Executive oversight and management of drug theft can be achieved system-wide with machine learning
In 2014, the Centers for Disease Control (CDC) published a summary of how drug diversion by healthcare providers – the illegal use or acquisition of prescription medication – has been linked to severe patient harm. They revealed gaps in prevention, detection and response in U.S. healthcare facilities, and called for stronger security measures.
Fast-forward several years, and we’re still seeing high profile cases of drug abuse and theft, as well as continued drug diversion in hospitals and specialty settings. In a 2017 study conducted by Porter Research, 96 percent of surveyed providers acknowledged that drug diversion is still occurring in healthcare, with 65 percent stating that most diversion is going undetected.
The impact of drug diversion to healthcare facilities cannot be overstated. Diversion severely undermines the safety of patients and health care providers. It leads to costly financial fallout for healthcare systems in the form of fines and lawsuits, damages health system reputation, and leads to revenue loss in the form of missing drugs. Finally, diversion often leads nurses, pharmacists, anesthetists, and other healthcare providers to addiction, the loss of their licenses, and, all too often, overdose and death.
In May of 2017, the National Institute on Drug Abuse (NIDA) within the National Institutes of Health (NIH) awarded a research grant to combat this key component of the national opioid and drug abuse epidemic. The grant utilizes supply-chain visibility and analytics with the goal of providing end-to-end visibility of drug inventory along with machine learning to detect the behavior patterns of healthcare providers who are diverting drugs.
Undetected Threats
Diversion is a two-headed monster. First, medication management in a typical healthcare system is a labyrinthine supply chain with thousands of touch-points and end users entering millions of transactions that makes visibility and diversion detection a daunting task. Secondly, current diversion technology, such as anomalous usage reports, fails to detect most diversion, while often flagging innocent healthcare providers as potential diverters. Porter Research concluded a majority of investigations found that no diversion occurred and took upwards of eight hours per investigation to reach that conclusion – essentially wasting valuable time investigating healthcare workers who are not diverting drugs.
It is estimated that a mid-size 500-bed hospital can assume that approximately 25 to 75 people are at risk for diverting at any time. However, most hospitals only investigate an average of five or fewer diversion cases annually. A shocking 22 percent of respondents reported their healthcare facility had no established diversion prevention program at all.
When you look at how diversion is currently being detected, it’s often as subjective as a behavior change recognized by a colleague or a patient long after the theft began – establishing an extremely risky pattern for patients and healthcare providers. In many cases, these subjective changes led to costly investigative review of documentation that verified diversion occurred.
A Better Way
The NIH-funded research identified the ability to detect the occurrence of diversion as one of the first priorities during the research project. To accomplish this, the pilot hospital and technology partner compiled a list of 12 confirmed diversion cases and extracted historical data from five health IT systems at the hospital over a two-year period.
The research hypothesis was that if you combined Automated Dispensing Machine data with data from Electronic Medical Records (EMR), Employee Time Clocks, Wholesaler Shipments, and Internal Inventory Systems, you could establish a holistic visibility over medication transactions. Once end-to-end visibility had been established, you can then apply machine learning and advanced algorithms to assign a “risk score” to every dispense – enabling hospitals to detect diversion earlier and with far greater accuracy.
The research hypothesis was proven; the technology flagged risky transactions and detected all 12 cases of blinded and confirmed diversion, weeks to months before they were detected with traditional methods. In addition, the solution detected previously unknown cases that were already undergoing investigation. This initial success has now led to a Phase II question: How effectively can machine learning detect unknown diversion while also improving the efficiency of the process?
Executive Response
Awareness alone is a major initial milestone. Executives have the opportunity to better understand what their hospital is doing, how it measures up, and seek ways to solve the problem.
The Porter study found that 66 percent of respondents felt their programs were neither efficient nor effective, and yet 90 percent believe their strategy is on par or better than their peers. This contradiction is the key misunderstanding that undermines preventing drug diversion as a healthcare priority and have led us to where we are today.
Some critical first steps:
• Get Help. Most facilities dedicate less than one full-time employee (FTE) on diversion prevention, regardless of size. It’s critical to increase the efficiency of diversion investigators through better tools.
• Embrace Innovation. Advanced analytics and machine learning technology is now available to detect diversion effectively and efficiently. Let the data do the heavy lifting in spotting the smoke before the fire.
• Let Information Flow. Multi-department collaboration is vital to ensure informational flows between tradition healthcare silos. This is not a “nursing” problem, nor a “pharmacy” problem, nor a “warehouse” problem; it’s a healthcare problem.
• Change the Culture. All the tools in the world won’t make a dent if the culture remains static. Executive leadership is required to prioritize patient and healthcare worker safety against diversion and ensure all healthcare workers are aligned within this goal.
Best practices are starting to form that enable tangible steps in detecting diversion – hopefully eliminating major threats to patients, employees and hospitals.
Tom Knight is the founder and CEO of Invistics, a leading provider of inventory visibility and analytics.