Illicit drug use is a pressing public health issue. And detecting cocaine use, or any type of drug for that matter, typically relies on either inaccurate self-reporting or collecting urine samples—none of which provide precise timing of cocaine use.
But under the auspices of the Dartmouth-based Northeast Node of the Clinical Trials Network (CTN), which is part of the National Institute on Drug Abuse (NIDA), a new project is underway to detect cocaine use via a smart watch. Lisa Marsch, PhD, the Andrew G. Wallace Professor of Psychiatry and director of the Center for Technology and Behavioral Health at Geisel School of Medicine, and Santosh Kumar, PhD, a professor and the Lillian & Morrie Moss Chair of Excellence in Computer Science at the University of Memphis, are leading an endeavor to develop and test a smart watch with a suite of sensors to detect inter-heartbeat data. Their study is funded by NIDA.
Kumar, who heads up a 12-institution consortium, Mobile Data to Knowledge, or MD2K, together with Emre Ertin, PhD, a research associate professor in Electrical and Computer Engineering at The Ohio State University, and Kenzie Preston, PhD, a senior investigator in NIDA’s Intramural Research Program, developed a few years ago a method that showed it is possible to detect cocaine use from its response on heart rate—captured with a wearable sensor called AutoSense.
“The benefit of that initial step was the creation of a computational model that was critical in distinguishing cocaine use and how it differs from other things that might stimulate inter-rate heart intervals,” Marsch explains.
Kumar and his team refined and tested the data model at different stages with both cocaine users and non-cocaine users resulting in promising data.
“Sensor-based method of cocaine detection nicely complements the existing methods of urine assessment and self-report by pinpointing the timing of cocaine use,” Kumar says. “NIDA-CTN is very interested in whether this method of can be applied in clinical trials using heart rate data collected by a conveniently wearable sensor such as a smart watch. They wanted a CTN Node with the capacity to do this so they talked to Lisa, who is very enthusiastic about the project, and we are quite happy she has taken on the responsibility.”
Marsch and Kumar are well acquainted and have collaborated on past projects—because of their natural synergy they have long been interested in further collaboration.
“The charge to Santosh and I, along with our team, is to take what he learned and refine it so the only point of data capture is on the smart watch itself,” Marsch says.
This depends on a smart watch that can collect good quality intra-beat data from the wrist. “Available devices don’t have the battery capacity to collect this high-frequency data for an entire day, and the data they do collect does not have the right granularity to support continuous detection,” Ertin says.
In order to meet that need, Ertin is building a smart watch sensor with a long battery life and a sensor array to collect inter-beat data continuously, which will be sent to a smart phone.
Joining Marsch, Kumar, and Ertin in the study is August Holtyn, PhD, a research associate in Psychiatry and Behavioral Sciences at Johns Hopkins University who will test the watch with active cocaine users in her Baltimore, MD research clinic. Preston and her colleagues at NIDA are also partnering on the project.
“This is very exciting even though we are in the early stage of development,” Marsch says. “We are hopeful we will be able to accurately capture the heart rate data on these watches in a way that will allow us to refine the algorithm in the computational model to enable reliable detection of cocaine use—this is a two-year project, but based on what we learn, a larger clinical trial may follow.”
Though Marsch acknowledges there are other important research questions around getting active cocaine users to wear a smart watch, for now the question is whether or not they can achieve a level of sophistication with the technology to precisely capture cocaine use data in real time.
“You could think of a model where if we learn enough about people’s pattern of cocaine use, it is possible to begin developing a predictive model of risk—which could trigger a mobile-based intervention to help the user,” Marsch says. “It could lead to the linkage of measurement to intervention delivery over time and this could also inform similar approaches to using sophisticated sensing technologies to measure other substance abuse as well.”