Artificial Intelligence Enhances Emergency Room Admission Predictions for More Efficient Patient Care

Artificial intelligence (AI) is emerging as a transformative tool in emergency departments (EDs), capable of predicting which patients are likely to require hospital admission much earlier than traditional methods. A comprehensive multi-hospital study conducted by the Mount Sinai Health System demonstrated that AI models, trained on data from over one million previous patient visits, can significantly improve patient flow management.

The study involved collaboration with more than 500 ED nurses across seven hospitals, evaluating a machine learning model's ability to forecast hospital admissions. Over a two-month period, predictions made by the AI system were compared with nurses' triage assessments. Results indicated that the AI could reliably identify patients likely to need admission shortly after their arrival, often days or hours in advance.

This advance notice has the potential to optimize resource allocation, reduce overcrowding, and decrease patient wait times in the ED. It also helps avoid "boarding," where patients are admitted but remain in the emergency department due to lack of available beds.

Lead researcher Jonathan Nover highlighted the importance of predictive analytics in healthcare, comparing it to demand forecasting in industries like airlines and hotels. The goal was to develop a system that not only performs well in theory but can be integrated into real-world clinical workflows to enhance decision-making.

Remarkably, the study found that AI predictions alone were as accurate as combining human assessments with machine learning, emphasizing the robustness of the system. The model's training on extensive data allowed it to recognize meaningful patterns related to patient admissions, supporting clinicians in their planning and clinical decisions.

While the research was conducted over a short time and within a single health system, the promising results lay the groundwork for broader application. Future steps include real-time implementation to measure impacts such as reduced boarding times, improved patient throughput, and operational efficiency.

Chief Innovation Officer Robbie Freeman emphasized that AI is a practical tool, not an replacement for clinicians. Instead, it aims to empower healthcare providers by offering timely insights, enabling more personalized and compassionate patient care. The study underscores how combining human expertise with AI-driven insights can transform emergency care delivery.