How AI is Stopping Drug Diversion and Saving Healthcare Workers' Lives – Healthcare Informatics

By | December 27, 2018

Using artificial intelligence to better understand how controlled substances move through one’s organization is an important step for healthcare organizations to take

Lisa Marie Jones first started taking painkillers to deal with a dental surgery, but soon she was using them to manage other pain in her life. She had a disabled family member to care for and suffered from significant childhood trauma. Unlike most people dealing with addiction though, Jones had free and essentially limitless supply of opiates at her disposal. Jones was a practicing nurse.

Her addiction did not just hurt herself, it put patients at risk too. Jones worked at a Denver-based healthcare organization, and had access to their supplies of hydromorphone, morphine, and fentanyl. She would bring the vials to the bathroom, take the contents, refill them with saline, and seal them up again with skin glue. Jones not only diverted pain relief away from patients who may have critically needed it; by unsealing the vials, she could have exposed patients’ blood to bacteria, fungus, and infectious diseases.

Sadly, Jones’ case is far from isolated. It is called clinical drug diversion, and by conservative estimates, it led to the loss of over 18.7 million pills and $ 164 million in the first half of 2018 from reported incidents alone. In Jones’ case, the vials were found before they could be administered, but not all incidents have been so harmless. From 2007 to 2012, a traveling radiology technician with Hepatitis C named diverted drugs for his own use and exposed his blood to thousands of patients in hospitals across eight states. Hepatitis C causes more deaths per year in the U.S. than HIV, and the technician only stopped spreading the disease after a 32-patient outbreak in a New Hampshire hospital raised suspicion.

When healthcare workers divert drugs, it is often their employers who end up in legal hot water. This year, a Georgia health system paid a $ 4.1 million penalty after it failed to adequately oversee tens of thousands of oxycodone pills the DEA had found stolen. Kwiatkowski’s case itself incited several lawsuits against the staffing agencies and institutions who hired him, some of which are still ongoing.

Healthcare providers need to protect patients from the risk of clinical drug diversion and prevent caregivers from misusing pharmaceuticals, but that is easier said than done. It is infeasible, for example, to have someone stand over the shoulder of every doctor, nurse, and medical technician. Not only would it be a prohibitively expensive, but it would sow seeds of distrust and slow down the whole organization.

Webinar

Through real use cases and live demo, Frank Lee, PhD, Global Industry Leader for Healthcare & Life Sciences, will illustrate the architecture and solution for high performance data and AI…

Another option would be to have a team go through the logs of every time someone administers a drug and look for suspicious behavior, but this, too, is impractical. Combing through War and Peace-sized logs is a tedious, labor intensive task that could at best review only a fraction of everything going on.

A third option would be some sort of software solution, and on its face, even this seems insufficient. Computers are cold and unthinking, and clinical drug diversion is a tenuous issue. Even intelligent systems are still often dumb—think of how often Siri fails at the most basic tasks and accidentally calls the most obscure person in someone’s contact list.

But artificial intelligence (AI) does have two key strengths. First, it is good at quickly churning through mountains of data, a task too slow and monotonous for humans. Second, it is even better at sifting through that data to find anomalies. That is why AI systems today are used in detecting everything from fraudulent financial transactions to manufacturing malfunctions to spam email.

Clinical drug diversion can be broken down into just this kind of problem—by understanding the historical data of drug transactions associated employee behavior, AI can learn when care providers are giving patients the pain relief they need and when they are taking extra for themselves.

The U.S. is in the midst of an opioid epidemic, unprecedented in the lives it has taken and the diseases it has spread. Clinical drug diversion, 95 percent of which involves at least one type of opioid, is just one part of this story. By better understanding which employees are at most risk of addiction and diversion is one of the first steps healthcare organizations can take to begin to better understand this critical challenge in healthcare. Organizational leadership can then begin to educate their workforce on the signs of diversion and the steps they can take to protect their colleagues and patients.

Using artificial intelligence to better understand how controlled substances move through the organization is another important step for healthcare organizations to take. This new insight allows organizations to identify abnormal usage patterns that may indicate an instance of clinical drug diversion. Healthcare organizations that proactively detect these instances can quickly intervene, protecting their employees, themselves, and most importantly their unsuspecting patients from the risks that come with addiction.

Robert Lord is the co-founder and president of Protenus, a compliance analytics platform that detects anomalous behavior in health systems.  He also serves as a Cybersecurity Policy Fellow at New America.

"Health|HealthCare" – Google News