Beyond self‐reports

Ways to obtain more comprehensive insights into substance use events

Bibliographic Data

ID21597071
AuthorsEmmanuel Kuntsche (0000-0001-8931-0790, Centre for Alcohol Policy Research La Trobe University Melbourne Australia, corresponding author), Cassandra Wright (0000-0001-9751-4005, Centre for Alcohol Policy Research La Trobe University Melbourne Australia), Johannes Thrul (0000-0001-8929-9579, Centre for Alcohol Policy Research La Trobe University Melbourne Australia)
Year2021
Volume40
Issue7
Pages1108-1111
Publication date2021-11-01
Peer ReviewedYes
Open AccessYes
TypeARTICLE
VenueDrug and Alcohol Review (JOURNAL)
Journal identifiersISSN: 0959-5236 • E-ISSN: 1465-3362
PublisherWiley (PUBLISHER • GB)
DOI10.1111/dar.13398
PMID34761835
OpenAlexW3213180997
LanguageEN
References cited25

For more than a century, most information on personal substance use have been collected by asking participants to report their behaviour retrospectively in questionnaires [1]. While cost-efficient and convenient, this approach has major disadvantages. For example, in alcohol research, survey estimates reproduce only 30–70% of the total per capita alcohol consumption compared with sales data [2, 3]. Among several biases (e.g. social desirability, sampling) that may be at play, this mismatch between alcohol survey and sales data is in part attributable to memory deficits, as people tend to remember only part of the amounts consumed [3-6]. Recording behaviours in or close to real time or matching objective data with discrete alcohol or other substance use events may produce more precise estimations of substance use [7]. Fortunately, the development of sensor technology (e.g. using breath, perspiration or wastewater samples) holds promise to overcome the limitations of retrospective recall. Moreover, although they are not novel data sources, death registries and other administrative data can also be used as an alternative method to self-report. However, the possibilities with all of these methods come with challenges related to analysing and interpreting enormous amounts of data and linking these data to the real-life experiences and (substance use) behaviour of individuals in meaningful ways. In a field in which the use of sensor and registry data is on the rise, this Special Section showcases what is possible today. It includes data obtained from different kinds of transdermal alcohol monitors [8-12] and breathalysers [11, 13, 14], in wastewater analyses [15, 16], from death registries [17-19], identification (ID) scanners [20] or pictures from participants' smartphones [21]. The 14 studies included span eight countries, comprising Australia, the Czech Republic, Lithuania, New Zealand, Russia, Switzerland, the UK and the USA. While transdermal alcohol monitors have been available for more than a decade, their performance and application in research settings remain understudied [22]. The articles included in the Special Section make an important contribution in this area of research by revealing both advantages and challenges of using transdermal monitors. The finding that transdermal alcohol concentration measures correspond only roughly to alcohol consumption measured with breathalysers, as reported by Norman et al. [11], or with self-reports, as demonstrated by Croff et al. [9], highlights some of these problems. In these studies, limitations of comparison measures are identified as potentially contributing to the mismatch. For instance, breathalysers usually obtain single-point measures and thus may miss important peaks in consumption that transdermal sensors can measure [11], while cognitive biases may make retrospective weekly self-reports an unsuitable match for event-level alcohol measurement [23]. Van Egmond et al. [10] demonstrated limitations to transdermal sensors, with only moderate correspondence evident between sensors used to measure alcohol concentration in the same individual. They found that individual factors such as sex [10] or temporally changing factors such as stomach content, movement (e.g. walking) and changes in ambient conditions (e.g. moving from inside to outside) [8] were unrelated to intraindividual differences in transdermal alcohol readings. These differences instead appear to be due to skin thickness and implementation factors (tightness of monitor strapping [8, 10] or potential contamination of devices). The latter suggests that transdermal devices may be challenging to use in naturalistic settings where these factors are more difficult to control. Croff et al. identified that transdermal sensors appear to be insensitive to low levels of alcohol consumption [9]. This means that low-level drinking episodes often remain undetected, particularly among participants with high body mass index [10], which makes transdermal sensors problematic for measuring drinking frequency [9]. In contrast, transdermal alcohol sensors appear to perform well when measuring consumption patterns in individuals who drink heavily [12] or heavy drinking events, for example events occurring at music festivals [11]. This is particularly important because increasing intoxication levels and alcohol-related consequences such as vomiting or blackouts make other data collection approaches, including self-reports and breathalysers, unsuitable or even impossible. While the articles included in this Special Section all report evidence obtained from the more ‘traditional’ and relatively bulky transdermal monitors, that is SCRAM-CAM [8-12] and WrisTAS [9], sleek new-generation devices, such as the BACtrack Skyn and the Milo ION, are becoming increasingly available and offer exciting possibilities, particularly in terms of user-friendliness and performance [24, 25]. While these devices are still relatively expensive, they may become more affordable in the future, which is important to facilitate their use in research studies with large samples. While it is unclear whether cost considerations are the only reason for this, it is nonetheless striking that the number of participants included in the transdermal alcohol device studies described in the Special Section is relatively small, that is ranging from 2 [8] to 59 [9] participants. In contrast to these relatively small samples in the included transdermal alcohol sensor studies, the paper of Stumbrys and Tamutiene [13] demonstrates that it is possible to use sensor devices to collect alcohol use data at scale. Their study includes almost 2.5 million breath tests collected over the course of 1 year (i.e. in 2018) in 153 workplaces in Lithuania. This enormous sample size allowed for detection of rare events, in the case of this study whether employees came to or were at work while under the influence of alcohol. Fortunately, detected quantities of blood alcohol concentration (BAC) were relatively low for the majority of cases with alcohol-positive test results, which may indicate heavy drinking the day before or drinking of small amounts of alcohol just before starting work [13]. Monds et al. [14] demonstrate the necessity of objective assessment (such as with breathalysers) when both the individual who is drinking and an observer fail to identify higher levels of alcohol intoxication. This may be due to individual tolerance or a ceiling effect (i.e. signs of intoxication such as blurred speech and staggering occur at a relatively low level of intoxication and do not increase linearly with increasing BAC levels). The two wastewater analysis studies included in this Special Section [15, 16] showcase the potential of this technology for detecting or confirming regional and temporal patterns of substance use. Trowsdale et al. [15] found that per-capita nicotine and alcohol use in New Zealand was higher in rural than in urban areas, with community deprivation linked to nicotine but not alcohol use. While these results from New Zealand are consistent with and can be used to corroborate survey data and sales statistics, the contribution of Rozhanets et al. [16] from Russia demonstrates that wastewater analysis can also provide information in addition to surveys and sales or prescription data. In the case of this study, it was the detection and distribution of use of illegal substances such as morphine, as well as temporal and regional variations in their use. Wastewater analysis can also help to identify otherwise unrecorded or unidentified consumption such as of illegally produced or imported alcohol [16]. Three studies included in the Special Section demonstrate the usefulness of analysing death registry data [17, 19, 20] to provide insights into substance use or related outcomes as an alternative to self-report or in circumstances where self-report is not possible. Bailey et al. [17] analysed 1722 decedents in England who died by deliberate self-injury and found a higher likelihood of cocaine use prior to suspected suicide with increasing levels of BAC. These results emphasise the need to include substance use and co-use in suicide prevention approaches, which is often not done today [17]. A relatively simple, yet efficient and powerful method of obtaining longitudinal data without any response burden or cognitive distortion due to self-reporting is to link existing client information with routinely collected health care records, which is particularly useful among hard-to-reach populations. Linking data on release from residential substance use treatment in Queensland, Australia, with deaths registered at one-year follow-up, Tisdale et al. [18] provide evidence for the need of continuing care during the high-risk period directly after release from residential treatment. The contribution of Petruželka et al. [19] highlights the importance of careful attention to data collection approaches and quality when using registry data. Their study analysed drug mortality data and found structural breaks in time series, which are most likely due to improvements in coding or reporting practices. The authors argue that reporting of drug-induced mortality may have largely underestimated the extent of drug problems in the Czech Republic prior to these changes in coding and reporting practices [19]. Two other studies included in this Special Section [20, 21] used a novel type of objective data to investigate substance use-related circumstances on weekend nights. De Andrade et al. [20] analysed an impressive number of ID scans – more than 14 million – at licensed venues in Queensland, Australia, collected over almost 2 years. This used a novel dataset which provided useful information about patterns of venue attendance and demographic characteristics of nightlife patrons including banned patrons attempting entry, who were mostly male and between 18 and 24 years old. The authors highlight the value of the networked ID scanner system which may be used to prevent entries of individuals with court or police mandated bans to potentially reduce alcohol-related violence in nightlife venues. In an attempt to obtain objective information and to reduce the high response burden commonly occurring in event-level studies that rely on participant self-reports, Labhart et al. [21] asked adolescents and young adults in Switzerland to take close-up pictures with their smartphone of each drink they consumed on weekend evenings. Unfortunately, the results showed that it was often challenging for researchers to extract and annotate information from these pictures such as beverage type, container size and filling level to assess the number of standard drinks being consumed. The authors conclude that, to date, this method cannot replace more standard methods of collecting information on drinks consumed in event-level studies. The contributions to this Special Section demonstrate, first, the usefulness of sensor, wastewater and registry data in providing ‘objective’ evidence and confirming what can be gathered from self-reports. Examples are regional and temporal patterns of substance use based on wastewater analysis [15] or patterns of night-time attendance in premises based on ID scans [20]. Second, sensor, wastewater and registry data can be a powerful source to provide additional information to complement self-reports. For example, wastewater analysis can help identify otherwise unrecorded or unidentified consumption, including illegally produced or imported alcohol [16], while smartphone pictures can provide additional contextual information on drink containers (shape, size, environment, etc.) and thus on drink size [21]. Third, sensor and registry data can generate entirely new insights and evidence when obtaining self-reports is very difficult or even impossible. Examples include death registries to examine the co-occurrence of alcohol and cocaine in suicide decedents [17] or continuous transdermal alcohol monitoring during a multiday music festival [11]. However, the use of sensor and registry data comes with several challenges. For instance, low sensitivity and considerable intraindividual differences in transdermal alcohol readings raise questions about the reliability of the devices and the way they measure consumed alcohol [8-10]. Changes in coding or reporting practice over time (or in different locations) can bias registry data and, for example, lead to an underestimation of drug problems [19]. While many of the biases of self-report methods are known and can be taken into account in study design and analyses, at this point we know considerably less about the biases in novel data sources such as from sensors and wastewater and biases within registries can be very specific to each dataset, making them difficult to account for. Being mindful about their limitations, conducting more research on their biases, or combining these data sources with other or more established methods (where biases are better understood) is an important way to enhance our understanding of the strengths and weaknesses of each, so that we can better apply them to the contexts that they are best suited to. The use of the most appropriate comparison measure is of course important. For instance, in event-level studies, ecological momentary assessment data are more suitable a match to sensor data than retrospective recall. As our knowledge on substance use measurement strengthens, it is likely that there are some circumstances in which self-report will remain our strongest method [21], while other methods, including the emerging methods used in papers within our Special Section, may be established as preferential. Given the potentially powerful (complementary) evidence that sensor technology, wastewater, and registry data can provide, but also considering current challenges and limitations of these methods documented in the papers included in this Special Section, we look forward to seeing future studies in this area. We are optimistic that, by combining sensor technology, wastewater, and registry data with self-reports we will gain a more comprehensive understanding of substance use behaviours, event-level contextual factors and relevant harm outcomes. The authors would like to thank Professor Paul Dietze and Professor Robin Room for their helpful comments to improve this work. None to declare

Alcohol · Alcohol consumption · Cognitive psychology · Data science · Environmental health · Internet privacy · Per capita · Psychiatry · Recall · Social science · Sociology · Statistics · Substance use · Survey data collection · Advanced Chemical Sensor Technologies · Computer Science · Data Stream Mining Techniques · Forensic Toxicology and Drug Analysis · Mathematics · Medicine · Psychology

Tools

Open DOIOpen Access