Challenges to forecasting: new paper shows how infectious disease prediction is likely to succeed within moderate temporal horizons

The global community of scientists, public health officials, and medical professionals studying infectious diseases places a high value on predicting when and where outbreaks will occur. Considering that these outbreaks emerge from the multi-level interaction of hosts, pathogens, and environment, forecasting them requires an integrative approach to modeling. In a new paper out in Nature Communications, ISI Foundation Research Leader Giovanni Petri and ISI Fellow Samuel V. Scarpino (Northeastern University) adopt permutation entropy as a model independent measure of predictability to study the predictability of a diverse collection of outbreaks.

Petri and Scarpino identify a fundamental entropy barrier for disease time series forecasting. However, they find that for most diseases this barrier is often well beyond the timescale of single outbreakes, implying that prediction is likely to succeed. Researchers show how the forecast horizon varies by disease and demonstrate that both shifting model structures and social network heterogeneity are the most likely mechanisms for the observed differences in predictability across contagions.

Results support the utility and accuracy of embracing dynamic modeling approaches, especially those that leverage myriad data streams and are iteratively calibrated as outbreaks evolves. They also suggest challenges for performing model selection across long time series and may relate more broadly to the predictability of complex adaptive systems.

“On the predictability of infectious disease outbreaks”, Samuel V. Scarpino and Giovanni Petri, Nature Communications, 22nd February 2019