The Department of Health and Human Services has declared it a national health emergency. Others have called it a crisis or an epidemic. But whatever name you give it, the explosion of opioid abuse continues to be a plague on the nation.

Consider that in 2016, the latest year for which there are figures, there were 32,445 deaths involving prescription opioids, which works out to roughly 89 deaths per day. That’s a 44% increase in deaths in just one year, up from 22,598 in 2015.

More striking is that those numbers don’t even include overdoses from illicit drugs such as heroin, which drive the total up to over 64,000.  That’s more deaths per year than at the height of the AIDS epidemic in 1995. Additionally, according to the Centers for Disease Control and Prevention (CDC), despite the supposed “war on opioids,” emergency department (ED) treatments for opioid overdoses rose 30%  in all parts of the U.S. between July 2016 and September 2017.

Clearly, the trendline continues to rise. What we’re doing as a nation to address the issue isn’t working. So what will it take to gain control over this crisis and set the U.S. on a path toward recovery?

One important step is to begin using the power of next-generation analytics and machine learning to identify purchasing and prescribing patterns that offer a high probability of abuse so they can be addressed at the source. Success requires a two-pronged approach that focuses on members/patients seeking opioids as well as reducing fraud, waste, and abuse (FWA) among prescribers.

Uncovering member/patient drug-seeking behaviors

A good place to start is with claims data from the health payer and pharmacy benefits manager (PBM). Some opioids can be expensive so members/patients are likely to file an insurance claim for each prescription. And while they may have several providers, most members/patients only have one health insurance payer and one PBM. Claims data has the advantage of seeing across different providers and health systems to deliver a holistic view of each member’s/patients total opioid purchases.

Once the data has been analyzed, patterns have been identified through machine learning, and algorithms are established, the organization can develop risk scores and a color-coded dashboard that shows the likelihood that a particular member is committing FWA around opioid use. Risk scoring can be tightly controlled to minimize false positives, enabling investigators to focus on those members who are displaying the highest propensity for opioid abuse, collusion, or diversion.

For example, the organization can establish pre-set thresholds such as elevating the risk scores of members/patients who are seeing more than 10 physicians or filling prescriptions at more than 10 pharmacies. The thresholds can be set based on industry benchmarks or payer/pharmacy benefit manager (PBM) preferences. They can also take special circumstances into consideration, such as an oncology patient who is receiving different prescriptions from multiple specialists; or those members receiving a shorter day’s supply with no overlap.

The analytics can be further refined by bringing in additional data, such as geospatial analytics that show the locations of prescribers, pharmacies, and the member’s/patient’s home on a map. A normal pattern will tend to have all of the locations clustered in one area. Filling several prescriptions at pharmacies that are far away from the member’s/patient’s home, on the other hand, is often a strong indicator that doctor-shopping or FWA is occurring.

Bringing in additional behavioral data about members/patients, such as age, gender, income, education levels, and other socioeconomic and psychographic factors based on Zip+4 and other sources can help further refine the analytics to deliver greater accuracy when developing risk scores.

This level of data analysis is simply not practical using manual inspection of spreadsheets or even basic analytics. But by automating the process through the intelligent application of advanced analytics, behavioral patterns that indicate a high likelihood of FWA will surface quickly. Healthcare organizations can then focus their efforts on the highest risk members/patients to help reduce the danger of overdose or even death, improve their long-term health, and reduce costs. All without alienating members/patients through false positives.

Discovering FWA at pharmacies

Winning the battle against opioids also requires a way to discover and shut down pharmacies that are intentionally committing FWA. Close down the supply and it will be far more difficult for opioid abuse to continue.

Of course, these pharmacies are very good at hiding their activities, even from organizations actively investigating FWA. Here again analytics can do a lot of the legwork to help investigators focus their attention where it’s most-needed.

It starts by establishing a benchmark of dispensing patterns for all pharmacies over a specified time period. In most cases, analyzing a year’s worth of historical data will be enough to create that baseline. Once it is in place, analytics can monitor dispensing activities against that benchmark on a weekly basis.

If there are any significant deviations from the benchmark, color-coded dashboards can again quickly call them out. The organization can use those dashboards to determine where action is required – and how urgent that action needs to be. The analytics can also be used to help payers and PBMs comply with Centers for Medicare and Medicaid Services (CMS) monitoring of “watch or risk lists.”

Next-generation analytics can monitor a broad range of metrics, including:

• Rate of “new billing”
• Reversal rate (very high and very low)
• Percentage of member co-pays
• Average ingredient cost
• Average paid per prescription
• Average number of prescriptions per member (stratified by age)
• Percentage of controlled substances
• DAW-1 percentage
• Average dollars paid per member

The dashboard view makes it easy to spot overall trends for these metrics. It will highlight pharmacies that require corrective interventions such as pending claims or withholding payment. In cases where an onsite visit or more severe actions are needed, it will highlight those as well.

Adjusting for Variables

One of the greatest challenges in working with data about opioids is that there is not only a high volume of it, but there are many variables to consider—including the organization’s preferences and standards. Since a “one-size-fits-all” approach won’t work, the analytics must be highly flexible to ensure they can be directed to the areas of greatest need.

Take a retail pharmacy that occasionally experiences a one- or two-week spike in sales of controlled substances in certain locations. It may want its analytics to be configured to compare performance in those locations with peer or like-type pharmacies in that area. This extra data point will help the organization determine whether the spike is part of a common pattern or something that requires further investigation

Healthcare organizations that adjust what they’re measuring and how the results are being displayed are in a better position to focus their limited resources where they will deliver the best ROI or other benefits.

Victory in sight

Despite the continuing upward spike in opioid deaths and ED visits, the trend can be reversed. Next-generation analytics are a powerful weapon in the battle against opioid FWA.

By uncovering the greatest areas of risk, they can help healthcare organizations lower the cost of prescription pharmaceuticals for them and their members/patients. They will also ensure that members who do require opioids receive them as their providers prescribed so they can achieve better health outcomes.

Rena Bielinski is  senior vice president at SCIO Health Analytics. She can be reached at [email protected].

Keep up with latest coverage on digital healthcare by signing up for Internet Health Management News today.