23 Jun Commuter Patterns Can Help Predict Influenza Spread
MedicalResearch.com Interview with:
Population Biology, Ecology, & Evolution Program
MedicalResearch: What is the background for this study?
Response: Previous research at the global scale has shown that air travel is important for the spread of disease. For example, much work has focused on the recent Ebola epidemic in Africa, identifying where this disease emerged and then using air travel networks to predict the path of spread from there.
At a more local scale, other modes of transportation may be more important to structuring pathogen populations. We were interested in investigating seasonal influenza in the United States. Previous research has shown that once the winter influenza epidemic starts, it spreads very rapidly across the continental states, suggesting that the US may act as one large, well-mixed population. Previous work using genetic data to look for spatial structure at this scale didn’t identify any patterns. However, these studies used geographic proximity to define the distance between states; we wanted to see whether similar patterns existed at this spatial scale if we instead used movement data as a proxy for the distance between locations. Commuter movements have previously been shown to correlate with influenza timing and spread based on influenza-like-illness and mortality data.
MedicalResearch: What are the main findings?
Response: We found that spatial structure is detectable within the US. We used data on the genetic distance between sequences collected from different states and compared that to different measures of ‘distance’ between states—geographic proximity, the daily number of people flying between states and the daily number of commuters traveling between states using ground transportation—to see whether any correlations were present. Further, we did this for two different subtypes of seasonal influenza: A/H3N2 and A/H1N1. These subtypes have different epidemiological properties, so there was reason to believe that the observed patterns might differ depending on subtype.
We found that some correlations were present for all the distance metrics studied, but that they were observed a greater proportion of the time when looking at commuter movements, and when looking at the A/H1N1 subtype. Since A/H1N1 is generally milder and spreads more slowly throughout the US compared to A/H3N2, we interpret this to mean that spatial structure is likely more easily detected in this subtype. If A/H3N2 spreads rapidly from coast to coast, any signature of spatial structure is likely obscured before we have a chance to observe it.
MedicalResearch: What should clinicians and patients take away from your report?
Response: The detection of network structure implies that patterns of epidemic spread are, to some extent, predictable. The absence of predictability is problematic for the design of targeted surveillance and control strategies, since it suggests that the annual seasonal spread of influenza within countries is highly variable and depends heavily on chance events. In that case, broad scale surveillance and mitigation strategies must be utilized, and these generally perform worse than more directed approaches. We showed here that there are underlying spatial patterns in the genetic data, and that these are dependent on how the ‘distance’ between locations is being measured. This information can then be used to target surveillance and control to certain geographic locations or host groups.
MedicalResearch: What recommendations do you have for future research as a result of this study?
Response: Our study demonstrates the importance of incorporating host movement data when trying to predict how diseases will spread. Humans can move long distances very rapidly so the idea that geographic proximity is key to determining disease spread doesn’t always hold.
Our findings based on genetic data are in agreement with previous research using influenza-like illness case data; both show that patterns do exist, with strongly connected states having similarly timed epidemic peaks and similar genetic variants circulating. The patterns we found are likely influenced by states with many commuters, and the identification of these states, as well as network pathways that contribute substantially to influenza spread, is an important next step for epidemiological research.
Brooke Bozick, Ph.D. Candidate, Population Biology, Ecology, & Evolution Program, & Emory University (2015). Commuter Patterns Can Help Predict Influenza Spread