‘AI doctor’ helps hospitals predict readmissions

WEDNESDAY, June 7, 2023 (HealthDay News) — Doctors and hospital officials at New York University are using an artificial intelligence (AI) computer program to predict whether a newly discharged patient will soon become sick enough to be hospitalized. Can be recruited again.

The AI ​​program “NYUTron” reads doctors’ notes to estimate a patient’s risk of death, the likely length of their hospital stay, and other factors important to their care.

Tests showed that NYUTron could predict four out of five patients who would require hospital readmission, according to a report published online June 7 in the journal Nature.

NYUTron is what its developers call a “large language model” that can read and understand the creative, personal notes that doctors often take.

The researchers said this is an improvement over previous health care computer algorithms that required data to be specially formatted and organized into ordered tables.

“Our findings highlight the potential of using large language models to guide physicians in patient care,” said lead researcher Lavender Jiang, a doctoral student in NYU’s Center for Data Science and lead author of the study.

“Programs like NYUTron can alert health care providers in real time to factors that could lead to readmissions and other concerns, so they can be addressed quickly,” Jiang said in a news release from the school. Can be achieved or avoided.”

Jiang and his colleagues trained NYUTron to scan unformatted text from electronic health records and make useful assessments about a patient’s health status from what it learned.

The results of the study showed that the program could predict about 80 percent of those who were readmitted, which was about a 5 percent improvement compared to a standard computer program that required reformatting medical data. Is required.

By automating basic tasks, such technology could give doctors more time to spend with their patients, Jiang said.

Large language models work by predicting the best word to complete a sentence, based on the likelihood that real people will use a particular word in that context.

Jiang explained that the more data you feed the computer to teach it to recognize those word patterns, the more accurate its predictions will become over time.

Researchers trained NYUTron using millions of clinical notes collected from the electronic health records of 336,000 men and women who received care within the NYU Langone hospital system between January 2011 and May 2020.

The study’s authors reported that this resulted in a language “cloud” of 4.1 billion words, which included any records written by the doctor, such as radiology reports, patient progress notes and discharge instructions.

Importantly, there was no standardized language in clinical notes, forcing programs to learn to interpret abbreviations and terms unique to a particular author.

The researchers reported that in the test, NYUTron identified 85 percent of people who were going to die in the hospital (7 percent improvement over standard methods) and predicted the actual length of stay for 79 percent of patients (12 percent improvement over standard models).

The tool also successfully assessed the likelihood that a patient might have additional conditions in addition to their primary disease, as well as the likelihood that insurance would deny coverage.

“These results demonstrate that large language models make the development of ‘smart hospitals’ not only a possibility, but a reality,” said lead researcher and neurosurgeon Dr. Eric Orman. “Because NYUTron reads information taken directly from electronic health records, its predictive models can be easily built and quickly deployed into the health care system.”

Orman said future studies could explore the model’s ability to extract billing codes, predict infection risk and identify the right medication to order.

However, Orman emphasized that NYUTron is supposed to be a helpful tool for healthcare providers, not a replacement for doctor’s judgment tailored to an individual patient.

Funding for the study was provided in part by the US National Institutes of Health.

more information

The Brookings Institute provides more information on the major language models.

Source: New York University Grossman School of Medicine, press release, June 7, 2023

Source link

About Admin

Check Also

An oral interleukin-23 receptor antagonist peptide for plaque psoriasis.

Original message: In this Phase II trial, individuals with moderate to severe plaque psoriasis were ... Read more

Leave a Reply

Your email address will not be published. Required fields are marked *