NUHS now use AI to estimate patients’ hospital stay

The National University Health System in Singapore has recently developed its own large language model called RUSSELL-GPT.

Healthcare staff at the National University Health System can now use AI to estimate the duration of a patient’s hospitalisation, as well as use RUSSEL-GPT to boost their productivity.

Associate Professor Ngiam Kee Yuan, Group Chief Technology Officer at NUHS, explained NUHS’s version of ChatGPT as a fine-tuned model called RUSSELL-GPT, which operates within a secure internal environment. “This AI model assists in tasks such as summarization and reference letter writing, streamlining administrative tasks for healthcare staff,” he said.

He also introduced the Patient Trajectory Prediction AI Model, which significantly differs from traditional patient care analyses. “This model uses text data from doctors and key clinical notes, as well as blood test results, to predict a patient’s hospital stay duration. It offers consistent, hospital-wide analysis, differing from the variable interpretations of individual doctors,” Ngiam explained. 

He added that this systematized approach enables more uniform care and anticipatory planning across various medical conditions.

Ngiam mentioned that before launching any AI tool, NUHS conducts extensive tests to verify its functionality and accuracy with real-world data. 

“We also ensure that the tool does not exhibit bias against any patient groups and that it operates securely within the hospital’s environment,” Ngiam emphasized.

Looking ahead, Ngiam is optimistic about the role of AI in healthcare. “Generative AI opens up new possibilities for improving the efficiency and safety of healthcare tasks. These tools, available to NUHS staff, hold the potential for advancing care for both doctors and patients,” he noted.