Blog

Fighting the Opioid Crisis with Machine Learning

Paul Derstine

January 11, 2019

BLOG_Fighting the Opioid Crisis-2

In March 2017, President Trump established the President’s Commission on Combating Drug Addiction and the Opioid Crisis. In October 2017 Eric D. Hargan, Acting Secretary of Health and Human Services, declared that a national public health emergency exists as a result of the consequences of the opioid crisis affecting our nation. In September 2018 the U.S. Department of Health and Human Services awarded over $1 billion in opioid-specific grants to help combat the crisis.

The Scope of the Opioid Crisis

The statistics on this crisis are staggering and according to the CDC, drug overdose deaths continue to increase in the United States:

  • From 1999 to 2016, more than 630,000 people have died from a drug overdose.
  • Around 66% of the more than 63,600 drug overdose deaths in 2016 involved an opioid.
  • In 2016, the number of overdose deaths involving opioids was 5 times higher than in 1999.
  • On average, 115 Americans die every day from an opioid overdose.
  • The rate of drug overdose deaths involving synthetic opioids other than methadone (which include drugs such as fentanyl, fentanyl analogs, and tramadol) increased by 88% per year from 2013 to 2016 (Figure 1).

Age-adjusted drug overdose death rates by opoid category in the US from 1999-2016

Figure 1. Age-adjusted drug overdose death rates by opioid category in the US from 1999-2016 (Source: CDC)

In a recent article published on Concord Monitor U.S. Department of Labor Secretary Alexander Acosta detailed an aggressive strategy pursued by the U.S. Department of Labor to prevent the over-prescribing of opioids for the approximately 200,000 federal employees who receive workers’ compensation benefits per year.

Some of the dramatic results achieved by this strategy include:

  • A 22% drop in new opioid prescriptions
  • A 43% decline in new opioid prescriptions lasting more than 30 days
  • A significant drop in the number of claimants prescribed higher dosages of opioids
  • A 45% drop in claimants with a morphine-equivalent dose of 500 or more
  • An 18% drop in users with a morphine-equivalent dose of 90 or more
  • Nearly 15 times more fraud referrals so far this year than all of 2016

Detecting Provider and Claimant Fraud

Elder Research is honored to have provided advanced analytics to help the Department of Labor (DOL) achieve these dramatic results in support of the President’s Initiative to Stop Opioid Abuse and Reduce Drug Supply and Demand. The DOL contracted with Elder Research to build and deploy analytics solutions to:

  • Detect problematic trends and identify anomalous billing patterns to combat fraud
  • Understand and monitor the size and nature of the DFEC claimant population that was using opioids and the number of “New Entrant” (NE) Cases for opioids

The Department of Labor Office of Workers’ Compensation Program, Division of Federal Employees’ Compensation (OWCP-DFEC) division handles almost 200,000 claimants who are currently receiving medical compensation for services rendered by tens of thousands of medical providers. Department of Labor Office of Inspector General (DOL-OIG) contracted with Elder Research to create a predictive model to detect fraud in Office of Workers’ Compensation Program (OWCP) data. The goals of this project were to highlight abnormalities in the claims data that could be used to form the basis of future audits and to create visualization tools to enable auditors to explore model results.

Elder Research implemented a healthcare case fraud predictive model and RADR data visualization tool. The supervised risk model was based on the attributes of known fraud cases from the Office of Labor Racketeering and Fraud Investigations (OLRFI) case management database. DFEC’s program integrity and fraud analysts use the Elder Research RADR analytics platform to efficiently explore, analyze, and surface risky and highly-suspicious behavior in the sea of DFEC data. RADR fuses data from multiple data systems that would otherwise take inordinate amounts of time to connect — and would require technical expertise more appropriate for IT personnel than analysts and investigators — creating a unified, intuitive view with the context that’s required for experts to make important decisions.

In RADR, an analyst can explore aggregated data on services, service providers, and claimants, as well as drill down to transactional details. They can view charts of data over time, geographic maps, and networks of providers based on common claimants or services. Integrated statistical analytics create risk scores that highlight unusual changes in billing behavior, abnormal patterns of services provided compared to peers, and other factors that give analysts data-driven starting points that do not rely on tips or other outside information (Figures 2 and 3).

DOL Provider List View-Figure 1

Figure 2. RADR provider list view showing providers with the highest fraud risk scores

Network Graph of Providers Connected to Pharmacy 4364- Figure 2

Figure 3. Example RADR provider network. Three of the high-risk pharmacies (3283, 4656, 2874) and the connected provider (39807) were investigated and indicted on fraud as a result of this tool.

Analysts have found that investigations and forensic analysis that took hours can now be completed in minutes, making the most efficient use of limited and valuable resources. Data fusion and presentation in a variety of useful visualizations allows analysts to find emerging fraud schemes. 

Identifying Opioid Population and New Entrant Cases

The primary goals for the DFEC opioid strategy in fiscal year 2018 were to understand and monitor the size and nature of the DFEC claimant population that was using opioids and create risk metrics to indicate any pharmacies/physicians/patients that were potentially involved in or aiding the hazardous use of opioids.

The Elder Research team assisted with these goals in the following ways:

  • Data Pipelining and Optimization: Elder Research examined tools and implemented programs to streamline the extract/transform/load (ETL) data pipeline process and optimize the reporting and analytics workflows.
  • Business Intelligence (BI) Tool Development: Elder Research developed self-service visualizations and dashboards to address the DFEC’s opioid reporting needs. The project team reviewed the available BI tool options in DFEC’s OWCS environment and recommended using MicroStrategy as their BI platform.
  • New Entrant Modeling: Elder Research developed a model to predict the likelihood a new entrant (NE) claimant would continue using opioids. Based on the predicted likelihood, DFEC planned to use different intervention tactics such as sending letters to the NE claimants and/or the prescribers of these opioids. These tactics were able to reduce the number of new entrants, as shown in Figure 4 (and on the monthly NE Policy Goal Monitoring report on the DOL Opioid Action Plan web site).

Fig 4 - Monthly New Entrant cases

Figure 4. Monthly New Entrant cases as reported by the DFEC from January 2015 to March 2018.

 


Watch our on-demand webinar on Best Practices for Deploying a Fraud Analytics Solution. 


Related

Download the case study Detecting Fraudulent Healthcare Claims

Read the blog Improving Unemployment Insurance Claim Fraud Detection

Learn more about Fraud and Risk Analytics


About the Author

Paul Derstine As Director of Marketing, Paul Derstine works with clients to understand their data analytics goals and how Elder Research's vast experience with data engineering, data analytics, and data visualization can deliver solutions that enable return on their analytics investment. Prior to joining Elder Research, Paul worked for 18 years for GE Intelligent Platforms in Charlottesville where he worked with a broad spectrum of global customers to understand their business needs in order to deliver solution value and optimize profit and growth objectives. Paul has a B.S. degree in Electrical Engineering from The Pennsylvania State University.