Artificial intelligence (AI) has shown its potential in the medical field and has proven to be useful in reading medical images and even passing doctors’ licensing exams. A team at NYU Grossman School of Medicine has designed a new AI tool that can read physician notes and accurately predict patients’ risk of death, readmission to the hospital, and other outcomes that are important to their care. The software, called NYUTron, is currently in use at affiliated hospitals throughout New York, with hopes that it will become a standard part of healthcare. A study on its predictive value was published in the journal Nature.
NYUTron: The AI Tool
NYUTron was designed to use medical notes as its source of data and build predictive models on top of it. The large language model was trained on millions of clinical notes from the health records of 387,000 people who received care within NYU Langone hospitals between January 2011 and May 2020. These notes included any records written by doctors, such as patient progress notes, radiology reports, and discharge instructions, resulting in a 4.1-billion-word corpus. One of the challenges for the software was interpreting the natural language that physicians write in, which varies greatly among individuals, including in the abbreviations they choose.
Researchers tested the tool in live environments, training it on the records from one hospital and seeing how it fared in another hospital with different patient demographics. NYUTron identified 95 percent of people who died in the hospital before they were discharged and 80 percent of patients who would be readmitted within 30 days. It outperformed most doctors on its predictions, as well as the non-AI computer models used today. However, the most senior physician had better predictions than the model.
NYUTron also correctly estimated 79 percent of patients’ actual length of stay, 87 percent of cases where patients were denied coverage by insurance, and 89 percent of cases where a patient’s primary disease was accompanied by additional conditions.
The software’s success in predicting patient outcomes is due to the vast amount of data it was trained on and the ability to interpret natural language. While non-AI predictive models have been around for a long time, they were not widely used in practice because the data they needed required cumbersome reorganization and formatting. NYUTron’s ability to use medical notes as its source of data has the potential to make it a standard part of healthcare.
AI will never be a substitute for the physician-patient relationship. However, it can provide more information for physicians seamlessly at the point-of-care so they can make more informed decisions. NYUTron has demonstrated the potential to bring up the baseline of predictions for patient outcomes and become a useful tool in healthcare.
Leave a Reply