In 2020, nearly 92,000 Americans died due to a drug overdose, and three-quarters of those deaths involved an opioid, according to the Centers for Disease Control and Prevention. Of those treated for nonfatal opioid overdoses in the Emergency Department (ED), 1 in 20 die within the year, many within the next two days. Lives can be saved by prescribing naloxone, or Narcan, an opioid rescue drug, when patients who experienced opioid overdoses are discharged from the ED. However, naloxone is prescribed in just over 7% of cases.
That is a huge missed opportunity to help these patients, said MUSC Health ED physician Lindsey Jennings, M.D., who has spearheaded other ED-based initiatives to address the opioid crisis.
“Take-home naloxone is a key component of harm-reduction strategies for patients presenting with opioid overdose,” said Jennings. “Studies have shown that increased access to naloxone in communities can reduce mortality.”
One unexpected finding during the rollout of a new Medical University of South Carolina (MUSC)-led initiative known as the Opioid Overdose Network (O2 Net) is that a novel physician-decision tool dramatically increases the rate at which ED physicians prescribe these patients take-home naloxone.
“Any intervention that improves naloxone distribution has the potential to save lives,” said Jennings, who is also part of O2 Net, when asked about the tool.
For real advances to be made in the care of these patients, approaches that are successful at the local level must be tested on a national scale in clinical trials. The mission of O2 Net is to make it easier to conduct opioid overdose trials by helping researchers to identify patients for their studies using the electronic health record (EHR). These efforts, funded by a grant from the National Center for Advancing Translational Sciences, are detailed in a recent article in JAMIA Open. The initiative is led by Leslie Lenert, M.D., director of the MUSC Biomedical Informatics Center (BMIC) and associate principal investigator for the South Carolina Clinical & Translational Research Institute.
Instead of trying to “reinvent the wheel,” the O2 Net team is adapting existing clinical trial infrastructure – the Accrual to Clinical Trials (ACT) network – to this purpose. The ACT network enables Clinical and Translational Science Awards (CTSA) hubs to share EHR information on more than 130 million patients.
“A foundation is just that. It’s something to build on,” said Lenert. “If you don’t use the foundation you have, then you are going to have to pour a new one, and then that always takes time.”
And there’s no time to waste when it comes to the opioid crisis.
Lenert and his team at MUSC are working on adaptations for the ACT network that will make it more suitable for the study of opioid overdose. As the adaptations become available, they are rolled out to the team’s collaborators at Dartmouth, the University of California San Diego and the University of Kentucky, all CTSA hubs. If this new, focused O2 Net proves successful in this pilot project, it could be adopted by more CTSA sites in the ACT network.
The first needed adaptation was to develop a way to identify ED patients treated for opioid overdose, using EHR data. The O2 Net team, which included experts in biomedical informatics and biostatistics as well as clinicians, achieved this goal using machine learning and natural language processing techniques, both of which are types of artificial intelligence (AI). “This involves meticulous reviews of charts to validate the AI models,” said O2 Net team member Jihad Obeid, M.D., Endowed Chair for Biomedical Informatics Associated with Clinical Effectiveness & Patient Safety. The team also recruited medical students to help with the validation process.
“The collaboration with the clinical team, ED doctors, psychiatrists and addiction science experts in this study is essential for successful informatics approaches,” said Vivienne Zhu, M.D., of BMIC, who headed up the natural language processing team. “As we observed in this study, incorporating clinical experts’ knowledge into algorithms significantly improved the performance.”
The next step was to make the opioid-related data in the EHR more robust. Bioinformatics experts on the O2 Net team worked closely with clinical experts, such as Jennings, to build a template for the EHR that would encourage physicians to collect more data around opioid use and overdose: For instance, was the overdose intentional or accidental? Such detailed data is necessary for the clinical research needed to find new approaches to opioid overdose.
Finally, the O2 Net team wanted to build a physician-decision tool that would improve outcomes for opioid users presenting with indications of an overdose to the ED. Unlike typical decision tools, such as pop-up best-practice alerts, which can interfere with the clinical workflow, this tool is designed to be used after the patient visit has ended and the physician is finalizing documentation.
“While it’s super important to remember to send someone home with naloxone, it’s probably not important enough to let that thought interrupt what you’re doing to save his or her life,” said Lenert, explaining why the O2 Net team chose the timing it did for its physician-decision tool.
The physician can either choose to disregard the treatment guidance, such as the importance of providing take-home naloxone, or answer a set of questions to have the notation automatically added to the patient’s record and the prescription for take-home naloxone automatically ordered.
The rollout of this tool at MUSC dramatically improved the prescription rate of take-home naloxone for ED patients who overdosed on opioids. That rate jumped from 16% without the tool to 66% after its implementation at MUSC. Going forward, Lenert would like to see whether this novel physician-decision tool could have a similar impact in other diseases.
If O2 Net is successful, it would have ramifications beyond the opioid epidemic. It would show that the ACT network could be rapidly adapted to address pressing health concerns as they arise.
“We’re always going to have new priorities,” said Lenert. “What we need is a flexible architecture to extend a strong foundation like ACT so that we’re ready for the next problem. We could be working on monkeypox, on social determinants of health or whatever the challenges are. What we need are modular ways to expand our work rapidly from a strong foundation instead of inviting delays by trying to build something from scratch.”