Lee lecturer connects climate and disease

In another instance of your mother being right, the presenter of this year’s Richard V. Lee, MD, Lectureship in Global Health showed how the weather — specifically absolute humidity — can affect our health.

Jeffrey Shaman is director of the Climate and Health Program at Columbia University’s Mailman School of Public Health and faculty chair of Columbia’s Earth Institute. He studies how infectious diseases survive and transmit, including how things like climate and weather can affect those processes and human health.

Shaman discussed his recent work looking at the connection between the flu and humidity and temperature. The fact that flu outbreaks peak in wintertime has led researchers over the years to hypothesize why that might be. Among the theories is one related to the lower temperatures and lower humidity of winter air, an idea that interests Shaman.

“Influenza variants are subject to those environmental conditions,” Shaman explained, “some of which may be more conducive to the survival and transmission of the virus and others less so.” His studies in that arena indicate that in winter conditions of low absolute humidity — the amount of water vapor in the air — flu survives longer and transmits better. That helps explain why flu season occurs when the weather is colder, as opposed to during the summer. Shaman said his next question was whether he could use observed humidity conditions at population levels to mathematically model a flu season — and thus predict seasonal flu outbreaks.

Shaman developed his model and populated it with data from the past 31 years on humidity conditions in New York, Washington, Florida, Arizona and Illinois. The result reproduced “the cycle of influenza seasonality” in all five states. The next step was finding out whether the model could predict individual flu outbreaks. He discovered that the model “would do a terrible job” due to the highly irregular nature of outbreaks.

But here’s where Shaman’s expertise in climate comes in. He wondered if weather prediction could inform a better model. Weather, he noted, “is another system where the dynamics of it are non-linear and highly irregular. Yet, we generate weather forecasts with reasonable lead times … that have real reliability.” Could Shaman build a system that mimics weather prediction, but apply it to influenza prediction?

To find out, Shaman combined three ingredients analogous to those used in weather prediction:

• A simple model describing the transmission dynamics of flu in a population, including the effect of absolute humidity.
• Real-time (or nearly real-time) observations — in this case, flu-infection rates based on reported and estimated cases of flu and flu-like illnesses in Salt Lake City in 2013.
• Statistical data-simulation methods that allowed Shaman to combine the observations with the model to correct for errors in the model.

Numerous simulations showed a system that almost consistently predicted the peak of flu outbreaks in Salt Lake City five weeks into the future. Further refinement of the system allowed it to show how certain the forecast would be, akin to when weather forecasters “tell you there's a 90% or … a 20% chance of rain tomorrow.” And, just as with weather predictions, Shaman found the likelihood of good forecasts eroded over time.

Another aspect of Shaman’s work has significance for the COVID-19 pandemic. He and his colleagues conducted a field study called the Virome of Manhattan, in which their aim was to improve their flu forecast by understanding the nitty-gritty of respiratory illnesses. They studied a group of 200 people who reported daily their common cold symptoms like runny nose, cough and chills. The group also had weekly tests for common respiratory viruses.

Ultimately, the study found that about a quarter of people with flu saw a doctor; the numbers were lower for RSV and coronaviruses. It also found that most people had no symptoms, yet they were shedding detectable virus and probably contagious. In effect, he added,” the fact that most infections are undocumented and mild or asymptomatic means that these viruses can get around. It’s why common respiratory viruses are common.”

Shaman and his team in January 2020 got word of a newly emerging virus coming out of Wuhan as its epicenter and saw it very quickly spreading throughout China.

“Then it is hopping on airplanes and going to Thailand and Japan and South Korea and the U.S.,” he said. Because of their work on the Virome project, the team immediately realized that COVID-19 was behaving like a common respiratory virus. “Most of the people who are infected probably don't know they have it,” he said.

They built a system to determine if that was, indeed, the case, coupling travel records from a common GPS app in China and observations of confirmed cases of COVID-19 from 375 Chinese cities. Their system estimated that 86% of COVID-19 infections were falling in the undocumented category, “just like a common respiratory illness.” They also saw evidence of a two-to-three-day period before symptoms when people were already contagious. Their conclusion? “This virus is not going to be stopped. It has epidemiological properties consistent with common respiratory illnesses. And that characteristic of doing all this undocumented, undetected infection means that you're not going to be able to constrain its transmission.”

Building a true forecasting model for COVID, however, has proven difficult, as guidelines and regulations for social distancing and mask-wearing continue to change.

Happily, Shaman’s flu forecast model is now in use with organizations like the Centers for Disease Control and Prevention, municipalities and even other countries, helping to make a difference in preparing and protecting people from illnesses that are as inevitable as the weather.

The Richard V. Dr. Lee, MD, Lectureship in Global Health is held annually to honor the former UB faculty member. It's supported by an endowment established by a longtime friend of Lee. It is presented by the School of Public Health and Health Professions.