way concerning the information gained from the EPHT analysis poses additional challenges, especially when attempting to set out clearly any uncertainties and their implications (IPCC 2005).

Previous health impact assessments of air pollution conducted in Europe, the USA, and Canada have grappled with these issues (Cohen et al. 2004; Le Tertre et al., 2005; USEPA 2005b; Burnett et al. 2005; Medina et al. 2009). Considering the lessons learned from these efforts, other relevant methodological work, and the context of EPHT programs based in state and local health departments, workshop participants developed the conclusions and recommendations summarized below. A full version of the workshop report and recommendations is available at http://www.cdc.gov/nceh/tracking/default.htm.

Major workshop conclusions and recommendations

  • Future progress in public health protection will likely require improved understanding of local air pollution sources and control measures that require the engagement of local stakeholders. By providing timely and locally relevant data on air pollution health impacts, the EPHTN can become a motivating factor in such engagement while making an important contribution to public health protection.

  • Analyses of PM2.5 and ozone health impacts that use only air pollution and health data from a single geographic area are unlikely to provide robust estimates of the relation between air pollution and acute and chronic health effects, although such analyses may be appropriate for other research or for surveillance applications (Talbot et al. 2009). Therefore, when making such estimates, various methods of “borrowing strength” from other evidence are essential, especially when quantifying local public health impacts (Fuentes 2009).

  • The EPHT program should incrementally develop tracking of air quality health impacts (Fig. 1).

    Fig. 1
    figure 1

    Conceptual model for staged development of air pollution health impact assessment for environmental public health tracking

  • An initial goal should be the development, testing, and application of a methodology for local health impact assessment at selected locations and that methodology should use quantitative estimates of the concentration–response relationships between air pollution exposure and health outcomes from the scientific literature (Hubbell et al. 2009). Clear operational guidance for applying the method and communicating results should be provided.

  • In the longer term, a network of EPHT programs should develop analyses that draw strength from pooled evidence across locations to produce cross-sectional estimates of local CR functions (Fuentes 2009). Once established, this type of network would enable longitudinal analyses that track impact over time, which may identify changes in CR functions attributable, for example, to changes in PM2.5 composition. Such analyses could support “accountability” measures (Shin et al. 2009; HEI Accountability Working Group 2003) of the effectiveness of local, regional, and national air quality management initiatives.

  • Developing local CR function estimates that describe, at the state and sub-state levels, the relationship between air pollution exposure and health will initially require considerably more methodological groundwork, as discussed by Fuentes (2009), than will health impact assessments using published CR function estimates. In addition, this approach requires a process that ensures standard methods for data preparation and analysis across states while addressing the confidentiality requirements of data stewards in each state.

  • The EPHT should review the current experience with regard to choice of air pollution health impact metrics (e.g., counts of attributable deaths and other adverse health outcomes vs. attributable impact on life expectancy or healthy life expectancy) and the way in which they are communicated, with the goals of (1) achieving consensus on the best approaches for the EPHTN and (2) identifying critical knowledge gaps that could be addressed with additional research or methods development. Currently, the scientific community has yet to resolve which metric(s) best quantify the impacts of air pollution and which are most meaningful to diverse stakeholders, including the public (McMichael et al. 1998; Rabl 2005; Brunekreef et al. 2007; Wartenberg 2009).

  • A critical but challenging goal for EPHTN will be to provide a complete and straightforward account of the uncertainties in local air pollution health impact estimates. The CDC Environmental Public Health Tracking Program should review the current experience with regard to efforts to communicate uncertainty in estimates of health effects and health impact assessments of environmental hazards (e.g., IPCC 2005), with the goals of (1) achieving consensus on the best approaches for the EPHTN; (2) identifying critical knowledge gaps that could be addressed with additional research; and (3) supporting the overall transparency and credibility of the tracking network’s results.

  • Communication among different agencies conducting health impact assessments is also needed. Currently, in the USA, health impact assessments of exposure to air pollution are carried out by agencies at different levels, including US EPA, states, and municipalities. The tracking program should work with involved agencies to avoid, if possible, methodologic inconsistencies that could produce artifactual differences in impact estimates. A well-developed communication strategy regarding health effects and impacts of air pollution, coordinated with other relevant agencies such as US EPA, should be an integral part of the EPHT.