Researchers utilize holistic approach to predict severity of influenza season

U.Va. researchers at the Biocomplexity Institute and Initiative are developing computational models to estimate and manage the biological, social and economic impacts of influenza

hs-flu-eklein

Spread by the influenza virus, the flu is a contagious respiratory illness associated with symptoms such as fever, cough, runny nose, vomiting and headaches. 

Emma Klein | Cavalier Daily

The winter months are often associated with an increase in influenza cases, and U.Va. researchers are working to track the spread and control of the infectious disease. The new initiative and similar projects at the Biocomplexity Institute and Initiative take a holistic approach to solve complex societal issues. They integrate the social, economic and biological aspects of these problems into computational methods that aid in management and planning.

“Our overall goal is to develop the mathematical and computational foundations to study the epidemic process and develop associated technologies to plan, respond, detect and intervene before and during seasonal and epidemic outbreaks of infectious diseases,” said Madhav Marathe, division director of the network systems science and advanced computing division of the BII. 

The project is associated with a CDC challenge to predict and forecast influenza. Marathe cited nascent collaborations with U.Va. Health and the Data Science Institute. Additionally, the Institute has general ties with the School of Medicine and faculty from other departments at the University. 

Furthermore, collaboration is an essential piece for this project at the Biocomplexity Institute and Initiative. Infectious diseases can be classified as a societal problem and, in turn, require attention from all scientific disciplines. Their teams encompass a variety of backgrounds rather than depending on one singular department.

“Our institute is a transdisciplinary team science-oriented organization, [as] we all work on a variety of different projects,” said Bryan Lewis, computational epidemiologist and co-principal investigator of the influenza project. “For this influenza initiative, for example, we have several different teams tackling different aspects of the work.” 

Influenza or the flu is a contagious respiratory illness caused by the influenza virus. It spreads through droplets, and a healthy individual can acquire the virus by touching an infected surface and then touching their mouth, eyes or nose. Symptoms include fever, cough, muscle or body aches, vomiting and headaches. In order to prevent illness, the CDC recommends that individuals get vaccinated, avoid contact with those who are sick and frequently wash their hands. 

Flu season is common during the fall and winter months. Although the virus occurs year-round, there is usually a peak in activity around December and February. According to the CDC, 3 percent to 11 percent of the population are reported to have the flu, depending on the season.

Researchers at the Biocomplexity Institute and Initiative are developing methods to analyze the patterns associated with flu season. In particular, they are creating computational models to forecast and control outbreaks. According to Marathe, the models integrate the influence of social networks on the spread of influenza by creating representations of cities. 

“Our group was one of the first groups that even articulated the role of social networks in understanding diseases such as influenza,” Marathe said. “[They] spread because of social contacts…. These networks capture how in a city might be moving around, meeting others and doing their day to day task.” 

Srini Venkatramanan, computer scientist and co-principal investigator of the influenza project, detailed the process of creating the models. First, researchers break down their research question into multiple hypotheses. The first phase of the project includes data collection, and then researchers create computational models in which they build synthetic representations of society. Finally, they run the model to validate it. 

There are many applications for the developed computational models. For example, Venkatramanan detailed work with AccuWeather in influenza forecasting. This feature would allow short term, realistic predictions about impacted populations. Marathe added that the application would inform individuals of the prevalence of flu in the region. 

“These are projects on influenza forecasting,”  Venkatramanan said. “This one is a short-term realistic way of how seasonal influenza is going on here and what would happen in the next four weeks … We look at the spread of it and short term forecasts, and we make these forecasts on a weekly basis.” 

Additionally, the models can be accurate measures for public health measures. They can be used to manage vaccine allocation and distribution. Also, predictions can be used for hospital management.  

“If you were managing the hospital system and anticipate a surge in flu in the next work, you can manage the use of face masks, ventilators or beds,” said Marathe.

Furthermore, researchers have applied their findings to identify patterns regarding the 2019-20 flu season. In fact, based on their project associated with AccuWeather, the current season has been more active than normal. Influenza B has been the leading strand of the virus nationwide.

Influenza consists of different strands of the virus and strands A and B result in seasonal epidemics. Influenza A can cause pandemics and exhibits rapid genetic changes. Influenza B’s genetic and antigenic properties change more slowly.

“Flu is raging strong at this point,” Marathe said. “This season saw unusually high activity in December.” 

While conducting different phases of the projects, researchers faced many challenges related to computational models. In particular, they have had difficulty acquiring accurate data sets that are not noisy. Venkatramanan noted challenges with effectively translating anecdotes and field studies into computational quotes. Also, there are complications when developing representative social networks that maintain anonymity and protect individual privacy. 

In addition to improving forecasts impacted by these challenges, Marathe noted future plans for the project. He hopes to focus on additional epidemics around the globe. Additionally, he plans to improve the resolution of forecasts and incorporate artificial intelligence and machine learning computational techniques into the process. 

related stories