The Data Antidote
Claims data proves a valuable weapon in the fight against opioid misuse.
- Kate Smith
- September 2018
- Capture the Flag: Workers’ comp writers have used claims data to identify red flags that indicate patients who are at high risk for opioid misuse.
- Managing Meds: By intervening earlier in the life of a claim, insurers are helping patients find alternatives to opioids.
- Controlling Cost: Workers’ comp pharmacy costs have dropped by nearly a third in the past eight years, largely due to a decline in opioid spend.
Dan Hunt remembers the day the posters went up in his hospital. “Pain is the fifth vital sign,” they read.
It was 2001 and the Joint Commission, a nonprofit organization, had just launched its Pain Management Standards, which made treatment of pain a key focus for medical providers. The Visual Analogue Scale—a series of faces ranging from smiling to frowning to crying—was introduced as a tool to measure the severity of a patient's pain, and doctors were judged on how quickly they alleviated this highly subjective new vital sign.
“If a patient's pain was over a 5 on the scale, you needed to address it,” recalled Hunt, a former surgeon who is now the medical director at AF Group. “You had to document what your intervention was to fix pain, because the Joint Commission audited charts to make sure people's pain was being addressed in a timely fashion.”
“Consequently, every resident is going to learn that if a patient says they have pain, you should give them a narcotic. It's no wonder doctors transitioned from careful judicious use of opioids to everybody gets opioids.”
That quest to ease pain ultimately contributed to the rise of a painful epidemic—the opioid crisis.
Opioids are a class of highly-addictive drugs that include prescription pain relievers such as oxycodone (e.g., OxyContin, Percocet), hydrocodone (Vicodin), codeine and morphine. According to the Centers for Disease Control and Prevention, 61.8 million people in the United States received one or more opioid prescriptions in 2016 and 11.5 million misused prescription opioids. The U.S. Department of Health and Human Services (HHS) reported more than 42,000 people died of an opioid overdose in 2016 and 2.1 million had an opioid use disorder.
Last year, the HHS declared the opioid problem a national health emergency. The insurance industry, however, declared it a business emergency years ago.
“We saw the opioid problem a lot earlier than the general population because most of our injured workers in the mid-2000s were being prescribed opioids, even for relatively minor injuries,” said Mary Nasenbenny, senior vice president of Group Benefits and Workers' Compensation Claims at The Hartford. “So we in the comp setting have been more aware of the problem, particularly the problem of addiction.” The Hartford Insurance Group is the second-largest U.S. workers' comp writer based on direct premiums written, according to A.M. Best.
Insurers, particularly workers' compensation writers, have been battling the addiction problem ever since. Their most effective weapon in this war on drugs has been claims data.
By applying data analytics to claims information, insurers have been able to spot dispensing trends, monitor opioid usage, compare treatment outcomes and identify red flags that could predict whether opioid use is likely to become misuse.
“Insurers are actively trying to use claims data to identify risk earlier and intervene earlier,” said George Van Antwerp, senior manager at Deloitte Consulting and co-author of a Deloitte report on strategies for stemming the opioid crisis. “Everybody is doing something to address this crisis. Some are more progressive in their use of predictive analytics, but they are all looking into it.
“This has moved from being a clinical issue to, in many cases, a CEO type of issue. They see the impact this is having and are trying to take more forward action.”
Workers' compensation insurers have been quick to turn their claims insights into action, creating programs and protocols to intervene early and keep patients on a path to quick recovery. These initiatives have also helped rein in costs.
Workers' comp writers have decreased pharmacy spend by 31.5% or $1.1 billion over the last eight years, according to CompPharma, a consortium of workers' compensation pharmacy benefit managers. Opioids account for nearly a quarter (24.2%) of the roughly $3.7 billion spent on total pharmacy, but opioid spend has dropped dramatically thanks to targeted efforts to reduce usage. Opioid spend decreased 18.2% in 2017.
Experts say the credit goes to the data.
“We're seeing the results of what the data can do for us, and it is exciting,” Nasenbenny said. “This is the way to drive much better outcomes.”
Mary Nasenbenny, The Hartford
"We’re seeing the results of what the data can do for us, and it is exciting. This is the way to drive much better outcomes."
Every night, 25 machine learning models scan the 130,000 workers' comp claims files The Hartford receives each year. If a new piece of information is added to a file, it will be analyzed the same day.
In isolation, a notation by a claims handler or a comment by an injured worker may appear innocuous. The data has shown, however, that harmless statements aren't always harmless.
“Through data mining, natural language processing and text mining of claim handler notes, voice-recorded transcripts and third-party data, we have identified certain characteristics or flags that suggest an individual is susceptible to a delayed recovery,” Nasenbenny said.
By continuously scanning files, The Hartford hopes to spot problems before they arise. Its data analytics program, called Vista, combs through all of the workers' compensation claim information to identify claims that could become volatile in the future.
“We may have a claim that comes in with a low back strain that looks sort of vanilla, and we would expect a typical duration,” Nasenbenny said. “But there could be other factors lurking that we will find through the data that our claim adjuster might not have or might miss.
“So our Vista program, through data analytics, will identify claims that could become much more severe than they initially look. It will then send those claims to higher level adjusters immediately—adjusters that have more expertise or have experience dealing with individuals who have comorbidities or complications. It may also assign a clinician to that file very early, because we know it has the potential to be much more complex than it looks today, and we don't want that claim sitting with a less tenured individual who might not recognize the flags right away.”
One of The Hartford's machine learning models looks for psychological and social flags. It searches text for indications that an injured worker might have social or psychological stressors that could affect recovery or lead to opioid addiction.
“We might be having the initial interview with the injured worker and they may mention they just were divorced or they just lost a spouse or they just moved to a new place and don't know anyone. These are things that might suggest a social challenge for them,” Nasenbenny said.
“The psychological piece is that they're demonstrating an inordinate amount of anxiety or depression. Not clinical depression, but appearing really down or fearful. That might not appear for the first couple of weeks of the claim. But once it does—if the injured worker suggests they're afraid they may get hurt again or afraid they may never be fully restored to function—the fear is damaging to recovery, our data suggests.”
When it analyzed its claims data, The Hartford found that 11% of injured workers who miss time from work demonstrate these psychological and social flags, but those 11% account for more than 50% of claims costs.
“Claims from this small group of injured workers—because of the potential for not only longer recovery but also additional surgeries, relapses and opioid use—tended to snowball if they didn't get additional support early on in the claim.
“They might start with conservative treatment, but if it's not working it will morph into more aggressive surgical treatment and ultimately opioids. Those cases go down a bad path.”
Two years ago, The Hartford began offering coaching to injured workers who exhibited these psychological and social flags. The 10-week coaching sessions teach injured workers how to manage different aspects of recovery. High-risk individuals also are informed of alternative treatments for pain relief and, if warranted, offered cognitive behavior therapy.
“In cases where we have successful coaching—and by that I mean, we invite them and they participate in the program—we have double-digit improvement in durations (i.e., shorter claims) and better overall claim outcomes,” Nasenbenny said. “Usually they avoid the use of opioids. So it's good for us, but more importantly it's really good for our injured workers.”
The numbers behind the problem.
|61.8 million people receive one or more opioid prescriptions.
|11.5 million people misuse prescription opioids.
|2.1 million people have an opioid use disorder.
|42,000 people died from an opioid overdose in 2016.
|116 people die every day from opioid-related drug overdoses.
Sources: Centers for Disease Control and Department of Health and Human Services. Based on 2016 data.
Liberty Mutual began running predictive analytics a decade ago on its 300,000 annual workers' comp claims.
“Originally what was observed were slow emerging cases that looked like relatively minor claims and were getting the attention associated with a minor injury,” said Frank Radack, vice president and manager of workers' compensation managed care at Liberty Mutual. “But over time, return to work didn't happen or we saw the emergence of opioids on cases.” The Boston-based company is the seventh-largest U.S. workers' comp writer based on direct premiums written, according to A.M. Best.
Through claims analysis, Liberty Mutual identified some of those early indicators and used them to flag cases that needed intervention.
“We have a set of data I would refer to as 'additional data capture' that is outside the normal course of claims handling. It helps identify risk factors—everything from comorbid conditions to the injured worker's attitude, whether that's toward the employer or their own recovery,” Radack said. “Those different data points can weigh into the predictive trajectory of a claim.
“If we see something that either warrants attention based on severity or potential to escalate from where we internally have it set, those flags go to our case owners so that a clinician can intervene or other resources can be offered. So what the predictive models allow us to do is intervene earlier in the life of the claim. If you can do that, you can change the trajectory of claims.”
When the CDC released its opioid use guidelines in 2016, Liberty Mutual embedded them into its drug and data strategies.
“We changed some of our alerts that happen at point of sale to engage a clinician if we see short-acting opioids prescribed beyond seven days' supply or if we see a morphine equivalent dosage that's high or if we see opioids prescribed with benzodiazepines, which are central nervous system depressants,” Radack said. “So we changed our strategy to engage earlier in the life of the claim, and we saw some positive results.”
Since then, Radack said, Liberty Mutual has seen a 9% reduction in patients who are receiving opioids and an 18% reduction in the average day supply that injured workers who did receive opioids were getting in the first six months. Opioids as a percentage of overall pharmacy spend also has decreased by 12%, Radack added.
“We're really encouraged by the results,” Radack said. “We've been able to get very granular and get insights on how and where we deploy clinicians. It's made us a smarter, better partner.”
Workers’ comp writers are spending less on pharmacy costs.
|$3.7 billion, approximate annual pharmacy spend by workers’ comp writers.
|24.2% of total pharmacy spend is directed toward opioids.
|18.2% decrease in opioid spend in 2017.
|9.8% decrease in total pharmacy spend in 2017.
|$1.1 billion (or 31.5%) decrease in pharmacy spend by workers’ comp writers in the last eight years.
|$1 out of every $6 spent on opioids was cut in 2017.
Avoidance of opioids is seen as critical in workers' compensation. Not only are these drugs addictive, but they are also costly—and not just in terms of dollars. Opioids complicate the path to recovery, with side effects ranging from depression and anxiety to gastrointestinal issues and sleep problems.
AF Group wanted to get a full picture of the impact opioids had on claims. The insurer teamed with Johns Hopkins University in 2012 to study the correlation between opioid usage and claims duration. What they found was that claims duration tripled when an injured worker received a long-acting or extended release opioid, such as OxyContin.
That insight spurred AF Group to create a dedicated pharmacy team managed by nurses and a pharmacist. The company also enlisted its predictive analytics team to build a program to monitor prescriptions in real time.
“You can't look at every script. You need analytics,” Hunt said. “Our Early Detection program runs in the background and looks for a whole host of things to identify those claims that may be having some challenges. It brings forward prescriptions with certain characteristics that we think are important for a nurse to take a closer look at.
“Our flags are not to stop people from getting pain meds; they're a chance for us to look at the claim and make sure the medication is appropriate.”
AF Group's efforts have led to a decrease in opioids. Hunt said the company's opioid spend was down 7.6% last year and 17% over the last three years, and the number of injured workers receiving an opioid script dropped from 34% in 2016 to 31% in 2017.
The Hartford implemented similar measures after analyzing its historical claims data and finding claims were twice as expensive when opioids were prescribed.
“Because we were able to look at like claims, we realized the opioids were a significant factor in the delayed recovery we were seeing,” Nasenbenny said. “We also saw that in individuals who were prescribed and taking opioids, there was an addiction problem. Research has shown that one in five who start a 10-day supply of opioids become long-term users. One in four or one in five will become addicted. And individuals taking opioids for 30 days are highly likely to become addicted.
“So we saw that, but we also saw a host of other issues that complicated the path to recovery. A lot of it related to things like depression and other complications directly resulting from taking opioids—nausea, digestive issues, muscle issues and a host of other problems.”
Convinced of the problem by its data, The Hartford in 2014 launched a series of programs to address the prescribing of opioids on new claims. In addition to having adjusters educate injured workers on the seriousness of opioids and involving clinicians in drug review activities, it also enlisted peer review doctors to have conversations with physicians who were prescribing opioids.
“Through our pharmacy benefit manager, we tightened up our formulary and would reject any opioid prescription for more than that first handful of days post-injury,” Nasenbenny said. “That's not to say further prescriptions for more than a two-, three-, or seven-day supply of an opioid post-procedure would not be approved, but we would reach out to the physician and ask whether there were any alternatives that could be used for pain relief, and if they had created a weaning plan for the individuals.”
Since 2015, when its programs were up and running, The Hartford has seen opioid prescriptions drop 39%, Nasenbenny said.
“A great outcome for us is our claimant recovering, returning to function, and feeling good,” Nasenbenny said. “When the data can help us do that, it's a win for all of us.”