An artificial intelligence tool developed by researchers at NYU Langone Health has demonstrated the ability to accurately predict which patients will require skilled nursing facility care after hospital discharge, according to a new study published in npj Health Systems.
The study focused on improving hospital discharge planning by using AI-generated summaries of doctors’ notes. Researchers found that these concise summaries enabled the AI model to predict with 88 percent accuracy whether a patient would need continued care in a skilled nursing facility. The approach could help hospitals plan earlier for complex cases and avoid situations where patients are ready for discharge but lack appropriate post-hospital care options.
“Our two-step approach acts like a fast, careful reader, turning a complex medical note into a simple summary of what matters most for discharge planning,” said Yindalon Aphinyanaphongs, MD, PhD, director of operational data science and machine learning for NYU Langone and research professor at NYU Grossman School of Medicine.
The team analyzed electronic health records from 4,000 patients admitted to general medicine services at NYU Langone. They used an AI model to extract information related to seven risk factors from each patient’s admission note—such as living situation and ability to perform daily tasks—and organized this data into an “AI Risk Snapshot.” This snapshot was about 94 percent shorter than the original notes, making it more manageable for AI processing.
Nine different AI models were tested against both full-length notes and these snapshots. The results showed that models using the summarized information outperformed those using the raw notes due to length limitations in processing.
To validate the system’s reasoning, nurse case managers reviewed the AI-generated summaries without access to its predictions. Their assessments closely matched the AI’s risk scores; notably, patients flagged as high-risk by the model were 13.5 times more likely also to be identified by nurses as needing skilled nursing care.
“Our next step is to test this model in a real-world clinical setting to see if it helps our care teams plan discharges more effectively across all patients,” said William R. Small, MD, clinical assistant professor in the Department of Medicine and first author of the study. “We will also monitor the system to ensure it is fair and safe and helps to improve patient care.”
Other contributors included Ryan Crowley; Kevin P. Eaton, MD; Lavender Yao Jiang; Eric K. Oermann, MD; Chloe Pariente; and Jeff Zhang from MCIT Department of Health Informatics. The research received support from grants provided by the National Institutes of Health and National Science Foundation.
NYU Langone Health operates an integrated health system with multiple inpatient locations in New York and Florida as well as over 320 outpatient sites. It has been recognized for achieving some of the lowest mortality rates nationally (https://www.vizientinc.com/), with top rankings from Vizient Inc., which placed NYU Langone No. 1 among comprehensive academic medical centers nationwide for four consecutive years (https://www.usnews.com/best-hospitals). U.S. News & World Report recently ranked four clinical specialties at NYU Langone as best in the nation (https://www.usnews.com/best-hospitals).
NYU Langone’s system includes two tuition-free medical schools located in Manhattan and Long Island alongside its Perlmutter Cancer Center.
