In 2015, there were over 2.5 million Americans addicted to opioids; 33,000 of them died of overdose. Nationwide, that rate is over ten deaths per 100,000 people, and the rate is over 30 in some states, such as West Virginia and New Hampshire. At the height of the crack epidemic, in comparison, crack was “only” causing about four deaths per 100,000 people. The vast majority of opioid abusers are addicted to legal prescription pain killers such as Vicodin or Percocet, and many of those eventually become addicted to illicit drugs like heroin. As analytics experts, we outline ways that technology can help the Federal Government reverse this trend and curtail this destructive epidemic.
Part of the crisis stems from an over-correction in the medical field after years of what is now believed to be under-prescription and under-treatment of patient pain. To achieve the correct level of pain prescription, use of bioinformatics and evidence based medicine will be critical; enhanced patient targeting will allow us to fight under-prescription and over-prescription at both ends. However, medical changes are very slow; we can achieve faster effects by also applying financial oversight.
There is a natural incentive by pharma companies to market opioids in order to drive profits. Doctors are often persuaded by pharmaceutical companies’ ‘face to face’ marketing. Many times drugs are the easiest solution to pain since insurance is more likely to cover drugs than other treatments like physical therapy, and national guidelines surrounding prescription levels are easily ignored. The result is a huge influx of opioid prescriptions doled out annually, compared to what is medically necessary. In fact, recently over 3,000 patients were prescribed opioids even after an overdose. Tightening the purse strings could help to correct that bias in the system.
Federal Government Uniquely Suited to Take the Lead
The Federal Government’s health care programs — including Medicare, Medicaid, Multi-Health Systems (MHS), Veteran’s Health Administration, the Health insurance Marketplace — is the nation’s largest healthcare provider, accounting for nearly a third of all health care spending. This centralization makes the government uniquely suited to take the lead in developing analytics solutions to combat opioid spending.
This oversight has to remain patient-centric. It would be counterproductive to clamp down too drastically on pain killer prescriptions, as there would be too much collateral damage to medically necessary pain treatment. Precise interventions with analytics acting as their laser-guidance system are necessary. Risk models focusing on patients and medical providers are a great place to start. Actions of these entities can be graded on scales of risk ranging from legitimate medical necessity to dubious behavior or outright fraud.
Solutions Guided by Big Data
On the patient side, analytics can be used to examine key indicators such as:
- Trips to multiple prescribers or pharmacies
- Early refill of prescriptions
- Higher than normal volumes of pain killers
- Longer than normal use terms
- Traveling long distances to pharmacies
- Switching between multiple types of controlled drugs
Predictive analytics can also be used to analyze the expected term length for each patient, and identify which cases are likely to turn into a long-term dependence. These models can guide the review or investigation into a patient’s history. The most extreme cases may be investigated for fraud; other risky ones may need to have payments stopped until further information can be gathered, or may need a referral to a treatment center or a pain management team. Human judgment can help determine the actions to take for cases flagged by the models as risky. The expert domain knowledge of those in the field can take into account the second-order effects of the actors and select the likely move leading to the best long-term outcome. For instance, a patient abusing pain killers could have a “cold stop” order on pain killers immediately issued. But this may have the negative consequence of inducing the patient to turn to a cheaper street drug like heroin. Perhaps the first action then would be to lower dosage and wean the patient off of the drug. Many government providers also have employment information for patients, so can identify potentially dangerous situations such as a patient on opiates whose job involves operating heavy machinery. In that case, the first action might be to immediately remove the patient from such responsibility.
On the provider side, patterns over the course of many patients can be examined. For example, doctors or pharmacists in “script mills” prescribe pain killers at an excessive rate when compared to the nationwide standard for similar cases, signaling lax medical requirements for a prescription. Similarly, pharmacies known as “pill mills” dispense high volumes of pain killers compared to other drugs, signaling looser requirements on the legitimacy of prescriptions. It’s also important to analyze the connections between these entities with link analysis. Risky or known fraudulent claimants, pharmacies, and doctors tend to be linked together.
Working Together – Economies of Scale
Some government agencies are already taking action, employing many of the above tactics. Elder Research recently worked with two different government organizations to monitor opioid use. We used many of the risk factors described above to prioritize manual reviews of patient opioid history and re-evaluate whether continued payments should be authorized or other medical recommendations should be made (e.g. obtaining a second doctor’s opinion). Elder Research also worked to target risky providers who consistently run higher pharmacy bills than expected for a given patient, given the patient’s injury. In addition to targeting these providers for investigation, Elder Research created a visualization platform in our Risk Analysis Data Repository (RADR) tool (shown below) which helped analysts explore the relationships between pharmacies and providers, allowing the organization to uncover clusters of risky entities operating in concert. The link graph showsd the following information:
- The coloring for the nodes is based on the masked anomaly score. (black is no score in the data).
- The icons represent the primary provider type: hospital (ambulance), pharmacy (shopping cart), physician (doctor).
- The providers are linked based on the number of shared claimants weighted by the number of total claimants for each.
The next evolution for the government should include resource sharing. This starts with sharing knowledge, insight, and techniques for analyzing and combating the problem, but it eventually must go deeper. Shared data across government organizations will enable a holistic view of opioid use, leading to more precise analytical models. For example, if a patient’s care is being paid for by multiple government agencies, then a shared pool of data could be used to more quickly identify a pattern of abuse. A single oversight group could detect such a problem patient and plan an intervention. Or a pill mill could be universally blacklisted by the government for issuing excessive drug prescriptions, saving the individual agencies from each having to discover that risky entity.
The more information can be centralized, the more program administrators can begin to deploy strategic analytics. Much like an actual epidemic, drug use tends to spread geographically. Geospatial and network analytics can detect “hotspots” of use and trigger clean needle programs or similar health interventions.
Even high quality fraud detection algorithms are best at catching only the most extreme offenders. But, eliminating those outliers can remove a lot of the excessive drug supply from the system. Moreover, attention to fraud and abuse tends to deter other actors, and inspire a cultural shift away from excess. This shift will be critical to aiding the millions of sufferers in this epidemic while the medical community learns how to consistently give patients the most efficient treatment for pain. The opioid problem is large and complex, but armed with the emerging capabilities of healthcare analytics and the immense knowledge stored in national patient data repositories, our government is well-positioned to lead a robust counter-offensive.
