An artificial intelligence program can predict hospital readmission and other patient outcomes based on physician notes, according to research published June 7 in Nature from the New York University Grossman School of Medicine.
The research was conducted using a large language model nicknamed NYUTron that was trained using millions of clinical notes, according to a June 7 news release. These notes were pulled from 336,000 patients who had received care within the NYU Langone hospital system in New York City between January 2011 and May 2020.
Unlike other technology that can only read data that's been reformatted into standardized files, NYUTron can read any patient notes or records as written, even when a physician uses unique abbreviations.
NYUTron correctly identified 85 percent of those who died in the hospital and 79 percent of patients' length of stay. The AI program was also able to identify the likelihood of additional conditions alongside a primary disease and the potential of an insurance denial.