Artificial intelligence approach to optimise safety for hospitalised patients with dementia (3 September 2025)

Summary

An AHRQ-funded study published in BMJ Open Quality found that machine learning tools can provide insight into the causes of patient safety events for hospitalised people living with dementia.

Researchers evaluated 1,387 dementia-related patient safety event records from a diverse 10-hospital health system from January 2018 to July 2023.

After categorising the free-text reports by contributing factors, the team developed two machine learning models to identify whether the events were caused by situational factors, such as patient-related care challenges, or active failures, such as those caused by staff. The models were able to identify situational factors—such as agitation, wandering, or mobility challenges—that drive most dementia-related safety events with over 70% accuracy, but detected just over 8% of errors caused by staff.

These findings suggest that AI has strong potential to inform targeted training, medication management, and systemwide interventions to enhance safety for hospitalised patients with dementia.