It’s no secret that big data can combat the problem. Here’s an inside look at what experts are actually doing.
MY COUSIN DAVE WAVED HIS LIGHTER beneath a folded sheet of tinfoil, staring at the crushed blue pill as it melted into brown ooze along the crease. He dipped his head and inhaled the vapors through a hollowed pen pressed between his lips. The foil and pill snapped and fizzled as he slid the flame and his makeshift pipe down the line of chunky powder. Seconds later, he exhaled and sank into a serious opioid high.
Dave, a pseudonym to protect his identity, had become dependent on “blues,” slang for Roxicet, the prescription pain medication composed of oxycodone and acetaminophen. He smoked them every day, at first for fun and then to ward off the withdrawals that kept him from attending class and going to work. Several times, he came close to what he assumed were overdoses, smacking his own face to stay awake. Other addicts ransacked his family’s home for drugs, and a man once pointed a gun at him in a deal gone wrong.
Dave’s opioid abuse, with its undesirable effects, began about a decade ago, making him an early victim of the opioid crisis.
He also played a role in its rise. Dave needed a constant supply of pills, which sold for about $25 apiece on the black market. He began selling opioids and other prescription narcotics to fund his consumption. At first, he sold to other users, doctor-shopped, and stole the occasional blank prescription pad. Networking through the suburban underworld, he built more substantial connections. Soon he was taking weekly trips to a sketchy pain clinic, where gangsters took his cash and an elderly physician asked about fictional back pain and wrote a prescription for a Herculean bottle of opioids. When Dave left, he smoked some of the pills and sold the rest to smaller dealers.
Dave eventually got clean. But whether he could game the system today as he did then is unclear, if unlikely. If he were using now, a prescription drug monitoring program might nab him for doctor shopping, or a health insurer could notice the dirty doc’s peculiar behavior. Perhaps Dave’s prior medical history would have tipped off healthcare providers to his potential for opioid abuse before it ravaged his life.
Since the early days of the opioid epidemic, big data and analytics have rocketed in both power and prevalence. Stakeholders across healthcare, government, and academia are crunching numbers and merging previously disparate pieces of information to map and thwart its spread. Their efforts focus on everything from harm reduction and drug abuse prevention to scanning for alarming activity. Healthcare leaders and public health officials consider the use of data a key tactic in the larger fight.
The precise scope of the problem is blurred, but nearly everyone agrees it is overwhelming. Some expect opioids to kill half a million Americans over the next decade. In 2016, more than 42,000 people died as a result of opioid overdoses, 5 times more than in 1999, according to the CDC. The agency also claims that 2 million people in the United States misused or were addicted to prescription opioids in 2014. These numbers are expected to increase, a sign of a crisis so devastating that late last year, the president declared it a “public health emergency.” Healthcare professionals have labeled it “unprecedented.”
The effects that big data and tech will have on the opioid crisis remain to be seen. But campaigns have begun to shape how this country targets the issue, and more innovations are on the way. For hospitals and other healthcare organizations, it is crucial to see what is happening on the ground. Experts reached for this story said the industry needs total buy-in, and only by understanding these data-driven initiatives can hospitals prepare to strike partnerships or craft their own responses.
IN HIS OFFICE at the University of Virginia, economist Christopher Ruhm, PhD, was trying to solve a mystery. Over the years, he had shown that physical health improves when the economy stutters, but about 5 years ago, that dynamic was changing. In his quest to understand why, he cornered a suspect: a rise in poisoning mortalities. “I had no idea what poisoning deaths were,” he said. “I was thinking rat poisoning.”
It turned out that drug overdoses had caused more than 90% of poisoning deaths. The revelation surprised Ruhm, who decided to dig deeper. “But as I got into it,” he recalled, “I discovered that the data were more problematic than certainly I had known.” For example, roughly a quarter of overdose cases compiled by the CDC included no information on which drug caused the death. Ruhm settled on his next step: “Can we come up with better numbers?”
Computing his way through data sets, the investigator learned that drug-related deaths often involve multiple substances. In 2015, that was true about half the time, a statistic compounded by mortalities involving cocaine and fentanyl. “Drug cocktails” made ascribing deaths to a single culprit difficult, and the many possible narcotic combinations further complicated the picture.
Establishing baseline information is important, Ruhm said, because it provides the foundation from which healthcare providers, public health officials, and civic leaders act. If they do not know what exactly is killing people, they might stumble to provide relief. “Just understanding the dimensions of the problem—clearly, good data are critical for that,” Ruhm added. “We have to have good data to make good policy and, in this case, to make policies and interventions that could help to address the problem.”
Ruhm’s peers in social science are exploring policy options, how changes to public health insurance programs like Medicaid influence drug use, whether the greater availability of Naloxone (the medication that reverses the effects of an opioid overdose) encourages use, and to what extent physical changes to Oxycontin have pushed people to move to heroin. Ruhm, meanwhile, continues to analyze drug death data and advocate for more actionable data points, examining whether the economy or a change in opioid pricing and access is driving the epidemic.
What does it all mean for healthcare organizations and clinicians? They, too, must be on the lookout for bad data. And the more investigators like Ruhm toil behind the scenes, the better informed and prepared hospitals might be in the future—and that means patients could see better outcomes.
IN THE 1930s, something revolutionary was brewing in California. Eager to curb criminal drug abuse, lawmakers and law enforcement devised a plan to track Schedule II controlled substances through a paper-based system. When a patient received a drug like cocaine, morphine, or methadone, the prescription information would be copied 3 times by the doctor and pharmacist before entering a state database. The California Triplicate Prescription Program went live in 1939 and ran for roughly 60 years, until a new revolution changed everything.
The digital era enabled California and states that followed in its footsteps to keep tabs of prescriptions over the internet, providing more convenient and faster access to records. Although some areas took more time before committing, all 50 states enacted laws creating some type of prescription drug monitoring program by last fall.
“Now, it’s not just a question of whether it can be done, but it’s also a matter of getting the right information to the right people at the right time to get the right value,” said Daniel Castro, MS, vice president of a think tank called the Information Technology and Innovation Foundation and director of its Center for Data Innovation. If you ask him, state-level prescription drug monitoring programs are the most prominent public uses of data in the fight against opioids.
The initiatives function like California’s legacy system, in that they track a prescription through the healthcare system, often incorporating providers, pharmacies, state insurance programs, licensing boards, state health departments, and law enforcement agencies.
Ideally, the first success a prescription drug monitoring program might have is catching someone trying to fill the same prescription twice. Then doctors try to help the patient, who learns that this path is unsustainable, Castro said. The programs may also catch people using fake bulk prescriptions, reducing black market supply. Over the long term, the programs could empower physicians in emergency rooms, primary care settings, and elsewhere to recognize troubling patterns among patients who receive opioids after an injury and then repeatedly ask for refills. From there, the programs could monitor the progress of patients undergoing substance abuse disorder treatment, Castro said.
“A lot of states are doing different things, and some are doing it better than others,” he added. New York, for instance, established a mandatory reporting program in 2012 and saw a 75% drop in doctor shopping the next year, according to the CDC, which publishes guidelines and provides funding to nearly every state for data surveillance and prevention initiatives. The federal government has recognized Ohio, Kentucky, Tennessee, Florida, and Oregon for their versions of the program.
Despite guidance from the CDC, prescription drug monitoring programs face considerable challenges, Castro said. For starters, it is difficult to build complete databases. “You want to have all the data, not just some of them,” he said. “You want all doctors reporting and all pharmacies reporting and looking up to see what kinds of transactions are taking place.” Voluntary reporting programs impede this goal, along with privacy concerns, a lack of longitudinal record keeping, and insufficient interoperability. “The data aren’t there yet,” Castro said. “We’re still in the early stages of collecting those data, and that is a necessary first step.”
If states and healthcare stakeholders collaborate to nurture prescription drug monitoring programs, they may one day become capable of intervening when a patient exhibits early signs of addiction, before they lose their jobs or their support systems, and when a physician prescribes opioids too often. States may come to monitor intervention programs and then scale up, modify, or even quash certain plans.
WHEN CERNER, the nation's largest electronic medical records (EMR) vendor, decided to commission an opioid data task force, it assembled 7 employees. By the end of the group’s second meeting, 5 had shared stories of how the crisis had directly affected them. Two said they had lost loved ones to overdoses. “As we started to delve into that and have broader conversations, it became extremely personal to us,” said Jennifer Conner, Cerner’s senior director and solution executive of state and public health and the project lead. Before long, the size of the task force had grown to 32 people.
They brought different skills, areas of expertise, and perspectives but shared a common goal: to leverage a big data platform to gain actionable insights for clients like Medicaid programs, public health agencies, and individual healthcare organizations. Although the epidemic was obvious, few knew its true scope or how to target resources based on information. Cerner believed its team of providers, strategists, and data scientists could unearth that information. “Quite frankly,” Conner said, “it’s too important not to do something.”
First, like the economist Ruhm, the task force set out to determine the magnitude of the opioid crisis. Cerner built analytical tools and data sets for a state public health agency vying to probe the nuances of its population. The EMR firm pulled in claims data, which spotlighted the actions of providers and patients who were obtaining opioids through legitimate means. To examine more illicit activities, the team turned to prescription drug monitoring program data for insights regarding access, discretionary levels, and types of use. But that wasn’t enough.
Next, Cerner added data on crime, social determinants of health, and treatment access points. Conner and her colleagues learned about the environments in which patients lived, black markets and all, though they have yet to complete this work. “But we just try to layer our strategies around ‘What’s the right data to tell the story?’ and to give our clients the data at their fingertips,” she said.
Early on, Conner came across a shocking sign that her effort was worthwhile. After gaining access to a large data set, she opened the dashboard and saw that a person with a broken arm had visited 4 providers in 1 week and got an opioid prescription from each. The state had installed gates to prevent such exploitation, but a breakdown occurred anyway, to the client’s surprise. “That, to me, was the chills-down-my-spine, hairs-on-my-neck-standing-up moment,” Conner said.
Cerner also identified high-prescribing doctors, an issue it is trying to address at the point of care, through its EMR platform. But the task force must also be careful to explore context—a flagged provider, for example, had an online reputation as a pill shop, an assertion supported by the data, but left out of the narrative was the fact that it was a pain clinic. Its job was to prescribe opioids.
The EMR task force is working on another big project: predictive analytics. It is melding machine learning and behavioral health expertise to build a model that tackles grim relapse rates. The data may show the flow across pain populations, from treatment to outcome to the point at which they must taper off opioids, Conner said.
Artificial intelligence might also discover gaps between people with opioid abuse disorder and treatment facilities, she noted. Where are the beds? Are these facilities open to people who live nearby? Is the population growing? What’s more, is a given population at high risk for substance abuse, and is funding available? Cerner hopes to predict and solve these problems before they occur. Public and private facilities, for instance, may one day lean on analytics to obtain grant funding. Providers, meanwhile, could use the technology to forecast outcomes and vulnerabilities in people who receive long-term opioid prescriptions.
“I’m sorry, but down the road is too late,” Conner said of the need for predictive analytics. “I don’t want anyone else to become addicted to this or become a death statistic.”
THE DOCTORS WERE NOT THRILLED. A year and a half ago, the health insurance behemoth Aetna mapped all its prescribers by how often they recommend opioids. Each provider in the top 1% received a personalized letter, signed by the company’s executive vice president and chief medical officer, laying bare their standing in a race few want to win. Next, Aetna sent similar notifications to the highest 5% of oral surgeons who write opioid prescriptions for more than a 1-week supply. Emails came rolling in, along with 24 phone calls. “Doctors were skeptical, to say the least,” recalled Dan Knecht, MD, Aetna’s head of clinical strategy. “Some were angry. Some were curious.”
But he had the data to back it up. His team reviewed the numbers with the prescribers, looking at each scenario, fielding feedback, and talking through the unwelcome revelations. “If we had taken a different track and kept the discussion of the analytics high-level, I think the conversation would have been a lot less friendly,” he said. It’s too early to tell whether they changed their habits, though Knecht said education drove the campaign. His staff also recommended an alternative to opioids to oral surgeons.
Aetna and other health insurers fight for yet another unit in the war against the opioid crisis. Knecht said he views the rise of data here like the advent of radar in World War II. Analytics allow him to better understand, and respond to, incoming strikes.
Aetna aims by 2022 to increase access to evidence-based treatments for chronic pain, reduce inappropriate opioid prescribing for its customers, and boost the percentage of members with opioid use disorder receiving medication-assisted treatments and other proven therapies—all by 50%.
Data soldiers monitor progress and other trends through an internal dashboard. They have extracted insights regarding patients who overdosed in the past or are receiving benzodiazepines, tracked those with an opioid prescription spanning more than 1 week, and retooled their approach based on the metrics.
The movement has also prompted secondary interventions. By analyzing Naloxone prescription data, for example, Aetna found that as many as 30% of the prescriptions had gone unfilled in high-need areas. The insurer then donated some of the drug to places like Kentucky and Pennsylvania and, as of this year, removed co-pays for the anti-overdose drug. It has also pointed to CDC data in instituting a 7-day opioid fill limit. “People get super excited about various sophisticated analytics, and I am, too, but I think we shouldn’t try to overengineer a solution,” Knecht said.
Even so, his team is working on a tool that could identify patients who are going in for surgery and have a risk of opioid abuse, alerting stakeholders to the danger.
As in traditional wars, new technologies are emerging from the opioid crisis. “This is, unfortunately, a form of war we find ourselves in,” Knecht said. “Hopefully, these barriers will come down, and we will integrate better solutions to impact population health.”
IN ITS FIGHT AGAINST OPIOIDS, the nation has drafted many data experts, from insurers and governments to EMR vendors and academics. Their individual contributions and greater collaborations may pinch the problem in any number of ways, but the important thing to the experts is that these stakeholders are coming together. Even if some projects fail, at least a few are bound to succeed on some level.
“There’s really nothing in modern times that is anything close to this,” Aetna’s Knecht said. “Because it’s unprecedented, the traditional silos have been broken down, and there’s much more collaboration.” Conner, of Cerner, agrees that it is time to come together, a call to action that also applies to the hospitals and clinics in the trenches.