Population Study Shows Flu Flourishes When Air Dries Out

For decades, scientists have been searching for factors that may explain the timing and seasonality of common flu outbreaks. Some of the proposed factors include temperature, school terms and sun exposure. Recent reanalyses of laboratory data led by Jeffrey Shaman of Oregon State University showed that influenza viruses survive longer and transmit better when the air is dry, as measured by absolute humidity—the quantity of water vapor in the air. In temperate regions such as the United States, humidity is highest in the summer, when flu outbreaks are least common, and lowest in the winter, when flu season peaks. Because these trends parallel laboratory results, Shaman and Marc Lipsitch, professor of epidemiology and director of the Center for Communicable Disease Dynamics at HSPH, predicted that the humidity cycle could be a major factor in the onset of seasonal flu epidemics.

Low humidity heralds illness. In the United States, increased flu-related deaths are coupled with low humidity in the preceding weeks. In the graph, variations in absolute humidity (daily value minus daily average) are shown during the six weeks before and four weeks after an influenza outbreak. Onset was designated as the date after which excess influenza-related deaths had been above a particular threshold for two weeks. The onset date was defined by a different mortality threshold in each of the datasets (solid lines). A mathematical model was able to reproduce 1,181–1,420 epidemics, depending on the threshold used, out of a possible 1,470 (30 winters; 48 contiguous states plus D.C.). Image courtesy PLoS Biology.

In a paper published online Feb. 23 in PLoS Biology, Shaman, Lipsitch and collaborators showed that absolute humidity alone can drive the onset of influenza at the human population level. The researchers generated a mathematical model of flu transmission, in which the risk of transmission depended on absolute humidity, and fit it to data from five states. This model was able to reproduce the observed influenza epidemics of the other 43 contiguous states and the District of Columbia over a 30-year period. The researchers noted that in both the fitted model and real data, the onset of flu season was associated with periods that were drier than expected for the time of year (see figure).

Shaman and Lipsitch plan to determine whether a similar model could explain flu seasonality in the tropics, where humidity variation is less pronounced, and whether useful short-term forecasts of flu activity could be made based on weather patterns.

For more information, students may contact Marc Lipsitch at mlipsitc@hsph.harvard.edu.

Conflict Disclosure: Marc Lipsitch discloses consulting income from the Avian/Pandemic Flu Registry (Outcome Sciences, funded in part by Roche) and from Novartis Vaccines and Diagnostics. The other authors declare no competing interests.

Funding Sources: The National Institutes of Health; the Bill and Melinda Gates Foundation; the RAPIDD program of the Science and Technology Directorate, U.S. Department of Homeland Security; and the Fogarty International Center of the NIH. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.