The current roles of Artificial Intelligence in public health: A scoping review
This scoping review of high-income country public health practices from 2020 to 2025 reveals a significant scarcity of evidence regarding the real-world implementation and mixed effectiveness of AI tools, underscoring the urgent need for rigorous evaluation and dissemination of results to guide future adoption.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer
Imagine the world of public health as a massive, busy library. The librarians (public health professionals) are trying to keep the community healthy, but they are often short-staffed, running on a tight budget, and drowning in paperwork. Enter Artificial Intelligence (AI), a new, super-fast robot assistant that promises to do the heavy lifting, organize the shelves, and even chat with the visitors.
This paper is a "scoping review," which is basically a giant treasure hunt. The authors wanted to see: "How are these robot assistants actually being used in the library right now?" They didn't just ask what the robots could do; they looked for proof of what they are actually doing on the job.
Here is what they found, broken down simply:
The Treasure Hunt Results
The team searched through four massive digital libraries (databases) looking for stories about AI being used in public health between 2020 and 2025. They sifted through 5,400 potential clues and read 155 full stories, but in the end, they only found 8 reports that showed AI actually being used in real-world public health settings.
It's like looking for a specific type of rare bird in a forest the size of a continent, but only finding eight sightings. The authors realized that while everyone talks about AI, there is very little written proof of it actually working in the day-to-day jobs of public health.
The Eight Stories They Found
The eight reports they found were like eight different tools in a toolbox, each used for a specific job:
- The Helpful Guide (The WIC Chatbot): In the US, a chatbot named "Maya" helped moms find local food programs and report lost cards. People liked it because it was faster than calling a hotline. Verdict: Helpful.
- The Confused Librarian (The WHO Chatbot): The World Health Organization tried a chatbot named "S.A.R.A.H" to fight fake news. Unfortunately, it often gave bad answers, ignored user concerns, or sounded robotic. It was so ineffective that the WHO eventually turned it off. Verdict: A potential threat to trust.
- The Fast Sorter (Evidence Review Software): In Canada, a tool called "DAISY" helped researchers sort through thousands of scientific papers to find the ones about COVID-19. It was incredibly accurate, finding papers humans missed. Verdict: Very useful for sorting.
- The Friendly Caller (CareCall): In South Korea, an AI called "CareCall" made phone calls to lonely elderly people to check if they were okay. It worked well to reach more people, but it created a new problem: when the AI couldn't reach someone, human workers had to call them anyway, adding to their workload. Verdict: A mixed bag of good reach but extra work.
- The Weather Forecaster (Machine Learning): In the UK, AI was used to predict when hospitals would get flooded with sick children (specifically with coughs and bronchitis). It was pretty good at predicting the "peak" times, helping hospitals prepare. Verdict: Useful for planning.
- The Symptom Checker (COVID Chatbot): In Canada, a chatbot named "Chloe" helped people check if they had COVID-19 symptoms. It was decent at answering questions, but it's no longer active. Verdict: Mixed accuracy, now retired.
- The Duplicate Finder (Review Tools): In the UK, AI tools helped researchers remove duplicate files when reviewing evidence. It saved a lot of time, even if it missed a few duplicates. Verdict: A time-saver.
- The Image Detective (Mpox App): In Singapore, an app looked at photos of skin rashes to see if they were Mpox. It was quite accurate at spotting the rash. Verdict: Effective for image detection.
The Big Takeaway: "The Gap Between Hype and Reality"
The main lesson from this paper is that we don't have enough proof yet.
Think of AI in public health like a new, expensive engine for a car. Everyone is talking about how fast it could go. But this review found that very few people have actually driven the car on a real road and written down the mileage.
- The Good: Some tools (like the ones that sort papers or predict hospital crowds) seem to work well.
- The Bad: Some tools (like the WHO chatbot) failed so badly they actually made people lose trust in the organization.
- The Ugly: Sometimes, using AI creates new problems, like making human workers' jobs harder because they have to fix the AI's mistakes.
Why Is This a Problem?
The authors point out that most of these reports came from huge, well-funded organizations (like the WHO or national health agencies). Smaller, local health teams—who might need the help the most—might be using AI too, but they aren't writing about it.
Because there is so little evidence, it's hard for a public health team to know: "Should I buy this AI tool? Will it save us money? Will it actually help people?"
The Conclusion
The paper ends with a simple piece of advice: Don't just buy the shiny new robot because it's trendy.
If a public health team wants to use AI, they need to test it first to make sure it actually works and doesn't cause harm. And if they do test it, they need to share their results with everyone else, so we can finally build a real "instruction manual" for using AI in public health. Until then, we are mostly flying blind.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.