Dr. Amy Wesolowski
Welcome to FOCUS In Sound, the podcast series from the FOCUS newsletter published by the Burroughs Wellcome Fund. I’m your host, science writer Ernie Hood.
In this edition of FOCUS In Sound, we meet a Burroughs Wellcome Fund grantee who is innovating in methods of detecting infectious disease.
Dr. Amy Wesolowski is an assistant professor in the Department of Epidemiology at the Johns Hopkins University Bloomberg School of Public Health. She holds a BA from College of the Atlantic, and earned her PhD from Carnegie Mellon in 2014.
She completed her postdoc at the TH Chan School of Public Health at Harvard University.
Amy received a 2016 Career Awards at the Scientific Interface, or CASI, grant from the Burroughs Wellcome Fund to further her work on the impact of human travel on infectious disease dynamics. She has studied those elements associated with malaria, dengue fever, rubella, measles, Ebola, and most recently, COVID-19. She uses data generated from mobile phone calling records to quantify travel patterns.
Amy Wesolowski, welcome to Focus in Sound…
Thanks for having me.
Tell us about the overall approach your research employs…
Sure. The majority of my research is really focused on trying to understand how people travel and ways that we can quantify human mobility patterns, and then how that relates to infectious disease dynamics. So that’s sort of the focus of the Burroughs Wellcome grant that I have. It’s really trying to use particularly mobile phone data and other sources of data in low and middle income countries to try and quantify and model human travel patterns, then look at how those patterns can help inform models of disease spread.
How did you use that approach to study COVID-19 patterns, as you published recently in Nature Communications?
That Nature Communications paper is trying to look at how we might be able to use mobile phone data to monitor and evaluate different aspects of the pandemic. So in general, mobile phone data is often used to try to look at how people are traveling or if there’s aggregations or congregations of people in different places. And given that most of COVID is transmitted, a lot of it happens in enclosed places and things, we’re trying to figure out different ways that you can measure these things. So, if you put in travel restrictions, do people travel less? If you put in additional social distancing, do people go to grocery stores less? And so mobile phone data can provide a real time estimate of those measures and metrics, so we can try to evaluate, are people actually traveling less and going fewer places. And so in this paper, we’re just trying to review and outline different aspects on different epidemiological questions and how mobile phone data might be used, not just in terms of contract tracing apps, which is what they are often thought about, but also thinking about population mobility patterns too. And then what are going to be some of the biases by using mobile phones, and if you are using things that rely on smart phones, for example. What are the biases in terms of who is the actual population at risk, and what are these different kinds of data and these kinds of methods able to measure.
What were some of the results of that study?
I think some of the results are that there’s a lot of promise in using mobile phone data, because a lot of people own a mobile phone. It’s possible now and companies and telecom regulators are becoming much more open to being able to utilize some of these data for public health purposes, and there’s a lot of work and pipelines on how you can actually extract these data and analyze these data. I think one of the main points that we wanted to raise in this was that ultimately you really want to be able to show that you are capturing patterns that are relevant to disease transmission. For example, the mobile phone data might be able to say that two people are within X meters within one another, but there’s a lot of variability in that. They could be in completely separate buildings, they could both be wearing masks, and as you need to get a finer and finer measure of behavior, it’s important to think about, are these data actually able to capture those patterns, and are they able to capture those patterns for the population that you’re most interested in. With COVID-19, there’s a lot of disparities in who actually has the most severe disease and who actually is getting infected. But thinking about how those biases, if you’re really interested in mortality, that’s predominantly in older age groups, who might be people who are less likely to own a smart phone, for example. So these kinds of data might not be as relevant for some of those questions, even if they’re still able to estimate population level clustering and mobility patterns.
Do you think there are further opportunities to use your methods in relation to COVID-19?
Yeah, I think so. I think increasingly what they might be able to be used for is trying to look at how populations are going back to normal travel patterns and behavioral patterns and aggregation. So I think there’s a lot of different studies that show mobility, measures of clustering, aggregation, all these things from different mobile phones that have all dropped as the pandemic started. But they’re all going back up and we’re also seeing there’s a lot of other factors that are not just related to mobility that are probably coming into play. How well can people isolate? How well can people quarantine? So I think there is still going to be a useful measure alongside other types of data and metrics, so being able to get some sort of real-ish time estimate about how populations are behaving.
What about issues of privacy related to mobile phone data? How do you safeguard against inappropriate breaches?
It’s a really big concern, and particularly as you’re trying to get information on a much, much finer scale. So most of the work that we have done has been a lot of aggregated mobility and population level mobility patterns, so looking at how many people are traveling on a given day between counties or states or something, which affects thousands and thousands of people, often, or mobile phone subscribers. And I think that increasingly there’s going to be two really interesting and fine-scaled behaviors, and the finer you get the more issues there are with privacy. And so oftentimes we will analyze data that if it’s only one subscriber who has made a trip or something because that could be identifiable. We’ve been working a lot with mobile phone operators and regulators to try to push and have a platform where aggregated data and aggregate mobility patterns are able to be shared more broadly, and that they’re aggregating away those who have privacy concerns. But they’re still there. But I think the issue with SARS-CoV-2 as you want a really fine understanding of behavior and individual level behavior, a lot of those privacy concerns have not been really addressed. So I think that it’s still an issue, and it still hasn’t been really fleshed out, and the regulators aren’t really sure, and the companies aren’t. So I think there are things that can be done about sort of like still making the data aggregated or like summary measures and those sorts of things, but I think that part of it is about that everyone is sort of aware about the public health utility and the public health good about these kinds of data, and ways that it’s still getting information to policymakers, who often don’t actually want very, very individual level details, they want to have summaries and ideas about behaviors. So I think it’s always going to be a play between what is actually useful to inform decision making, and what is actually protecting the privacy of subscribers. But I think the more that there are general discussions or frameworks about how these data are being used and processes, and everyone aggregating it and analyzing them similarly, I think that helps a lot of these discussions along.
But I think it is a different question about the contact tracing apps, which are sort of a different aspect versus these mobile phone companies collect data from their subscribers, aggregating and analyzing that. So I think there are still differences between the applications.
Amy, I know you’ve used this approach in other studies for the past several years in various places and various diseases…would you give us some examples, and tell us more about the evolution of your ideas and career…
Sure. We’ve used it for a bunch of different other pathogens. My favorite pathogen is malaria, and so we’ve used it a lot for looking at malaria control and elimination. So one of the issues with malaria is if you are able to reduce it in a particular location, there still might be a lot of mosquitos and a lot of people who could be getting malaria, so if people are infected and they come into a location, they might be able to re-introduce the pathogen and then cause additional transmission events related to those introduction events. And so in a lot of areas, the malaria in these lower transmission areas, where they find transmission down a really, really far way but there is still possible that they could have secondary cases, we’ve used these data to try to estimate and look at those patterns; where those parasites might be coming from, where they’re going to, what sort of the rate of those, and trying to identify where additional control measures should be targeted. So a lot of the work that we’ve done has been using mobile phone data to inform malaria control.
The other application that we’ve done a number of times is looking at measles outbreaks. So people can introduce measles if vaccination rates aren’t really high, sort of the same story where it’s like, oh, you can have an outbreak of the disease. So trying to use these data to look at those different patterns of mobility and how that can impact your control efforts.
And then we’ve also been sort of working on developing models of mobility. You can’t always get these data everywhere, it’s a lot of effort and a lot of work, and there’s going to be issues. So can we better model mobility if you don’t have these data or if you have other kinds of data, and then also trying to provide evidence and proofs of concept that these data are able to be, we’re able to aggregate them in a way that’s not infringing on the privacy of subscribers, where it’s making individuals identifiable from these data. But also you’re able to sort of like provide like pipelines that operators can use where these data can get aggregated and anonymized and analyzed more broadly for public health.
What has the CASI grant meant to your scientific career?
The CASI grant is definitely the best grant that I have had, and that everyone I know who’s had them has also really, really valued it. I mean, I think the best part of it is that it gives you a huge amount of freedom and support. If I want to start thinking about how this might apply to Ebola or some other pathogen, the Burroughs Wellcome Fund has always been super-supportive and very excited about sort of like, they just want to figure out the best way to support you to do the science that you want to do. There’s very, very few other grants that combine that kind of like freedom of being able to think about new problems and sort of expand the research that you’re working on, with very few restrictions about things. They’re never coming at you being, “you said you would do this very specific, very exact thing.” It’s more about supporting you and what you’re interested in, and sort of helping you, particularly in that really early stage where you’re transitioning and starting some new kinds of projects, and working on maybe some different things with different collaborators. It’s just like a really flexible, really supportive grant to have. It’s the best grant, I highly recommend it to everyone!
Amy, it’s been great pleasure meeting and chatting with you. We wish you the best for continued success. Thanks for speaking with us.
OK, great, thanks so much.
We hope you’ve enjoyed this edition of the FOCUS In Sound podcast. Until next time, this is Ernie Hood. Thanks for listening!