In February 2009, Lauren Lollini, a psychotherapist from Colorado, had a kidney stone removed at Rose Medical Center in Denver. It was a routine procedure, but for weeks after her surgery, Lollini experienced flu-like symptoms, including nausea and extreme fatigue. When she went back to the hospital to find out what was going on, doctors diagnosed her with Hepatitis C, a tough-to-treat liver infection that is most often spread among intravenous drug users who share needles.
Lollini, who was then in her early 40s, didn’t present any of the other usual risk factors for Hepatitis C. She hadn’t gotten any tattoos, and as far as she knew, hadn’t been exposed to contaminated blood. Lollini’s first thought was that she may have contracted it during her surgery, but her doctors dismissed that idea.
Some months later, Lollini got a phone call from the Centers for Disease Control and Prevention (CDC), which told her it was investigating an outbreak. Unbeknownst to her, she was one of two patients who had surgeries mere days apart at the same facility, and had both contracted Hepatitis C. Within weeks, at least 18 patients who had undergone procedures at Rose Medical were diagnosed with a virtually genetically-identical strain of Hepatitis C.
An investigation revealed they had all been infected by Kristen Parker, an IV drug user who worked as an operating room technician at Rose Medical between October 2008 and April 2009. During this period, Parker allegedly stole syringes filled with Fentanyl, a powerful opioid-based painkiller, injected herself with the drug, then refilled the syringes with saline and placed them back on surgical trays. Parker, who reportedly gave the disease to as many as three dozen people, was later sentenced to 30 years in federal prison, where she remains today.
Drug diversion—or put simply, stealing drugs—in healthcare facilities is an increasingly serious public health problem in the United States that affects patients, hospitals, law enforcement and healthcare systems more broadly. It harms patients who may not receive the pain medication they need, or in extreme cases like Lollini’s, may get sicker than they were before. It affects hospitals, which take on the costly burden of investigating and reporting these cases, in addition to hefty fines from the Drug Enforcement Agency (DEA). It also puts additional stress on healthcare systems, for instance by diverting Medicare and Medicaid funds from patients who need them most.
One investigator conservatively estimates there are as many as 37,000 instances of diversion in the United States each year, a figure he calls “low at best.” Other studies claim that up to 10% of pharmacists, nurses, and anesthesiologists are now stealing drugs at work. Earlier this month, a former Veterans Affairs hospice nurse was indicted for allegedly ingesting liquid morphine meant for patients in her care. In another recent case, a Virginia doctor was arrested for allegedly prescribing drugs to a patient on the condition he split the pills with the doctor’s wife.
“Kristen Parker was one person and they could paint her out to be the one rogue employee who…slipped through the cracks,” Lollini told Quartz. “But there are a lot of Kristen Parkers out there.”
In response, technology companies have sought to fill a gap by selling increasingly-sophisticated software aimed at helping hospitals and pharmacies track controlled substances and catch drug diverters. One example is Kit Check, a company that sells software called Bluesight for Controlled Substances, which uses a machine-learning algorithm the company claims can identify risky prescription and dispensation patterns among healthcare staff. Another, Invistics Corp., offers cloud-based controlled substance tracking software it calls Flowlytics, which is billed as a diversion detector. A third, BD HealthSight, uses artificial intelligence to analyze “multiple dispensing behaviors.”
But experts say the technology is not perfect and even when it works, hospitals don’t always act upon the information the algorithms provide. That not only exacerbates the overall problem but also potentially exposes healthcare facilities to costly civil lawsuits. This raises an obvious question: Can technology provide solutions to the fundamentally human problem of addiction?
Becoming a priority
Before retiring in 2015, John Burke served as the commander of the Cincinnati Police Department’s Pharmaceutical Diversion Squad. These days he leads the International Health Facility Diversion Association (IHFDA), a nonprofit that advocates for solutions to drug diversion by healthcare personnel. (Kit Check is a corporate sponsor of IHFDA.)
During Burke’s tenure, he said the Cincinnati police arrested one nurse a week on average for drug diversion. If there is a “typical” case, it is people like Kristen Parker—doctors, nurses, pharmacists, and surgical techs who steal controlled substances from their place of work to feed an addiction.
Among the cases US attorneys general have prosecuted this year are a nurse in Montana who swapped her patients’ pain pills for over-the-counter medication; a pharmacist in Kansas who unlawfully gave her husband more than 20,000 tablets of hydrocodone that he used or sold; and a nurse working in a nursing home in Iowa who stole painkillers from her elderly patients.
“They’re just like anybody else,” James Geldhof, a former DEA agent who now serves as a consultant on lawsuits related to the opioid epidemic, told Quartz. “You have a bad marriage, you got a bad divorce, issues with the children, money issues, whatever. And one thing leads to another and these drugs are available pretty much on a regular basis.”
These kinds of cases don’t usually attract the attention of the DEA, which uses its limited resources primarily to go after large-scale drug traffickers. And prosecutions aren’t always the answer. Many hospitals prefer to help a cooperative employee get into rehab and avoid a criminal record.
But priorities are changing in the wake of large-scale national diversion scandals. The DEA has recently levied fines against prestigious institutions like Massachusetts General Hospital and the University of Michigan Health System for violating provisions of the Controlled Substances Act, the main tool that allows the DEA to go after individuals and institutions mishandling controlled substances.
The MGH case
In 2015, Massachusetts General Hospital (MGH)—which is one of the world’s premier hospitals—paid the United States $2.3 million to settle allegations that weak oversight allowed hospital employees to improperly divert drugs for their own use.
The hospital was accused of violating sections of the Controlled Substances Act in 10 different instances (pdf, p. 2). In one instance, it failed to appropriately report two cases where nurses stole about 16,000 pills of oxycodone and other drugs. When DEA investigators conducted an audit of MGH pharmacies, they found 16,681 missing or extra pills at the inpatient pharmacy, and 7,177 missing or extra pills at the outpatient pharmacy.
At the time, the settlement was the largest of its kind, and it forced the hospital to enter into a three-year “Corrective Action Plan” (pdf, p. 12) with the DEA. Among other things, the DEA required MGH hire a drug diversion compliance officer and establish a drug diversion team, restrict access to its pharmacy vault, buy controlled substance surveillance software, and send the reports to department heads.
Christopher Fortier, MGH’s chief pharmacy officer, started working for the hospital after the diversion scandal broke. He was tasked with finding a solution. Thanks to the new DEA-mandated controls, he said the hospital identified numerous diverters right away. But when the software alerts “leveled off” and reports from other employees of potential diverters were still coming in, he knew the hospital needed a better system.
“We are very much still somewhat looking for the needle in the haystack,” Fortier told Quartz. “Even with the surveillance systems that are out there currently, we’re still doing a lot of manual work.”
MGH, which now uses a non-AI-powered automated system to track its drugs, is working to implement an AI system by the beginning of 2020. But, conscious that there’s only so much technology can do, Fortier said MGH has also created a “comprehensive controlled substance surveillance and compliance program.”
“It’s not just putting [in] a surveillance system and saying, ‘We’re good,’” he said. “It’s about a drug diversion team that is reviewing data and always looking to identify where there [are] gaps, and working to close those gaps. There’s a response team in terms of when we identify somebody, how are we investigating it and what are we doing for that employee. It’s human resource policies…It’s education of our staff.”
Technology is only “half the equation”
On average, hospitals discover instances of diversion between 24 and 33 months after they occur. Technology, however, could improve that average by spotting trends and outliers more quickly. Did a nurse sign a controlled substance out at noon but dispose of it at 3 pm? Did a pharmacist dispense a drug while they weren’t on shift? Automated programs could quickly identify those kinds of discrepancies and alert hospital officials.
“Even though it didn’t explicitly say in the record that they had diverted that pill, there’s a bunch of other metadata around it that we can use to get to the heart of diversion anyway,” Kevin MacDonald, Kit Check’s CEO, told Quartz.
Kit Check didn’t start out with a focus on drug diversion. First launched commercially in 2011, it was initially meant to help hospitals become more cost efficient. But after hearing from hospital staff that diversion was a growing problem, and having amassed a huge dataset from multiple facilities over several years, in 2018 Kit Check introduced Bluesight, its diversion-control platform.
The system uses data feeds from different systems, including electronic medical records and dispensing cabinets. An algorithm then gathers and interprets the information generated throughout the dispensing process. It assigns risk scores to each doctor, nurse, technician, and staffer in the facility based on their dispensing patterns and associated behaviors. By establishing a baseline of “normal” patterns and behaviors, administrators can look for those that differ. Outliers are flagged and their individual transactions analyzed. Kit Check does not include the social media postings of hospital staff in its analyses, but hasn’t ruled out that possibility in the future, MacDonald said in an interview last year, raising some questions about individual privacy.
Situations and circumstances—as well as people—are always changing, however, and the risk scores are dynamic. MacDonald concedes that a high score doesn’t always mean that person is stealing drugs. An alert may instead encourage hospital officials to speak with that particular employee and give them a chance to explain themselves.
Other companies offer systems that work in similar ways, generally using a combination of physical tracking and behavioral analysis to monitor staff. Kit Check, in fact, was recently sued by a competitor for patent infringement. (That case has been settled.) The company is also suing a former supplier for allegedly doing the same thing.
Programs like Kit Check’s are not a silver bullet. Artificial intelligence is limited by the information it receives, and operator error is always a factor. Surveilling employees can be a sensitive issue, which must be performed with appropriate controls in place. Further, accurate spotting of trends doesn’t guarantee the hospital will act on the information—bad press and lawsuits are two things most hospitals try at all costs to avoid.
That’s why at MGH, Fortier is looking to implement a system that combines the new technology with human oversight. He recognizes that there is no perfect solution, and that diverters are always finding new ways to get around the system. “We track month to month, quarter to quarter, year to year. We’re always looking at trends,” he said.
Geldhof, the former DEA agent, agrees. He says technology only solves “half the equation.”
“The other half is, what kind of action are you going to take when you determine an issue exists?”
In a world where machine learning and artificial intelligence track drug diversion, it’s all too easy to start looking at the problem in terms of numbers—incidence rates, volume of stolen pills, risk scores. But addiction is a human problem, so it’s necessary to design systems that take that into account. As just one example, employees are more likely to report a colleague if they believe that colleague will be treated compassionately.
“What I’ve learned over the years is that addicts are human beings,” Burke said. “And that there’s been times that we have not done everything we can to assist them.”