Study finds AI can speed up health software design

Artificial intelligence is helping clinicians design diabetes prevention software faster, a new study finds.Publication in Medical Internet Research JournalThe study investigated the capabilities of a type of artificial intelligence (AI) called generative AI or GenAI. It predicts likely choices for the next word in a sentence based on how billions of people have used the word in context on the internet.

A side effect of this next-word prediction is that generative AI “chatbots” like chatGPT can generate realistic language responses to questions and produce clear summaries of complex text.

This paper, led by researchers at NYU Langone Health, examines the application of ChatGPT to the design of a software program that uses text messages to encourage patients to eat healthy and exercise. Masu.

The researchers tested whether AI-powered interactions between doctors and software engineers could facilitate the development of such personalized automated messaging systems (PAMS).

In the current study, 11 evaluators from fields ranging from medicine to computer science spent over 40 hours successfully creating a version of a diabetes tool using ChatGPT. The original work without AI took the programmer more than 200 hours of her time.

“We found that ChatGPT improves communication between technical and non-technical team members and facilitates the design of computational solutions to medical problems,” said lead author of the study and New York University Langone School of Population Health. said Dr. Danissa Rodriguez, assistant professor. , Healthcare Innovation He is a member of the Bridging Research, Informatics, and Design (HiBRID) Laboratory.

“Chatbots have brought about rapid advances across the software development life cycle, from capturing the original idea to deciding which features to include to generating the computer code. If proven, this could revolutionize the design of healthcare software.”

AI as a translator

According to the study authors, generative AI tools are sensitive, and if you ask them questions in two slightly different ways, you may get different answers. The skill required to structure the questions you ask a chatbot in a way that elicits the desired response is called prompt engineering, and it combines intuition and experimentation. Because doctors and nurses understand delicate medical situations, they can design strategic prompts that improve communication with engineers without having to learn how to write computer code.

But these design efforts, in which healthcare providers who will be the users of new software try to advise engineers on what the software should include, are often challenged by attempts to converse using a “different” technical language. It may be damaged.

In the current study, clinical members of the team type their ideas in plain English into chatGPT and ask the tool to translate that input into the type of language needed to guide the team's software engineers in their coding efforts. I was able to. The authors say that while AI-powered software design could only go so far as to require a human software developer to generate the final code, the overall process has been significantly accelerated.

“Our study shows that chatGPT can democratize healthcare software design by empowering doctors and nurses to drive the creation of healthcare software,” said senior author of the study , said Devin Mann, MD, director of the HiBRID Lab and strategic director of digital health innovation at New York University Langone. Medical Center Information Technology (MCIT).

“GenAI-assisted development is committed to providing computational tools that are easy to use, reliable, and compliant with the highest coding standards.”