Latest / Public Health Review Morning Edition / 928: Approachable AI for Public Health
Transcript
- 0:03This is the award-winning Public Health Review Morning Edition
- 0:07for Monday, June 16th, 2025. I'm Summer Johnson.
- 0:12Now, today's news from the Association of State and
- 0:15Territorial Health Officials. At the end of the day, AI is
- 0:19simply just another tool in the public health toolbox, but it's
- 0:23one with enormous potential if we approach it thoughtfully.
- 0:27We don't have to wait for perfect systems.
- 0:29We can start small, learn and build the organizational
- 0:33capacity now. Public health leaders are
- 0:36interested in using AI effectively and ethically, so
- 0:40today we have a special episode to help departments approach AI
- 0:44with confidence. Trook Taylor is the Director of
- 0:47Health Data Analytics at Guidehouse, A founding member of
- 0:50Asto's Innovation Advisory Council, a new multi sector
- 0:54collaborative of private sector partners keeping Asto's
- 0:58membership and leadership informed and nimble on emerging
- 1:02topics affecting public health. Druke, we have a lot to cover
- 1:05today on what a lot of people assume is a very complicated
- 1:09topic, so let's get right into it.
- 1:11What do you believe is really the most compelling reason to
- 1:15use AI as public health leaders? I often tell public health
- 1:19leaders to think of AI not as a mysterious black box, but as a
- 1:23new colleague, so one who can shift their massive amounts of
- 1:27data in seconds, spot patterns we may overlook, and suggest
- 1:31options that help us make better decisions.
- 1:33In my opinion, the hook is simple.
- 1:35AI can give us better, faster situational awareness, whether
- 1:39that's in tracking outbreaks, predicting resource needs, or
- 1:43identifying vulnerable populations.
- 1:45It can even help us do our jobs faster, like grant writing or
- 1:49automating work flows. When you realize that AI isn't
- 1:52about replacing expertise, but about amplifying it, the
- 1:55technology can become less intimidating.
- 1:58How important is it to keep a human in the loop, as we've
- 2:02heard a few times before? In public health, the human in
- 2:06the loop isn't optional. It's essential.
- 2:09AI can process complex data and suggest insights, but public
- 2:12health decisions have real human and population consequences.
- 2:16Context, ethics, equity considerations, and cultural
- 2:20factors require human judgment that AI cannot replace and
- 2:24replicate. And critically, many of the data
- 2:27sets we rely on in public health are messy and complete or
- 2:30contain historical biases, whether that's in case
- 2:33reporting, social determinants, or access to care.
- 2:36Without human oversight, AI models risk perpetuating or even
- 2:40amplifying those biases. Think of AI as a decision
- 2:44support tool that can help surface correlations, flag
- 2:48anomalies, and predict likely scenarios.
- 2:51But it's the public health professional who must assess
- 2:54whether the patterns make sense, whether additional data is
- 2:57needed, and whether there may be any unseen confounding factors.
- 3:02The most effective implementations of AI and public
- 3:04health are ones where humans and machines collaborate, and where
- 3:08humans remain ultimately accountable for interpretation
- 3:12and action. How can organizations build
- 3:16competence in AI systems, and what can we do to ensure that
- 3:20the answers provided by AI are reliable and accurate and
- 3:24trustworthy? Trust is earned through
- 3:28transparency, validation, and ongoing oversight.
- 3:31Period. Public health organizations
- 3:34should prioritize explainable AI models that allow practitioners
- 3:39to understand why a recommendation is being made,
- 3:41not just what the outcome is. But even explainable models need
- 3:45human validation, right? Data sets may contain biases,
- 3:48such as under representation of certain communities or
- 3:51historical inequities and access to care.
- 3:54And if those biases aren't recognized and controlled, AI
- 3:58can inadvertently reinforce them.
- 4:00That's why multidisciplinary oversight is so important,
- 4:04bringing together data scientists, epidemiologists,
- 4:07ethicists, and even community representatives to catch blind
- 4:11spots and course correct when models make mistakes, which they
- 4:15inevitably will. AI systems also need ongoing
- 4:18recalibration as public health conditions and data sources
- 4:22evolve. What works today may not work
- 4:25next year, especially as novel public health events emerge.
- 4:29Right? And this isn't just theory.
- 4:32So according to a recent study that Guide House and CDO
- 4:35Magazine conducted a public and commercial technology leaders In
- 4:40that survey, 76% of organizations support that they
- 4:43are not fully equipped to govern and oversee AI effectively.
- 4:48Now, that gap highlights why agencies need to build
- 4:51governance structures first. Do you have any examples of
- 4:55recent public health work that has successfully incorporated AI
- 4:59into its processes, and what made those examples successful?
- 5:04Absolutely so. One strong example is how
- 5:07several health departments have used AI driven syndromic
- 5:10surveillance during COVID-19 to detect early signals of
- 5:14outbreaks from emergency department visits, even using
- 5:17social media and wastewater surveillance data as well.
- 5:21So these systems helped identify local surges earlier than
- 5:24traditional reporting did and allowed faster public health
- 5:27insights and reactions. Now, even CDC's Legionnaires
- 5:31Disease team is using AI to automatically detect cooling
- 5:34towers from aerial imagery to accelerate their ability to
- 5:38respond to outbreaks. Pretty cool, right?
- 5:41Another example is the use of AI powered Natural Language
- 5:45processing, or NLP for short, to extract insights from
- 5:49unstructured clinical notes or social service records to
- 5:53identify social determinants of health.
- 5:55If you're speaking to someone or an organization that wants to
- 5:59incorporate AI into their public health processes, what are some
- 6:02steps that you tell them they can take?
- 6:06Great question. I start by saying begin with the
- 6:10problem definition, not the technology.
- 6:13Look for areas where you have complex data rich challenges
- 6:17that really strain from your current resources, whether
- 6:20that's improving case investigation workflows,
- 6:23optimizing resource allocation, or identifying populations at
- 6:28higher risk. So 1 great example is what
- 6:31Kansas local health departments are doing.
- 6:34They've developed an AI Integration Road Map that lays
- 6:37out a deliberate process for adopting AI responsibly,
- 6:41building internal capacity, addressing workforce
- 6:44development, ensuring strong governance, and keeping equity
- 6:47and Community Trust at the center.
- 6:50What would you like to leave your colleagues with today?
- 6:53At the end of the day, AI is simply just another tool in the
- 6:57public health toolbox, but it's one with enormous potential if
- 7:01we approach it thoughtfully. We don't have to wait for
- 7:04perfect systems. We can start small, learn, and
- 7:08build the organizational capacity now.
- 7:10The real opportunity is not in the technology itself, but in
- 7:14how we use it to strengthen our ability to serve our
- 7:17communities, address inequities, and respond more quickly to
- 7:21emerging threats. AI won't replace public health
- 7:24or its practitioners, but it can help us do public health better.
- 7:29If you'd like to read Guidehouse's State of Gen.
- 7:32AI Today report, we have a link right to it in the show notes.
- 7:36We'll also have a link to more information about Asto's
- 7:39Innovation Advisory Council. That'll do it for today.
- 7:43Meet us right back here tomorrow morning with more ASTO news and
- 7:47information. I'm Summer Johnson, you're
- 7:50listening to the award-winning Public Health Review MORNING
- 7:53Edition. Have a great day.