How did your research on this topic begin?
I was working on my dissertation in the doctoral program in the Environmental and Occupational Health Department and my advisor and one of my dissertation committee members was Beth Carlton, who got pulled into the Colorado modeling team for the COVID response efforts.
Very quickly it just became clear that we needed to have some way to incorporate weather and the anticipated effects of seasonality into the Colorado models for hospital demand and hospitalization admissions. We were looking into what is already known to date about other viruses that are similar – RSV, SARS 1, MERS and influenza are the main examples that we had to draw on. We know that these viruses tend to be more viable and more infectious when temperatures and humidity are on the lower end.
For COVID, the earliest preliminary studies suggested that it was going to have a similar seasonal response. We decided as a team to put together a small-scale grant to look at weather and its impacts on COVID-19 transmission.
As the pandemic continued, it became clear there were some untested assumptions that still existed around season and weather-related transmission patterns. So, parsing this out: Do spikes in respiratory virus transmission in the winter result from the direct effects of weather phenomenon like temperature and humidity and their influences on virus transmission potential? Or is it that weather influences transmission-related behaviors, such as spending more time indoors during cool winter weather? It meant we were really trying to dive into the causal pathways, which wasn't something that was super clear even from the existing literature on diseases that have been around for centuries.
What were your findings? Why are they important to the discussion around understanding this specific time period of the pandemic?
We went into this with the assumption that the human behaviors would probably be driving the changes in transmission dynamics more so than these smaller-scale changes to viability or survival of the virus.
We specifically wanted to use an analysis framework that could kind of dive less into correlation and more into causal pathways – that’s where this mediation analysis model came in. This just allowed us to look at the very specific direct effects of weather on COVID hospitalizations and then distinguish that from the mobility piece. Just being able to parse out each piece and look at how much is each contributing.
The big takeaway: The indirect effects of weather via changes to social mobility on COVID-19 hospitalizations – that is, weather’s impact via changes to the amount of time people spent indoors away from home – were limited. It’s an interesting finding because it challenges the general assumption that winter spikes in COVID-19 and other respiratory viruses primarily result from people spending time indoors, away-from-home during the winter (presumably in confined spaces and in close contact with others). Instead, we found that weather-induced changes in time spent indoors away from home had little to no effect on COVID-19 hospital admissions.
What was different with our findings from previous research studies was that we found that the direct effects of weather on transmission probability, rather than the indirect effects of weather via changes in mobility behaviors, was actually the stronger of the two pathways. So weather had a more direct influence in our study, which was surprising to us. We were expecting to see the mobility piece really drive what we found overall, and that's just not what we found here.
Is it especially surprising given the interventions and distancing interventions that were going on at that time in Colorado?
Yeah, I think with all of the interventions going on at the time, we were very much assuming that the social pathway was going to be the more important of the two. And so it was just very surprising to find that weather directly was the key player in our analysis.
I think with that said, this entire study was conducted with interventions going on. Recommendations of changing your behavior and where you spend time with people, social distancing, creating a small social circle. It's an interesting finding that weather so directly affects transmission, but we really need to look into it in a more recent context too, to draw conclusions of ‘what does this mean?’ in a non-pandemic setting.
Did you run into any challenges with having relatively smaller data sets?
In a way. Our study was definitely restricted in how we came up with those five counties – Boulder, Denver, Douglas, El Paso and Mesa – because those are the ones with enough data in general to be able to draw any conclusions.
Still, it is important to look at microclimates. We know that just small-scale changes in a climate can have pretty profound impacts on how infectious disease dynamics work. That's definitely an important angle to take. But then, of course, you're more limited in your power when you dive in and come down to this very fine scale.
Given the interesting results that you all found from this, but also acknowledging in the paper that your data availability ends in 2021, do you all have a sense of what the natural follow up to this research would be and where you go from here?
Yeah, I think that we would love to run some version of this. The challenge is we have this very short-term access to the specific mobility data – we'd have to find a new source. That's just a roadblock for us in particular. But I think that there are a few different avenues in an ideal world to pursue, especially with respect to the time periods we examine. Specifically: looking at weather, mobility and infection again in the “new normal” – post-public-health-emergency period of COVID-19.
