The National University Health System in Singapore (NUHS) is tapping into artificial intelligence technology to improve clinicians’ productivity. At the heart of this goal lies its newly-launched GPT tailored for healthcare providers.
Reports on Monday revealed NUHS launched a ChatGPT-like tool that can help healthcare providers summarise patient case notes in seconds. Called “RUSSELL-GPT,” the large language model is also able to write coherent referral letters for doctors from the summaries.
RUSSELL-GPT has been specifically trained to provide assistance and answers to questions related to medical conditions and clinical practice guidelines, according to the report. The AI model is already being used in NUHS’ Patient Trajectory Prediction AI model to analyze individual patient healthcare journeys from historical data.
“From synthesising precise local medical knowledge to reducing the administrative work of our doctors and nurses, this LLM will bring benefits to both healthcare workers and patients,” says the associate professor and group CTO at the NUHS, Ngiam Kee Yuan.
NUHS Builds ChatGPT-like to Facilitate Medical Workflows
The new AI tool is reportedly built atop NUHS’ petabyte-scale Prescience supercomputing infrastructure, which became fully operational in July. The supercomputing facility also hosts SMILE AI, another AI-powered 3D teeth charting model that can show the position of teeth condition in the mouth.
Speaking on the supercomputer use case in AI, the director of Strategy, Planning, and Engagement at NSCC, Bernard Tan, said:
“The launch of the NUHS supercomputer is timely because of the all-time interest in [g]enerative AI, and especially with the worldwide demand crunch for the computing resources needed to drive such AI-driven technologies.”
NUHS is currently looking to expand the use case of RUSSELL-GPT by deploying it across its cluster progressively. In the future, the healthcare institution reportedly intends to use the tool to analyze and predict the severity and trajectory of common health conditions, like urinary tract infections.
RUSSELL-GPT now adds to the short list of large language models designed to provide assistance to medical practitioners.
In June, Cryptopolitan reported that UNC Health, in collaboration with Epic and Microsoft, is piloting an LLM to reduce clinicians’ burnout and enhance the patient-health-provider relationship. The AI tool achieves this by helping clinicians draft responses to patient communications, among other capabilities.
Last month, Singapore’s Integrated Health Information Systems also signed a memorandum of understanding with Microsoft to build smart applications with generative AI, one of which includes a GPT that can automate tasks and transform clinician workflows.
The Issue with AI Models in Healthcare
Artificial intelligence is believed to have enormous benefits in the healthcare sector and beyond. The development of LLMs is only one of the most-explored applications of the technology. However, in spite of the promising use cases, some researchers have found that AI-generated medical information could pose some risks.
Researchers from MIT recently learned that AI models have the potential to exacerbate inequities and biases in medical diagnosis and treatment, especially for underrepresented subgroups. Biases can meddle with how certain people or groups are treated.
In their analysis, the researchers discovered four types of subpopulation shifts causing biases with AI models. These include spurious correlations, attribute imbalance, class imbalance, and attribute generalization. According to the team, these shifts need to be addressed to achieve equitability with AI models in the medical field.
