1023: AWS Examines AI In Public Health

JANSON SILVERS: 

This is the award-winning Public Health Review Morning Edition for Friday, October 31, 2025. I'm Janson Silvers. Now, today's news from the Association of State and Territorial Health Officials.

This morning, we have a special episode that examines AI in public health. We are joined today by Jim Daniel, the public health leader for state and local government at Amazon Web Services. AWS is a founding member of ASTHO's Innovation Advisory Council, a new multi-sector collaborative of private sector partners helping keep ASTHO's membership and leadership informed and nimble on emerging topics affecting public health.

Jim, I'd love to jump right in. Tell me about how generative AI has assisted with electronic case reports.

JIM DANIEL: 

A couple of years ago, we were approached by many public health departments to help them process the large volume of electronic case reports that were coming in. And originally, our thoughts were, these are standard data feeds with X, standard XML tags that we could parse the way we normally parse public health data. We could look for the tags, parse them into a data lake, and look at the data that is in those various fields. What we quickly found out, though, that there was not a lot of standardization from provider organization to provider organization, basic demographic data might have different field names with different XML tags. They might be using different vocabulary standards within those fields. So, instead of traditional methods of parsing the data, using the XML tags as a way to identify the information critical for public health, we instead turned to GenAI and just started using natural language to ask those questions, is this person pregnant? Is this person a food handler? So that we can take that data from the electronic case reports, transform it with GenAI into a format that is actually then used directly in the electronic disease surveillance systems that public health uses.

SILVERS: 

And then tell me how you see AI improving our ability to handle forms.

DANIEL: 

Yeah, I don't think that public health will ever get away from paper forms, where even today we see some electronic lab reports coming in, not electronically, but on paper. And one of the great ways that generative AI can help there is actually looking at those paper-based forms and automatically figuring out what are, the fields that you're looking at, and what are the various options for those fields. With GenAI, it's actually able to look at a very complicated form, whether it's a fill-in-the-blank, circle the right answer, x's, it can look at all of that and automatically figure out exactly what you're trying to do with that form. You can also have a human in the loop with these where you can get confidence intervals for whether or not the GenAI is confident about the answer that is chosen. And then you can choose the scores where a human might want to review those that are less than 70%, perhaps. And it can bring up the form and show how GenAI has proposed that the answer is. And then you can edit that directly in the user interface before it gets extracted and put over into the system of record.

SILVERS: 

I also know that AI chatbots have been successful. Can you expand?

DANIEL: 

You know, many times public health is overwhelmed with different stakeholders calling in or emailing with questions whether it's a provider who might have a question about an immunization information system, a person who might have questions about whether or not they're eligible for WIC, these are the kinds of questions that public health departments spend many, many hours on answering via phone or email. And when it comes to those kind of simple types of questions, we can actually help with an integrated virtual agent or chat bot that can be placed directly on the website, where we can use our tools to consume the information that has all of these answers to the questions, and then have a chatbot that can actually help answer those questions. That chatbot can be available on a website. It can be available via phone number where people can call and ask their questions and get an answer verbally. You can ask questions verbally in multiple languages, and even though your source documents are only in English, you can still interface with the people who are asking questions in multiple languages. You can also be very specific with the chatbots about only looking at your documents and not trying to look at the rest of the world's information to find an answer, and then it's just the harder, more complicated questions that end up coming to a live agent, saving public health time and money.

SILVERS: 

And when it comes to telling a story, that's also where AI can be instrumental, right?

DANIEL: 

Yeah, so one of the, I think, key pain points for public health departments is the lack of highly skilled data analysts. And public health is always getting questions coming in from leadership, from their governor's office, where they have to immediately come up with visuals and graphs, and then we can use QuickSight, for a business user to come in and start asking questions with natural language. So, you can start asking questions like, can you describe the relationship between obesity and cardiovascular disease? And our GenBI tools will start suggesting the types of graphs and charts that you should build, and through the user interface, help guide you to creating those. And then once they're created, it can actually create the draft story for you, talking about the relationships between those risk factors and diseases, and even start suggesting other types of risk factors you might want to look at in your data, or you know, what your interventions or what you might want to do with that information once it's created the story. So, it's just a really nice way to kick start that process and make it easier for business analysts to start interfacing with the data and getting the information that they need without a data analyst having to always create those visuals for them. Again, this is not about replacing jobs. Data analysts are still very much needed, especially in this case, to help set the data up the right way, but then it makes it easier for everyone to interface with the data.

SILVERS: 

Lastly, what would you say the biggest obstacle to adopting AI is?

DANIEL: 

One of the biggest challenges still with GenAI is the policy issues that go around implementing GenAI and working with your state policy to actually be able to do that. We do have a lot of examples from public health departments that have been successful there, and we're working on getting a set of shareable tools to help others get through the policy aspects. The technology aspects are not always as difficult to implement, but when they are, if your internal IT department doesn't have the resources to deploy generative AI resources on AWS, we have several partners that we can work with, depending on the use case, that I have expertise in doing that and can help with that.

SILVERS: 

Thank you so much for your time, Jim.

You can find more information on AWS, including an on-demand webinar with other industry partners on some of their successes in the show notes.

That'll do it for today. We're back on Monday morning with more ASTHO news and information. I'm Janson Silvers. You're listening to the award-winning Public Health Review Morning Edition. Have a great weekend.