AI Tool Accurately Predicts Post-Discharge Skilled Nursing Needs

An artificial intelligence (AI) tool has demonstrated the ability to predict which patients will require a skilled nursing facility after hospital discharge. This significant finding is based on a new study led by researchers from NYU Langone Health.

The study suggests that promptly identifying these patients could support hospitals in earlier planning for complex post-discharge care. The research, published in the Nature-family journal npj Health Systems, found that a model using short, AI-generated summaries of doctor notes was more accurate than models relying on the original, lengthy notes.

The research found that a model using short, AI-generated summaries of doctor notes was more accurate than models relying on the original, lengthy notes.

How the AI Method Works

The innovative method employs a two-step AI process:

• A generative AI model first summarizes key risk factors from a patient's admission notes, producing an "AI Risk Snapshot."
• A second AI component then utilizes this snapshot to predict the need for skilled nursing care with an impressive 88 percent accuracy as inpatient hospitalizations conclude.

The AI-generated summaries were 94 percent shorter than the original doctor notes, addressing a challenge where original notes were often too long for AI models to process effectively.

Validation and Future Steps

To ensure the AI's predictions were sound, human experts thoroughly reviewed the AI-generated summaries. Assessments from nurse case managers aligned strongly with the AI's risk scores. Notably, a high-risk score from the model made it 13.5 times more likely that a nurse would independently flag a patient as needing skilled nursing care.

Researchers now plan to test this model in a real-world clinical setting. The aim is to assess its effectiveness in actual discharge planning and to continuously monitor the system for fairness and safety.