About a year ago, anesthesia staff at Erlanger Baroness, Chattanooga’s largest hospital, noticed that a nurse had slurred speech and was struggling to stay awake while working in the surgery center, according to a consent order from the Tennessee Board of Nursing.
Over the next few days, the nurse failed drug tests and was fired, the order states. The nurse later admitted that, as ordered, she stole and abused fentanyl left over after surgery, sometimes on a daily basis for several months.
Under most circumstances, this would be a routine case of so-called “drug diversion,” the illegal removal of controlled substances from a medical facility. This is so widespread that it is believed to occur in nearly every hospital in the United States.
But the Erlanger case stands out because it should have alerted high-tech watchdogs.
The hospital uses Sentry 7, the latest line of defense against drug diversion. Sentry 7 is drug monitoring software that leverages artificial intelligence and is designed to detect lost drugs faster than humans. But for several months at Erlanger, Sentry 7 failed to issue alarms and overlooked missing medications and other “inconsistencies” that “should have been alerted to,” the nursing board’s order states.
The Erlanger incident, previously unreported, is a rare glimpse into the glaring flaws in AI drug repurposing software used in hundreds of U.S. hospitals with little transparency or oversight. Healthcare facilities are not obligated to disclose their implementation of this type of software or report malfunctions to anyone, so there is no complete explanation of how widely these programs are used or how often they fail.
Baroness Erlanger (also known as Erlanger Medical Center) declined to comment on the use of Sentry 7 or the repurposed drug. André Rebelo, a health department spokesperson for Wolters Kluwer, the Dutch technology company that developed Sentri7, declined to answer questions about what happened with Erlanger, but said the company remains “confident in our software.”
low transparency
David Lustall, a neurologist and AI researcher at Johns Hopkins University, said AI technology is often proprietary and hospital officials often don’t understand how it works, so a lack of transparency means errors can go uncorrected. This means the mistake could be repeated at other hospitals.
“Ideally for patients, caregivers, and hospital systems, when it turns out that the AI is making some sort of error, it would be very transparent and public,” Rustall said.
The Drug Enforcement Administration requires hospitals to confidentially report lost or stolen medications. Hospitals can also report stolen medicines to state health agencies, which license medical professionals and investigate fraud.
However, these reports do not need to include details about the AI software involved, according to interviews with three anti-drug trafficking experts. In interviews, all said they had never seen an obvious AI failure like the one that occurred with Erlanger publicly documented.
“I personally have never seen these technologies specifically referred to as such,” said Jacob Smith, a Johns Hopkins drug security pharmacist, regarding Sentry7’s apparent failure. “I don’t understand how anyone could miss that.”
Smith and other experts said the Erlanger case also raises questions because theft of leftover drugs is one of the most well-known methods of diversion. And the painkiller fentanyl, which is 50 times more potent than heroin, is one of the most common targets.
Terri Vidals, founder of Rxpert Solutions, questioned whether the Erlanger incident was the result of user error rather than a malfunction.
“This is the fundamental of the fundamentals of this software,” says Vidalus. “I think it’s interesting that they say it wasn’t flagged by the software. I think there’s probably more to this story.”
The apparent failure of Sentry 7 at Erlanger was revealed in a state disciplinary order issued by the Tennessee Department of Health in December. Those records also included a Board of Nursing order summarizing a state investigation into nurse anesthetist John Stevenson, who settled the lawsuit against him by signing a document in November.
Mr. Stevenson, through his lawyer, declined to comment. He has not been charged with any crime related to the Erlanger incident. The Board of Nursing placed his license on probation while he received drug counseling.
Bill Christian, a spokesman for the Department of Health and the Board of Nursing, declined to comment on the Erlanger case or Sentry 7. In response to public records requests, the Department of Health and the Tennessee Health Care Facilities Commission each said they had no other documents regarding the apparent failure of the Sentry 7 at Erlanger.
Erlanger spokesman Charlie Milburn said earlier this year that the hospital prepared a written statement regarding its use of Sentry 7 in response to questions from KFF Health News.
That statement was never released.
“Our legal team is debating whether this is even worth talking about,” Milburn said in a March email, but later declined to answer any questions.
Kristi Drollinger, a Wolters Kluwer executive who spoke generally about Sentry7 on KFF Health News in March, said the software is in high demand because so many hospitals are struggling to secure medicine.
Dollinger said Sentry 7 monitors about 60 “attributes of risk” that identify red flags for further investigation by hospital staff.
“It’s pretty scary,” Dollinger said of the drug theft epidemic. “Every health system, every health care facility has experienced diversion at some point and is probably still diversioning.”
“Road to the future”
Pharmaceutical diversion is a widespread challenge in U.S. healthcare facilities. Patients may not receive their medication or may receive medication contaminated with a blood-borne disease. According to the nonprofit Healthcare Diversion Network, it is estimated that as many as 15% of all healthcare workers have diverted medicine at least once.
Since 1985, diversion has been linked to at least 13 disease outbreaks, causing more than 200 infections, mostly due to hepatitis C, according to the U.S. Centers for Disease Control and Prevention.
To prevent this, hospitals are trying to track each pill or vial from the time it is dispensed to the time it is given to the patient by comparing data from the electronic medication cabinet and the patient’s health record.
Hospital staff once performed this arduous process manually, but over the past decade diversion prevention software has largely automated this task. After years of mergers and acquisitions, two programs now dominate the industry: Wolters Kluwer’s Sentri7 and Bluesight’s ControlCheck. Both are equipped with AI.
“This is definitely the way of the future,” says Luke Overmyer, owner of Diversion Specialists.
According to Bluesight, more than 1,500 hospitals use ControlCheck. Another 700 companies are using the Sentri7 Clinical Surveillance program, which may include the company’s drug repurposing software, Wolters Kluwer said.
Neither company has disclosed the price of their software. Smith, the drug safety officer at Johns Hopkins University, said hospitals are buying these “expensive technologies” because they could face millions of dollars in fines from the DEA if a disastrous diversion incident were to occur.
“Return on investment is not guaranteed,” Smith says. “They promise cost avoidance.”
In 2022, a peer-reviewed study funded by the National Institutes of Health found that Sentri7, then known as Flowlytics, could detect drug repurposing faster than existing methods. The study’s lead author previously worked for Invistics, the company that owned Sentri7.
According to the study, researchers tested the software by combing through drug data spanning two years and 10 hospitals to find 22 nurses who were already known to have diverted drugs.
Not only did the program discover them all, but it did so faster than humans by as much as a week and as much as a year and a half, according to the study.
In Erlanger, humans first spotted signs of trouble.
According to the Board of Nursing’s order, co-workers reported that on or about June 30, 2025, Stevenson appeared to be impaired while “working at the surgery center.”
According to the order, Stevenson “had slurred speech, appeared extremely tired, was seen standing unsteadily with his eyes closed, nodded while standing upright, and appeared to have difficulty keeping his eyes open.”
Under questioning by state investigators, Stevenson admitted in March 2025 that he began diverting “unused fentanyl that would have been disposed of following surgical procedures,” according to the order. Stevenson said he initially used fentanyl waste once or twice a week, then by June of that year “he was using it daily,” the order states.
Erlanger audited Stevenson’s dispensing records over that four-month period. The order says it found about five instances where Sentri7 did not flag missing drugs.
The hospital added that it had discovered “further discrepancies between drug dispensing and waste records that should have been flagged by the automated monitoring system.”
One possible explanation is provided by the Board of Nursing, which said Sentry 7 is in an “early learning phase” at Erlanger, but the board did not provide further details.
In an interview, Drollinger did not mention Erlanger specifically, but said Sentri7 does not have a “learning phase” because when it is introduced into a new hospital, it is trained on nine to 12 months of historical data.
Johns Hopkins’ Smith had a different theory.
Smith said in an interview that his experience with AI drug diversion software has convinced him that while it is effective for monitoring emergency rooms and intensive care units, it is less effective in operating rooms, where drugs are dispensed and recorded differently.
These areas are difficult for AI to track and require careful human monitoring, Smith said.
“We have people whose only job is to work with this software,” Smith said. “Software is part of it, but if you rely on software for all your signals, you’re going to miss something. It’s not 100%.”
