Enhanced informatics tool could help better identify nonfatal opioid overdose cases in the ED

A team of researchers at the Medical University of South Carolina (MUSC) has been awarded $3.75 million by the National Center for Advancing Translational Sciences to enhance an existing informatics tool known as "informatics for integrating biology and the bedside" or I2B2. The MUSC team will enhance I2B2 so that it can better identify Emergency Department cases of nonfatal opioid overdose.

The enhanced tool will use a form of artificial intelligence known as natural language processing, which helps to make human language intelligible to computers. With that added capability, the tool will be able to look for clues suggesting opioid overdose in ED physicians' clinical notes in the electronic health record.

In addition, the enhanced tool should help researchers understand what constellation of clinical traits, discoverable in the electronic health record, reliably predicts for opioid overdose.

It will also provide real-time information on opioid addiction. That will help researchers design more intelligent clinical trials and improve patient recruitment into those trials.

This project can help us leverage our existing data assets to design more intelligent trials addressing the major issues of the opioid epidemic."

Leslie Lenert, M.D., the principal investigator for the award

Lenert is assistant provost for data science and informatics and chief research information officer at MUSC. He is also associate principal investigator and informatics director of the South Carolina Clinical & Translational Research (SCTR) Institute. SCTR is the Clinical & Translational Science Awards Program (CTSA) hub headquartered at MUSC.

The SCTR team comprises biomedical informatics experts, addiction science specialists, ED physicians and biostatisticians/methodologists. Biomedical informatics experts include Lenert, who leads the team, SCTR informatics co-director Jihad S. Obeid, M.D., and Vivienne J. Zhu, M.D., who heads up efforts around natural language processing. The addiction sciences experts are MUSC vice president for research and SCTR director Kathleen T. Brady, M.D., Ph.D., and Jenna L. McCauley, Ph.D. MUSC Health ED physician Lindsey Jennings, M.D., and methodologists Annie Simpson, Ph.D., Ralph Ward, Ph.D., and Jillian Harvey, Ph.D., round out the team.

The team will test how well a prototype of the enhanced tool identifies nonfatal opioid overdose cases in the ED at MUSC and three collaborating CTSA hubs that include the University of California San Diego, Dartmouth University and the University of Kentucky.

Like MUSC, the collaborating hubs belong to the Accrual to Clinical Trials (ACT) network. The ACT network is a large consortia of CTSA hubs that have agreed to share electronic health record data using I2B2 to improve recruitment to clinical trials.

Why a better tool is needed

Researchers need up-to-date information on the opioid epidemic because it is constantly evolving. Once driven primarily by prescription opioids, the epidemic has now shifted more to fentanyl and other synthetic opioids. If successful and widely shared, the enhanced informatics tool could provide real-time data on how the epidemic is changing. It could also enable researchers to learn more about the situation in their region or at their institution so that they can design clinical trials appropriately.

"Right now, if I wanted to do a clinical trial where I needed to recruit patients who had had nonfatal overdose in the ED, the best, most up-to-date data would be about a year old," said McCauley. "Developing a tool like this allows not just researchers, but from the surveillance perspective, it allows clinicians to stay on top of trends."

Jennings, who is a clinician, agrees. "From a public health perspective, the tool could give us real-time information on what is happening in our community, allowing us to tailor population health interventions," she said.

Currently, the data on opioid overdose in the electronic health record, which is based on discrete diagnosis codes, can be inaccurate or incomplete.

"When the presenting condition is something else, such as a respiratory disorder, the OD might not be recognized or coded properly," said McCauley. "Sometimes when patients are delivered to the hospital, it's unclear what substance they've overdosed on or whether they've overdosed or are just in a comatose state."

Even when coded data in electronic health records do identify cases of opioid overdose accurately, they are not very informative on what treatment plans were recommended and whether the patients followed through.

"We don't always have good data on what the treatment plans are and where people are and how successful they were in getting into those treatment plans," said Lenert. "We don't know what the handoff was between the hospital and the treatment facility."

What the enhanced tool will do

The tool being enhanced by the SCTR team will look for clues in a physician's narrative clinical notes that a case could have been a nonfatal opioid overdose. For instance, if the emergency medical responders administered Naloxone (Narcan®), that would not be in the hospital's coded data, but the physician might have mentioned it in the clinical notes.

"It is often difficult to retrospectively identify the chart of a patient who presented to the ED with an opioid overdose, because the diagnosis is not coded in a discrete, easily searchable field," said Jennings. "Having a way to locate these charts that relies on how ED providers naturally document will improve research efforts, ultimately improving patient care."

The team will also create smart forms for electronic health records. These forms will encourage physicians to enter more information about patients with opioid overdose and their treatment plans.

By the end of the five-year project, Lenert hopes the participating sites will use the tool to begin planning clinical trials.

With the tool, researchers will be able to enroll patients at any of the participating sites in their trials. They will also know which trials make the most sense for their own patients.

"I think we're going to have a number of clinical trials that are launched based on the findings that we glean from these databases," said Lenert. "That's really our plan."

Future plans

If the prototype is successful, the MUSC team will share the technology with the rest of the ACT Network. The number of members who adopt it will be one sign of the project's success.

"We will be interested in how many of the ACT sites we can get to adopt this expanded set of functionalities and join the network to work on opioid overdose," said Lenert.

If accepted widely, the tool could have nationwide implications for improving our understanding of the opioid epidemic and which treatments most effectively target it.


The opinions expressed here are the views of the writer and do not necessarily reflect the views and opinions of News Medical.
You might also like... ×
Machine learning-based prediction model may enable early diagnosis of opioid use disorder