Drug Diversion is Complicated: 5 Data-Related Challenges to Overcome
COVID-19 thoroughly dominated American life last year, contributing heavily to the kind of despair that drives drug overdose deaths, which soared last year as well.
This worrisome trend has intensified long-running concerns about how drugs like opioids find their way to abusers’ hands in the first place. One point of vulnerability stems from our healthcare systems, including employee diversion, a problem so big and complex only sophisticated computing can address it.
Over 81,000 drug overdose deaths occurred in the U.S. in the 12 months ending in May 2020, the highest number of overdose deaths ever recorded in a 12-month period, according to the CDC. Although the number is a fraction of the nearly 600,000 American COVID-19 deaths to date, it’s still a lot of grief.
“The disruption to daily life due to the COVID-19 pandemic has hit those with substance use disorder hard,” then-CDC Director Robert Redfield, M.D., said in December. “As we continue the fight to end this pandemic, it’s important to not lose sight of different groups being affected in other ways. We need to take care of people suffering from unintended consequences.”
Drug abuse plagues healthcare
One group that doesn’t fit the stereotypical substance abuse profile is health care workers, about 10% of whom are abusing drugs, according to figures from both the U.S. Substance Abuse and Mental Health Services Administration (SAMHSA) and the American Nurses Association (ANA). And where would health care workers get drugs? At work, of course.
That’s harmful in several ways. “Drug diversion,” as the misappropriation of prescription medicines is called, jeopardizes the well-being of 1) workers who ingest medicines not prescribed for them, 2) pain-stricken patients deprived of treatment they need, 3) patients who are neglected or otherwise harmed by clinicians under the influence, 4) third parties who buy and consume the diverted drugs, and 5) any person or organization that pays for healthcare, including patients, employers, insurers, hospitals, and governments.
The problem of drug diversion is complicated. At-risk are hospitals, hospital systems, pharmacies, clinics, nursing homes, and emergency centers. Each one has multiple medication-lifecycle processes that are vulnerable to diversion, including wholesaler procurement and delivery, central pharmacy inventory management, pharmacy distribution, dispensing to doctors and nurses, and, finally, administration to the patient.
Healthcare organizations often dedicate teams of employees to detection, but all the time-consuming, labor-intensive meetings and manual record-keeping in the world rarely flag potentially deadly misuse cases that offenders are determined to keep under the radar. Nor does manual record-keeping offer safeguards against false accusations. Done right, drug diversion prevention powered by advanced computing can save lives, improve patient care, reduce health care costs, and eliminate wasted effort.
The data drives the cure – five challenges
The solution, as with so many thorny problems in healthcare, is data analytics, artificial intelligence, and machine learning. New technologies are consuming all the medication data that organizations can provide to detect anomalies in usage and flag potential diversion. Even though the technology is more effective than manual record-keeping, it’s not a simple undertaking. Companies dedicated to solving the problem are wrestling with five data-related challenges:
1. Lack of historical data – Many providers have scant data to normalize, enrich and purify even if they could. That means that anomalous events can fatally skew modeling. For example, suppose a health care system has only three or five years of historical medication-process data. In that case, the models will be grossly distorted by the pandemic, during which there was a major decline in elective procedures. And the end of the lockdown, there will likely be a major spike in those procedures. These whipsaw events will confuse models that are based largely on recent data. But for organizations whose models are driven by decades of data, the events will be considered in context; peaks and valleys will effectively be smoothed out. For them, advanced drug diversion detection can be truly intelligent and, therefore, accurate.
2. Changing technologies – Historical data drives artificial intelligence quality, and the more foundational data you have, the better you can train AI to make reliable predictions. Unfortunately, health care organizations are constantly updating and replacing information systems to keep pace with the state of informatics. So even in the rare cases where decades of medication-process data exist, it’s typically incompatible, fragmented, and useless in its native form. Solution providers are working hard to find, gather, normalize, enrich and purify this data.
3. Differing data labels – To amass meaningful data across locations and steps in the medication lifecycle, organizations need to gather apples with apples. The problem is that one vendor’s pharmacy information system, automated dispensing system (ADS), or electronic health record system likely uses different data formatting from another’s. For example, one vendor’s system may store a patient’s name in four fields—first name, middle name, last name, and the suffix—while another system may use a single field to store the name. Organizations also need to combine data for branded drugs with generic equivalents since there’s no meaningful difference in the context of diversion. There are similar integration issues with units of measure.
4. Human factor – Although advanced drug diversion detection systems do what humans cannot, they still need human input to become reliable. If a drug diversion detection system flags a clinician for anomalous behavior, a trusted user needs to tell the software whether the alert was right or wrong so it can learn. To be effective, that validation process needs to be embedded in the workflow. For the user – e.g., the diversion team member or nursing supervisor – the process should be as easy as clicking yes or no to indicate whether the anomaly is a confirmed diversion, a false positive, or in rare events, a false negative.
5. Updating – Few drug diversion detection systems bring real-world, as-it-happens data streams on drug flow back into the system in near-real-time. But that’s what’s needed to refine the baseline. User feedback also needs instant updating integration.
Major medical centers across the country have pulled these practices together and detected behavior by individuals who’d long eluded accountability for diversion. As these organizations improved their monitoring and detection, they also reported dramatic time savings over manual processes.
Most importantly, human lives are at stake when prescription drugs are diverted from the patients to whom they’re intended. Having the ability to solve and overcome the aforementioned challenges distinguishes the adequate from great in the realm of drug diversion solutions. Although the technology challenge for reducing diversion is formidable, manual processes are effectively futile. Data analytics, monitoring, and accountability are the answer.
About David Brzozowski Sr