ChatGPT can speed the design of healthcare software, with research showing that the AI tool helped clinicians and software engineers work together on a personalized diabetes messaging system.
The findings point to a way in which healthcare can usefully deploy generative AI, a form of AI that uses complex algorithms to create new and unique content that powers chatbots such as ChatGPT to create realistic conversations.
ChatGPT produced a version of the automated diabetes prevention tool over 40 hours versus what initially required more than 200 hours of work from content and technical experts.
The research is published in the Journal of Medical Internet Research.
“We found that ChatGPT improves communications between technical and non-technical team members to hasten the design of computational solutions to medical problems,” said researcher Danissa Rodriguez, PhD, an assistant professor at NYU Langone.
“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.”
Previously, the team created a non-AI-powered novel communications platform to promote patient-provider communication and patient engagement in a commercial digital diabetes prevention program.
The personalized, automatic system used text messages to address diabetes by encouraging patients to eat more healthily and exercise. Its development included early prototyping and user testing, a technical development phase, and a randomized controlled trial.
The core content and user experience features were identified, prototyped, and evaluated using the well-established design thinking “discover, define, design, and test” approach.
In the current study, researchers compared the above system with one created using ChatGPT-4 using 11 evaluators, each of whom had at least 10 years’ experience in areas ranging from medicine and implementation science to computer science.
All were either familiar or had prior exposure to the original personalized automatic messaging system intervention.
The evaluators rated the ChatGPT-produced outputs in terms of understandability, usability, novelty, relevance, completeness, and efficiency.
The researchers reported that ChatGPT served as a usable facilitator for researchers engaging in the software development life cycle, from product conceptualization to feature identification, and user story development to code generation.
GenAI technologies aided effective communication and understanding within the multidisciplinary team by providing well-described features and supporting the role of a software engineer.
The researchers noted there were significant limitations to the output of ChatGPT. Specifically, while it was able to provide general guidance in tool design, it was unable to provide evidence to support its rationale for these choices.
Nonetheless, senior study author Devin Mann, MD, also at NYU Langone, maintained: “Our study found that ChatGPT can democratize the design of healthcare software by enabling doctors and nurses to drive its creation.
“GenAI-assisted development promises to deliver computational tools that are usable, reliable, and in-line with the highest coding standards.”
