Opioid abuse is a national epidemic. Some 2 million people abuse prescription pain relievers and 591,000 use heroin, according to the American Society of Addiction Medicine. There were more than 20,000 deaths related to pain-killer abuse in 2015 and almost 13,000 heroin overdoses.
Technology leaders in Indianapolis have been using information to attack the problem since 2013, and have lately become increasingly sophisticated in how they apply data science to combat the scourge.
“Data is how you know whether something is really a problem or whether it is just a blip,” said Tom Arkins, chief of IT and Informatics with Indianapolis Emergency Medical Services (IEMS), the city’s ambulance service. “If we have five incidents in a day and no more for a month, that’s a blip. But when you have five every day for a month, that’s a trend.”
The trend in this case — and IEMS’ big data breakthrough — involves naloxone, an emergency medication used to check the effects of a potentially fatal opioid overdose. “We had been hearing anecdotal reports from our crews that we were using a lot more naloxone and I wanted to dig in and see what was really going on,” Arkins said.
Crews aboard the city’s 30 ambulances record any time the drug is given, so the data was already available, and the anecdotes turned out to be correct. From doling out 565 doses in 2011, the city’s emergency medical teams had ramped up to dispense more than 1,800 doses in 2016, and 2017 is on the same trajectory, with 468 doses administered by late March.
The data has helped the city to better organize its resources. “From a budgetary perspective, we can plan to have a greater supply of this so we don’t run short,” Arkins said.
A data-based approach also allows emergency medical personnel to coordinate their efforts more closely with law enforcement, in an effort to address the addiction at ground level.
“We started looking at it geographically to see what parts of the city were having problems,” Arkins said. “With our data, we can see down to the actual address, so we know there is a certain hotel where there is a problem, there are certain blocks that are problems. We can’t give the police an actual address because of HIPAA [Health Insurance Portability and Accountability Act] concerns, but we can aggregate it and make a heat map to show an area.”
IEMS now produces a monthly map, and is looking to develop a real-time tool. The ambulances’ data reporting systems don’t attach geographic coordinates to incidents: That information has to be coded manually, which slows down production of maps. Arkins’ team is exploring various off-the-shelf software tools that would allow it to geo-code on the fly in order to streamline production of maps, which in turn could facilitate more effective law enforcement efforts.
The data has helped to drive big-picture changes as well. Spurred in part by IEMS reporting, the state legislature last year passed a law that empowers police and fire crews to administer naloxone even before the arrival of medical personnel on scene.
In addition to teaming with emergency responders from other agencies, IEMS has leveraged its data to forge partnerships with the social service community.
“Now every time somebody gets a dose of naloxone, a message goes to the social workers and that triggers them to go to the hospital to start the follow-up process with that patient,” Arkins said. IEMS has been building up that capacity over the past year. “It’s about actually getting to people at the moment of impact. We want the social worker to get to them when everything is still fresh in terms of what just happened to them.”
Health officials also have begun to drill down more deeply into the data in search of other avenues to address the crisis. In so doing, they noticed that about 10 percent of patients who receive the counter-narcotic will get it more than once. “It tells us that there is a small subset of people who are having a larger problem, and so we want to think about how to target those people,” Arkins said.
Demographic data also has surfaced that could help to drive policy around the issue. It turns out 83 percent of people who get naloxone are from the local county, while 15 percent come from out of county and 1 percent from out of state.
“Where are these people coming from? Is there a way we can feed that information back to those police departments and those jurisdictions so they can look at it alongside what they are already seeing?” Arkins said. “This is epidemic across the entire country, and everyone’s data can contribute to the solution.”
While the data has helped to move the needle — engaging law enforcements, driving policy, building a social service response — Arkins says that such an approach must be handled with some care. In a complex landscape, subtle cues can be easy to miss.
Suppose police see a cool spot in a heat map of heroin activity. One might suppose that’s an area without a problem — but it could be an indicator of a zone where a drug dealer operates. “You have to look at the whole picture,” he said. “You can’t just look at what you are seeing on the surface, because there may be something else there that is telling you an important part of the story.”
Nor should medical services be looking at the data alone. The information may originate with ambulance crews, but an effective use of the data requires multiple sets of eyes. “This is not an EMS problem, this is not a police department problem," he said. "This is a city of Indianapolis problem. We need to be working together. We need to be sharing our data. That’s how you fill in the story."