You are here

Assessing adverse health effects of long-term exposure to low levels of ambient pollution

Principal Investigators: 

Harvard T.H. Chan School of Public Health


Harvard T.H. Chan School of Public Health

This study will examine health effects of low levels of air pollution in the US using data from about ~56 million people enrolled in Medicare and Medicaid. In addition, they will develop new causal modeling methods to characterize the shape of the exposure-response function. See also this Program Summary of HEI's research program on low levels of air pollution. 

Funded under

Abstracts for the 2019 HEI Annual Conference. Please scroll down to view two abstracts and posters. 

Assessing the short-term effect of PM2.5 on cardiovascular hospitalizations in the Medicaid population: a case-crossover study

Priyanka deSouza1, Danielle Braun2,3(Co-presenter), Francesca Dominici2, Marianthi-Anna Kioumourtzoglou4 (Co-presenter)

1Massachusetts Institute of Technology, Cambridge, MA, USA; 2Harvard T.H. Chan School of Public Health, Boston, MA, USA; 3Dana-Farber Cancer Institute, Boston, MA, USA; 4Columbia University Mailman School of Public Health, New York, NY, USA

Background. Many studies have demonstrated an association between short-term exposure to fine particulate matter (PM2.5) and cardiovascular disease (CVD). This association has been shown to vary by socioeconomic status (SES), in space (e.g., by region) and time (e.g., by season). We conducted a national epidemiologic analysis using US-wide Medicaid data to estimate the association between short-term PM2.5 and CVD hospitalizations for the years 2010 and 2011 among low-income and disabled Americans.

Method. We applied a time-stratified case-crossover design to estimate the association between short-term PM2.5 (measured as the average of PM2.5 on the day of the event and the preceding day), and the rate of total and cause-specific CVD hospitalizations. We also examined this association for days with PM2.5 levels lower than 25 μg/m3, the current guideline for daily PM2.5 concentrations by the World Health Organization (WHO). We used the Medicaid eligibility criteria to identify the subpopulation with a disability, and report the association between PM2.5 and CVD hospitalization for this subpopulation. Finally, we compared the association between PM2.5 and CVD in the Medicaid population with the association in the Medicare population for the same years.

Results. We observed an increase of 1.4% (95% CI: 0.8, 2.0%) in total CVD hospitalizations,  1.1% (95% CI: -0.8, 3.1%) in acute myocardial infarctions, 2.6% (95% CI: -3.1, 8.7%) in ischemic heart diseases, 2.6% (95% CI: 1.1, 4.1%)  in congestive heart failure, and 1.3% (95% CI: -0.5, 3.2%) in ischemic strokes for each increase of 10 μg/m3 in PM2.5. The estimated association between PM2.5 and CVD hospitalizations among Medicaid enrollees with a disability was 1.3% (95% CI: 0.4, 2.3%).  The estimated association between PM2.5 and CVD hospitalizations, only considering days where PM2.5 levels were less than 25 μg/m3 was 1.7% (95% CI: 1.0, 2.4%). The estimated association between PM2.5 and CVD hospitalizations for the Medicaid population over 65 years of age was 2.2% (95% CI: 1.1, 3.3%) The corresponding estimated association in the Medicare population was 1.0% (95% CI: 0.7, 1.3%).These associations were robust in sensitivity analyses.

Conclusions. Our analyses showed an increased rate of CVD hospitalizations associated with short-term PM2.5 exposure in the Medicaid population. This association in the Medicaid population older than 65 years old was greater than the association in the Medicare population, albeit the confidence intervals of the two estimates were overlapping. This finding indicates increased vulnerability among the low-income and disabled elderly. In addition, we found that this association was still statistically significant at PM2.5 levels below the WHO guidelines for 24-hour PM2.5 concentrations.


Poster by deSouza, Dominici et al., 2019 HEI Annual Conference

An Ensemble Model of PM2.5 Concentration across the Contiguous United States with High Spatiotemporal Resolution

Qian Di1 (Presenter), Heresh Amini1, Itai Kloog2, Rachel Silvern3, James Kelly4; M. Benjamin Sabath1, Christine Choirat1, Petros Koutrakis1, Alexei Lyapustin5, Yujie Wang1, Joel Schwartz1, Francesca Dominici1

1Harvard T.H. Chan School of Public Health, Boston, MA, USA; 2Ben-Gurion University of the Negev, Beer Sheva, Israel; 3Harvard University, Cambridge, MA, USA; 4U.S. Environmental Protection Agency, Research Triangle Park, NC, USA; 5NASA Goddard Space Flight Center, Greenbelt, MD, USA

Background. Various approaches have been proposed to model PM2.5 in the last decade, with satellite-derived aerosol optical depth, land-use variables, chemical transport model predictions, and several meteorological variables as major predictor variables. The training methods have been evolving from simple linear regressions to more complex machine learning algorithms.

Methods. Our study used an ensemble model that integrated multiple machine learning algorithms and predictor variables to estimate daily PM2.5 at a resolution of 1 km ´ 1 km across the contiguous United States. We used a generalized additive model that accounted for geographic difference to combine PM2.5 estimates from neural network, random forest, and gradient boosting models. The three machine learning algorithms were based on multiple predictor variables, including satellite data, meteorological variables, land-use variables, elevation, chemical transport model predictions, and several reanalysis datasets.

Results. The model training results from 2000 to 2015 indicate good model performance with a 10-fold cross-validated R2 of 0.86 for daily PM2.5 predictions. For annual PM2.5 estimates, the cross-validated R2 was 0.89. Our model demonstrated good performance up to 60 mg/m3. Using the trained PM2.5 model and predictor variables, we predicted daily PM2.5 from 2000 to 2015 at every 1 km ´ 1 km grid cell in the contiguous United States. We also used localized land-use variables within 1 km ´ 1 km grids to downscale PM2.5 predictions to 100 m ´ 100 m grid cells. To characterize uncertainty, we used meteorological variables, land-use variables, and elevation to model the monthly standard deviation of the difference between daily monitored and predicted PM2.5 for every 1 km ´ 1 km grid cell.

Conclusions. Our PM2.5 prediction dataset, including the downscaled and uncertainty predictions, will allow epidemiologists to accurately estimate the adverse health effects of PM2.5. Based on the model performance of individual learners, we also conclude that model performance of PM2.5 models is based on context, and the best training algorithm to fit PM2.5 globally does not exist.


Poster by Di, Dominici et al., 2019 HEI Annual Conference



Di Q, Wang Y, Zanobetti A, Wang Y, Koutrakis P, Choirat C, Dominici F, Schwartz JD. Air Pollution and Mortality in the Medicare Population. N Engl J Med. 2017 Jun 29;376(26):2513-2522. doi: 10.1056/NEJMoa1702747.

Di Q, Dai L, Wang Y, Zanobetti A, Choirat C, Schwartz JD, Dominici F. Association of Short-term Exposure to Air Pollution With Mortality in Older Adults. JAMA. 2017 Dec 26;318(24):2446-2456. doi: 10.1001/jama.2017.17923.

Dominici F, Zigler C. Best Practices for Gauging Evidence of Causality in Air Pollution Epidemiology. Am J Epidemiol. 2017 Dec 15;186(12):1303-1309. doi: 10.1093/aje/kwx307.

Makar M1, Antonelli J, Di Q, Cutler D, Schwartz J, Dominici F. Estimating the Causal Effect of Low Levels of Fine Particulate Matter on Hospitalization. Epidemiology. 2017 Sep;28(5):627-634. doi: 10.1097/EDE.0000000000000690.

Xiao Wu, Danielle Braun, Marianthi-Anna Kioumourtzoglou, Christine Choirat, Qian Di, Francesca Dominici. Causal inference in the context of an error prone exposure: air pollution and mortality. Submitted to Annals of Applied statistics. Version June 28, 2018.