Artificial Intelligence Can Detect Patients at Risk of Delirium

The researchers focused on the fact that delirium occurring in hospitalized patients is often not detected by doctors in time — or at all. If left unnoticed, this dangerous condition can significantly prolong a patient's hospital stay. Moreover, the occurrence of delirium increases the risk of mortality and worsens long-term treatment outcomes.

Real-Time Calculation

The study’s lead author, Joseph Friedman, founder and director of Delirium Services for the Mount Sinai Health System, said the researchers were the first to try to change the unsatisfactory situation by creating an AI model capable of accurately calculating the risk of delirium in real time. Moreover, it integrates smoothly into clinical workflows, helping hospital staff identify and treat even those patients with delirium who might otherwise be overlooked.

Several studies on predicting delirium have already been conducted. For instance, in April 2025, a study was published in which scientists developed a machine learning model to predict delirium in ICU patients using only continuous physiological data. Another study focused on a model capable of identifying hospitalized patients at increased risk of delirium within the next 24 hours. However, none of these studies dealt with real-time delirium prediction.

A Practical Model for Clinical Use

The study published this May is also unique because the authors chose not to create the AI model in isolation and only test it in a hospital setting afterward. Instead, they closely collaborated with doctors and hospital staff from the outset. This strategy allowed them to refine the model in real time and ensure its effectiveness and practicality for clinical use.

The researchers used the AI model to analyze data from more than 32,000 patients admitted to Mount Sinai Hospital. These were notes and records from doctors and other healthcare staff taken from electronic health records. Using these notes and insights, the model identified data patterns associated with a high risk of delirium. Thanks to entries on even subtle changes in patients’ mental states, the model's accuracy improved over time.

Alerts for At-Risk Patients

According to the researchers, delirium can affect up to one-third of hospitalized patients. After integrating the AI model into Mount Sinai’s hospital system, scientists observed a dramatic improvement in the detection of delirium, leading to a 400% increase in identified cases without increasing the time spent screening patients.

The model can identify a patient at high risk of delirium and then alert a specially trained team to assess the patient's condition and create a suitable treatment plan if necessary.

Identified patients can begin treatment earlier and receive lower doses of sedatives, potentially reducing side effects and improving overall care.

More Efficient Work for Physicians

The study was the first to demonstrate that an AI-generated delirium risk model can not only function well in a lab setting but also bring improvements to clinical practice in a real hospital environment. A major advantage is the improved efficiency of physicians, as the analysis of vast amounts of patient data is entirely within the domain of artificial intelligence.

The model has shown good results at Mount Sinai Hospital. However, it will be necessary to integrate and validate it in other hospital systems to reliably assess its performance and make adjustments if needed.

Editorial Team, Medscope.pro

Sources:

1. Friedman J. I., Parchure P., Cheng F. et al. Machine Learning Multimodal Model for Delirium Risk Stratification. JAMA Netw Open 2025 May 1; 8 (5): e258874, doi: 10.1001/jamanetworkopen.2025.8874.

2. Park Ch., Han Ch., Jang S. K. et al. Development and Validation of a Machine Learning Model for Early Prediction of Delirium in Intensive Care Units Using Continuous Physiological Data: Retrospective Study. J Med Internet Res 2025 Apr 2; 27: e59520, doi: 10.2196/59520.

3. Shaw K. M., Shao Y-P., Ghanta M. et al. Daily Automated Prediction of Delirium Risk in Hospitalized Patients: Model Development and Validation. JMIR Med Inform 2025 Apr 18; 13: e60442, doi: 10.2196/60442.