Health systems must ensure they're working to better understand and prevent safety risks among patients with disabilities, said Tejal Gandhi, MD, chief safety and transformation officer at Press Ganey.
At present, disabilities are not a field hospitals are required to denote in safety reports, making it difficult to identify and understand safety trends for this patient population. As part of ongoing efforts to integrate equity into routine safety work, Press Ganey sought to determine whether AI and language processing models could identify patients with disabilities in its Patient Safety Organization database. The Press Ganey PSO operates the nation's largest database of patient safety events, featuring data from 129 health systems.
For the pilot project, Press Ganey analyzed data on 150,000 patient safety events reported to its PSO between December 2022 and January 2023. Researchers used AI to scan patient safety reports for words and phrases that could identify patients in various disability categories, including mobility impairments, communication disorders, visual impairments, intellectual disabilities and mental health disorders.
Press Ganey identified 1,900 safety events involving patients with disabilities, which were validated by human review. Overall, patients with mental health disorders had the most safety events, followed by patients with communication disorders and visual impairments.
A majority involved precursor safety events, which were defined as deviations from generally accepted performance standards that reach the patient and result in minimal or no detectable harm. Thirteen involved serious safety events that resulted in severe to moderate harm or death. These were most likely to occur among patients with communication disorders.
The study's findings suggest that AI can be used to effectively identify patient populations at risk of preventable harm.
For health systems, the first step is examining their own data and implementing mechanisms to identify patients with disabilities. As health systems conduct this work, Dr. Gandhi said it's also crucial to engage patients and families in conversations around safety improvements.
"You can find and identify these patients, and then you can start going deeper to try to understand what were the types of events they had and what caused them," she told Becker's July 29.
Press Ganey plans to embed the AI tool into its safety-event reporting system and conduct additional studies using larger sample sizes. Dr. Gandhi noted these capabilities can be used to not only identify patients with disabilities but other characteristics that may put them at higher risk for safety events, such as low English proficiency.
"This has opened up a whole realm of possibilities and opportunities for us to go deeper, especially as we try to embed equity into our safety work," she said.