May 08, 2017 Using Analytics to Make a Difference: Finding and Preventing Opiate Abuse
Tackling an Epidemic in the Volunteer State
Based on notes from Teradata PARTNERS Conference 2016 taken in a presentation by Mike Upchurch, COO Fuzzy Logix, and Leigh McCormack, Manager, Predictive Analytics, BCBST
Data has the power to solve many of today’s most critical issues, and PARTNERS strives to bring these successes to light. In the United States, one of those issues is the abuse of drugs, specifically opiates.
The growth of opiate abuse and addiction has hit epidemic proportions in the United States. The CDC recently estimated the total economic burden to be $78.5 billion a year. To say that the cost to lives, directly and indirectly, has been huge would be an understatement. Several states count more opiate prescriptions than people. Tennessee is one of those states. Serving more than 3.4 million Tennesseans, Blue Cross Blue Shield of Tennessee(BCBST) has been thrust into the middle of the epidemic.
While opiates are an important component of many treatments, their abuse can have devastating consequences. In Tennessee, currently 1 in 6 residents is either misusing or abusing opioids, or in active treatment. Overdose deaths continue to climb, with more than 60 percent of overdose deaths attributed to an opioid—and more than half of those overdose deaths involving a prescription opioid. Something has to change, so patients can get the treatment they need while minimizing and eliminating the misuse of dangerous drugs. Could data provide the answer? BCBST thought so.
A Picture of the Problem
The Centers for Medicare and Medicaid Services (CMS) has created an Overutilization Monitoring System (OMS) to ensure that prescribers with a contract have systems in place to help prevent overutilization of prescribed medications. They developed clearly defined definitions of potential overutilization issue types. While this was a great start, its overall scope was limited by the CMS list itself: the problems were well developed by the time CMS could reveal them, and bandwidth—just two people were dedicated to the program.
Image source: http://partners.teradata.com/resources/archives
They needed to do more. Looking for areas of improvement, BCBST decided on four big questions to explore that would have an impact and that could possibly be solved with data. At the PARTNERS Conference, they outlined the following opportunities for improvement.
- Can we identify patterns of overutilization earlier than CMS?
- Can we use the significant, analytical factors to prescribe how to best engage and alter behaviors?
- How do we integrate into existing workflows?
- How do we continue to refine and learn as this epidemic changes?
While pharmacy data is great, they knew it would take more data and models to track and predict opioid abuse. To answer those questions, they looked at the collection of data sources they had available, including:
- Identifying clinical symptoms through medical claims
- Using prescriber information to determine types and quantity of providers
- Leveraging pharmacy data to define abuse and determine historical patterns of prescription fills—looking for trends and paths
- Determining which facilities and pharmacies are being used to prescribe and dispense drugs
- Calculating the distance traveled to obtain prescription pain medications
Sharpening the Focus
The challenge was to take a huge data set, shrink it, and use models to get to the real factors that lead to abuse of the drugs. The team, using Fuzzy Logix, was able to process 25,820 records with 1,788 known “over-utilizers.” From there, they narrowed the predictors from 742 to a manageable 44 using correlation analysis to cull non-predictive variables. The data modeling was done in just two days. The influential variables that remained enabled BCBST to make significant strides toward tacking opiate abuse. Some of the most telling variables they found were the following:
Prescription Intensity – Higher costs per supply day, numbers of prescriptions filled, and the number of claims helped identify abusers.
Diagnosis – Was their prescription tied to a specific condition like a surgery, or a more subjective symptom like back pain?
Region – How far were members traveling to receive their drugs? Non-abusers would visit their regular doctor and local pharmacists. Abusers often needed to travel some distance to find a willing prescriber.
The result has had multiple business applications. One of the first was to support company and community efforts where the effects are felt most directly. A program was started to provide secure drop-offs for unused opioids, preventing them from falling into the wrong hands. Physical drop boxes were created and placed in the areas of greatest need as determined by the geo-spatial data they obtained. The Count It! Lock It! Drop It!™ campaign helped to build community awareness while imploring everyone to not “be an accidental drug dealer.”
Image source: http://countitlockitdropit.org
The additional data helped to enhance workflows to meet regulatory standards and measures required by CMS more efficiently. Importantly, the data served to complement internal clinical and pharmacy management efforts. Arming the patient-facing caregivers with information enables them to be more aware of the patient’s situation. For example, knowing how many visits the patient had in the last few months can help determine if they are doctor shopping or actually suffering from back pain. With that information, professionals can make informed decisions about the patient’s care and whether to consider alternative treatments.
While the battle against opioids is far from over, BCBST’s efforts show that data will be one of our most effective tools in the fight against addiction.
Want to Know What’s Next?
Did you attend Teradata PARTNERS 2016 in Atlanta? If so, you can view the presentation slides in the archives.