Generative artificial intelligence like the famous ChatGPT is changing the design of health software, according to a recent study. Researchers from New York University Langone Health have shown that generative AI, specifically ChatGPT, can significantly reduce the development time for diabetes prevention software by enabling doctors to be part-time developers — no actual software development skill required.

When doctors can become software developers overnight

The study investigates how generative AI can enhance the design process of a personalized automatic messaging system (PAMS) for diabetes prevention. These AIs based on the transformers architecture (GPT stands for Generative Pre-trained Transformer) predict the next word in a sentence based on extensive training data and millions or even billions of parameters (individual variables), enabling them to generate natural-sounding language and summarize complex texts effectively.

But ChatGPT doesn’t just ‘chat’. You can also write code, and it’s pretty good at it too. It generates code snippets, sample code, and copy-pasteable actual programming code in natural language programming.

The research team, led by Danissa Rodriguez, tested the use of ChatGPT in facilitating communication between clinicians and software engineers. Their goal was to accelerate the development of a text-based app encouraging healthier eating and exercise among patients.

The results were very promising. Eleven evaluators from various fields ranging from medicine to computer science successfully used ChatGPT to create a version of the diabetes tool in just 40 hours. That’s a 5X improvement over the traditional method, which required over 200 hours of programming without AI assistance.

“We found that ChatGPT improves communications between technical and non-technical team members to hasten the design of computational solutions to medical problems,” Rodriguez said in a press release.

The chatbot drove rapid progress throughout the software development life cycle, from capturing original ideas, to deciding which features to include, to generating the computer code. If this proves to be effective at scale it could revolutionize healthcare software design.”

Bridging the Gap

One of the study’s key findings was ChatGPT’s ability to improve communication between technical and non-technical team members. This facilitated a smoother and faster design process. Doctors and nurses don’t know how to code and software developers don’t know how medical practice works. In this context, AI is bridging the gap by enabling physicians and nurses to essentially act as prompt engineers thanks to their ability to frame medical problems in a nuanced and precise manner.

However, the authors note that while AI can streamline much of the design process, human software developers are still essential for final code generation. You still need someone who can perform version control, continuous integration and continuous deployment (CI/CD), code reviews, and regression testing. The latter is crucial in software development because it ensures that new code changes do not adversely affect the existing functionality of an application. By systematically re-running existing tests on the updated codebase, developers can identify and fix bugs that may have been introduced inadvertently during enhancements or bug fixes.

The study suggests that generative AI can democratize the design of healthcare software by enabling medical professionals to directly contribute to its creation. As a result, AI-assisted development promises to produce reliable, user-friendly tools that meet high coding standards.

“Our study found that chatGPT can democratize the design of healthcare software by enabling doctors and nurses to drive its creation,” said senior study author Devin Mann, strategic director of Digital Health Innovation within NYU Langone Medical Center Information Technology (MCIT).

“GenAI-assisted development promises to deliver computational tools that are usable, reliable, and in-line with the highest coding standards.”

If these findings hold at scale, they could revolutionize the field of healthcare software design. The ability to rapidly develop effective health tools could lead to more timely and personalized patient care. Additionally, reducing the time and resources needed for software development could make it more accessible to a broader range of healthcare providers.

Generative AI’s potential extends beyond diabetes prevention. Its application could transform various areas of medical software development, from patient management systems to diagnostic tools. As AI technology continues to evolve, its role in healthcare is likely to expand, offering new ways to enhance patient outcomes and streamline clinical workflows.

The findings appeared in the Journal of Medical Internet Research.