Disease experts team up with Florida Museum of Natural History to create a forecast for West Nile virus

(Press-News.org) Key points

State and local officials in Florida maintain hundreds coops with what are referred to as sentinel chickens, which act as an early alarm system for the presence of mosquito-borne illnesses in an area. This alarm system just got an upgrade. An interdisciplinary team of experts, including a zoonotic disease specialist, a museum data scientist and a salamander biologist, have combined their skills and created a statistical model that accurately predicts the activity of West Nile virus in an area up to six months in advance. The model was trained using two decades of sentinel chicken data. The original data files were destroyed in a flood, and the study was only possible because a University of Florida professor kept a personal copy in his lab. GAINESVILLE, Fla. --- A new study has significant bearing on the hackneyed joke about chickens and their numerous reasons for crossing roads. In Florida, there’s a good chance that the chicken crossed the road because it had completed its year-long conscripted service as a disease sentinel, a sort of early alarm signal for mosquito virus activity across the state.

Mosquito control programs maintain hundreds of chicken coops from the Panhandle down through Miami. Once a week, throughout large portions of the year, officials take a small blood sample from each of several chickens and send them to the Florida Department of Health, where they’re tested for antibodies to common mosquito-borne diseases, such as West Nile virus and eastern equine encephalitis. If the results are positive, the state will warn residents to be especially wary of mosquitos until the danger is past.

There’s no way to know with certainty, but the state’s sentinel program, which has been ongoing for more than 40 years, has likely saved many lives. But the program has its limitations, the most obvious being that warning can only be given after the confirmed virus activity, by which time people may have already been exposed.

Researchers at the University of Florida want to change that. In the study, published in the journal Science of the Total Environment, they demonstrate that they can reliably predict elevated West Nile virus activity in chickens up to six months before it occurs, which they accomplished by training and testing a statistical model on 20 years’ worth of sentinel chicken data. Their results are an important first step toward forecasting West Nile virus across Florida.

“People have spent years testing these chickens, and one of the main reasons is to get to the point where you can make predictions,” said study co-author Rob Guralnick, curator of biodiversity informatics at the Florida Museum of Natural History. He described the results as an exclamation point to the work that’s been done up to this point.

With this and future models added to their toolkit, health officials may one day soon be able to supplement their reactionary mitigation responses to disease outbreaks with proactive preventative measures.

When health officials want to predict the next big arbovirus outbreak, they look at birds

The study is partially the product of a 2023 Research Opportunity Seed Fund awarded by the University of Florida to forecast a variety of mosquito-borne illnesses in the state, and it almost didn’t happen.

The sentinel records, which had been stored in a basement at the Florida Department of Health, were destroyed by a flood. Luckily, Jonathan Day — a scientist at the University of Florida who studied viruses transmitted by arthropods (called arboviruses) — had the foresight to keep his own copy in big, unwieldy binders.

“One of my colleagues had stored these weekly reports all through his career. They were called arbograms,” said senior author Lindsay Campbell with a straight face before losing her composure and laughing at the pun.

Campbell is an assistant professor at the Florida Medical Entomology Laboratory in UF’s department of entomology and nematology. After hearing about the flood, she offered to share the records she’d inherited from her colleague. She also gave the data a digital face-lift with help from Guralnick, whose specialties include efficiently digitizing avalanche-sized reams of analog information.

“I knew from working with the Florida Museum that there were tons of advances in digitizing collections and other types of data,” Campbell said. Together, they created what’s referred to as an informatics pipeline, which automatically converted and organized the chicken data. Manual checks were performed afterward to ensure the transition had gone smoothly.

It’s hard to understate just how important these data are. The sentinel program was already in place when West Nile virus first arrived in Florida in 2001, meaning scientists who study the disease have a detailed record of how the epidemic unfolded. Other states had to piece together how the disease spread from several disparate and less reliable sources.

This was true of New York City back in 1999. The first intimation residents of the city’s Queens borough had that anything was amiss was an unusually large number of dead and dying crows in June and July. A veterinarian directly observed a crow that was unable to walk straight, suggesting it suffered from a neurological illness.

Things happened quickly after that. Within a few months, several types of animals began exhibiting similar symptoms: a wild goose in Queens; multiple bird species — including flamingos and a bald eagle — at the Bronx Zoo; a cooper’s hawk in Connecticut.

Wildlife pathologists hastily worked to find the cause while, unknown to them, humans started getting sick as well. Eight people who lived within a 4-square-mile radius developed severe fevers, brain swelling, muscle weakness and paralysis. By late September, medical doctors and wildlife experts had an answer. The cases represented the first known outbreak of West Nile virus in North America. By then, several more people had contracted the illness, with 59 reported cases in all at summer’s end.

The same pattern soon played out in other states. Dead birds presaged the arrived of West Nile virus throughout 10 counties in New York, Connecticut and New Jersey the following year. The year after that, it had spread to 38 counties across 10 states. In 2002, West Nile virus exploded across the U.S. in what was then the largest outbreak of the disease ever recorded. Thousands of people got sick and hundreds died.

West Nile virus waxes and wanes with the seasons, but it’s remained in the United States since its initial arrival at the tail end of the millennium.

Once a zoonotic virus has become established in a population, it’s nearly impossible to get rid of it. Rather than attempting full-blown eradication, scientists and civil servants have worked hard to track the disease. In Florida, the same animals that once augured imminent outbreak now stand guard as sentinels, only this time it’s chickens instead of crows. Chickens are a good choice because although they can become infected with West Nile virus — and their cells produce antibodies in response — they do not develop symptoms, nor can they transmit the virus to others.

Even with this information at their disposal, the ability to accurately predict where a virus will be active at any point in the future is a difficult task. This is particularly true of zoonotic arboviruses and other zoonotic diseases, because they can often infect multiple hosts.

“West Nile virus is a good example of that, because there are so many birds and mosquitos that can be infected and likely play a role in the transmission cycle. That’s why it does such a good job when it moves into a new area,” Campbell said.

West Nile virus has been documented in 65 mosquito species in North America alone, and though not all of them transmit the virus in nature, it can be difficult to distinguish the active transmission routes from the dead ends. Plus, only some mosquitos feed on humans, but all of them have to get their blood from somewhere.

That means it’s theoretically possible for a mosquito that is only a menace to birds to infect, say, a crow, which could then pass on the pathogen to another species of mosquito that has fewer reservations about where it feeds. These generalist mosquitos, called bridge vectors, are responsible for the spillover of West Nile virus to humans and other animals.

Any model attempting to account for each individual component in this web of complicated interactions would invariably stall in a quagmire of intractable data. Instead, the authors devised something simpler to cut through the noise and find the signal.

Refined model accurately predicts past viral activity

Had the authors been doing this work in the early 2000s, they would have used a simple model that included components for where a disease was active and when that activity occurred. According to study lead author J. Alex Baecher — who formerly modeled the distribution of salamanders before turning his attention to arboviruses for a postdoctoral research appointment — these types of models are useful but limited, because they do not account for interactions.

“They’re essentially snapshots, but the world isn’t static,” he said. “Ecological systems are dynamic, and modeling these systems requires accounting for how conditions in one location and time influence another.”

To connect the dots, ecologists took a page from quantum mechanics and fluid dynamics, in which interaction and randomness factor heavily, and added a component to their models that accounted for these variables.

“This tool has dramatically expanded our capabilities,” Baecher said. “We’re now able to model disease dynamics with higher biological realism, incorporating nonlinear interactions and latent variables that better reflect real-world complexity.”

The authors fed their model with data that included precipitation and minimum and maximum temperatures from the years 2001 to 2019, along with information about land cover across Florida. Then they ran the numbers, and the model spat out its predictions, or — rather — postdictions, since it was actually computing the where and when West Nile virus activity would have shown up in the past.

The chicken data was a crucial component of the analysis. Rather than predicting the future and waiting around for the next outbreak to test their model’s accuracy, they made a 20-year disease hindcast, then used the chicken data to assess its performance.

The results did not disappoint. The retrospective predictions closely aligned with the number of chicken infections and accurately predicted elevated virus activity in general areas where cases occurred in not only humans but horses as well, which number among the many casualties of West Nile outbreaks.

The monthly predictions were the most accurate and had less uncertainty, but the seasonal predictions were still reliable. The combined results gave the authors a sense of which environmental conditions are immediately conducive to elevated virus activity and which work further out in time to set up the stage.

For the monthly predictions, overall higher minimum temperatures and precipitation in an area two months before officials tested chickens had a strong positive association with West Nile virus activity, and high maximum temperatures during the testing month had a negative association. Intermediate levels of precipitation six months prior to testing had the strongest signal at the seasonal scale.

Altogether, this information may give officials a clear picture of what to look for at multiple points ahead of time to help inform management decisions.

The authors said their results are a milestone, but not one to stop at. Rather, they see the model as a tool that can be further refined and added to.

“The bird part is a missing piece in the model, and human behavior is another missing piece, and mosquitos and where they are is another. What we’re really trying to do is wrap these together to understand the mechanics all the way up and down the chain, from the weather patterns that favor mosquitos to the transmission cycle in birds and how that affects humans,” Guralnick said.

Disease forecasting is one of the many ways natural history museums are branching into the health sciences. Pamela Soltis, a distinguished professor and curator at the Florida Museum, recently led a series of workshops funded by the National Science Foundation to encourage and empower museum collection managers and curators to form stronger ties with health officials and disease experts. Efforts to do so are currently underway through the aegis of One Health, an interdisciplinary framework adopted by the Centers for Disease Control and Prevention and multiple other agencies in the U.S. and internationally.

“The concept is that there should be integrated information from all sources, including wildlife, agricultural species and cultivated plants,” Soltis said.

When zookeepers, veterinarians and wildlife experts began to track down the cause of the 1999 West Nile outbreak, there were few direct lines of communication between them and the health professionals who recorded the same outbreak in humans. Both groups launched independent lines of inquiry and arrived at their conclusions at different times roughly two months after the outbreak began. Had they worked together, it may not have taken them as long.

Collectively, the world’s natural history museums hold about 3 billion specimens from across the tree of life, which have the potential to help track pathogens and trace the origin of outbreaks before they develop into pandemics, Soltis said.

She gave Sin Nombre virus, which causes hantavirus pulmonary disease and was first discovered in the U.S. in 1993, as an example. Rodents are the primary host for hantavirus, but humans can become infected by inhaling aerosols from rodent droppings and urine. Researchers tested tissue from newly collected rodents and those preserved at the Museum of Southwestern Biology, 27% of which were positive for the virus. They used their results to determine which species most likely acted as vectors and where the disease was most prominent.

According to Soltis, this is one of the many unanticipated uses of museum collections. Researchers initially began collecting specimens during the Renaissance to document Earth’s biodiversity, but developments like the discovery of DNA, the advent of CT scanning and advances in data science have since vastly expanded the utility of collections and resulted in a variety of seemingly unlikely interdisciplinary collaborations. Their use in disease prevention is an ongoing paradigm shift.

If you live in Florida, one way you can support disease forecast programs is to adopt your very own sentinel chicken! After remaining stalwart for a year in the service of human health, they deserve a little rest and relaxation.

Additional authors of the study are Ashay Anand, Amy Bauer, Yasmin Tavares and Yesenia Sánchez of the University of Florida; and James Thorson of the Alaska Fisheries Science Center at the National Oceanic and Atmospheric Administration.

END