Polysubstance Use: A Broader Understanding of Substance Use During the Opioid Crisis

Abstract

Objectives. To understand important changes in co-occurring opioid and nonopioid drug use (i.e., polysubstance use) within the opioid epidemic in the United States.

Methods. We analyzed survey data on the past month co-use of prescription and illicit opioids and 12 nonopioid psychoactive drug classes from a national sample of 15 741 persons entering treatment of opioid use disorder.

Results. Past-month illicit opioid use increased from 44.8% in 2011 to 70.1% in 2018, while the use of prescription opioids alone dropped from 55.2% to 29.9%, yet overall remained high (94.5% to 85.2%). Past-month use of at least 1 nonopioid drug occurred in nearly all participants (> 90%), with significant increases in methamphetamine (+85%) and decreases across nonopioid prescription drug classes (range: −40% to –68%).

Conclusions. Viewing opioid trends in a “silo” ignores the fact not only that polysubstance use is ubiquitous among those with opioid use disorder but also that significant changes in polysubstance use should be monitored alongside opioid trends.

Public Health Implications. Treatment, prevention, and policymaking must address not only the supply and demand of a singular drug class but also the global nature of substance use overall.

The substantial public health impact of the opioid epidemic in the United States is well documented. According to the Centers for Disease Control and Prevention (CDC), 47 600 individuals died from an opioid overdose in 2017,¹ with countless others revived by the timely use of the opioid antagonist naloxone (Narcan). The origins of the current epidemic can be traced back to the proliferation of opioid prescriptions to manage chronic pain in the 1990s and the subsequent misuse of these drugs for nontherapeutic purposes, resulting in dramatic increases in treatment admissions for opioid use disorders (OUDs) as well as overdose fatalities, particularly in opioid-naïve individuals.²

As these outcomes reached epidemic proportions, a number of mitigating steps were taken to deter abuse and diversion, such as prescription drug monitoring programs,^3,4 abuse-deterrent opioid formulations and associated legislation,^5,6 new prescribing guidelines,⁷ pill-mill legislation, and increased physician awareness of the appropriate use of opioids.⁸ However, as these supply-side interventions began to realize some success in reducing availability of prescription opioids,^9,10 a considerable proportion of those with OUD substituted or shifted entirely to using other opioids, primarily heroin,^11,12 as that market has expanded. More recently, the increased use of heroin as an initiating opioid,¹³ along with the proliferation of the highly potent synthetic opioid, fentanyl—either sought out specifically or unknowingly mixed with heroin—has led to a dramatic increase in overdose fatalities.^14,15

The continued growth in the opioid epidemic has had the effect of creating a silo, so to speak, in that researchers and policymakers, particularly when it comes to treatment, increasingly are focused almost exclusively on OUD (e.g., medication-assisted treatment that targets only opioids) and its nuanced components (e.g., prevention of doctor shopping through prescription drug monitoring programs), rather than taking a more global view of substance use disorders per se. This opioid-centric perspective ignores a substantial, though often overlooked, empirical body of research that has observed that the majority of those with a substance use disorder are polysubstance users.^16–19

In a recent study that used data from the National Survey on Drug Use and Health, those with a prescription OUD had greater prevalence of other substance use disorders than misusers or general users of prescription opioids.²⁰ While many may prefer a specific drug or drug class, use of multiple substances is commonplace, making treatment or policy directed at a single drug less effective than those focusing on substance use overall. Understanding polysubstance use applies not only in the context of the opioid epidemic itself (e.g., use of both prescription and illicit opioids), but also in consideration of co-occurring use of opioids with a wide array of other substances (e.g., benzodiazepines and stimulants) that have the potential to increase risk for adverse events, relapse following a treatment regimen, or overdose fatalities.^21–23 For example, recent overdose mortality data provide evidence of a relationship between of methamphetamine and opioid use.^24,25

For opioid treatment–related policy and programs to be effective in the long term, polysubstance use among opioid-addicted persons needs to be better assessed and understood, particularly over time. In an effort to fill this void, we utilized data from a long-running national opioid surveillance system on treatment-seeking opioid users to evaluate (1) temporal trends in opioid drug use, (2) temporal trends in nonopioid drug use overall and as a function of opioid drug use, and (3) the proliferation of polysubstance use among persons with an OUD.

METHODS

The Survey of Key Informants’ Patients (SKIP) Program database is composed of individuals who have entered treatment of an OUD at any one of the participating treatment centers from across 49 states and Washington, DC, and has been validated against other opioid and substance surveillance systems and shown to be nationally representative.^10,26 Surveys are administered at intake to individuals presenting with an OUD at participating treatment centers and who are new admissions to sites (all of whom were not previously enrolled, to minimize repeated participation). This analysis included data from 270 treatment centers (42.6% private, 36.3% public, 16.7% private and public, 4.4% unspecified) recruited by using the Substance Abuse and Mental Health Services Administration Behavioral Health Treatment Services Locator. The SKIP program has collected information regarding co-occurring (i.e., past-month) nonopioid drug use since the second half of 2011. In the event that a participant did not complete the nonopioid drug section, we pursued imputation adjustments to help distinguish between participants who had skipped the section versus those whose nonresponse could reasonably imply nonuse (i.e., “did not use [drug x] in the past month”). To be included in this study, participants must have had to (1) complete the survey between the second half of 2011 and the first half of 2018 and (2) provide information about their past-month nonopioid drug use. The resulting sample included 15 741 participants.

Opioid Drug Use

Participants were given a list of opioids wherein they were instructed to select all of the opioids that they had used “to get high” in the past month. The list consists of 14 prescription opioid classes and only 2 illicit opioids—heroin and nonprescription fentanyl. To convey the absolute prevalence of prescription and illicit opioids in our sample, we stratified the 15 741 participants into 2 nonexclusive groups (i.e., any prescription opioid and heroin/nonprescription fentanyl; Figure 1a). However, for the sake of future analyses, we created more discrete groupings, stratifying the 15 741 participants into 3 exclusive groups (i.e., prescription opioids only, heroin/nonprescription fentanyl only, or prescription opioids and heroin/nonprescription fentanyl; Figure 1b). Comparative statistics include the absolute percent change and Cochrane–Armitage tests to evaluate significance.

FIGURE 1—

Percentage of Survey of Key Informants’ Patients (SKIP) Sample Who Indicated Past-Month Use “to Get High” of (a) Any Prescription Opioid and Heroin or Nonprescription Fentanyl, and (b) Prescription Opioids Exclusively, Prescription and Illicit Opioids, or Illicit Opioids Exclusively: United States, Second Half of 2011 to First Half of 2018

Note. q = quarter; Rx = prescription.

^aDoes not include nonprescription fentanyl.

Polysubstance Use

We defined polysubstance (i.e., nonopioid drug) use for our SKIP sample as the co-occurring, nonmedical use of any of the following drug classes: nicotine, marijuana, excessive alcohol use defined as having more than 4 drinks in a single day, antidepressants, anxiolytics, muscle relaxants, prescription sleep medications, prescription stimulants, crystal meth, crack or cocaine, hallucinogens, and MDMA. Participants were instructed to indicate which of these drugs they had used for “recreational use, to get high, or for any other non-medical reason” in the past month; with the exception of prescription stimulants, which did not appear on the survey until the first half of 2015.

To demonstrate temporal differences in drug use patterns among study participants, we compared prevalence rates from the first time point with the last time point (Figures 2 and 3). Figure 2 conveys these prevalence data as a bar chart. To illustrate temporal variation across the analysis period, capping each column is the juxtaposition of each drug’s interpolating trend line. Figure 3 illustrates the absolute percent change in prevalence of use for each drug.

FIGURE 2—

The Prevalence Rates of Past-Month Use of Nonopioid Drugs Among the Survey of Key Informants’ Patients Sample in the First Analysis Period (Second Half of 2011) and the Final Analysis Period (First Half of 2018): United States

Note. q = quarter; Rx = prescription. Trend lines indicate the prevalence rates across the entire study period.

^aData first made available in 2015q1,2.

FIGURE 3—

The Percent Change in Past-Month Use of Nonopioid Drugs Among the Survey of Key Informants’ Patients Sample From the Initial Analysis Period (Second Half of 2011) to the Final Analysis Period (First Half of 2018): United States

Note. Rx = prescription.

^aData first made available in 2015q1,2.

Next, we sought to combine data from Figures 1, 2, and 3 to explore group-specific trends in polysubstance use among the 3 exclusive opioid subgroups (i.e., prescription opioids only, heroin/nonprescription fentanyl only, or prescription and heroin/nonprescription fentanyl). Specifically, we wanted to assess (1) interclass variation (i.e., between opioid subgroups) over time by examining whether opioid subgroup had any association with the total number of nonopioid drug classes used in the past month and (2) intraclass variation (i.e., within opioid subgroups) over time by examining whether there were any changes in the number of nonopioid drugs used in the past month over time. To show these nuances, we created a line (Figure 4a) and bubble chart (Figure 4b) showing temporal variation in polysubstance use both between and within opioid subgroups. Because each bubble represents the proportion of individuals at each time point (located on the x-axis) who had used the respective number of different nonopioid drugs (located on the y-axis), the bubble chart allows an apples-to-apples comparison of nonopioid drug use both within groups (as a function of time) and also between groups (as a function of opioid subgroup and time). We have also added group-specific trend lines to convey the (arithmetic) mean number of nonopioid drugs used, depicting standard error with the shaded gray region (Figure 4a).

FIGURE 4—

Past-Month Opioid Use Showing the Relative Frequency of Nonopioid Polysubstance Use Among Each Opioid Subgroup at Each Time Point by (a) Line Chart and (b) Bubble Chart: United States, 2011–2018

Note. Rx = prescription. Past-month opioid use included prescription poioids exclusively, prescription and illicit opioids, or illicit opioids exclusively. In part a, the trend line represents the mean number of unique, nonopioid drugs used at each time point, with standard error shown in shaded region. In part b, the size of the bubble corresponds to the relative proportion during each time period.

RESULTS

The SKIP (n = 15 741) sample was primarily White (79.1%), with a mean age of 33.5 years (SD ±10.3) and fairly split between males (53.9%) and females (46.1%), and in terms of urbanicity (49.3% urban residents). The majority had no health care coverage (42.0%) followed by Medicare or Medicaid (35.6%), and the types of treatment included inpatient (60.0%), outpatient (41.4%), counseling (33.3%), and medication-assisted treatment (8.5%).

Temporal Changes in Opioid Drug Use

Figure 1a shows that from 2011 to 2018, the past-month use of heroin/nonprescription fentanyl increased from 44.8% in the second half of 2011 to 70.1% in the first half of 2018, an increase of 57%. It is worth noting that the increase in heroin/nonprescription fentanyl was not mirrored by a comparable decrease in the past month use of prescription opioids; on the contrary, this group maintained a high prevalence in our sample, decreasing roughly 10%, from 94.5% in 2011 to 85.2% in 2018. Figure 1b breaks past-month opioid use down further into 3 mutually exclusive groups: use of prescription opioids only, use of heroin/nonprescription fentanyl only, or the indiscriminant use of prescription opioids and heroin/nonprescription fentanyl. We observed a marked decline in the prevalence of individuals who used prescription opioids only in the past month, from 55.2% to 29.9% (–46% change from 2011). Conversely, the prevalence of heroin/nonprescription fentanyl–only use increased from 5.5% to 14.8% (+169% change from 2011), and the indiscriminant use of prescription opioids and heroin/nonprescription fentanyl increased from 39.2% to 55.4% (+41% change from 2011). All changes over time were significant at a P level of less than .001.

Temporal Changes in Nonopioid Drug Use

Figure 2 shows the prevalence of past-month use for each nonopioid drug surveyed at the initial and final analysis periods, including past-month use of at least 1 nonopioid drug. To add context, a trend line caps each bar dyad, depicting the standardized interpolations in prevalence throughout the entire analysis period. To summarize these data, Figure 3 shows the percent change from the first time point to the last time point. Past month use of at least 1 nonopioid drug remained substantially high, with 95.8% meeting this criterion in 2011 and 96.4% in 2018. Past-month use of nicotine, alcohol, and marijuana were common among our sample and saw little change over the analysis period, as did the use of crack or cocaine. With the exception of prescription stimulants, whose relative percent change perhaps belies its nonmonotonic growth pattern (Figure 2), the co-occurring use of all other nonopioid drugs changed demonstrably over time, including a significant increase in prevalence of methamphetamine use (+85%) and significant decreases in anxiolytics (–40%), antidepressants (–46%), prescription sleep medications (–68%), hallucinogens (–44%), MDMA (–57%), and muscle relaxants (–61%). All changes were significant at a P level of less than .001 except for marijuana (P = .18), nicotine (P = .12), crack or cocaine (P = .46), and alcohol (P = .07).

Polysubstance Use as a Function of Time and Opioid Type

Figures 4a and 4b present data that allow us to compare changes in polysubstance use over time both within and between opioid subgroups. While the trend line is helpful in showing the centrality of the data, the size and scarcity of the bubbles help to show the spread of the data. Overall, individuals classified as indiscriminant users of prescription and illicit opioids also had the highest overall mean number of unique, nonopioid drugs used in the past month (4.3; 95% confidence interval [CI] = 4.27, 4.33), followed by users of prescription opioids only (3.5; 95% CI = 3.47, 3.53) and then users of heroin/nonprescription fentanyl only (2.6; 95% CI = 2.56, 2.64). This order notwithstanding, over time the mean significantly decreased in the prescription-only group, from 3.7 in 2011 to 3.0 in 2018, and the prescription–heroin/nonprescription fentanyl group, from 4.3 in 2011 to 3.7 in 2018 (both at P < .001), while the mean significantly increased for the heroin/nonprescription fentanyl–only group, from 2.3 in 2011 to 2.5 in 2018 (P = .007). It should be noted that, because prescription stimulants did not enter the survey until the second half of 2015, the maximum number of different nonopioid drug categories before this time was 11.

DISCUSSION

In agreement with other studies,^1,20 we found that, among treatment-seeking opioid users, although past month use of heroin/nonprescription fentanyl has dramatically increased, the use of prescription opioids nonetheless continues to be a significant part of the opioid epidemic, with a prevalence rate higher than that of heroin/nonprescription fentanyl, though decreasing slightly (–10%) over our analysis period. It seems clear from our data that prescription opioids are no longer primarily used exclusively by those with an OUD, but are now most often used in tandem with illicit opioids. This may suggest that while supply-reduction efforts (e.g., prescription drug monitoring programs, prescribing guidelines, pill-mill legislation) may have had their intended effect of decreasing overall physician prescriptions for opioids,⁹ they may have had the unintended effect of pushing those with an OUD to more potent and inherently riskier drugs, particularly heroin and fentanyl. In particular, this shift has been described in the literature as a function of not only reduced supply of prescription opioids but also practical factors such as price and availability of heroin and fentanyl.^11,27,28 As seen with previous supply-reduction efforts (e.g., the Prohibition Era), our data confirm the construct that as long as there is a demand for a product, there will remain strong efforts to provide a supply, in one form or another, to meet demand.

The present study also reinforces the conclusion that polysubstance use is common in those with an OUD; indeed, our results indicate that polysubstance abuse is the norm, not the exception, with nearly the entire sample endorsing at least 1 nonopioid drug used in the past month. In fact, those who used both prescription opioids and heroin/nonprescription fentanyl, who made up the largest proportion of our sample in 2018, used, on average, 4 other nonopioid drug classes while also using a variety of opioids. What is also striking is that there appear to have been dramatic shifts in nonopioid drug use over time. Specifically, of all the nonopioid drugs we tracked, only 1 had increased significantly from 2011 to 2018—methamphetamine. While increased production, distribution, and access to methamphetamine has led to its increased use across the country in general,²⁹ it has been suggested that the co-use of opioids and methamphetamine may establish an equilibrium between the stimulation produced by methamphetamine and the sedation produced by opioids so as to be able to function as normally as possible, or the use of both drugs produces a “roller coaster ride” of a high with 2 entirely different pleasurable sensations.³⁰ These relationships may also help explain why those who used heroin/nonprescription fentanyl exclusively endorsed fewer nonopioid drugs on average: the more potent the drug, the less need there is to supplement use with a multitude of other substances.

Initially we hypothesized that, as the supply of prescription opioids continued to shrink, many other nonopioid psychoactive drugs—such as benzodiazepines and other sedative–hypnotics—would increase among advanced opioid users, much like what was observed with methamphetamine. Given that these drugs share sedative-like properties with opioids, it seemed reasonable to posit that they would be useful to those with an OUD as either a substitute for an opioid in short supply or as a means of suppressing opioid-withdrawal symptoms.

However, we found no evidence to suggest that this hypothesis has merit. In fact, changes in anxiolytics (e.g., benzodiazepines) represent the most surprising shift as more than half of our sample had used them in the past month in 2011, which steadily dwindled to less than one third in 2018. The reason for this unexpected difference is not completely understood, but it is possible that the implementation of supply-side interventions to limit the diversion and nonmedical use of prescription opioids, as noted previously, has had a spillover effect that has led to reduced diversion and nonmedical use of other prescription drugs, resulting in the observed decreases we observed for all categories of prescription drug types analyzed here, as well as reductions in the mean number of nonopioid substances used by those who had exclusively used prescription opioids.

The implications for these data are important for substance use treatment and policies, which must take into account a more global understanding of substance use, lest the intense focus on opioids lead to mistaking the forest for the trees. Policies aimed at expanding mental health care need to take into account that the links between mental health issues (e.g., depression or anxiety, trauma, life stressors) and substance use (i.e., the Khantzian notion of self-treatment^31,32) are likely applicable across substances. More importantly, while providing medication-assisted treatment is an important component of treating OUDs, the reliance on this alone fails to treat both mental health antecedents and the use of nonopioid substances, which appear to be ubiquitous among those entering treatment of an OUD. Understanding these components and including them as part of a comprehensive treatment regimen is vital in improving treatment success and preventing the high rates of relapse that are common among substance users.

There are important limitations in our study. Most significantly, ours is a treatment-based sample with survey language (i.e., “use to get high”) that may not be representative of those who use opioids “recreationally,” nonmedically, or do not meet criteria for an OUD. Furthermore, differences in the factors influencing the decision to enter treatment, such as family or court pressures and financial ability, could also limit the heterogeneity of our sample. However, these data make a compelling argument that the treatment of OUDs should include the treatment of substance use disorders on a broader level.