Elena Austin

A framework to spatially cluster air pollution monitoring sites in US based on the PM2.5 composition

BACKGROUND Heterogeneity in the response to PM2.5 is hypothesized to be related to differences in particle composition across monitoring sites which reflect differences in source types as well as climatic and topographic conditions impacting different geographic locations. Identifying spatial patterns in particle composition is a multivariate problem that requires novel methodologies. OBJECTIVES Use cluster analysis methods to identify spatial patterns in PM2.5 composition. Verify that the resulting clusters are distinct and informative. METHODS 109 monitoring sites with 75% reported speciation data during the period 2003-2008 were selected. These sites were categorized based on their average PM2.5 composition over the study period using k-means cluster analysis. The obtained clusters were validated and characterized based on their physico-chemical characteristics, geographic locations, emissions profiles, population density and proximity to major emission sources. RESULTS Overall 31 clusters were identified. These include 21 clusters with 2 or more sites which were further grouped into 4 main types using hierarchical clustering. The resulting groupings are chemically meaningful and represent broad differences in emissions. The remaining clusters, encompassing single sites, were characterized based on their particle composition and geographic location. CONCLUSIONS The framework presented here provides a novel tool which can be used to identify and further classify sites based on their PM2.5 composition. The solution presented is fairly robust and yielded groupings that were meaningful in the context of air-pollution research.

A Framework for Identifying Distinct Multipollutant Profiles in Air Pollution Data

BACKGROUND The importance of describing, understanding and regulating multi-pollutant mixtures has been highlighted by the US National Academy of Science and the Environmental Protection Agency. Furthering our understanding of the health effects associated with exposure to mixtures of pollutants will lead to the development of new multi-pollutant National Air Quality Standards. OBJECTIVES Introduce a framework within which diagnostic methods that are based on our understanding of air pollution mixtures are used to validate the distinct air pollutant mixtures identified using cluster analysis. METHODS Six years of daily gaseous and particulate air pollution data collected in Boston, MA were classified solely on their concentration profiles. Classification was performed using k-means partitioning and hierarchical clustering. Diagnostic strategies were developed to identify the most optimal clustering. RESULTS The optimal solution used k-means analysis and contained five distinct groups of days. Pollutant concentrations and elemental ratios were computed in order to characterize the differences between clusters. Time-series regression confirmed that the groups differed in their chemical compositions. The mean values of meteorological parameters were estimated for each group and air mass origin between clusters was examined using back-trajectory analysis. This allowed us to link the distinct physico-chemical characteristics of each cluster to characteristic weather patterns and show that different clusters were associated with distinct air mass origins. CONCLUSIONS This analysis yielded a solution that was robust to outlier points and interpretable based on chemical, physical and meteorological characteristics. This novel method provides an exciting tool with which to identify and further investigate multi-pollutant mixtures and link them directly to health effects studies.

Health effects of multi-pollutant profiles.

BACKGROUND The association between exposure to particle mass and mortality is well established; however, there are still uncertainties as to whether certain chemical components are more harmful than others. Moreover, understanding the health effects associated with exposure to pollutant mixtures may lead to new regulatory strategies. OBJECTIVES Recently we have introduced a new approach that uses cluster analysis to identify distinct air pollutant mixtures by classifying days into groups based on their pollutant concentration profiles. In Boston during the years 1999-2009, we examined whether the effect of PM2.5 on total mortality differed by distinct pollution mixtures. METHODS We applied a time series analysis to examine the association of PM2.5 with daily deaths. Subsequently, we included an interaction term between PM2.5 and the pollution mixture clusters. RESULTS We found a 1.1% increase (95% CI: 0.0, 2.2) and 2.3% increase (95% CI: 0.9-3.7) in total mortality for a 10 μg/m(3) increase in the same day and the two-day average of PM2.5 respectively. The association is larger in a cluster characterized by high concentrations of the elements related to primary traffic pollution and oil combustion emissions with a 3.7% increase (95% CI: 0.4, 7.1) in total mortality, per 10 μg/m(3) increase in the same day average of PM2.5. CONCLUSIONS Our study shows a higher association of PM2.5 on total mortality during days with a strong contribution of traffic emissions, and fuel oil combustion. Our proposed method to create multi-pollutant profiles is robust, and provides a promising tool to identify multi-pollutant mixtures which can be linked to the health effects.

Estimating Causal Associations of Fine Particles With Daily Deaths in Boston

Many studies have reported associations between daily particles less than 2.5 µm in aerodynamic diameter (PM2.5) and deaths, but they have been associational studies that did not use formal causal modeling approaches. On the basis of a potential outcome approach, we used 2 causal modeling methods with different assumptions and strengths to address whether there was a causal association between daily PM2.5 and deaths in Boston, Massachusetts (2004-2009). We used an instrumental variable approach, including back trajectories as instruments for variations in PM2.5 uncorrelated with other predictors of death. We also used propensity score as an alternative causal modeling analysis. The former protects against confounding by measured and unmeasured confounders and is based on the assumption of a valid instrument. The latter protects against confounding by all measured covariates, provides valid estimates in the case of effect modification, and is based on the assumption of no unmeasured confounders. We found a causal association of PM2.5 with mortality, with a 0.53% (95% confidence interval: 0.09, 0.97) and a 0.50% (95% confidence interval: 0.20, 0.80) increase in daily deaths using the instrumental variable and the propensity score, respectively. We failed to reject the null association with exposure after the deaths (P =0.93). Given these results, prior studies, and extensive toxicological support, the association between PM2.5 and deaths is almost certainly causal.

PM2.5 and survival among older adults: effect modification by particulate composition.

Background: Fine particulate (PM2.5) air pollution has been consistently linked to survival, but reported effect estimates are geographically heterogeneous. Exposure to different types of particle mixtures may explain some of this variation. Methods: We used k-means cluster analyses to identify cities with similar pollution profiles, (ie, PM2.5 composition) across the United States. We examined the impact of PM2.5 on survival, and its variation across clusters of cities with similar PM2.5 composition, among Medicare enrollees in 81 US cities (2000–2010). We used time-varying annual PM2.5 averages, measured at ambient central monitoring sites, as the exposure of interest. We ran by-city Cox models, adjusting for individual data on previous cardiopulmonary-related hospitalizations and stratifying by follow-up time, age, gender, and race. This eliminates confounding by factors varying across cities and long-term trends, focusing on year-to-year variations of air pollution around its city-specific mean and trend. We then pooled the city-specific effects using a random effects meta-regression. In this second stage, we also assessed effect modification by cluster membership and estimated cluster-specific PM2.5 effects. Results: We followed more than 19 million subjects and observed more than 6 million deaths. We found a harmful impact of annual PM2.5 concentrations on survival (hazard ratio = 1.11 [95% confidence interval = 1.01, 1.23] per 10 &mgr;g/m3). This effect was modified by particulate composition, with higher effects observed in clusters containing high concentrations of nickel, vanadium, and sulfate. For instance, our highest effect estimate was observed in cities with harbors in the Northwest, characterized by high nickel, vanadium, and elemental carbon concentrations (1.9 [1.1, 3.3]). We observed null or negative associations in clusters with high oceanic and crustal particles. Conclusions: To the best of our knowledge, this is the first study to examine the association between PM2.5 composition and survival. Our findings indicate that long-term exposure to fuel oil combustion and power plant emissions have the highest impact on survival.

Ozone trends and their relationship to characteristic weather patterns.

Local trends in ozone concentration may differ by meteorological conditions. Furthermore, the trends occurring at the extremes of the Ozone distribution are often not reported even though these may be very different than the trend observed at the mean or median and they may be more relevant to health outcomes. Classify days of observation over a 16-year period into broad categories that capture salient daily local weather characteristics. Determine the rate of change in mean and median O3 concentrations within these different categories to assess how concentration trends are impacted by daily weather. Further examine if trends vary for observations in the extremes of the O3 distribution. We used k-means clustering to categorize days of observation based on the maximum daily temperature, standard deviation of daily temperature, mean daily ground level wind speed, mean daily water vapor pressure and mean daily sea-level barometric pressure. The five cluster solution was determined to be the appropriate one based on cluster diagnostics and cluster interpretability. Trends in cluster frequency and pollution trends within clusters were modeled using Poisson regression with penalized splines as well as quantile regression. There were five characteristic groupings identified. The frequency of days with large standard deviations in hourly temperature decreased over the observation period, whereas the frequency of warmer days with smaller deviations in temperature increased. O3 trends were significantly different within the different weather groupings. Furthermore, the rate of O3 change for the 95th percentile and 5th percentile was significantly different than the rate of change of the median for several of the weather categories.We found that O3 trends vary between different characteristic local weather patterns. O3 trends were significantly different between the different weather groupings suggesting an important interaction between changes in prevailing weather conditions and O3 concentration.

The impact of multipollutant clusters on the association between fine particulate air pollution and microvascular function.

Background: Prior studies including the Framingham Heart Study have suggested associations between single components of air pollution and vascular function; however, underlying mixtures of air pollution may have distinct associations with vascular function. Methods: We used a k-means approach to construct five distinct pollution mixtures from elemental analyses of particle filters, air pollution monitoring data, and meteorology. Exposure was modeled as an interaction between fine particle mass (PM2.5), and concurrent pollution cluster. Outcome variables were two measures of microvascular function in the fingertip in the Framingham Offspring and Third Generation cohorts from 2003 to 2008. Results: In 1,720 participants, associations between PM2.5 and baseline pulse amplitude tonometry differed by air pollution cluster (interaction P value 0.009). Higher PM2.5 on days with low mass concentrations but high proportion of ultrafine particles from traffic was associated with 18% (95% confidence interval: 4.6%, 33%) higher baseline pulse amplitude per 5 &mgr;g/m3 and days with high contributions of oil and wood combustion with 16% (95% confidence interval: 0.2%, 34%) higher baseline pulse amplitude. We observed no variation in associations of PM2.5 with hyperemic response to ischemia observed across air pollution clusters. Conclusions: PM2.5 exposure from air pollution mixtures with large contributions of local ultrafine particles from traffic, heating oil, and wood combustion was associated with higher baseline pulse amplitude but not hyperemic response. Our findings suggest little association between acute exposure to air pollution clusters reflective of select sources and hyperemic response to ischemia, but possible associations with excessive small artery pulsatility with potentially deleterious microvascular consequences.

Use of mobile and passive badge air monitoring data for NOX and ozone air pollution spatial exposure prediction models.

Air pollution exposure prediction models can make use of many types of air monitoring data. Fixed location passive samples typically measure concentrations averaged over several days to weeks. Mobile monitoring data can generate near continuous concentration measurements. It is not known whether mobile monitoring data are suitable for generating well-performing exposure prediction models or how they compare with other types of monitoring data in generating exposure models. Measurements from fixed site passive samplers and mobile monitoring platform were made over a 2-week period in Baltimore in the summer and winter months in 2012. Performance of exposure prediction models for long-term nitrogen oxides (NOX) and ozone (O3) concentrations were compared using a state-of-the-art approach for model development based on land use regression (LUR) and geostatistical smoothing. Model performance was evaluated using leave-one-out cross-validation (LOOCV). Models performed well using the mobile peak traffic monitoring data for both NOX and O3, with LOOCV R2s of 0.70 and 0.71, respectively, in the summer, and 0.90 and 0.58, respectively, in the winter. Models using 2-week passive samples for NOX had LOOCV R2s of 0.60 and 0.65 in the summer and winter months, respectively. The passive badge sampling data were not adequate for developing models for O3. Mobile air monitoring data can be used to successfully build well-performing LUR exposure prediction models for NOX and O3 and are a better source of data for these models than 2-week passive badge data.

Correction: Laboratory Evaluation of the Shinyei PPD42NS Low-Cost Particulate Matter Sensor.

The y-axis label for Fig 5 incorrectly states mean rather than difference. The correct y-axis is “Difference of Shinyei and APS Mass (μg/m3)”. The authors have provided a corrected version of Fig 5 here. Fig 5 Bland Altman Plots (4 sensors pooled together).

Using the design-feedback iterative cycle to improve the usability of a personal air pollution monitor (Preprint)

Background There is considerable evidence that exposure to fine particulate matter (PM2.5) air pollution is associated with a variety of adverse health outcomes. However, true exposure-outcome associations are hampered by measurement issues, including compliance and exposure misclassification. Objective This paper describes the use of the design-feedback iterative cycle to improve the design and usability of a new portable PM2.5 monitor for use in an epidemiologic study of personal air pollution measures. Methods In total, 10 adults carried on their person a prefabricated PM2.5 monitor for 1 week over 3 waves of the iterative cycle. At the end of each wave, they participated in a 30-minute moderated focus group and completed 2 validated questionnaires on usability and views on research. The topics addressed included positives and negatives of the monitor, charging and battery life, desired features, and changes to the monitor from each previous wave. They also completed a log to record device wear time each day. The log also provided space to record any issues that may have arisen with the device or for general comments during the week of collection. Results The major focus group topics included device size, noise, battery and charge time, and method for carrying the device. These topics formed the basis of iterative design changes; by the final cycle, the device was reasonably smaller, quieter, held a longer charge, and was more convenient to carry. System usability scores improved systematically across each wave (median scores of 50-66 on a 100-point scale), as did median daily wear time (approximately 749-789 minutes). Conclusions Both qualitative and quantitative measures showed an improvement in device usability over the 3 waves. This study demonstrates how the design-feedback iterative cycle can be used to improve the usability of devices manufactured for use in large epidemiologic studies on personal air pollution exposures.