Tip Sheets

Covering artificial intelligence in health care: A primer

By Karen Blum

With more and more stories being reported about the applications of artificial intelligence platforms in health care, it may be inevitable that most journalists writing about health care will touch on AI at some point. Here is a primer on what’s happening. 

What is artificial intelligence?

Artificial intelligence is a branch of computer science that works to simulate human intelligence. AI programs or algorithms can be built into machines to perform certain tasks such as searching radiology images to identify aberrations or potential diseases. Through a process known as machine learning, the programs incorporate and learn from the data they are being fed and, therefore, get better and more accurate at their tasks over time.

How is AI being used in health care?

AI is being incorporated and studied in health care in a multitude of settings including helping diagnose cancers and other conditions through analyzing radiology images, analyzing patient data to detect sepsis, detecting diseases through voice analysis and breathing patterns, assisting with patient and staff scheduling and more.

What are the benefits of AI?

AI can analyze tons of data at a speed much faster than a human and without the fatigue humans experience that can contribute to errors. It can be used as a useful tool by physicians, researchers and other health professionals to strengthen or complement their work. Experts have said the tools will not replace human work; people will still be needed to create and feed the programs and review the results that the programs find.

What are the limitations or drawbacks?

One of the concerns surrounding AI is about the data that is fed into them. AI programs learn as they go, incorporating all data points, so if any biased data or studies are introduced to the program, those biases will be incorporated into the program’s findings. 

Several groups have stepped up to study this phenomenon and push for more fair and equitable programs. The Coalition for Health AI (CHAI), a group of academic hospitals, government agencies and private companies, aims to promote the adoption of credible, fair and transparent health AI systems. The National Institutes of Health’s Bridge to Artificial Intelligence program is working to generate data sets and best practices that researchers can adopt in their own AI applications. The National Science Foundation and Amazon have launched the Fairness in AI program that provides research funding for AI and machine learning program studies that have an equity focus. These are just a few examples.

Another concern is how to properly get consent from patients to include their data in larger pools. As such, you may start to see more comprehensive patient release forms make their way into clinical care.

Where can I find unbiased sources to interview?

Many health care organizations today are using or evaluating some type of artificial intelligence. Depending on your story angle, you can find numerous sources through searches in the National Library of Medicine’s PubMed database collection of peer-reviewed journal articles. Other avenues are contacting universities, hospitals or government agencies such as the National Institutes of Health, the Food and Drug Administration, the Department of Health and Human Services, or the Office of the National Coordinator for Health Information Technology (ONC). Some universities and academic medical centers have started their own AI centers. These include the Artificial Intelligence in Medicine division at Cedars-Sinai Medical Center, the Massachusetts AI & Technology Center, and the Clemson-MUSC AI Hub.

Health care associations also are heavily focused on artificial intelligence. Experts can be found through the American Medical Informatics Association (AMIA) and Healthcare Information and Management Systems Society (HIMSS), as well as physician groups like the American Medical Association and those that focus on particular specialties, like the American Heart Association.

Tip: Always ask your interviewees if they have any relevant disclosures such as receiving research funding from any organizations or companies for AI-related work. If they do have such relationships, mention them in your story.

What questions should I ask when reporting on AI projects or initiatives?

You can ask about the program or task being studied: What can it do? How does it work? Why is it needed or how can it help solve a problem? How is it beneficial over a human performing the same task, etc.? But because bias is a concern, you also should ask about what type of data sets were fed into the program, and how the program was trained. How will they continue to feed the program so it becomes more accurate over time? 

This is a huge area. How can I get up to speed or find story ideas?

Read, read, read. Follow the news. See the journals and news sources section under the resource links tab in the Health IT core topic area for some ideas. Check in with your local hospitals and universities to see if there are any experts there who can offer some guidance on trends to watch. AMIA and HIMSS (organizations mentioned above) have periodic meetings throughout the year where physicians and others present new research. Even if you don’t go to the meetings, you can see the agenda online or possibly search the posters and learn of new topics and potential sources. Similarly, physician associations also have larger conferences at least once a year where you might be able to find story ideas by looking at the agenda or contacting the press office.

 

You can join the email list for any of these organizations and be informed of news or when conferences are occurring. 

 

Additional resources from AHCJ: