Leveraging Analytics and Disease Forecasting in Public Health

Mati Hlatshwayo Davis: [00:00:12] Hello and welcome to Let's Talk ID. I'm Dr. Mati Hlatshwayo Davis, director of health for the City of St. Louis and member of the IDSA board of directors. Really excited about today's episode. Joining me today are Dr. Dylan George, director of the Centers of Disease Control and Prevention, Center for Forecasting and Outbreak Analytics, or the CFA. And Dr. Lior Rennert, associate dean for health sciences and director of the center for Public Health Modeling and Response at Clemson University. Welcome to both of you.

Dylan George: [00:00:46] It's a pleasure to be with you. Thank you for having us.

Mati Hlatshwayo Davis: [00:00:49] So they are joining me today to discuss the CFA's work in assisting public health leaders respond to disease outbreaks by integrating cutting edge science into real time decision making. You know, I don't always get to lean into my people in public health, but I am biased in my excitement about today. So let's get started. So I'll come to you first, Dr. Dylan George. CFA's vision and mission, can you start by telling us about the broader mission here and how Insight Net fits into that vision?

Dylan George: [00:01:19] The Center for Forecasting and Outbreak Analytics is one of the newest centers within the CDC, and our mission is to harness cutting edge analytics to improve response capabilities for public health emergencies. You know, our unique contribution is to use data science to get people the information they need when they need it. And so we live in this world of AI. Now, if you think about rideshare capabilities of various, Uber and Lyft, and we can know by the minute when we arrive at a destination because of the capabilities that are enabled, because of data analytics, to get us that information and allow us to do that capability. This shows that these kinds of capabilities can be put into place. Now we need to apply that same sort of superpower, that same sort of capability to public health. And that's what we're doing at the Center for Forecasting and Outbreak Analytics. And one thing that I do want to really emphasize here, though, too, is that public health is local. The critical decisions of public health, especially with outbreak responses, happen at the local level. Now, Insight Net is a really important capability, a really important effort that we put into place to help us use advanced analytics and disease modeling at the local level. We launched this effort in late 2023. 13 primary partners across the United States academia, health departments and the private sector. Now, today, there's over 120 partners, including 51 state, tribal and local public health departments. In short, what we're trying to do is we're trying to use advanced analytics to support local health departments and health care systems during responses, and we're working with Insight Net to advance our capabilities to use that much more effectively at the local level.

Mati Hlatshwayo Davis: [00:03:08] I mean, I couldn't be more excited because I am uniquely aware of the challenges that we have at the local level, having both capacity and even budget to do exactly what you're describing that is critical. So you are speaking my language, and can you just say once again for the people in the back? Public health is what? It's local? Did you say it's local?

Dylan George: [00:03:31] Public health is local. Yes. We need to have those capabilities at the local level.

Mati Hlatshwayo Davis: [00:03:35] Yes.

Dylan George: [00:03:35] That's why we're leaning in with Insight net to try to build capacity and capability at the local level, to actually make sure that we have a much more robust capability there.

Mati Hlatshwayo Davis: [00:03:45] That really excites me. So how exactly does CFA integrate cutting edge data science into real time decision making? And then what unique challenges come with applying predictive models to public health? I'll come to you first, Dylan, but Lior, feel free to weigh in.

Dylan George: [00:04:02] Analytics and disease forecasting can improve health outcomes during and outside of public health emergencies. Now, for example, it's public health data really helps us know what has happened in a community. Think about it from this perspective, though too. It's like many people have to realize they're sick. They have to get an appointment with a doctor. The doctor has to actually request a test. That test has to go to the lab. The lab has to run the test and then report back to the doc. The doc then has to report to the local health department. If it's a reportable disease, then it has to be reported up to the CDC. Now, there's lots of efforts that are making that go much faster, but fundamentally that's a long lead time process. So what we see in public health data is what has happened in the past. And we're getting better at making that actually more real time. But it's still the past. What we're trying to do with the analytics and the modeling is what we're trying to trying to project that forward to help us anticipate what's coming at us. If you think about driving a car, you use both your rear view mirror and your windshield to make sure you get to your location safely.

Dylan George: [00:05:06] What we've been doing largely in public health is we've been looking through the rear view mirror, and we're very good at that. Now, we need to add the windshield to help us actually drive forward, to know how we're actually going to arrive safely in our destination. So these tools can be used much more effectively and much more efficiently, not only within public health, but also in the health care systems. And so we're very excited about the efforts that we're doing. And again, this is to try to help us develop response capabilities during an outbreak. But it also can be used outside of emergencies. Analytics that can be tailored to a local community's data to identify and address recurring challenges in that community are going to be critically important. And that's one of the reasons why we're so excited about working with Clemson, because they're doing exactly that. They're trying to actually help us not only build capabilities that will be useful in an outbreak, but also capabilities that will provide insights to address challenges that are happening in healthcare right now.

Mati Hlatshwayo Davis: [00:06:07] Love that. Lior, anything you want to add?

Lior Rennert: [00:06:09] It is true public health is local and healthcare is also local.

Mati Hlatshwayo Davis: [00:06:14] Oh, tell me more, tell me more. [laughs]

Lior Rennert: [00:06:18] Well, and with what we are trying to do and Dylan described the challenges of data integration and data collection is we have to have more sophisticated modeling and strategic data collection strategies to complement some of the challenges or overcome some of the challenges that we have. So one of the things we are doing now is using artificial intelligence and machine learning to integrate across various data sources, not just electronic health records or whatever eventually gets to the CDC, but wastewater data, digital trace data, census data. So community level demographic and socioeconomic characteristics into the forecasting framework and traditional models kind of struggle to actually do this. So with artificial intelligence, we can actually now estimate disease trends in regions with insufficient coverage.

Mati Hlatshwayo Davis: [00:07:08] What really excites me about what you just said is seeing what this was like for us in the very real sense, even here in St. Louis during Covid and how we had our hands tied in the very real sense. Mpox came after that. Ebola kind of started to rear its head. We're currently dealing with avian flu. The ability to empower us to use our local data, which we're doing a good job of, but be able to have that higher level analytical approach truly, truly excites me. So thank you for that. Can you give some tangible success stories, maybe a specific story about how Insight Net has made a tangible difference, particularly in underserved areas? For me in St. Louis, you know, the north side of St. Louis repeatedly is the area that is hit the hardest. So I'd be interested in hearing any specific success stories you've had.

Dylan George: [00:07:54] One of the reasons why we're very excited about working with Lior and Clemson in working in South Carolina is it is a great example of how advanced analytics can improve public health today and safeguard communities against future disease threats. Now, the Clemson team is doing what they're doing, is really working to get more details here, but it's like it's really exciting and we're in the early stages of developing this. But what they're doing is they're using models to identify gaps, who is not being reached. They will use that information to guide deployment of mobile health clinics to make the best use of those valuable resources, and expanding access where it is most needed. Now, as we all know that there's been a retreat of clinical services and hospitals in rural settings, including infectious disease specialists. They're often very scarce in rural settings, and being able to identify these gaps and the need to drive better response will improve not only health outcomes right now, but it will also help us be much more prepared in these rural, underserved communities. This innovative use of models and partnerships with hospital systems will leverage the public health data to improve health outcomes in those communities. So we're really excited about growing that partnership with not only with our academic partners, but also with the healthcare systems and infectious disease professionals to adapt and use analytics to actually improve health outcomes going forward.

Mati Hlatshwayo Davis: [00:09:24] Fantastic.

Lior Rennert: [00:09:26] Yeah, and I just wanted to give one specific recent example as to how we use this in practice. And one of the things with mobile health clinics is they can't just go anywhere and you can't just park them anywhere. We need community partners to help secure locations and ensure our patients actually know about our services, so they utilize the mobile health clinic and its services. What we have found through our modeling, just this past year, is there are several communities with large infectious disease outbreaks that were candidates for delivery of the mobile health clinics, but our health systems didn't have the community partnerships established there. So thanks to our modeling insights, they're now working to get community partners in those areas so that we will be ready to quickly send a mobile health clinic when needed.

Mati Hlatshwayo Davis: [00:10:07] I can't tell you how excited I am about this and how the wheels are already turning. So you've messed up because you're about to have another partner as soon as this podcast is taped.

Lior Rennert: [00:10:15] We would love to have you.

Mati Hlatshwayo Davis: [00:10:17] Listen, be careful what you ask for. So the reason that I'm excited is two things. Number one, people don't understand how deprioritized public health has become. Right. I inherited less than 1% of the city's budget when I took over in 2021. I have a health department that's 3 to 4 times smaller than it should be for a city its size, which means I don't have the workforce to do this work and especially during outbreaks. So what you are describing is what we urgently need right now, not just in the future. And then I think about my partners in rural settings and how critical this is as well. When you talk about mobile van access here in an urban city. That's a large city. They de-prioritize the budget that we don't even have clinical services. They were sunsetted a while ago. So mobile clinics are truly the center of how I have to operate. And so this example you gave speaks directly to what I'm saying. But I can imagine it's so for our rural partners for whom access is such a challenge as well. So really excited by what I'm hearing. Dylan, you referred to scaling. And so can you talk about what the biggest challenges in scaling tools like Insight net are to ensure all communities, regardless of their resources, can benefit equally?

Dylan George: [00:11:28] Yeah, no. First off, you know, like public health and data in public health, as you well know, has been historically siloed. There's this thing in manufacturing, medical countermeasures. It's like one bug, one drug sort of approach. It's like in public health, it's been one bug, one data system. So it's like sharing data across those silos or sharing data with different jurisdictions has been a challenge. And there's lots of efforts that are going on right now with data modernization to try to improve upon that and move that forward. And so there's some exciting things that are happening in that space. But it's a challenge to do that sort of work. And that's why it's incredibly important for us to try to work with local health departments, because they have that access uniquely to those data, and they can actually understand what's happening at the local level during an outbreak. And so as you've pointed out as well, there isn't a one size fits all approach to an outbreak. Even understanding it or even responding to it, that's just not effective. We need to find local solutions to those problems. And so each community is different their needs, their resources, their priorities. And so we need data driven tools that can be applied in different communities to generate insights about what's happening in that area. And we want to support the leadership so that they can, quote unquote, right size actions for their communities to keep the people, their neighbors and their communities safe in a time of crisis. And that's why we're trying to lean so much more heavily into supporting local decision makers and working on projects like what Lior is doing.

Mati Hlatshwayo Davis: [00:12:56] Incredible. So, Lior, you've worked closely with communities in South Carolina. Can you explain how models developed by Insight Net are adapted to meet the unique needs of rural areas?

Lior Rennert: [00:13:07] So one of the biggest challenges facing rural and other medically underserved communities during infectious disease outbreaks is susceptibility to severe health outcomes, and that's primarily due to greater exposure risk, underlying chronic conditions, and inadequate health care access. So we are developing modeling tools to both forecast outbreaks and allocate essential resources to these areas, including mobile health clinics for infectious disease screening, treatment and vaccination. And since these communities are typically underrepresented in electronic health records, we have adapted our models to integrate non-traditional data sources, including digital trace data. So, you know, Twitter/X, Google trends, along with wastewater and demographic and socioeconomic data to estimate disease trends in these areas. And we're also developing GIS models for once we identify a high risk community, where can we place it within that community so that we capture as many people as possible, as quantified by driving or walking distance.

Mati Hlatshwayo Davis: [00:14:10] And we haven't said this explicitly, but let me say this to everyone listening. This is the work of health equity, right? You're hearing the intentionality around social and structural determinants of health, around access issues, the intentionality that both of you have prioritized in every single answer around making sure that all, not just some, benefit, which I really appreciate. So, Lior, how do you involve local stakeholders such as community leaders or public health workers in implementing data driven solutions? This is central to my leadership style. I believe that while I'm the incumbent director of Health, I've got to prioritize and even ahead of me, trusted messengers. What's your approach here?

Lior Rennert: [00:14:51] Yeah so we are working with Mobile Health Clinic and other public health decision makers and their network of community partners, along with some of our own, to identify and prioritize high risk, rural and underserved communities for delivery of these essential but limited resources. And just so I can give some background about how amazing these mobile health clinics are, these are vehicles that deliver health care to medically underserved communities. They're typically staffed by clinicians. And they can provide essential services from prenatal care to diabetes management to HIV screening and treatment. They're also extremely cost effective. It is estimated that for every dollar invested, you get $12 back in savings. And that's through prevention of severe health outcomes. It's estimated that every mobile health clinic prevents an average of 600 emergency department visits each year. So as the old public health saying goes, an ounce of prevention is worth a pound of cure, and mobile health clinics are the perfect example of this, so it shouldn't come as a surprise that these were widely used during the Covid 19 pandemic to deliver infectious disease screening and vaccines to underserved communities. Obviously, the challenge is that these are a limited resource and they can't be everywhere all at once. Lack of data driven approaches to guide allocation poses daunting challenges for decision makers and leads to inefficient distribution. So, to increase the number of individuals that are served through the mobile health clinics and ultimately save more lives, we're developing modeling frameworks to identify and prioritize high risk communities for delivery of the mobile health clinics. And these models don't just forecast the size of the outbreak in each community, but integrate information on community resource needs to ensure that they go to the areas where they have the greatest impact. And of course, we have to work closely with our public health and community partners because they provide us with key insights into these models and our corresponding software to help guide the data driven decision making.

Mati Hlatshwayo Davis: [00:16:51] Right. So you've set yourself up because one thing that I always call out is how we operate in silos. Academics want to do their thing. Clinicians want to do their thing. And often we are ignored in public health. And I will always assert that we are the implementation arm. So you started to talk about this, but can you delve more into what role Clemson played in facilitating partnerships with the South Carolina Health Department and hospital systems, and what lessons you've learned from this collaboration?

Lior Rennert: [00:17:18] Yeah, so we have cultivated fantastic relationships and partnerships with South Carolina's Department of Public Health, our two largest health systems, Prisma Health and the Medical University of South Carolina. They serve about half the population in South Carolina and cover about two thirds of the geographic regions. And Clemson Rural Health and all of these institutions own and operate their own mobile health clinics. I would say the main lesson we have learned with infectious disease response is that the needs and constraints of each institution and our community are very different. Even within that institution, so even within health systems and health departments, there are several priorities. We talked about mobile health clinic allocation, but there's also issues with surge capacity and making sure you have enough essential resources and sufficient staffing, all while making sure you don't burn out your clinicians and other providers and communication campaigns and everything that goes into that. So one of the lessons we learned was early on was rather than trying to assist each institution independently, which would spread us very thin, we are instead starting with the underlying needs that they all share in common. So what we're doing now, and what we've really been doing since early on, is creating a unified statewide network for outbreak detection, forecasting and coordinating emergency response. This is done through strategic data collection and integration, innovative modeling, collaboration with our health partners and software development, and our software will be tracking outbreaks, identifying high risk communities, mapping available resources and assisting emergency response. And this software is actually being pilot tested in real world settings to track these outbreaks and guide interventions. And soon we will be integrating this into the workflows of our public health and healthcare partners. And this ensures that our modeling frameworks will ultimately be used in practice and save lives.

Mati Hlatshwayo Davis: [00:19:15] I want it, I want it, I want it now. I want it in St. Louis. This is so exciting. Let's talk about trust. So for both of you, how do you ensure that public health officials and communities trust and act on the insights generated by these advanced models? I'm particularly sensitive to this. I come from the Black community. I'm very aware of the valid mistrust and distrust that is historic and current. So let's talk about this a little bit. Dylan, what are you thinking here?

Dylan George: [00:19:43] First thing is like CFA is trying to be transparent in how we're doing things with our science, our code, our methods, our data. And so there's all broad. It's all publicly available. And so people can scrutinize that as they can in a very peer to peer kind of way going forward. But also it's like as I mentioned before, it's like public health is local and public health is scarce resources. So we need to find ways to use those scarce resources to show up. We need to be in the local communities, as you well know, and we need to find ways of stretching those scarce resources so they're most effective. These kinds of analytical capabilities will allow us to be able to right size the staff stuff and systems, so that we can do more good with what we're trying to accomplish and improving health and saving lives going forward. So showing up is a big deal. We need to stand with our neighbors, with our colleagues, and with our communities, and that's the way that we're going to show trust is by being there with them. And this is one of the reasons why we're trying to double down on supporting local communities in how they're actually using analytics and driving health outcomes in a better way.

Lior Rennert: [00:20:54] Yeah, that that was very well said. And you have to engage these communities and your health partners and engage them in this process in order to ensure that the tools you're developing actually meets their needs. And at the same time, we have to set expectations for ourselves, but also for our partners. These are, at the end of the day, decision support tools, and we have to be transparent about what they are. And I think that exactly how Dylan was saying it, that's how we have to build trust.

Mati Hlatshwayo Davis: [00:21:22] Couldn't agree more. And this is probably the most important part for me. I sat on a panel with local community leaders, predominantly Black, and I was talking about data as justice, and I just thought I was killing it. And then they started laughing and they said, you academics love your data, don't you? And they said something that has never left me. If you control the variables, then you control the outcomes. And I preach that to my team because it behooves us to come to them humbly as experts in their own right from the beginning of the process before we even start these things. So I'd loved what you said about engaging, about showing up, about being present. But for me, it's also about really respecting that there are experts in their own right and listening before we just plunge in with our well-meaning initiatives. So lastly, let's talk about the future. Looking ahead for both of you, how do you envision emerging technologies like AI, which you referred to earlier, further transforming public health forecasting and response efforts?

Dylan George: [00:22:21] I'll jump in right away. It's like AI is, as we all know, is already impacting our daily lives in multiple ways. Applying the strength of AI to public health is what CFA is built to do, and that's one of the reasons why we're funding different groups within Insight Net to actually try to capture the benefit of AI for public health going forward. Now, broadly speaking, AI in healthcare is being used for in a couple of different ways like early warning, optimizing health operations quickly, processing data sets like remote monitoring data. And they're trying to figure out how to do clinical decision support in a much more. And that's the most probably the stickiest wicket in all of those. But within CFA, we're working with AI to enable all of those components early warning, improved operations and data and code processing. That's what we're trying to use this superpower for. And we're already discussed with our example at Clemson and working in South Carolina to optimize this care delivery in these underserved areas. Of course, we need to make sure that these tools are responsibly and appropriately used. But regardless, we are well poised to help use emerging technologies like AI to benefit providers in a range of ways and to try to optimize or right size staff stuff and systems to be able to provide better health outcome in a range of communities.

Mati Hlatshwayo Davis: [00:23:40] While this has been an incredible discussion. I want to actually recap some of the gems that really moved me. I love that this entire approach, whilst being cutting edge, actually speaks to the central tenets of health equity. I love that we have the CDC partnering with a local university to really dig into that, at both the state and a local level. I really like that you touched upon cost effectiveness, because that speaks to the heart of some of the social and structural determinants of health. We know that economic mobility is something that we're driving into here in Saint Louis. So I love that you're taking that approach. The integration piece really stood out to me. I heard you talk about the fact that you're being uniquely aware about using your partnerships with academia, but pulling in local and state public health departments, community-based organizations and trusted messengers in those local community leaders. But once again, what did you say about public health? It's, what? It's what again

Dylan George: [00:24:43] It be local. [laughs]

Mati Hlatshwayo Davis: [00:24:44] It is local. So I love the fact that we are prioritizing local public health. I've long felt and have had a soapbox about the fact that we have far too long deprioritized public health, and it's shocking to me that we didn't learn enough lessons in Covid. So I want to thank you personally and professionally for making that a priority and for empowering incredible local health departments filled with civil servants, grossly underpaid, doing the work of 3 to 4 people without enough resources. This is truly exciting. And now you're going to be stuck with me. So this has been another episode of Let's Talk ID. Thanks so much for joining us. And thank you, Dylan and Lior. This has been a joy.

Lior Rennert: [00:25:25] Thank you for having us.

Mati Hlatshwayo Davis: [00:25:32] This episode was produced by the Infectious Diseases Society of America and edited and mixed by Bentley Brown.