By using website you agree to our use of cookies as described in our cookie policy. Learn More

TDWI Upside - Where Data Means Business

Q&A: Firm Uses Analytics to Curtail Prescription Drug Abuse

Analyzing data about injured workers can help reduce the rate of prescription drug abuse in workers' compensation claims.

Prescription drug abuse is a huge problem nationally -- over 100 people die each day in the U.S. from abuse of pain medications. Some 75 percent of those deaths are attributed to opioids like OxyContin, Percocet, and Demerol. A company called Helios uses predictive analytics to work closely with pharmacists to control pain medication abuse specifically among workers' compensation claims.

By gathering data and running analytics against an injured worker's age, location, demographics, socioeconomic status, and medications filled in a claim, Helios pharmacists can compare that data to historical models and better tailor a pain treatment plan to an individual. The result: The company has reduced opioid prescription rates by over 36 percent among its client companies.

In this interview, Joe Anderson, who is director of analytics at Helios, addresses the role of predictive analytics in the company's success in reining in pain medication abuse. "We have a long history of helping our clients identify ways to improve outcomes for injured workers," Anderson says.

BI This Week: What is Helios and how does the company use analytics?

Joe Anderson: Helios is a workers' compensation pharmacy benefits manager. Among other services, we work to control costs for worker's compensation. With respect to predictive analytics, we work with payers and worker's compensation in auto and other areas to help them identify what might happen with prescriptions using pharmacy management data. [A client] might be a large insurance company or a third-party [claims] administrator, for example.

How are predictive analytics being used in healthcare, specifically around prescription drug use and abuse and workers' compensation?

Before predictive analytics, the standard method that health care workers used to make decisions about improving patient outcomes was to use academic studies ... or information in a medical journal. Predictive analytics has become a way to supplement this. We can use data from large health care companies, both on the provider's side and on the payer's side, who have internal evidence about what's leading to the best outcomes for patients.

At Helios, we focus on worker's compensation and pharmacy and drug use and abuse. Better outcomes mean getting people back to work faster and avoiding dependence on drugs, especially opioid-class drugs.

Regarding workers' compensation and the use of medicine to treat pain in worker's compensation relative to Medicare, a much higher percentage of prescriptions filled for workers' comp claims are for pain medicine, including opioids. Opioids are high-risk medications for dependence. We're trying to use prescriptive analytics as a tool to combat the prescription drug use epidemic.

Speaking of Helios specifically, we have a long history of helping our clients identify ways to improve outcomes for injured workers. We have years of pharmacy data and workers' compensation [data], so with that, we've found an opportunity to use predictive analytics to identify which injured workers are at the most risk for longer-term prescription drug use and abuse.

Can you give an example of how that works?

Sure. [To build a current data model], we would take everything we knew about an injured worker in 2010, and correlate that to what prescription drugs they are using four years later, and we would see if there was something we could measure in 2010 that would have predicted what would happen later -- that's your standard statistical model. From that, we would use that information in conjunction with a pharmacist's ability to make recommendations to change the behavior of those claimants. We're seeing some success with using those measurements and that intervention to drive down opioid use. We can reduce opioid use by 20 to 50 percent, measured by morphine equivalents in our industry.

Helios says it has used predictive analytics to reduce opioid prescriptions by 36 percent. Can you delve into that number a bit?

Opioids are the No. 1 drug class prescribed in workers' comp, and it's been that way for years. Over 22 percent, roughly, of prescribed medications [in workers' comp cases] are opioids, but that almost understates the case because there will be injured workers who are taking opioids along with associated medications such as antidepressants. A lot of what we're doing is measuring long-term use of pharmacy [products] in workers' compensation cases focusing on opioid management. Our clients ask us to find ways to reduce use or weed patients off opioids or to help prevent its use in the first place.

That's specific to what Helios has done. One of the things we look at for an injured worker is, how long do they continue to use medication after they are injured? The whole idea behind predictive analytics is if we can do an intervention with them within the first six months of their injury, we can prevent some long-term opioid use.

It's a different way of thinking about it. Pharmacy managers look at the savings per claim of getting prescription drug use under control. This is a new way of preventing something from happening in the first place. There are challenges in that -- how do you show that this injured worker got of opioids, and that you actually prevented them from doing that as opposed to it going away by itself. It's been very successful.

What kinds of data are used to predict long-term opioid use? What do your models look like?

That's a great question. We spend a lot of time on that. Again, starting with the traditional way, we look at academic studies and morphine equivalents, a very common term in our industry. Someone who is using a certain level of morphine equivalents is at a higher risk for opioid dependence then another, so those are the pharmacy metrics that go into the models. We get everything about what the injured worker is currently using in terms of morphine equivalents, in terms of supply, costs of their medications, the types of medications they are getting, and so forth.

There's also demographics around the injured worker, including what type of injury they had [and] are they within a stage where we can direct their care. And then there are a lot of pieces around the prescribers that they are going to. We've really started to get a handle on this over the last couple of years. Often, we see that injured workers who go to pain management prescribers tend to have longer opioid use than those who to go general practitioners.

It kind of makes sense. There's a lot of discussion in our industry about the causality there. Are injured workers going to these pain management [clinics, who are] getting them hooked? That is not our position. Our position is that we don't know the causality. Maybe they have an injury that is going to require them to need opioids for a long time and that's why they came to this pain management prescriber in the first place. It's all data to help triage or determine where we need to focus our efforts first.

You said earlier that before predictive analytics, decisions about patient care were often made in a different way. When was the turning point in using analytics in health care? Did that happen that fairly recently?

I would say that it's still changing. Helios is a leader within our industry, and there's a lot of supplemental work that goes into using predictive analytics correctly. ... You have to approach it in the right way. Bringing predictive analytics in front of people who aren't familiar with it can be overwhelming sometimes. The approach we've taken at Helios is to use it internally with our pharmacists, who are trained to understand predictive analytics. We incorporate that into the intervention suggestions that they make.

What can be done to make this use of analytics more productive and more widespread? Is better data needed? Better tools? Better education of health care workers on how to use predictive analytics?

Again, I would say that education is most important. Many people [mistakenly] think they are using predictive analytics. That would be something along the lines of "we are taking a piece from this health care study that says we need to focus on everybody, and this level of opioid use, and we should intervene on all of these." They may think that is predictive analytics but it's not.

Predictive analytics is really a multivariant way of looking at things. It's not just about what level of opioids someone is on but who the prescriber is, what type of injury the injured worker has -- -- getting all of that together can link the injured worker to the percent likelihood that they are going to be a long-term [opioid user]. That's just an education piece so people understand how to do this.

What about the privacy issues that inevitably arise around healthcare and data discussions?

With respect to data and tools, because we're in healthcare, people are always concerned about privacy issues and around how the data is used. With what we do, each company has its own piece of the data. We're the experts on pharmacy. Some of our clients have less pharmacy data but have data in other areas. Everyone is very careful about sharing all that while still maintaining privacy and being fair to the injured worker. I don't see it as an issue of needing more data or better tools but rather just taking good care of the data we have and analyzing it better.

Do you make a distinction between predictive analytics and prescriptive analytics, which we might define as taking action based on what you've predicted?

At Helios, we take our predictions about who we think is going to be the most at risk for long-term prescription drug abuse and we give that to our pharmacists. The pharmacist makes the decision about what should be done to prevent abuse from happening. There is a lot of expertise within our pharmacies; we simply don't have the ability to program and predict every single drug combination and injury combination that could happen. That would be what is generally called prescriptive analytics.

I know people sometimes think of this as a continuum in which everyone will eventually get into prescriptive analytics. However, I think there is still a lot of opportunity to use predictive analytics to understand where the problems are before we even get to the point of trying to automate it. It just seems orders of magnitude harder.

[At Helios,] I don't know if we'll ever do prescriptive. There will always be so many new drugs coming out and so much to learn about something that might have been suggested today to change an injured worker's behavior might be totally different from what you would have suggested five years ago. As long as suggestions are backed by historical data, the pharmacist is going to be ahead of what the historical data can reveal. I don't know if that's the case in other areas of healthcare.

TDWI Membership

Accelerate Your Projects,
and Your Career

TDWI Members have access to exclusive research reports, publications, communities and training.

Individual, Student, and Team memberships available.