Doctors enjoy a prestigious reputation worldwide for their dedication to saving lives, often being revered as ‘living’ gods in certain cultures. However, even these esteemed professionals are unable to accurately predict a patient’s chances of death. In an interesting development, a team at NYU Grossman School of Medicine has created an Artificial Intelligence (AI) tool that claims to possess this very capability.
According to a report by AFP, the AI-based tool, known as NYUTron, has demonstrated promising results in analyzing physicians’ notes and accurately predicting patients’ risk of death, hospital readmission, and other important outcomes related to their care.
The team behind NYUTron envisions that it will eventually become a standard component of healthcare practices. Currently, NYUTron is being used by NYU’s affiliated hospitals in New York.
A study published in the journal Nature highlights the predictive capabilities of this AI tool. Lead author Eric Oermann, a neurosurgeon and computer scientist at NYU, explained that while predictive models in medicine have existed for some time, their limited usage can be attributed to the challenges of organizing and formatting the necessary data.
Recognizing that physicians consistently write detailed notes about their clinical observations and patient interactions, the team identified this as a valuable source of data for building predictive models. The team sought to leverage medical notes as the foundation for their data and construct predictive models on top of it.
To train NYUTron, researchers utilized millions of clinical notes from the health records of 387,000 patients treated at NYU Langone hospitals between January 2011 and May 2020. These notes encompassed various types of records written by doctors, resulting in a massive corpus of 4.1 billion words.
The software made predictions based on this data, which were retrospectively evaluated against real outcomes. Overwhelmingly, NYUTron demonstrated impressive performance, correctly identifying 95% of patients who died in the hospital before discharge and predicting 80% of patients who would be readmitted within 30 days.
Moreover, NYUTron accurately estimated the length of stay for 79% of patients, correctly identified cases where insurance coverage was denied in 87% of instances, and accurately recognized additional conditions accompanying primary diseases in 89% of cases. It outperformed non-AI computer models currently in use and surpassed the predictions made by most doctors.
However, the study revealed that the most experienced physician outperformed the tool, achieving results that exceeded the model’s predictions. Oermann emphasized that AI can never replace the physician-patient relationship. Instead, tools like NYUTron aim to provide physicians with more information seamlessly at the point of care, enabling them to make more informed decisions. The goal is to enhance the baseline of medical care rather than deliver consistently superior results.
Post Your Comments