At a Tennessee hospital, a nurse stole fentanyl and AI missed it, state records say
A year ago at Erlanger Baroness, the largest hospital in Chattanooga, a nurse was discovered to be struggling with drug diversion. The nurse, John Stevenson, was found to be pilfering and abusing fentanyl left over after surgeries, leading to his dismissal after failing a drug test. What made this case unique was the failure of the hospital’s AI drug diversion software, Sentri7, to detect the missing drugs and inconsistencies in Stevenson’s actions.
AI technology in healthcare, such as Sentri7, is often proprietary and lacks transparency. This lack of transparency can lead to errors being buried rather than fixed, potentially putting patients at risk. While hospitals are required to confidentially report lost or stolen drugs, they are not mandated to disclose details about any AI software involved in drug diversion cases. This lack of oversight and transparency raises concerns about the effectiveness of AI in preventing drug diversion.
Experts in drug diversion prevention question whether the failure of Sentri7 at Erlanger was due to user error or malfunction. The theft of leftover drugs, especially potent painkillers like fentanyl, is a common method of diversion and should have been flagged by the software. However, the case at Erlanger raises doubts about the reliability of AI technology in detecting drug diversion.
The Erlanger case highlights the challenges hospitals face in preventing drug diversion and the potential limitations of AI software in this area. While AI technology is seen as the future of drug diversion prevention, it may not be foolproof in all settings. Human oversight and vigilance are still crucial in detecting and preventing drug diversion in healthcare facilities.
As hospitals continue to invest in AI drug diversion software, the need for transparency and accountability in the use of these technologies becomes increasingly important. Patients’ safety and well-being should always be the top priority, and any failures or shortcomings of AI software must be addressed openly to prevent similar incidents in the future.
