Geriatrician urges careful use of AI predictions for older patients

Artificial intelligence tools are being added to electronic health records to flag patients at risk for conditions such as sepsis, falls or even death. The ease of integration has led many clinicians to accept the predictions without questioning how well the models work.

James Deardorff, a geriatrician and assistant professor at the University of California, San Francisco, has created several AI models that estimate outcomes for older adults, including mortality risk and the likelihood of needing nursing-home care. He emphasizes that the performance of these algorithms must be examined in detail, especially for sub-populations like the very elderly.

Deardorff warns that favorable predictions do not automatically translate into better patient outcomes. He calls for clinicians to understand the limitations of the models and to use the results responsibly, arguing that AI for older patients deserves careful scrutiny.