Moving away from piecemeal trends: The need for multiple data sources in drug use trend analyses

Trend analyses are essential to examining shifts in drug use and related morbidity and mortality across time to inform prevention, treatment and harm reduction efforts. However, trend analyses published in academic journals are typically limited to data from a single source or type of source. For example, in numerous manuscripts, authors estimate trends in use of a particular drug based on self-report, and in others, authors estimate trends in deaths related to use of that drug, but few manuscripts present trends in both use and deaths. This is understandable given that many researchers do not specialise in or have access to multiple types of data. There is nothing inherently wrong with presenting trends in cocaine use and trends in cocaine-related deaths in separate manuscripts, but when trends based on different sources diverge, this can lead to confusion among researchers, policymakers, public health experts and the public alike. For example, with recent increases in opioid-related overdoses and deaths in the USA [1], some people may incorrectly deduce that nonmedical use or misuse of prescription opioids is increasing, but it has actually decreased in recent years [2,3]. This divergence suggests severity of opioid use and/or exposure to more potent drugs, such as fentanyl, may be increasing, but not necessarily overall prevalence of opioid (mis)use. This should remind us that changes in death or hospitalisation rates do not always mirror changes in prevalence of use.

In recent years, more trend analyses have begun to focus on data from multiple sources. Specifically, trends have been compared in the same manuscripts for use and/or use disorder related to cocaine, opioids and alcohol, and deaths involving these drugs [4–8], and recent papers have examined trends in cocaine and opioid seizures in relation to hospitalisations and deaths related to use of these drugs [9–12]. Such manuscripts allow us to directly compare how rates of use and associated outcomes derived from various sources are predictive, concordant or discordant.

Most manuscripts that do present trend estimates from various sources focus on two data sources. However, in this issue of Drug and Alcohol Review, Man et al. [13] present a manuscript in which they estimate trends in cocaine use in Australia between 2003 and 2019 using many sources of indicator data. Specifically, they utilise data from: (i) triennial household surveys; (ii) annual surveys of sentinel samples of people who use stimulants; (iii) seizures; (iv) arrests; (v) hospitalisations; (vi) treatment episodes; and (vii) deaths. Utilising so many data sources appears to present a complete picture of cocaine use trends in Australia, over and above any analysis examining a single source of data.

Considering multiple data sources is important for both prospective and retrospective trend analyses. National and international drug monitoring organisations, such as the National Drug Early Warning System in the USA [14] and the European Monitoring Centre for Drugs and Drug Addiction [15], examine drug use trends using multiple indicators prospectively, and such monitoring can allow us to detect shifts in drug use and associated effects in almost real time. However, more attention is needed to focus on trends using multiple sources retrospectively, as these trends can help us acquire a deeper understanding of how trends emerge and diverge. This can also help us determine signals of outbreaks, by retrospectively determining predictors of shifts, to inform current surveillance efforts.

Each data source type has its own strengths and limitations, but utilising multiple sources of data helps present a complete picture of drug trends. More authors should consider examining multiple sources of data to better inform public health efforts.