Why It Matters
Tennessee’s healthcare system is raising difficult questions about the reliability of artificial intelligence tools designed to protect patients and prevent drug theft inside hospitals. A documented case at one of Chattanooga’s largest medical facilities suggests that widely used AI-powered software may carry significant blind spots — and that current disclosure rules leave the public with little ability to know when the technology fails.
What Happened
Erlanger Baroness Hospital in Chattanooga uses a drug diversion detection system called Sentri7, an AI-powered platform built to identify missing or improperly accessed controlled substances. For several months, however, the software failed to flag repeated thefts of fentanyl by a nurse anesthetist identified in state records as John Stevenson.
Stevenson admitted to taking leftover fentanyl after surgical procedures — sometimes on a daily basis — and using it himself. His behavior eventually drew attention from coworkers in the anesthesia department, who noticed him slurring his words and struggling to remain alert while on duty. Stevenson was subsequently tested for drugs, failed, and was fired. He later settled the matter by signing a disciplinary order with Tennessee’s Board of Nursing in November.
The board placed his license on probation and required him to complete drug counseling. As of the time state records were released, Stevenson has not faced any criminal charges.
The Tennessee Department of Health disclosed the board’s order in December as part of a routine release of disciplinary actions. The order explicitly states that Sentri7 overlooked missing drugs and other irregularities that regulators said “should have been flagged.”
By the Numbers
- Months: The duration over which Stevenson stole fentanyl before being caught
- Months: The period during which Sentri7 failed to generate any alerts
- ~60: The number of risk indicators Sentri7 is designed to monitor simultaneously
- 50x: How much more potent fentanyl is compared to heroin
- Hundreds: The approximate number of U.S. hospitals reported to use this category of AI drug diversion software
Experts Say Documented AI Failure Is Rare — and Troubling
Three specialists in drug diversion prevention said they were unaware of any prior instance in which an AI monitoring system’s failure had been documented so specifically in public records. The concern, they noted, is not just what happened at Erlanger — it is what remains hidden at other facilities.
Jacob Smith, a pharmacist at Johns Hopkins Medicine, said he found the oversight difficult to explain, noting that a case this clear-cut should have triggered the software’s detection mechanisms. “I’ve never myself seen these technologies be called out in that specific way,” Smith said. “It doesn’t make sense to me how you could miss it.”
David Rastall, a neurologist and AI researcher also affiliated with Johns Hopkins Medicine, argued the situation illustrates a broader problem with how AI failures in healthcare are handled. “The ideal for patients, caregivers, and hospital systems would be, when an AI is found to be making some type of error, that becomes very transparent and public,” Rastall said.
Erlanger Baroness declined to comment on its use of Sentri7. André Rebelo, a spokesperson for Wolters Kluwer — the company behind the software — declined to address the specific circumstances at Erlanger but said the company remains confident in its product.
A Regulatory Gap
Healthcare facilities operating in Tennessee are under no legal obligation to disclose when they deploy drug diversion software, nor are they required to report instances when that software malfunctions or fails to detect a problem. That absence of mandatory reporting means cases like the one at Erlanger may surface only by chance — in this instance, through a routine state disciplinary filing.
As AI tools become more common in hospital operations, the Erlanger case adds weight to calls for greater transparency and accountability standards around their use. The lack of public documentation on AI failures makes it difficult for regulators, patients, or competing healthcare systems to assess how often such tools fall short.
What’s Next
No legislation requiring disclosure of AI monitoring failures in Tennessee healthcare settings is currently in place. Whether the Erlanger case prompts regulatory attention at the state or federal level remains to be seen. Stevenson’s nursing license remains on probation, and no criminal proceedings have been announced. Wolters Kluwer has not indicated any plans to publicly address the software’s performance in this case.
For broader context on Tennessee policy developments, see recent coverage of Tennessee lawmakers and redistricting decisions that reflect the state’s evolving regulatory and legislative environment.