Q&A: Virtualizing HIV Prevalence in the United States
AIDSVu is an interactive online map visualizing the prevalence of HIV in the United States, presented by the Rollins School of Public Health at Emory University in partnership with Gilead Sciences, Inc.
Dr. Patrick Sullivan, AIDSVu’s principal researcher, discusses here the confluence of Big Data and public health.
Tell me about AIDSVu and your role there.
I’m the Principal Scientist for AIDSVu, which is a website that takes very rich public health surveillance data about HIV in this country and visualizes it in a format that’s more accessible to a broader range of people.
Surveillance data are traditionally disseminated mainly through reports with tables. Those are great for certain public health purposes, but we feel that putting data online in an interactive way is an opportunity to reach more people and raise awareness of the impact of HIV in our communities.
Where does the data in AIDSVu come from?
The data on AIDSVu come from two sources. One is from our partnership with the Centers for Disease Control (CDC). They have a national database that is collected from state health departments and includes information down to the county level. They provide that information in aggregate without individual identifiers.
We also map data from cities and the metropolitan statistical areas, and in those cases we work directly with the city health departments or the responsible state health departments. From there we can map the varying impact of HIV down to the ZIP code-level or sometimes down to the census tract level.
If this information is already being released, why visualize it?
The one thing that’s different about visualizing data is that most of us are used to a map interface as a way to find a place where we can receive medical care, or to find a restaurant or a bank. AIDSVu is an intuitive interface. Seeing data mapped, particularly down to the ZIP code and census tract level, allows people to see where the epidemic is concentrated relative to places that are important to them. And that’s a whole new dimension that’s very difficult to get from tabular reports.
Another reason this is particularly important is for the location of prevention services. As an example, in Philadelphia, where we work with the city health department to map the impact of HIV down to the census tract level, academic researchers from Brown University have used the data to figure out where they could best deploy door-to-door testing resources. By looking at the AIDSVu maps, researchers can find the areas most likely to have undiagnosed HIV infections and send testing staff out to the places where they’ll have the greatest impact. These uses of data simply aren’t possible when you’re looking at a table.
Tell me about some of the key takeaways for you in seeing this data visualized.
I would put my key takeaways in two segments. The first one is a little more process-oriented. Developing these maps has really been an issue of forming innovative partnerships to achieve something that is beneficial to everyone. We created AIDSVu with Gilead Sciences, Inc., and in doing so formed a unique academic-private-public health model that brings together health departments, the CDC, Gilead Sciences, and Emory University in a collaborative effort. By engaging with stakeholders that have different perspectives and strengths, we’re able to produce something that doesn’t feel like a typical public health product and that has a degree of innovation in presenting data. That can only grow out of collaboration with rich perspectives.
In terms of the second segment—the takeaways from the data themselves—I’ll mention three distinct points. One is the really striking impact of HIV throughout the southern United States, and particularly the extent to which a relatively high prevalence focuses in southern cities and beyond, into less urban areas in the south. The broad extent of high prevalence counties in the southern United States is something that the visual story of the maps illustrates in a way that might be surprising to many people.
The next takeaway for me has to do with our new diagnosis maps that show where new HIV diagnoses are occurring and represent more recent trends in the epidemic. These maps show that 92% of new diagnoses are occurring in just 25% of U.S. counties, and those 25% are largely reflective of urban and peri-urban areas.
The final observation for me is that when you look within cities there is a fair amount of difference in prevalence, even from census tract to census tract—these may be areas as small as several square blocks in a city. It’s striking. I think before I saw these maps I would have predicted that heavily impacted neighborhoods would be more consistently impacted. But getting down to finer levels of data looks almost like an impressionist painting. You have these spots of data, and as you back away from them you begin to see a different picture; as you zoom in, you also see a different picture. That picture is important with respect to providing the best services and making the most of our limited prevention funds.
One thing that your maps allow for is to compare the prevalence data with other statistics, such as race or particular determinants of health. When you make those connections, are there surprising correlations that you’ve seen?
How the HIV epidemic maps out relative to what we call the social determinants of health—like poverty, education, income inequality—had been described before at the broad population level. But I think what these maps created is an interactive way is to see how those relationships—those correlations—play out across smaller geographic units.
We find that counties that are highly impacted by HIV often have co-occurring high prevalence of poverty and lower educational attainment, for example. Another interesting feature is that we show health insurance coverage data at the county level. Here, the relationship with HIV is not so lockstep.
What we know in prevention science is that getting people linked to HIV care and sustainably on medication that controls the amount of virus in the body allows people with HIV to live longer lives. But that’s not the only thing it does. It possibly reduces the chances that those people will pass HIV on to others. Increasingly, the link between where people are living with HIV and accessibility to health insurance to support engagement in medical care is a pivotal issue as the United States thinks about how to achieve an AIDS-free generation.
You mention about where people live, and I know that AIDSVu allows you to look at state, county, or city level data. Since some of what the Foundation works on is city growth and competitiveness, I can't help but wonder about cities specifically. I'm especially curious about how AIDSVu has been used by specific cities, particularly by their public health authorities.
One of the ways that health departments use AIDSVu is as a way to leverage the rich data they collect, and to provide it in ways that people want to use them.
By working with health departments to understand the ways in which people want to use this data, we've been able to develop maps and city profiles that reflect a lot of the commonly requested pieces of information. That allows health departments to refer people to AIDSVu for informational requests, and frees up their time to work on more complicated data requests.
Often we hear about these stories after people have come to AIDSVu and used this data for a really smart purpose. One other “use case” reflects how Big Data can be used for social good. Medical AIDS Outreach in Alabama where they are trying to minimize the time that it takes people to access care for their HIV diagnosis. They used two publicly available data sets—AIDSVu and the Health Resources and Services Administration (HRSA) data on provider shortages—and overlaid the two sets on top of each other.
The mapping overlays were used to identify areas of provider shortages and high HIV impact, and set up telemedicine centers that would make care accessible to the most people in areas of shortage. Those telemedicine centers were then linked as spokes to hubs where there were HIV care specialists who could connect to those clients who needed it.
This is just one example of the thing that makes me happiest about AIDSVu: That somebody with a Google search can get a highly detailed level of information, laid out in a visually straightforward way. And people again and again will figure out smart ways to use this data that are not what we envisioned when we developed this site. The idea of this from a “Big Data” perspective is to make this data available in an intuitive way and a new way, make it broadly available, and know that people will use it for good.
And it's not simply experts using AIDSVu—it’s regular people as well, right?
When people who aren’t as engaged as HIV providers or researchers view AIDSVu, a common reaction is that people are surprised that they live in areas that are so heavily impacted by HIV. There’s a piece of human nature that says, I understand HIV is a big deal in, say, Atlanta, but I don't think that it’s so relevant to where I live. But when you look at the map of Atlanta and see the impact across the city it motivates people there to reassess their own situation. AIDSVu also provides testing and care service locators so people can find services closest to where they live or work.
I recently gave a talk in a university setting in the Atlanta metro area and I showed the Atlanta maps. Afterwards, I had a substantial line of students come up to me, each with individual questions. Many of their questions were about where they could get tested for HIV. They said they had no idea that right around where they live is an area that is so heavily impacted; they thought the HIV epidemic in Atlanta was mostly in other parts of the city.
The data remind us that HIV is still present and is still deserving of our attention. For many people who are not experts, they can look at the maps and say, I can't calculate the prevalence rate, but I can see that I live in area where HIV is still a problem. It motivates them to seek testing for HIV or to accept a test that's offered by their doctor. It helps them understand why their doctor is recommending, as the CDC does, that everyone between 13 and 64 get a test as part of routine medical care.
With all of that’s been learned, what's next for AIDSVu?
We see AIDSVu as an evolving platform. As the state of our understanding of HIV evolves, the platform will continue to grow to meet emerging needs. AIDSVu will continue to apply “Big Data” visualization to emerging issues in managing and preventing HIV. Right now, we’re working on a neighborhood mapping pilot by community area in Chicago and ward in Washington, DC, that details HIV at a level that aligns more naturally with how people identify in their community—so the data will be even more relatable to the user.
A current focus area for the U.S. is the HIV care continuum, which takes the perspective that our goal is to have everybody who’s living with HIV live in a state where the virus in their body is suppressed. That's better for their heath and it vastly reduces the chances that they'll transmit HIV to others.
The continuum process of getting people to an undetectable viral load involves several steps: people have to get diagnosed and learn of their HIV infection, link-in to medical care, stay in that medical care, and get the medications that control their viral load. If they learn that they have HIV, get into medical care, and stay in medical care most of them will achieve that viral load suppression. Right now, only about a quarter of people living with HIV infection in the United States have reached that goal of a suppressed viral load.
We've created a “Powered by AIDSVu” project, HIVContinuum.org, to begin to map the steps in the care continuum in Atlanta, Philadelphia, and Washington, DC. This lets public health practitioners and others see down to the ZIP code-level those hot spots where there may be particular problems along the HIV care continuum.
AIDSVu is focused on using data at finer geographic levels to really target how we can do prevention better. The HIV Continuum site focuses in on a key issue of our epidemic, which is how we use data to better inform our programs for prevention, for linkage to care and sustained treatment. It’s an example of how we see AIDSVu serving as a platform to meet new needs as the science of prevention evolves.
Patrick Sullivan, DVM, Ph.D., is a Professor of Epidemiology at Emory University’s Rollins School of Public Health, and Co-Director of the Prevention Sciences Core at Emory’s Center for AIDS Research (CFAR).
Interview conducted by Michael Hendrix, director of research and emerging issues for the U.S. Chamber of Commerce Foundation.