Breathing fine particles suspended in the air is harmful for everyone—and can kill those with cardiovascular or respiratory vulnerabilities, a fact known since the 1990s. Now a study of 95 million Medicare hospitalization claims from 2000 to 2012 links as many as 12 additional diseases, including kidney failure, urinary tract and blood infections, and fluid and electrolyte disorders, to such fine-particle air pollution for the first time. The research demonstrates that even small, short-term increases in exposure can be harmful to health, and quantifies the economic impact of the resulting hospitalizations and lives lost.
Fine particles (known as PM2.5 because they are smaller than 2.5 microns in diameter) can slip past the human respiratory system’s copious mucosal defenses in the nose and upper airways. These tiny byproducts of combustion, principally of fossil fuels such as coal and oil, land in the thin-membraned alveolar sacs deep in the lungs where oxygen exchange occurs. From there, they can pass into the blood. But the full extent of the systemic harm they cause is not well understood, explains principal investigator Francesca Dominici, Gamble professor of biostatistics, population, and data science and co-director of Harvard’s Data Science Initiative. Joel Schwartz, professor of environmental epidemiology and senior author of the BMJ (formerly the British Medical Journal) paper elaborates: “We wanted to shed further light on the risks of exposure to short-term air pollution by searching for links between such pollution and all diseases that are plausible causes of hospitalizations.”
To do so, the researchers used a “big-data” approach, aligning Medicare-patient hospital admissions by time and geography with known levels of PM2.5 pollution on the previous day. That information was modeled using satellite and temperature data, and verified with actual measurements from thousands of ground-based monitoring stations. The approach sounds simple, but Dominici explains otherwise: she built the data platform on which the study relies “one year at a time” during the past 20 years of her career. The construction of such a research tool, which harmonizes the data using standard values and definitions across multiple sources and data attributes, requires “enormous intellectual input” and many financial resources that are “hard to find,” she says, because most grant-making organizations, including the federal government, “don’t put their money” into data platforms. “They are not considered sexy or intellectually important.”
In fact, this foundational work is critical for gathering, linking, curating, and harmonizing the data (which come from more than two dozen different government databases). In this study, for example, satellite data for air pollution is provided in one-square-kilometer grids, whereas temperature and weather data cover areas of 32 square kilometers. And patient zip code records must be aligned with both these measures. Overcoming such hurdles so that the data sources can all talk to each other is challenging, but worthwhile, says Dominici: big-data approaches are “changing the paradigm for scientific investigation. Without the data platform, we would not have had the statistical power to run an analysis on all possible diseases” or “test many hypotheses at the same time.”
To the known effects on the cardiovascular, respiratory, and nervous systems, the team’s work has added pathologies of the blood, gut, skin, and kidney. “Identifying all of the diseases that fine-particle pollution is affecting,” Dominici continues, “will shed light on what additional experiments need to be done” to better understand the mechanisms of disease and methods for prevention.
The research demonstrates that every microgram-per-cubic-meter increase in PM2.5 within a 24-hour period has an incremental effect on human health. Even when starting from zero (perfectly clean air), each such increase of one microgram in concentration was associated with an annual increase of 634 deaths and 5,692 hospitalizations, as well as 32,314 patient-days in hospital. In the United States, such increases in pollution occur on more than 122 days a year in every geographic region. In lay terms, says Dominici, this represents “one additional hospitalization per day for every zip code for half of the year.” These data correspond to $100 million in annual inpatient and post-acute care costs, and an estimated $6.5 billion in lost value of human life. In the United States, where fine particle pollution began rising again in 2016 after a long-term decline, Environmental Protection Agency regulations specify that exposures greater than 35 micrograms of PM2.5 or less in a 24-hour period are unhealthy. (The target for annual, or long-term, limit of exposure, averaged over three years, is 12 micrograms per cubic meter.)
Dominici notes that these findings understate the economic impact of fine-particle pollution in several ways. The study captured only effects that led to hospitalization, and not prior visits to a doctor or the costs of previously prescribed medication, or costs after discharge (including readmission, outpatient, and drug costs). And it captured only fee-for-service patients, not those covered in HMOs. It didn’t capture effects on mental health, or in fact “any encounter with the healthcare system that did not lead to a billing record in the hospital.” And it did not capture data for anyone under the age of 65 (Medicare covers patients 65 and older). As a result, she says, “It is probably a very big underestimate.”
Older people may be more vulnerable to air pollution than young people with healthy immune systems, she says, but everyone is affected: “The entire body of epidemiological evidence for cardiovascular and pulmonary disease shows that these effects are present at all ages. Whether this will also be true for these new diseases, we don’t know yet.” Dominici and colleagues plan a similar study using Medicaid data (which covers patients of all ages).
Despite the study’s likely underestimate of economic impacts, the authors write that it “could provide information for assessing the cost effectiveness of air pollution interventions in other high income countries such as the United Kingdom, Canada, and Germany,” among others, “where the burdens of disease, population demographics, and healthcare use are similar to those of the U.S.” Dominici adds that there is another common benefit to controlling PM2.5 that is often overlooked in discussions of air pollution: “There is a direct linkage between the sources of fine particulate matter and the sources of greenhouse gases. Most are the same.”