Street drug monitoring with networked spectrometers powered by machine learning: a pilot study in Ontario, Canada.

Drug-checking services (DCS) provide people who use drugs (PWUD) with crucial information concerning the substances they may consume. The intent of DCS is to reduce harms potentially associated with those substances, by providing evidence-based information about the contents of said substances and possibly influencing PWUD consumption-related behaviors. This pilot project reports on a network of 10 dedicated Raman spectroscopy drug-checking devices, located at various organizations throughout the province of Ontario, Canada. The spectrometers were specifically developed for drug-checking analysis and use machine learning (ML) enabled software to provide participants with automated, real-time results. The same software was also used to collect participants' demographic data and self-reported consumption-related behavior changes. In this work, we report on the results provided by this network over a 14-months period from July 2023 to August 2024 on 7752 samples provided by 5083 participants. On select samples, high-performance liquid chromatography-mass spectrometry (HPLC-MS) measurements were also collected to probe the accuracy of the results and quantify variations between the two techniques. Voluntary feedback was also collected from a limited number of participants concerning the potential impact the drug-checking process may have on their consumption-related behaviors.