Use of AI to predict patient discharge times could ease hospital overcrowding, researchers say

Researchers at Université de Moncton are testing an AI tool designed to predict patient discharge times, potentially easing overcrowding in New Brunswick hospitals. Vitalité Health Network, which operates at a 95% occupancy rate, is collaborating on the project, as delays in discharging patients contribute to pressure on acute care beds. More than a third of patients in these beds could receive care outside hospitals, but limited community services create bottlenecks. The AI program, developed by computer science professor Moulay Akhloufi and researcher Oumeima Thaalbi, analyzes years of anonymized medical data—including admission times, symptoms, and discharge records—to identify patterns and forecast bed usage. It adjusts predictions dynamically if new symptoms or tests arise, offering real-time insights to staff. Early results suggest the AI’s accuracy matches that of experienced nurses, while reducing human error by providing objective data. The tool is not intended to replace clinical judgment but to assist nurses and managers in planning discharges more efficiently. Akhloufi emphasized that patient data is anonymized, with individuals identified only by numbers. The next step is tailoring the algorithm for Dr. Georges-L.-Dumont University Hospital Centre in Moncton, followed by a potential network-wide rollout if the pilot succeeds. Vitalité Health Network’s vice-president of clinical logistics, Jenny Toussaint, noted that hospitals struggle to maintain a safe occupancy rate below 85%. The AI could help address delays, particularly in emergency departments, while preserving care quality. Toussaint stressed the importance of discharge planning from a patient’s arrival to ensure smoother bed management across the system.