University of Maryland doctoral student Kenan Zamore ‘26, has won a grant from the American Public Health Association’s Injury and Violence Prevention Data Science Demonstration Project to use AI to rapidly identify overdose deaths. Opioids remain the most lethal drug in America, and yet a lag in toxicology surveillance data means it’s difficult for health care systems to identify overdose deaths and respond quickly.
“Most of the risk comes from fentanyl and fentanyl analogs, which have essentially replaced heroin on the street. So when there is an overdose, or a cluster of overdoses, we are able to send out messaging and resources fairly rapidly, to prevent further harm and connect people to life-saving resources and treatment,” says Zamore, who studies at UMD’s School of Public Health’s Behavioral and Community Health program.
“But right now we cannot tell with certainty that a fatality is drug-related without a toxicology report, which can take up to 3 months.”
Washington, D.C., currently reports fatal overdoses with this 90-day lag, says Zamore, making it impossible to detect overdose anomalies, such as spikes in overdose deaths, quickly enough to prevent more deaths. Zamore and his research colleagues, including advisor Dr. Amelia Arria, hope to tackle this by using machine learning to create a fast and reliable way to track opioid deaths, solving for the lag time inherent in relying on toxicology reports.
Drawing on several sources including 911 calls, Fire and EMS records and forensic scene investigation data, Zamore and his team will identify deaths by overdose and then create a rapid case definition for an opioid overdose. Modelling based on this will allow for early warning signals.
Zamore, who also works as an epidemiologist at DC Health, uses data to analyze trends in injuries and disease, including a recent project tracking gun violence in the capital. He hopes that this project will provide the timely data public health officials need in order to respond to fatal opioid overdoses and save lives.