2023-06-08 15:13:17
Artificial intelligence (AI) has already proven that it can analyze medical imaging and pass medical student exams. Recently, an AI-powered tool demonstrated its ability to read doctors’ reports and accurately predict the risk of death, hospital readmission and other possible complications.
The software was created by a team from the Grossman School of Medicine in New York. It is now being tested in several of the university’s partner hospitals in order to make it standard practice in the medical community in the future. A study on its possible interest was published this Wednesday in the journal Nature.
4.1 billion words analyzed
Predictive models not based on AI have been around for several years but they are little used because they require a lot of data entry and formatting work. The researchers had the idea of using medical notes and doctors’ assessments as a database.
The predictive model, called NYUTron, was formed from millions of medical observations from the records of 387,000 patients treated between 2011 and 2020. They included doctors’ written reports, notes on the evolution of the state of patients, x-rays, medical imaging and recommendations issued to patients upon discharge, i.e. a corpus of 4.1 billion words.
Stunning results
By looking at what happened to the patients, the researchers were able to measure the number of times the AI predictions turned out to be correct. Disturbing Finding: NYUTron identified 95% of patients who died at partner hospitals prior to discharge and 80% of those who were readmitted within a month of discharge.
His predictions therefore exceeded those of most doctors, like those of non-AI-based computer models currently in use. The software also successfully predicted 79% of patients’ length of stay in hospital, 87% of cases in which patients were denied reimbursement for care by their insurance and 89% of cases in which the patient suffered from pathologies additional.
AI will never replace the patient-doctor relationship, says Eric Oermann, lead author of the study. But it could make it possible “to provide more information […] physicians to make informed decisions.
1686270304
#predicts #patients #medical #complications #doctors