As recently as last summer, at the Springhill Medical Center in Mobile, Ala., clinical pharmacist Joe Adkins was struggling to keep ahead of possible cases of drug diversion among nursing and physician staff. Drug theft, which is when providers swipe narcotics, mostly for personal use or resale, occasionally plagues most hospitals.
By submitting your email address, you agree to receive emails regarding relevant topic offers from TechTarget and its partners. You can withdraw your consent at any time. Contact TechTarget at 275 Grove Street, Newton, MA.
Like many facilities looking to find efficiencies in managing the flow of pharmaceuticals, the 250-bed, for-profit community hospital uses 25 medication-dispensing machines customized to their various departments, such as medical/surgical, gastrointestinal, emergency room and cardiac.
Tracking, documenting and solving drug diversion is a concern not only to maintain the well-being of employees, as well as a patient safety tenet, but is also a Joint Commission accreditation standard that surveyors examine closely in their triennial visits. But manually examining drug use trends on the various floors was a tedious process, and picking out borderline cases wasn't at all straightforward.
"It was a lot of tracking and a lot of footwork to try and see what was happening, as far as usage patterns," Adkins said, adding that the "eye test" was a key component of detecting drug diversion: If an employee was acting out of character, the hospital tracked their drug use and looked for miscounts in narcotic pills or injectable doses. "We really didn't have a good tool to do that with."
Enter data analytics.
Analytics system detects drug diversion faster than humans
Since the 1980s, hospitals have turned to automated drug-dispensing machines, usually customized to each department. Some researchers argue that the machines not only help nurses work faster, but also can cut down on medication errors. While several vendors currently employ analytics, such as CareStream's Pyxis MedStation, to track drug diversion, Springhill employs Omnicell's G4 and its accompanying Pandora analytics package, which data-mines the Omnicell records and flags potential anomalies in drug dispensing for each employee if, for example, he's taking more hydrocodone pills than usual in the course of care.
If we did not have analytics as a prospective tool to monitor, it would have been easily months -- possibly a year -- before these [anomalies] were ever noticed, if ever.
Joe Adkins, clinical pharmacist, Springhill Medical Center
"If we did not have analytics as a prospective tool to monitor, it would have been easily months -- possibly a year -- before these were ever noticed, if ever," Adkins said, referring to three cases they've detected and taken action on so far, which involved hydrocodone tablets and injectable Demerol. "Usage patterns start slow and usually go unnoticed if you're manually monitoring this process, but if you're looking at statistics of usage, it makes it easier."
Adkins said that not only are the system's email alerts and graphical reports simple to read quickly, but the system also spots -- and flags -- exceptions in drug use patterns sooner than he was able to before, with his manual processes.
Have policies in place for responding to analytics red flags
Not every alert signifies a true drug diversion case, Adkins said, which necessitated Springhill to put policies and protocols in place for responding to them. For example, if a nurse typically works in the medical/surgical unit but covers a couple of shifts in orthopedics (where painkillers are used far more frequently for hip- and knee-replacement recipients), the system might flag her for accessing more narcotics than usual – which, in this case, would be perfectly acceptable and would not signify drug diversion.
Adkins said one policy that helps sort the false-positive alerts from the real thing is sharing analytics alerts with an employee's department head. Those managers, he said, can sometimes easily chalk up deviations in drug-dispensing patterns to the work the employee did. In other cases, they can put two and two together when it comes to drug abuse, such as remembering how an employee might have called in sick frequently during a time when the stats indicate diversion could have been happening, or had otherwise acted as if they were under the influence.
Springhill has protocols for when the system does detect true drug diversion, which usually involve treatment and eventually returning the employee to the workforce. There has yet to be a case that involves law enforcement, but Springhill would do that in obvious cases of an employee dealing hospital narcotics.
Using the software doesn't guarantee perfect results, Adkins said, but they have greatly improved since the hospital implemented it. "I think every system is prone to being gamed, but with the analytics it certainly makes it easier to find," Adkins said.