AI Tool Helps Predict Which Patients Need Continued Care After Leaving the Hospital
Forward Summit
Editorial Team
5 min read
January 31, 2026
NYU Langone researchers have developed an AI tool that predicts with 88% accuracy which patients will need skilled nursing care after hospitalization, enabling earlier discharge planning and reducing patient stress.
The Discharge Planning Challenge
Every year, approximately 15% of patients discharged from NYU Langone require care at skilled nursing facilities that provide short-term, intensive rehabilitation services. However, identifying these patients early enough to plan for complex care transitions has been a persistent challenge.
The consequences of delayed planning are significant: patients who are medically ready for discharge but have no safe place to go experience increased stress and uncertainty, while hospitals face operational challenges managing these extended stays.
Research Insight
"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," says Dr. Yindalon Aphinyanaphongs, Director of Operational Data Science and Machine Learning at NYU Langone.
How the Two-Step AI System Works
The research team, led by NYU Langone Health, analyzed electronic health records from 4,000 patients admitted to general medicine services. Their innovative approach uses two distinct AI components working in sequence.
First, a generative AI model reads lengthy admission notes and extracts information related to seven key risk factors, including the patient's living situation and ability to perform daily tasks. This information is organized into a brief "AI Risk Snapshot" that's 94% shorter than the original notes.
Second, a predictive AI model analyzes these condensed snapshots to forecast discharge needs with 88% accuracy. Remarkably, this compressed approach proved more accurate than models attempting to process the full, lengthy doctor notes.
Why Compression Matters
The compression step was critical because nearly all original admission notes were too long for AI models to process effectively. By distilling complex medical documentation down to essential risk factors, the system could both handle the data and make more accurate predictions.
Validation Through Human Expertise
To ensure the AI's reasoning was sound, researchers tested its outputs against human expert assessments. When nurse case managers reviewed the AI-generated summaries independently, their evaluations strongly aligned with the AI's predictions.
The validation was striking: a high-risk score from the model made it 13.5 times more likely that a nurse would independently flag the patient as needing skilled nursing care.
The study was published in the Nature-family journal npj Health Systems and supported by the National Institutes of Health and National Science Foundation.
Real-World Implementation Ahead
"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," explains Dr. William R. Small, clinical assistant professor and first author of the study.
The research team emphasizes their commitment to monitoring the system for fairness and safety as it moves from research to practice, ensuring it improves patient care without introducing bias or unintended consequences.
The Broader Impact
This research demonstrates how AI can enhance clinical decision-making without replacing human judgment. By identifying patients who need complex care planning early in their hospital stay, care teams can proactively arrange appropriate post-discharge support, reducing stress for patients and families while improving operational efficiency.
AI That Augments, Not Replaces
The NYU Langone approach exemplifies a crucial principle in healthcare AI: tools should augment clinical expertise rather than attempt to replace it. The system doesn't make discharge decisions; it provides care teams with early, accurate insights to inform their planning.
At Forward Summit 2026, we'll explore more innovations like this that enhance clinical workflows while keeping healthcare professionals at the center of decision-making. Join us in Orlando this September to discover how AI can support better patient outcomes without compromising the human touch in healthcare.
Ready to Shape the Future of Healthcare?
Join us at Forward Summit 2026 where healthcare professionals and technology leaders come together to drive innovation.