The medical field has been seeing the exciting integration of artificial intelligence (AI) in recent times. NYU Grossman School of Medicine has now introduced NYUTron, an AI tool capable of reading doctors’ notes and predicting patients’ health outcomes such as risk of death, readmission to the hospital, and more. Currently in use at NYU-affiliated hospitals across New York, the potential this tool holds for healthcare is enormous.
Research on NYUTron’s predictive capabilities was published in the journal ‘Nature’. Its creation stems from an understanding that while non-AI predictive models have been utilized in the medical field, their practical application is limited due to the need for arduous data reorganization and formatting. The team, led by neurosurgeon and computer scientist Eric Oermann, saw an opportunity in the wealth of data present in physicians’ notes, a common aspect of medical practice.
Also Read: Coldplay Lights Up Singapore with a 4-Day Extravaganza: Grab Your Tickets Now!
NYUTron was developed as a large language model trained on clinical notes from the health records of nearly 400,000 patients at NYU Langone hospitals, amassing a 4.1-billion-word corpus. These records included patient progress notes, radiology reports, and discharge instructions. A significant challenge for the software was interpreting the natural language used by physicians, which varied widely, including their use of abbreviations.
The tool’s effectiveness was assessed by looking back at the records and evaluating the accuracy of its predictions. Additionally, the AI was tested in live environments, trained on records from one hospital and then assessed in a different hospital with varying patient demographics.
NYUTron successfully identified 95% of in-hospital patient deaths before discharge and anticipated 80% of patient readmissions within 30 days, outperforming many doctors and non-AI computer models. The AI was also able to estimate patients’ length of stay, cases denied by insurance, and instances where a patient’s primary disease was accompanied by additional conditions with high accuracy.
However, AI tools like NYUTron aren’t designed to replace human doctors. As Oermann points out, they provide physicians with seamless, critical information at the point-of-care, enabling more informed decisions for patient care.
This news is based on The Star.
