A revolution is brewing in medicine. We knew artificial intelligence (AI) capable of analyzing medical imaging and ping medical student exams with flying colors… a new string has now been added to its bow.
A team from the Langone School of Medicine in New York, the Grossman School of Medicine, has developed an AI-based tool capable of anticipating, very precisely, the risks of death, readmission to hospital and other possible complications for patients.
4.1 billion words analyzed
The software, called “NYUTron”, was formed from millions of medical observations from the files of 387,000 patients treated between January 2011 and May 2020. First lesson: it identified 80% of people readmitted to hospital a month after their release. If prediction models already existed, they were less accurate – by 5% in this case – and difficult to use, as they required heavy data entry and reformatting.
This is not the case with this new software. NYUTron is able to interpret the language used by each doctor. And this, despite differences in how professionals take notes, especially in the abbreviations they use.
95% of deaths identified
Written reports from doctors, notes on the evolution of the patient’s condition, x-rays and medical imaging, but also recommendations given to patients when they leave the hospital… these documents were all read by NYUTron to make their estimates. . In total, some 4.1 billion words were analyzed.
The software also identified 95% of patients who died in hospital. NYUTron is also able to predict, in 79% of cases, the length of stay of patients in hospital. 12% increased accuracy compared to non-AI based computer models.
The tool was also able to identify 87% of cases in which patients would be refused reimburt for care by their insurance, and 89% of cases in which the patient suffered from additional pathologies. These results exceed the predictions of most doctors. This predictive model could ultimately allow them to make more informed decisions.
Alert in real time
“Programs like NYUTron can alert health care providers in real time to factors that could lead to readmission and other issues so they can be quickly addressed or even prevented,” said lead author Lavender Jiang. of the study and a doctoral student at the NYU Center for Data Science. By automating basic tasks, this technology could allow doctors to spend more time talking with their patients, she says.
The ambition is, in the future, to make NYUTron a common tool in the medical field. If AI is not intended to replace doctors, its use could still increase. The researchers’ predictive model could thus be used to produce bills, predict the risk of infection or even identify the right drug to order, says Eric Oermann, neurosurgeon and computer engineer from the New York Faculty of Medicine.