Exploring Potential Bellwethers for Drug-Related Mortality in the General Population: A Case for Sentinel Surveillance of Trends in Drug Use among Nightclub/Festival Attendees

Abstract

Background:

Drug-related deaths in the US continue to increase. Sentinel surveillance of high-risk populations can provide early warning for shifts in trends. Nightclub/festival attendees have high levels of drug use, so we explored whether use among this population can serve as a potential bellwether or indicator for use-related mortality in the general population.

Methods:

Trends in past-year cocaine and methamphetamine use were estimated from nightclub/festival attendees in New York City (NYC) and among NY residents, and trends were estimated for related death rates in NYC (2014/15-2019/20). Using national data from England and Wales (2010–2019), trends in past-year cocaine and ecstasy use (among the full population and among nightclub attendees) and related deaths were also estimated.

Results:

In NY/NYC, cocaine use remained stable in the general population, but use among nightclub/festival attendees and cocaine-related deaths doubled. Methamphetamine use among nightclub/festival attendees and death rates also more than doubled while use among the general population remained stable. In UK countries, increases in cocaine and ecstasy use were larger for infrequent/frequent nightclub attendees than in the general population, with 3.6- and 8-fold increases in related deaths, respectively. In UK countries, the association between nightclub attendance and death rates increased in a dose-response-like manner with larger associations detected when death rates were lagged by one year.

Conclusions:

Patterns of use among nightclub/festival attendees, more so than patterns in the general population, were similar to patterns of drug-related deaths. Use among this subpopulation could possibly serve as a bellwether for use-related outcomes. Continued surveillance is recommended.

Introduction

Now in its fourth wave of the opioid crisis, the United States (US) is seeing increases in deaths related to use of both opioids and psychostimulants such as methamphetamine (Ciccarone, 2021). For example, deaths related to synthetic opioids, primarily fentanyl and its analogs, have increased from 9,580 in 2015 to 56,516 in 2020, and deaths related to use of psychostimulants grew from 5,716 to 23,837 during the same period (Hedegaard et al., 2021). Provisional counts from 2022 suggest that the number for both drug classes has climbed further still (Ahmad et al., 2022). Continued increases in mortality, however, are not inevitable and likely preventable (Keyes & Cerdá, 2022). Given that characteristics of use, people who use, and the type of drugs co-used with synthetic opioids continue to shift (Palamar et al., 2022a), ongoing surveillance of drug use is essential to limit mortality and inform prevention, treatment, and harm reduction efforts.

National surveys—which sometimes are able to focus on specific cities or states—are often viewed as the gold standard for estimating patterns and trends in drug use, assessing the consequences of use, and identifying those at risk for use and those in need of treatment (Substance Abuse and Mental Health Services Administration [SAMHSA], 2020). However, such surveys have limited efficacy, particularly with respect to rarer or more stigmatized drugs. For instance, despite large increases in heroin-related deaths in the US beginning in 2010 (National Center for Health Statistics, 2020), prevalence estimates of past-year heroin use among national samples remained extremely low. Specifically, one study detected an increase in past-year use from 0.17% in 2002 to 0.32% in 2018 (Han et al., 2020), while a national study of high school seniors found that past-year use appeared to decrease from 2010 through 2019 (from 0.9% to 0.3%) (Miech et al., 2020). As such, national surveys are being criticized for severely underestimating use (Reuter et al., 2021). This problem may be largely attributable to underrepresentation of the at-risk populations most likely to use these drugs, ultimately diminishing the utility of population surveys for identifying and predicting shifts in use or related outcomes in a timely manner on their own. By the same token, it is possible that focusing on known high-risk populations may yield data that better reflect shifts in use and outcomes.

Sentinel surveillance can be a helpful means of quickly identifying outbreaks and early signs of change. By focusing on select populations that are thought to better signal shifting trends (World Health Organization, 2020), sentinel surveillance can detect changes more readily than national surveys or mortality data, which typically lag for many months post-collection (Ahmad et al., 2022; SAMHSA, 2020; Spencer & Ahmad, 2017). Additionally, utilization of multiple types of data sources (e.g., survey data and mortality data) can be advantageous. National indicators of drug use and related effects are often examined in a piecemeal manner, and patterns of use or related outcomes from one data source may not be generalizable to other populations of interest (Palamar, 2021). Indeed, it would be rather favorable if shifting patterns discovered in one data source could aid public health experts in predicting changes to come in other populations or the general population. To this end, some subpopulations who are at high risk for drug use may help serve as a bellwether for use or related deaths in the general population.

This analysis aims to determine whether a population at high-risk for psychostimulant use—nightclub/festival attendees—can serve as a bellwether for drug use-related trends in the general population, over and above population survey-based statistics. Here, “bellwether” is defined as an indicator or predictor of change in prevalence. High levels of drug use, especially of party drugs, have been established among nightclub and festival attendees, particularly those who frequent electronic dance music (EDM) parties (Black et al., 2020; Griffin et al., 2019; Grigg et al., 2018; Kelly et al., 2013; Ramo et al., 2011). A recent study of EDM party attendees in New York City (NYC) estimated that psychostimulant use among young adults ages 18–25 is highly prevalent, with 42.8% having used ever ecstasy, 26.0% having ever used cocaine, and 8.0% having ever used methamphetamine (Palamar et al., 2017). Moreover, the quantity and popularity of EDM and parties that feature this type of music have increased notably in recent years (Watson, 2018), and drugs like ecstasy have been shown to diffuse from EDM party-attendees to the general population (Palamar, 2020).

Methods

Procedure

Data were compiled to allow for the estimation and comparison of trends with respect to past-year drug use and drug-related deaths. Data sources were limited to New York (NY)/NYC and England and Wales (United Kingdom [UK]), which, to our knowledge, are the only areas where repeated cross-sectional data sources are available, allowing for examination of drug use among nightclub attendees in relation to local or national drug use data indicators. Cocaine and methamphetamine were the drugs of focus in NY and cocaine and ecstasy were the drugs of focus in UK countries based on data availability. For example, while most data sources in the US and UK provide data for cocaine, ecstasy-related mortality data are not available from the US and methamphetamine-related mortality data are not available from most UK sources. Analyses focused on NY were limited to 2015/16–2019/20 due to data availability, and analyses of UK data sources focused on data from 2010 to 2019, with 2019 being the most recent year of data being available.

NY/NYC

National Survey on Drug Use and Health

The National Survey on Drug Use and Health (NSDUH) is an annual nationally representative survey of noninstitutionalized individuals ages ≥12 in the United States (US) (SAMHSA, 2020). Samples each year are obtained through a multistage design and represent all 50 states and the District of Columbia. Surveys are administered via computer-assisted interviewing which are conducted by an interviewer using audio-computer-assisted interviewing. Participants are asked about past-year use of cocaine and, as of 2015, about past-year methamphetamine use in a section about illicit drug use (as opposed to previous years in which use was queried along with legal stimulants) (SAMHSA, 2015). Published state-level estimates for the state of NY were obtained for analysis (SAMHSA, 2021). NSDUH only provides state-level estimates of paired years with 2019/20 estimates being the most recent data available. As such, 2014/15, 2015/16, 2016/17, 2017/18, 2018/19, and 2019/20 estimates were used to represent 2015, 2016, 2017, 2018, 2019, and 2020, respectively. While NYC-level estimates are available, estimates are for groups of three years with some overlap (e.g., 2016–2018, 2018–2020) (Substance Abuse & Mental Health Data Archive, 2022). Estimates between NY and NYC in 2016–2018 were comparable, with larger differences detected between the Northeast region and the US overall (Supplemental Table 1). 2014/15 data on methamphetamine use are not available as NSDUH did not query methamphetamine use (in its own section separate from prescription drug misuse) until 2015 (Substance Abuse Mental Health Services Administration, 2015). Estimates were examined for the full population ages ≥12 and also for those ages 18–25.

Nightclub and festival attendees

Each summer, from 2016 to 2019, adults ages 18–40 about to enter randomly selected EDM parties and venues in NYC were surveyed using time-space sampling (Palamar & Keyes, 2020). Surveys were administered outside of 125 parties (37 in 2016, 39 in 2017, 24 in 2018, and 25 in 2019) including 16 days of recruitment outside of festivals. The aggregate sample size was 3,571 (1,084 in 2016, 954 in 2017, 1,029 in 2018, and 504 in 2019). Participants were asked about past-year use of cocaine and methamphetamine. Weights were calculated (based on self-reported level of party attendance and response rates) in order to make results more generalizable to the overall EDM party-attending population in NYC.

Deaths

The counts and rates of drug-related deaths in NY were derived from the US Centers for Disease Control Wide-ranging Online Data for Epidemiologic Research (CDC WONDER) database. Drug poisoning (“overdose”) deaths were defined as having an International Classification of Diseases 10^th Revision (ICD-10) cause-of-death code for drug-related unintentional death (X40–44). This analysis focuses on data available on deaths related to cocaine (T40.5) and other psychostimulants with abuse potential (T43.6) (NCHS, 2020). While data specifically on methamphetamine being involved in recorded deaths are not available, approximately 85–90% of death certificates involving psychostimulants other than cocaine mention methamphetamine (Han et al., 2021). Given that co-use of synthetic opioids (other than methadone), such as fentanyl, is now commonly involved in cocaine- and psychostimulant-related deaths in the US (Mattson et al., 2021), age-adjusted rates of deaths related to use of cocaine and other psychostimulants (per 100,000 persons) were tabulated for deaths: 1) regardless of synthetic opioid involvement, 2) involving synthetic opioid use (T40.4), and 3) not involving synthetic opioid use. Mortality data indicating cocaine-related deaths (both involving and not involving synthetic opioids) and other psychostimulants (independent of synthetic opioid use) were available for 2016–2020, but data for psychostimulants with and without synthetic opioid use were not available for 2015 as these numbers were suppressed.

UK countries

Crime Survey for England and Wales

The Crime Survey for England and Wales (CSEW), formerly known as the British Crime Survey, is a victimization survey of residents in households in England and Wales (Home Office, 2021). Structured interviews are conducted using computer-assisted personal interviewing, and self-completion modules are traditionally completed by those ages 16–59. These participants are asked about past-year use of cocaine and ecstasy, as well as frequency of attendance of nightclubs in the past month (with attendance coded by CSEW as attending 1–3 times in the past month and attending ≥4 times in the past month). Surveys are conducted from April through March the following year to represent one year of data collection, with prevalence estimates overall and by level of nightclub attendance published annually. For the purposes of comparability with other data sources, reported estimates were tabulated by calendar year in which the majority of data were collected (e.g., April 2019 through March 2020 was coded as 2019).

Deaths

Data on deaths related to drug-related poisonings or drug misuse are compiled each year for England and Wales (Office for National Statistics, 2020). Coding for cause of death is carried out via ICD-9 and ICD-10 criteria. Counts of cocaine- and ecstasy-related deaths and age-standardized rates (per 1,000,000 persons) are published each year and these estimates were tabulated for analysis.

Analysis

Data were examined to compare years 2015/16-2019/20 from all NY/NYC data sources and 2010–2019 for all UK data sources. For case-level data (i.e., NYC EDM party attendee data), past-year prevalence of use was estimated using available weights in order to generate population-level estimates (Heeringa et al., 2010). To estimate trends in use, logistic regression was used to estimate odds of use as a linear function of time as a continuous predictor (Ingram et al., 2018) using Stata SE 17 (StataCorp, 2021). For all other data (derived from published reports or publicly available aggregate count data), Joinpoint Regression version 4.8.0.1 (National Cancer Institute, 2020) was used to estimate trends. Also known as piecewise, broken line, multi-phase, or segmented regression, Joinpoint fits weighted least-square regression models to counts or rates on a log transformed scale (Ingram et al., 2018). Estimates (fitted values) from the best fitting Joinpoint models of trends in use and deaths, respectively, were plotted side by side for comparison.

To examine associations between prevalence of use and death rates, we further tested associations first using Spearman correlation (nonparametric) and linear regression. For both correlation and regression, we examined paired years, then re-examined associations with death rates lagged by one year (to determine predictive associations). We calculated standardized betas in regressions to allow for comparison between models. Results for the UK sample are formally presented, while results for the NYC sample were underpowered and hence presented in a supplemental table. This secondary data analysis was exempt from review by the New York University Langone Medical Center institutional review board.

Results

In NY/NYC, between 2015 and 2020, estimates of cocaine use among those ages ≥12 in NY remained relatively stable at 2.4–2.8% (p=.286), as did estimated use among those ages 18–25 (ranging from 5.2% to 6.9%; p=.766) (Figure 1). However, estimated use among nightclub and festival attendees in NYC doubled between 2016 and 2019, with estimated use increasing from 17.3% to 35.2% (a 103.5% increase; p<.001). Rates of cocaine-related deaths in general in NYC increased 191.2% between 2015 and 2020 (from 3.4 to 9.9 per 100,000 population; p=.014) and cocaine-related deaths involving opioids increased 1,650.0% (from 0.4 to 7.0 per 100,000 population; p=.016). Cocaine-related deaths not involving opioids, however, remained relatively stable between 2.8 and 4.0 per 100,000 population (p=.576).

Figure 1.

Trends in estimated past-year use of cocaine and related deaths in NY/NYC

With respect to methamphetamine use in NY/NYC, between 2016 and 2020, estimated use remained stable in NY at 0.2–0.3% among those ages ≥12 (p=.087) and those ages 18–25 (p=.552) (Figure 2). However, among nightclub and festival attendees in NYC, between 2016 and 2019, estimated use increased from 1.5% to 5.2%, a 242.1% increase (p=.027). Between 2015 and 2020, rates of deaths related to psychostimulant use (which were assumed to be primarily methamphetamine) increased by 220.0% (from 0.5 to 1.6 per 100,000 population; p=.002), and rates of psychostimulant-related deaths also involving opioid use increased by 412.3% (from 0.2 to 0.9 per 100,000 population between 2016 and 2020; p=.018). Rates of deaths not involving opioid use, however, did not significantly increase, remaining relatively stable at 0.4 to 0.7 per 100,000 population (p=.228).

Figure 2.

Trends in estimated past-year use of methamphetamine and related deaths in NY/NYC

Demographic characteristics of participants in the nightclub sample and of decedents in NYC are presented in Supplemental Table 2. We did not directly compare characteristics of people who use to decedents using statistical tests, but within decedents, the percentage identified as Black who used cocaine was 9.8-times higher than the percentage in the nightclub sample who used cocaine, and the percentage of descendants identified as Black who used other psychostimulants was 25.1-times higher than percentage in the nightclub sample reporting methamphetamine use. Further, while no participants in the nightclub sample were age ≥45, 62.3% of deaths related to cocaine use were in this age group and 33.9% of deaths related to other psychostimulant use were in this age group.

With respect to England and Wales, among the general population, estimated past-year use of cocaine increased from 2.1% to 2.6% between 2010 and 2019—a 23.7% increase (p=.004) while use increased from 1.1% to 1.6% (a 45.2% increase, p=.019) among non-nightclub attendees (Figure 3). However, increases were larger among nightclub attendees. Specifically, use increased from 6.5% to 10.5% (a 60.5% increase; p=.001) among those attending nightclubs 1–3 times in the past month, and use increased from 9.7% to 19.1% (a 95.8% increase; p<.001) among those attending nightclubs ≥4 times in the past month. The rate of cocaine-related deaths increased 360.7% between 2010 and 2019—from 2.6 to 11.9 per 1,000,000 population (p<.001).

Figure 3.

Trends in estimated past-year use of cocaine and related deaths in England and Wales

Estimated past-year ecstasy use in England and Wales remained stable among the general population overall at 1.2–1.7% (p=.390) and among non-nightclub attendees at 0.5–0.8% (p=.387); however, increases occurred among nightclub attendees (Figure 4). Specifically, use increased from 4.9% to 7.7% (a 57.2% increase; p=.003) among those attending nightclubs 1–3 times in the past month, and use increased from 8.9% to 14.4% (a 61.8% increase; p=.003) among those attending nightclubs ≥4 times in the past month. The rate of ecstasy-related deaths increased from 2010 to 2019 by 805% from 0.1 to 1.3 per 1,000,000, and two slopes were detected in which a significant increase occurred between 2010 and 2013 (p=.001) and then a non-significant increase from 2013 through 2019 (p=.086).

Figure 4.

Trends in estimated past-year use of ecstasy and related deaths in England and Wales

Finally, with respect to associations between level of nightclub attendance and mortality trends, in the UK sample (Table 1), correlation coefficients and standardized betas tended to be larger and more significant as level of nightclub attendance got higher, and coefficients in the lagged analyses tended to be slightly higher than in models examining paired years. Specifically, with respect to lagged death rates related to cocaine use, the association was smallest for non-attendees (β = 0.73, p=.021), and larger among infrequent (β = 0.97, p<.001) and frequent attendees (β = 1.00, p<.001). A somewhat similar increase was detected with respect to ecstasy: no attendance (β = 0.50, p=.154), infrequent attendance (β = 0.79, p=.018), and frequent attendance (β = 0.78, p=.010). We present these correlations for the NYC sample in Supplemental Table 3 with caution as these tests were under-powered. We also present correlations between prevalence of use between the nightclub-attending population and the general population for the UK in Supplemental Table 4 and for NYC in Supplemental Table 5.

Table 1.

Associations between level of nightclub attendance and rates of death related to use of cocaine and ecstasy use in the United Kingdom

	Cocaine-Related Death Rate		Lagged Cocaine-Related Death Rate
	Correlation	Regression	Correlation	Regression
Cocaine Use
General population	r = 0.85, p<.001	β = 0.83 (SE = 0.20), p=.003	r = 0.82, p<.001	β = 0.89 (SE = 0.16), p<.001
Non-attendees	r = 0.66, p<.001	β = 0.69 (SE = 0.26), p=.028	r = 0.70, p<.001	β = 0.73 (SE = 0.02), p=.021
Infrequent attendees	r = 0.94, p<.001	β = 0.94 (SE = 0.12), p<.001	r = 0.85, p<.001	β = 0.97 (SE = 0.17), p<.001
Frequent attendees	r = 0.92, p<.001	β = 0.95 (SE = 0.11), p<.001	r = 0.91, p<.001	β = 1.00 (SE = 0.12), p<.001
Ecstasy use
General population	r = 0.20, p=.583	β = 0.33 (SE = 0.33), p=.350	r = 0.70, p=.036	β = 0.56 (SE = 0.29), p=.101
Non-attendees	r = 0.23, p=.519	β = 0.37 (SE = 0.33), p=.295	r = 0.45, p=.224	β = 0.50 (SE = 0.31), p=.154
Infrequent attendees	r = 0.75, p=.013	β = 0.77 (SE = 0.23), p=.010	r = 0.87, p=.002	β = 0.79 (SE = 0.26), p=.018
Frequent attendees	r = 0.72, p=.018	β = 0.83 (SE = 0.20), p=.003	r = 0.79, p=.012	β = 0.78 (SE = 0.22), p=.010

Note. SE = standard error.

Discussion

Using estimates from various sources, the present study explores whether psychostimulant use among nightclub and festival attendees can possibly serve as a bellwether for drug use and related outcomes among the general population. Most of the analyses focused on cocaine due to availability of data. While the prevalence of cocaine use among the general NY population showed no significant changes, and only small increases in use were noted among the general population of England and Wales, estimated use among nightclub attendees significantly increased. In fact, estimated use more than doubled in NYC among nightclub attendees, while UK data revealed not only increases in use but a dose-response-like association whereby more frequent nightclub attendance was associated with greater increases in estimated prevalence. In both locales, cocaine-related death rates increased considerably. Associations between prevalence of use and mortality rates tended to be larger among nightclub attendees, and were often slightly higher when examining lagged death rates, suggesting that shifts in prevalence indeed appear to be somewhat predictive. Together, these results suggest that changes in estimated use among nightclub and festival attendees more closely reflected overall death rates than estimated prevalence of use among the general population.

The analysis of methamphetamine was based solely upon NY/NYC data. While methamphetamine use among the general population remained stable, estimated use more than tripled among nightclub and festival attendees. Increases in psychostimulant-related deaths (which were assumed to be mainly methamphetamine) also increased in NYC. Deaths related to use of cocaine and psychostimulants increased in NYC in general and among cases involving use of synthetic stimulants, but deaths not involving use of synthetic stimulants did not increase, which may limit the generalizability of findings. Indeed, nonmedical opioid use is prevalent among NYC EDM party attendees (Palamar et al., 2018); regardless, these analyses reaffirm that patterns of use among this population better align with patterns in mortality.

While death rates involving use of cocaine or other psychostimulants but not opioids did not significantly increase in NYC, it is important to note that co-use of such stimulants with opioids before death is becoming more common. In the US, age-adjusted rates of death involving synthetic opioids with psychostimulants and cocaine increased from 0.1 to 1.8 per 100,000 and from 0.1 to 3.2 per 100,000, respectively, between 2013 and 2019 (Mattson et al., 2021). Further, between 2013 and 2019, deaths involving use of psychostimulants and cocaine in absence of opioid use have also increased—from 1.1 to 3.2 per 100,000 and 1.5 to 1.7 per 100,00, respectively (Mattson et al., 2021). The increase in deaths related to co-use of opioids and stimulants in particular has become so substantial that it has been proposed that we are now in a “fourth wave” of the opioid crisis (Ciccarone, 2021). In this new wave, fentanyl co-used with stimulants now appears to be a major driver of drug-related deaths in the US. While the lack of increase in stimulant-related deaths in NYC may not be generalizable to other areas, deaths involving these drugs still increased—overall and among those co-using opioids—even if the opioids tended to be the primary driver of most deaths. Stimulants are now commonly used with opioids in attempt to prevent or reverse opioid effects or to alleviate withdrawal (Daniulaityte et al., 2022; Silverstein et al., 2021), so it is important to continue to focus on the involvement of stimulants when examining these deaths.

Ecstasy data in this analysis were limited to UK countries as ecstasy-related mortality data are not available in the US. While ecstasy use among the general population remained stable, prevalence of use increased among nightclub attendees, and in a dose-response-like manner as was the case with the aforementioned cocaine use. Rates of deaths involving ecstasy increased over 800% during the same period. As such, it appears once again that patterns in use among this high-risk population more closely reflect overall patterns in mortality.

It is worth noting that we do not necessarily believe that current deaths related to use are solely related to use among nightclubs attendees. Indeed, notable differences in age and race/ethnicity between the sentinel population and the decedents exist, and homelessness—a risk factor for overdose-related death (Fine et al., 2022)—is also underrepresented in these data. For clarification, then, we suggest that this sentinel population—regardless of the extent of overlap with the population of decedents—is a potential indicator or bellwether for such deaths, while also acknolwedging that there likely are other factors underlying the decedents. For example, it has been suggested that sentinal populations have played a large part in diffusing drugs such as ecstasy and even crack cocaine to the general population (Golub & Johnson, 1996; Hamid, 1992; Palamar, 2020), which may be contributing to the pool of decedents.

Although more research is needed, we primarily hypothesize that the link between prevalence of use among nightclub attendees and psychostimulant drug-related deaths among the general population may be explained by overlapping drug markets or distribution channels, with markets indicating drug availability. For instance, patterns in drug use are strongly linked to drug availability (including drug offers) (Lipari et al., 2017), and such availability is associated with lower disapproval toward use and increased willingness or intention to use (Palamar et al., 2012, 2018) in a manner that may be driving increases in use among nightclub attendees as well as increases in use-related mortality in the general population, which typically lag. In this way, we may actually be indirectly measuring drug availability in overlapping markets that affect both the general population and nightclub attendees.

Ultimately, our results largely suggest that drug use—or at least use of psychostimulants such as cocaine, methamphetamine, and ecstasy—among nightclub/festival attendees may in fact serve as a bellwether or indicator for shifts in drug use-related outcomes in the general population. This is important as population surveys likely include only a small number of such attendees—likely too few to signal changes within this high-risk subpopulation. Further, population surveys often underestimate use among important subpopulations due to underrepresentation of said populations. This can adversely impact the accuracy of estimates, particularly when use of a drug is not very prevalent (Reuter et al., 2021). Investigating high-risk subpopulations, such as nightclub/festival attendees, may also help provide information regarding polydrug use, new methods of use, and unknown exposure to drugs as adulterants or contaminants as trends may emerge in such populations before the general population. It is known that such trends can indeed diffuse into the general population (Palamar, 2020). Furthermore, trends in such subpopulations may provide added insight into new trends due to prevalent or highly concentrated use. For example, prevalence of detection of unreported exposure to new psychoactive substances in this population, such as synthetic cathinones (“bath salts”) or fentanyl and its analogs (Palamar et al., 2017), can serve as a warning for users among the general population who can be at risk for using adulterated or contaminated drugs such as ecstasy or cocaine. In fact, in NYC, the local health department has recently piloted a naloxone distribution program in nightlife venues given that recreational cocaine use is often concentrated in such venues and that cocaine can now be contaminated with fentanyl (Allen et al., 2020). It is currently unknown whether the potential nightclub bellwether effect is dependent upon a market in which there is a high level of drug adulteration or contamination so continued research is needed.

Widespread adulteration creates problems for surveillance when self-report is not accompanied with toxicology testing. While any testing of drugs and/or biospecimens would be a helpful addition to survey research, more advanced (and thus more expensive) testing is most accurate in testing for the presence of new analogs (Palamar et al., 2020). More research and surveillance using multiple modes of data collection (e.g., self-report plus toxicology testing) can provide more insight regarding what drugs people were exposed to and whether exposure was intentional or unintentional.

While the present study focused on nightclub and festival attendees, it should be noted that other subpopulations may also be worthy bellwethers. For example, psychonauts—people with extensive experience exploring psychoactive effects of various drugs—can shed light regarding trends in purity and related effects not only of new drugs but also of common drugs like cocaine (Assi et al., 2022; Deluca et al., 2012). Further, shifting drug use patterns (e.g., injection, fentanyl exposure), drug-related emergencies or deaths, or need for treatment among homeless populations (Hotton et al., 2021; Jones et al., 2020; Yoganathan et al., 2021) may serve as an indicator of current or future increases in use or related outcomes in the general population. Perhaps researchers can benefit from surveilling trends in these high-risk groups as such trends may occur in these populations first. In addition, this analysis only focused on psychostimulants, which are prevalent in nightclub scenes. We do not expect this population to serve as a possible bellwether for other types of drugs, but more research is certainly needed.

Limitations

National data on nightclub attendance are not available in the US, so analyses focused on NYC-level data as an indicator of drug use and use-related outcomes in NY and NYC. We relied on state-level population estimates for NY because NYC estimates were grouped by years of three, but we found that estimates were comparable between the state- and city-level. We relied on data that were available, so some data sources have fewer years of available data than others. The 2020 NSDUH also changed its methodology due to COVID-19 and this may effect comparability of estimates between years (Center for Behavioral Health Statistics & Quality, 2021). The England and Wales national survey did include questions about past-month nightclub attendance, but it is unknown to what extent such clubs featured EDM music. It should further be noted that the England and Wales survey did not query festival attendance, and that estimates regarding nightclub attendance were indeed derived from this national survey, not a separate survey. With regard to published data used in analyses, the lack of case-level data did not allow for more nuanced (e.g., subgroup) analysis. Also, use and deaths may have shifted over time in a heterogeneous manner. For example, in the US, drug-related deaths among older, racial minority, and urban populations in particular have been increasing (Lippold et al., 2019). Geographical variation could have also occurred. The size and characteristics of the population of nightclub attendees can also shift over time. Such differential risk and subgroup shifts may limit generalizability. Further, while it is unknown to what extent nightclub attendees are represented in national surveys, prevalence of drug use within such targeted samples is indeed much higher than in national surveys (Palamar et al., 2015, 2022b) which makes this an ideal sentinel population. Polydrug involvement was also common in deaths, and both reported use and deaths are largely limited to what drug or drugs were thought to be used. Drugs are commonly adulterated or replaced with other drugs, so unless data included toxicology testing, then such indicators are vulnerable to measurement error. Further, Joinpoint was used to examine aggregate data extracted from reports and Joinpoint often lacks efficacy to detect non-linear (i.e., quadradic or cubic) trends. Finally, deaths related to psychostimulant use had to be used as an indicator of methamphetamine-related deaths in this analysis. However, other studies have also had to rely on psychostimulant use as an indicator (Han et al., 2021; McKetin et al., 2021).

Conclusion

It appears possible that sentinel surveillance of nightclub and festival attendees may facilitate early detection of important trends in drug use and their adverse outcomes. National-level survey data may not capture true levels of drug use nor accurately reflect changes in prevalence of use-related outcomes, such as deaths. This by no means suggests that national survey data are not valuable. Rather, our findings indicate that other forms of data from various populations should also be considered when possible. Ongoing surveillance can not only inform prevention and harm reduction efforts in subpopulations of interest, but possibly in the general population as well. More research, however, is needed to examine such relations in a more nuanced manner. For example, subgroup analysis and closer examination to determine whether attendance can serve as a predictor more than an indicator of use and related outcomes is needed. Future studies could also combine data sources rather than examine them separately and determine the feasibility of using nightclub attendee data for purposes of triangulation. Regardless, this study provides initial evidence that drug use among this high-risk population might serve as a bellwether for drug use-related outcomes among the general population.