Fighting Flu With Math: Predicting Peak Season, Spread, and Vaccination Patterns

A computer simulation model developed by NYU investigators uses math to forecast influenza activity.

A team of investigators at New York University is using math to help fight influenza.

Flu activity in the United States continues to rise, with the percentage of people visiting their health care providers for flu-like illnesses up in the week ending February 5, 2019, according to the US Centers for Disease Control and Prevention’s (CDC) weekly FluView.

The mathematical model, developed by NYU professor Maurizio Porfiri, PhD, MSc, and 2 Italian investigators with visiting appointments at NYU, analyzes epidemiological and sociological factors to predict when the influenza season will peak, who should be vaccinated, when vaccinations should occur, and whether to quarantine infected patients, according to study published by the Society for Industrial and Applied Mathematics in the SIAM Journal on Applied Dynamical Systems.

“Human behaviors are characterized by nonhomogeneous temporal distributions. For instance, periods of high social activity typically alternate with periods characterized by a moderate social life. This phenomenon is called burstiness,” Lorenzo Zino, PhD, an investigator at Politecnico di Torino in Italy who worked on the study along with Dr. Porfiri and Alessandro Rizzo, PhD, also of Politecnico di Torino, told Contagion®.

“In our research, we have found that the burstiness of human behaviors plays a critical role in shaping the evolution of social systems. When modeling the inception and the evolution of an epidemic outbreak, we prove that neglecting burstiness in social interactions may lead to heavily underestimate the infection propagation and, consequently, the risk for the society,” Dr. Zino continued.

The investigators combined this “burstiness” metric along with the more traditional details about the infectious disease itself and the spatial and temporal properties of the network of social interactions and “activity-driven networks.” The result was a computer simulation that is able to act as humans do to predict flu trajectory.

But how can math predict human behavior when human behavior is, inherently, unpredictable?

“We totally agree with this statement. In our opinion, to predict human behavior at the level of a single individual is not possible. Also, it is not feasible to develop a model that tells whether an individual contracts a disease or not after an interaction with an infected one,” Dr. Zino said. “The goal of our mathematical approach, instead, is to predict the evolution of epidemics at the population level, using general features of human behavior, such as the heterogeneous tendency to be active in social life and the presence of common behavior such as burstiness, which can be captured by our mathematical framework.”

In 2016, the investigators used data from the 2014 Ebola outbreak in Liberia to prove the validity of their model. They are now hopeful their model will help health care providers increase the accuracy of flu forecasts.

“We hope that our research…will be used by health care providers to better prepare for flu season,” Dr. Zino told Contagion®. “For instance, the possibility to accurately predict the temporal and spatial evolution of flu outbreaks allows for providing areas with the vaccines that they need in a timely and optimized manner. This would avoid scenarios in which areas run out of the flu vaccines, while others are uselessly oversupplied.”

The research is funded by the Alexandria, Virginia-based National Science Foundation with Biagio Pedalino, MD, an international consultant in public health and former CDC epidemiologist, serving as adviser.