AI in Public Health

Technology and Emerging Trends

checking body temperature

COVID-19 was the first high-tech pandemic. Almost as soon as the first cases started to emerge, corporations and governments turned to technology to track, analyze, and mitigate the deadly disease. Applications were rolled out for symptom reporting and contact tracing; computer models simulated airflow in outbreak centers to determine how infectious the disease was; and through the miracle of the internet, entire industries were socially distanced overnight, undoubtedly saving countless lives.

If it was the first, though, it surely won’t be the last. As public health experts go through the after action reports from the COVID-19 response, they will have a laundry list of ways that technology can do even more in the next pandemic.

Artificial intelligence, which started to exponentially increase in capability even as the waves of disease swept the globe, will be a vital part of the public health toolbox the next time around. Even now, it is poised to change the game as public health officers face down more routine threats like:

  • Addiction and substance abuse
  • Obesity and malnutrition
  • Air pollution
  • Widespread antimicrobial resistant strains

Understanding machine learning in healthcare and how artificial intelligence in public health and epidemiology will impact the next epidemic is no longer optional in the field today.

AI and Machine Learning in Public Health

robot hand pointing to human hand

If artificial intelligence is a new term for many public health workers, machine learning is not. ML has been a key piece of data science for a decade are more. When it comes to analyzing the big data sets that come with disease surveillance, ML has already been a big help in connecting slender threads of evidence into a web of infection.

ML is technically a kind of artificial intelligence. Today, though, AI offers much more capability than straightforward analysis. Generative systems go further and faster in:

  • Predicting disease spread
  • Developing regional or even individualized treatments
  • Tying together information that human researchers might not otherwise connect in an ocean of data

That can include uses like:

  • Automatically detecting tuberculosis in chest X-ray imagery without a radiologist
  • Improving data collection on opioid-related deaths by parsing death certificate data to include various different terms and even misspellings
  • Identify subtle patterns in clinical data and develop predictive models for clinical outcomes

Each of those examples has already been put into practice at the Centers For Disease Control. More innovative uses of artificial intelligence in public health surveillance and research are sure to come.

Using AI and Data in Public Health

global virus

Data is in many ways the lifeblood of public health. From diagnostic reports to lab analysis, doctors on the front lines as well as officials in back offices rely on data to draw conclusions and make predictions in epidemics. Time is of the essence.

Sponsored Content

Data science has been hugely useful to organize and analyze all the information coming in around disease outbreaks. Especially at first, public health professionals may not even know what kind of information is important. So it’s necessary to build huge data stores with every potentially relevant data point.

Machine learning and AI are used to comb through that data to identify what’s relevant. In 2019, a solution was proposed to collect and process Tweets to vectorize and visualize information about disease outbreak clusters. Newer systems can tie together posts and other kinds of data for even more accuracy and information.

What Can AI in Public Health Achieve?


In some ways, AI and deep learning bring the capabilities of a human investigator to public health work. Operating faster and covering more territory, AI nonetheless comes to the work of surveillance and investigation with some of the inquisitiveness and creativity of human public health officers.

Going back as far as the Broad Street Pump and Dr. John Snow, epidemiology has relied on careful, rigorous analysis of data. And that’s not just healthcare data. Beyond the strict science of vital signs, lab results, and microbe analysis, public health work relies on insights into the behavior and daily routines of human beings.

Just as public health officers have been able to make connections between disease outbreaks and visits to wet markets or attendance at big conventions, AI can be used to connect patients in unexpected ways. And it can do so by using all the massive amounts of data people generate today in the course of their lives:

  • Social media posts
  • Selfies and other pictures
  • Audio recordings
  • Electronic health records

While traditional computes have offered ways to look at all this data quickly, it’s always taken a human being to put together the threads. AI can absorb all of that information instantly and draw even minor details out to put two and two together. A common background sound in video clips; a set of symptoms commonly mentioned in social media posts; timelines between interactions and outbreaks.

Algorithms can put all the clues together quickly to identify the source and spread of disease.

That all adds up to public health and medical professionals being able to respond and treat outbreaks faster than ever.

Using AI for Public Health Communications

Part of that process may include opening up AI-driven systems to direct public contact. Hotlines fill up and emergency rooms get jammed when a new health threat is in the wild. Misinformation and mystery are widespread. There are never enough public health professionals to take the calls, screen patients, or perform vital contact tracing duties.

NLP, or Natural Language Processing, offers an answer to all these challenges. Whether through voice recognition in phone systems or online chatbots, AI can carry on conversations fluently and naturally. And with logic and reasoning skills built in, it can answer basic questions or handle information gathering without involving a human healthcare pro.

These systems aren’t just a benefit to the public, either. They can tie in directly to other analytics to develop a real-time picture of what people are asking about and where problems are emerging. On top of freeing up public health workers from routine follow-ups, they can generate more data and interpret it more quickly than current systems.

Artificial Intelligence in Healthcare in General Will Be a Boon to Public Health

cutting edge healthcare

Much of the benefits of AI to the world of public health will come from contributions to healthcare in general. Artificial intelligence technology in healthcare is already well on its way to becoming a revolution of its own.

AI can help clinicians develop diagnostic screens quickly from existing electronic health records. It can identify patterns that humans would take longer to spot, and then it can select appropriate courses of action for confirmation or treatment to recommend.

AI NLP is also offering a solution to a healthcare system where providers are in short supply. Already, customer service and appointment booking systems are being built around AI that can communicate naturally and in multiple languages with patients. Before too long, similar systems will:

  • Conduct interviews to identify symptoms
  • Offer instructions when hospitals are overwhelmed
  • Conduct follow-up interviews persistently and automatically

The Risks of Artificial Intelligence Bias in Healthcare Extend to Public Health

doctor during covid

Like other uses of new technologies, building AI into a system as critical as healthcare requires thoughtfulness and oversight. While AI systems have proven to be miracle workers at many kinds of medical applications, they also show some disturbing tendencies that public health officials will need to watch out for.

First among these are hallucinations. Large language models (LLMs) aren’t truth-telling machines. They are purely representations of calculations of statistical likelihood of certain words appearing together. So they can’t tell fact from fiction and are as happy to serve up false information if it sounds good. That’s a big strike against using them to provide critical health information to the public.

They have also been shown to be susceptible to disgorging information from their training data that could be considered private or sensitive. So they represent a possible HIPAA violation, as well.

Even non-LLM models, like those used for research or investigation, represent a sort of risk. The most advanced machine learning systems today are a sort of black box. Their results can be amazing, but no one can really explain what is happening inside the system. For scientists, that’s a tough pill to swallow. Recommendations or analyses from AI will have to be double and triple-checked for error.

Finally, they are susceptible to subtle biases built into the training data itself. So hundreds of years of healthcare community discrimination against minorities can get baked into poorly trained models and be reflected back in abusive or even misleading results in the healthcare context.

That’s all on top of more specific uses in laboratory analysis and assistance in reading medical imagery.

It will also be accompanied by breakthroughs in medical manufacturing and pharmaceutical development. The breakneck pace with which COVID-19 vaccines were developed may seem like a long slog in future public health emergencies. AI is already being used to screen chemical compounds for likely drug development and to identify potential additional uses for existing drugs that could be useful in disease outbreaks.

Using Artificial Intelligence in Public Health Prevention Efforts

Tied to ordinary uses of AI in the healthcare system are applications of AI to the prevention of public health issues.

This can extend far beyond the immediate issues of detection and treatment. For example, the CDC is looking using AI to forecast trends in opioid overdose mortality by tying together multiple different datasets from around the country.

Ordinary applications of artificial intelligence in the healthcare world could have real and positive impacts on the state of public health even without being designed to do so.

Other examples of machine learning in healthcare include preventative care and screening  by combing millions of electronic health records. The problems that will head-off are likely to lead to a healthier population overall. AI-driven, individualized exercise programs could result in reductions of obesity and diabetes. In the realm of public mental health, AI chatbots and therapy programs could make treatments more available, accessible, and acceptable to new ranges of people in need.

It’s also possible that AI will draw new techniques in preventing outbreaks from the deep well of data it has access to. It’s impossible to know what these might be. But everything from currently unrecognized applications of drugs already on the market to discovering the causes of mysterious disabilities like autism could be in the cards.

Risks and Considerations for AI and Machine Learning in Healthcare Offers Lessons for Public Health Professionals

Using AI and ML in healthcare in general comes with plenty of caveats as well as opportunities.

The potential for bias in health screening and services isn’t unique to artificial intelligence. There’s pervasive bias in generative AI simply because there is pervasive bias in the data it is trained with—the world itself simply isn’t a fair or equal place yet.

Healthcare datasets for machine learning gathered over a span of decades include all the racism and sexism of those eras within them.

But in addition to basic issues of algorithmic fairness, public health officers also have to be concerned with bias leading to missed or mistaken diagnoses. In effect, machine learning and algorithmic fairness in public and population health isn’t just about discrimination against marginalized groups. It is also about the accuracy and effectiveness of the system as a whole.

Missing trends in disease spread doesn’t just impact minority populations. Unchecked, an epidemic can spread like wildfire. So it’s important that health datasets for machine learning in disease surveillance and artificial intelligence monitoring support absolutely clear-eyed assessment.

Machine Learning Technologies Create Privacy Concerns in Public Health

folder of protected health information

There are also real public health questions over privacy in the use of machine learning in epidemiology. Public health officials have always had to dig, sometimes uncomfortably, into the personal lives of individuals. The AIDS epidemic resulted in many gay men being inadvertently outed at a time when it was particularly dangerous for them.

But the depth of vision and inference available in machine learning may be even more intrusive. Weaving together so much data can reveal not just elements of exposure and infection, but other personal traits people may not wish to share.

Society as a whole has a lot of work to do in balancing these considerations, but public health official using AI are likely to be at the center of the controversy. A strong grasp of ethics will be required at every level of the system.

The Public Health Implications of AI Are Only Just Beginning to Unfold

doctor touching medical button

The various concerns and issues with AI in public health settings will require public health officers to develop a strong understanding of how AI works and what the risks are. In some cases, they’ll want to put the brakes on new developments. In others, they might see some new innovation that hasn’t even been imagined yet and pour on the gas.

On the flip side of the privacy problem, for instance, is the potential use of AI to automatically doublecheck notes and records for personally identifiable or protected health information that needs to be redacted. This time-consuming process is subject to many errors when performed by humans—screeners get tired or misread text.

AI can perform the same function tirelessly and consistently, which may actually help preserve privacy while speeding up the sharing of information crucial to public health services.

But the truth is that AI is still in early days. AI technology advancements will change rapidly in coming decades. So will the public healthcare system in general, as well as the threats of disease and disability it faces. The real necessity is having public health professionals who are well-educated enough to adapt to those realities.

Getting an Education to Understand Artificial Intelligence and Machine Learning in Healthcare Settings

sharing medical knowledge

Fortunately, many degree programs are emerging to help professionals prepare for the AI future in public health.

Of course, any Master of Public Health degree today will at least offer coursework in data science and healthcare analytics. Those disciplines have been tested and proven for public health use. Machine learning and public health are old friends there already. No competent professional in the field functions without them today.

Sponsored Content

In fact, there are specific degrees available that zoom in on such AI-adjacent specializations already. A Master of Science in Healthcare Analytics and Information Technology or a Master of Science in Computational Epidemiology are both courses of study that are very likely to be AI-heavy today and in the future.

For public health professionals whose college days are already well in the rearview mirror, short certificate programs can help bridge the AI gap. An Artificial Intelligence in Public Health and Healthcare Certificate comes with enough updates in math, ML techniques, and data analysis to bring anyone in the field up to speed on new developments.

Those developments will keep on coming. AI brings challenges as well as promise. But public health professionals have shown repeatedly they will do whatever it takes to keep the population safe.