Categories AI

AI Tool Predicts Post-Discharge Care Needs for Patients

Introducing an Innovative AI Tool for Hospital Discharge Planning

A new AI model has been developed to assist hospitals in identifying patients who may require skilled nursing care following their discharge. This cutting-edge tool aims to enhance planning processes and alleviate stress for both patients and caregivers, as highlighted in a recent study published in npj Health Systems.

This study is particularly relevant to skilled nursing facilities (SNFs), which provide essential short-term, intensive care and rehabilitation services. Notably, about 15% of patients at NYU Langone Health transition to skilled nursing facilities after their hospital stay.

Understanding the AI Tool

The researchers conducted an analysis of electronic health records for 4,000 patients admitted to general medicine services at NYU Langone. They concentrated on the history and physical admission notes, which contain valuable details about a patient’s health status, functional capabilities, and social circumstances.
To create their tool, the researchers developed a generative AI model that reviews each admission note, pinpointing and summarizing information related to seven specific risk factors into a streamlined ‘AI Risk Snapshot’. In their testing phase, they evaluated nine different AI models to determine which one most accurately predicted a patient’s discharge destination. Remarkably, they discovered that the model utilizing concise AI-generated summaries of doctor notes outperformed those using extensive, traditional notes.

The accuracy of the AI tool was validated by comparing its predictions with evaluations from human experts, resulting in a strong consistency with AI-generated risk scores. The researchers found that the AI tool can predict, with an impressive 88% accuracy, whether patients will require skilled nursing care post-discharge.

“Our two-step approach acts like an efficient, meticulous reader, transforming complex medical notes into simplified summaries that are crucial for discharge planning,” explained senior study author Yindalon Aphinyanaphongs, MD, PhD, who is the director of operational data science and machine learning at NYU Langone, and a research professor in the Departments of Population Health and Medicine at NYU Grossman School of Medicine.
“The next step involves evaluating this model in a real clinical environment to assess its effectiveness in helping our care teams facilitate smoother discharges for all patients,” stated first author William R. Small, MD, a clinical assistant professor in the Department of Medicine. “We will also oversee the system to guarantee fairness, safety, and improved patient care.”

The Importance of Predicting Discharge Destination

Early identification of inpatients needing ongoing support at skilled nursing facilities is vital for care teams. By recognizing these needs during a patient’s hospital stay, teams can efficiently coordinate with SNFs, arrange for transportation, and prepare necessary documentation while also keeping patients and families informed. This not only enables a smoother transition from hospital to post-discharge care but also minimizes the risk of discharge delays, ensuring patients receive appropriate care promptly upon leaving the hospital.

Leave a Reply

您的邮箱地址不会被公开。 必填项已用 * 标注

You May Also Like