Boston Children's Hospital's ChatGPT expert is teaching physicians and artificial intelligence how to understand one another better. And he is one of a "small handful of people on the planet" qualified to do so.
The medical center made headlines last year when it advertised a job posting for a large language model prompt engineer, hoping to harness the potential of generative AI in healthcare.
Boston Children's hired Dinesh Rai, MD, for the role in August. He hopped over from healthcare AI company Pieces, where he served as product lead of AI and clinical experience. Dr. Rai, an emergency physician by training, previously worked in clinical informatics, including for New Hyde Park, N.Y.-based Northwell Health.
So far, Dr. Rai has been helping develop prompts that generate the best responses from AI, extract the right data from documentation and unstructured text, build new applications, and guide medical residents through clinical cases via a chatbot.
Boston Children's has been one of the biggest users of ChatGPT since its launch in late 2022. It was an early adopter of the customizable ChatGPT Team program. It also recently launched its own internal ChatGPT platform (Renton, Wash.-based Providence has one as well.)
"Our view has been that these tools can play a significant role across all facets of healthcare, whether that's in clinical decision-making at the bedside, whether that's in the administration of healthcare or on the research front," John Brownstein, PhD, chief innovation officer of Boston Children's, told Becker's.
Over 1,000 employees accessed the private chatbot in the first few weeks after its late 2023 rollout, representing about a tenth of Boston Children's total workforce. "The demand was far greater than we had anticipated," Dr. Brownstein said.
Boston Children's developed the tool with OpenAI and Microsoft to be HIPAA-compliant so it could be safely and privately fed patient data.
Most of the early questions have involved documentation rather than anything clinical. Some popular prompts have included "How do I translate this for a patient?" or "Can you write a prior authorization?" The queries have been about a third each spanning clinical administration, administrative tasks and research.
Dr. Rai said his clinical background gives him a grasp of what providers need and want from a tech standpoint, as well as how to explain things to patients and prioritize diagnoses. "Building technology with an understanding of your end user is always helpful," he told Becker's.
Boston Children's Hospital didn't set out to hire a physician for the role; it just happened.
"It's a bit of a unicorn situation," Dr. Brownstein said. "Being able to speak both a computer science language and a clinical language — and having the authority to talk about these topics in a roomful of providers — is super unique and makes him uniquely qualified to be in this role.
"Maybe in the future, there will be legions of MDs with LLM experience. But for now, there's a very small handful of people on the planet with those credentials."
Dr. Rai said he hopes that in a few years, large language models will have opened up more time for providers to actually take care of patients, rather than be bogged down with documentation as many are today.
"There's so much more opportunity in front of us, and, of course, the holy grail becomes clinical decision-making and how these tools can support that," Dr. Brownstein said. "Right now, the key for us is building the infrastructure, building enthusiasm, and building trust."
It helps to have a "unicorn" like Dr. Rai on staff.