Global Influenza: The Hidden Interplay Among Viruses Behind Epidemics
Every flu season is different from the previous one. It’s not just the number of cases that changes, but above all the “mix” of viruses in circulation: A/H1N1, A/H3N2, and influenza B. Some years H3N2 dominates, often hitting the elderly hardest; other years H1N1 affects young people more. Understanding in advance which subtype will prevail means better planning for hospitals, vaccine distribution, and prevention campaigns.
A new international study, “Characterization and forecast of global influenza (sub)type dynamics,” published in Nature Health, attempts to answer a crucial question: is it possible to predict the composition of influenza subtypes a year in advance? The research was conducted by Francesco Bonacina, currently a researcher at the DONDENA Center at Bocconi University, and by Chiara Poletto, of the Department of Molecular Medicine at the University of Padua, in collaboration with Pierre-Yves Boëlle and Vittoria Colizza (Institut national de la santé et de la recherche médicale, France), Olivier Lopez (École polytechnique, France), and Maud Thomas (Université Claude Bernard de Lyon, France).
From apparent chaos to discernible trajectories
The starting point is technical but crucial: the data on subtypes are percentages that always add up to 100%, making it difficult to analyze them using traditional statistical methods. To overcome this obstacle, the researchers applied an innovative approach to epidemiology:
“We used Compositional Data Analysis to circumvent the problem and study trajectories of annual subtype compositions of countries.”
Instead of looking only at isolated percentages, the team reconstructed actual timelines for each country, from 2000 to 2023, studying the balance between H1N1, H3N2, and B in an analyzable mathematical space.
The years that changed everything
The results show that the mix of subtypes can vary greatly from one year to the next, and, in particular, certain seasons clearly stand out as global anomalies.
“We identified a few seasons that stood out for the strong within-country subtype dominance due to either a new virus/clade taking over (…) or subtypes’ spatial segregation (COVID-19 pandemic).”
In 2003, H3N2 dominated; in 2009, the pandemic H1N1 upended the global balance. During the COVID-19 pandemic, however, the subtypes “separated” geographically, driven by the collapse of international travel. And mobility itself emerges as a key factor:
“Geographical factors, most notably international mobility, contributed to shaping countries’ composition trajectories between 2010 and 2019.”
Countries with the most air travel connections show similar patterns in the alternation of subtypes. Influenza doesn’t just follow the climate—it follows the routes.
Predicting the coming winter
The most ambitious part of the study involves predicting subtype composition a year in advance. The researchers tested five models, ranging from the simplest—based on subtype frequency in previous years or the historical alternation between H1N1 and H3N2—to advanced statistical models capable of integrating temporal and spatial information.
“Incorporating the global history of subtype composition into a Bayesian Hierarchical Vector AutoRegressive model improved predictions compared with naive methods.”
The best-performing model, called HVAR (Bayesian Hierarchical Vector AutoRegressive), does not limit itself to looking at the past of a single country, but also “learns” from the trajectories of similar countries, for example those strongly connected by air travel. By integrating this geographic information as well, the model is thus able to compensate for the limitation of short time series (only a few years of stable data between the two pandemics).
The results show a significant improvement over less sophisticated methods, especially in predicting whether a subtype will be dominant (>50% of cases) or, conversely, marginal (<10%). In particular, the prediction of the dominance of A/H3N2 (often associated with more severe seasonal flu) is significantly more accurate.
Accurately predicting which virus will prevail remains difficult. However, the methodological leap is significant: for the first time, the global geographic structure of subtypes is formally incorporated into a predictive model.
The Importance of Reliable COVID-19 Forecasts
The composition of subtypes influences the age groups affected, the severity of the epidemic, and the strain on hospitals. Having more reliable forecasts means gaining precious months to prepare resources and strategies. The study demonstrates that behind the apparent unpredictability of global influenza lies a recognizable pattern and that, with the right statistical tools, the future may be less opaque than it looks.
