Abstract
Objectives. To characterize present-day polysubstance use patterns in the general adult population.
Methods. From a 2022 nationally representative survey in the United States, we defined polysubstance use as last 12-month use of 2 or more drugs (n = 15 800). Latent class analyses included medical (as indicated) and nonmedical (not as directed) use of prescription opioids, stimulants, benzodiazepines, and antidepressants; recreational use of cannabis, psilocybin or mushrooms, other psychedelics, cocaine, methamphetamine, and illicit opioids; and concomitant use with alcohol, cannabis, prescriptions, or recreational drugs.
Results. The national prevalence of polysubstance use was 20.9% (95% confidence interval = 20.5%, 21.3%), broken down into the following 4 latent classes: (1) medically guided polysubstance use (11.5% prevalence, 6.1% substance use disorder [SUD]): prescribed drug use, some cannabis, and no concomitant use; (2) principal cannabis use variety (4.0% prevalence, 31.9% SUD): high probability of cannabis use with various drugs concomitantly used; (3) self-guided polysubstance use (3.4% prevalence, 14.5% SUD): nonmedical use of prescriptions and concomitant use; and (4) indiscriminate coexposures (2.1% prevalence, 58.9% SUD): concomitant drug use with indiscriminate drug preference.
Conclusions. Different polysubstance profiles show adults with untreated SUDs, and there are 2 previously unrecognized classes. Prevention and treatment strategies addressing polysubstance use should take a personalized perspective and tailor to individuals’ use profile. (Am J Public Health. 2025;115(5):747–757. https://doi.org/10.2105/AJPH.2024.307979)
Polysubstance use of psychoactive drugs is a complex behavioral pattern described as use of multiple substances over one’s lifetime or over the same timeframe, with inclusion of both prescription and recreational drugs.1,2 Polysubstance use is a driver of morbidity and mortality in the United States. It is associated with increased risk of substance use disorders (SUDs).3 In 2022, an estimated 40.3 million people experienced an SUD while only 6.5% received treatment.4 Polysubstance use is also associated with unmet physical and mental health care needs,5 poor SUD treatment retention or relapse,2 increased risky behaviors (i.e., injection drug use or binge drinking5), and increased risk for overdose and death.6 Opioid- and psychostimulant-related mortality has received the most attention given its contribution to the rise in overdose deaths.7 However, an extensive review of US overdose death certificates indicated that an average of 2.4 drugs contributed to the cause of death, with many drugs involved (e.g., sedatives, antidepressants, antipsychotics).8 Some research has signaled that the fourth wave of the overdose epidemic will be defined by death caused by co-involvement of multiple drugs.9–11
To better understand the complexity of polysubstance use, previous research has largely focused on either tabulations of common combinations4 or latent class methods. Latent class methods are useful for identifying underlying subpopulations based on observed behaviors.12 Two systematic reviews of polysubstance latent class profiles were conducted among adolescents in 201613 and among young adults in 2022.14 These studies observed 2 common latent classes, a no use or single use of drugs or a predominately alcohol use with or without tobacco use.13,14 Most studies included found only a low-prevalence polysubstance use class. de Jonge et al. concluded that discerning polysubstance use patterns was difficult when a study included highly prevalent behaviors like no use or single drug use groups, or alcohol or tobacco use; the authors recommended that future studies consider removing these behaviors to capture more nuanced polysubstance use behaviors.14 Furthermore, few studies differentiated between prescription drug classes or nonmedical use (NMU) behaviors. This study directly addressed these previous limitations.
The objective of this study was to delineate polysubstance use patterns across psychoactive drugs and explore the associated risk profiles. No nationally representative study has looked at the general adult population to quantify and describe the totality of polysubstance use behaviors across multiple psychoactive drugs. The present-day patterns need to be more fully described to understand new populations that may be amenable to clinical or public health intervention if necessary. This research categorized polysubstance use behaviors by expanding the inclusion of drug types explored and differentiating prescription and recreational drug behaviors.14 We also independently assessed profiles among those with and without likely SUDs.
METHODS
This research utilized a nationally representative survey of adults (aged ≥ 18 years) from the Survey of Non-Medical Use of Prescription Drugs (NMURx) Program, part of the Researched Abuse, Diversion and Addiction-Related Surveillance (RADARS) System. Details about this survey can be found elsewhere, including its reliability and concurrent validity of drug use estimates compared to multiple national surveys.15 Briefly, NMURx is a semiannual, repeated, cross-sectional study consisting of a self-administered, confidential, online survey that measures drug use behaviors in all 50 states and Washington, DC. Importantly, it utilizes a calibration weighting scheme that corrects for any selection and nonresponse bias across both demographic and health measures to generate representative estimates. Data from 2022 (n = 59 041) collected across the 2 waves (April 15–June 3 and September 9–October 21) was used.
The survey collected behavioral information in the last 12 months across both prescription and recreational drugs. Respondents were asked about any use of 4 prescription drug classes (opioids, antidepressants, benzodiazepines, and stimulants) followed by NMU, defined as “use not directed by a health care professional.” Other behavioral questions surrounding prescription drug use were concomitant (simultaneous) use with other drugs and drug sourcing (e.g., from own health care provider). The survey collected cannabis use, including use of marijuana, hashish, or any other preparation of the plant, but did not differentiate reason for use. Finally, questions on recreational or illicit drug use were asked as well as concomitant use with other drugs.
Measures and Definitions
We defined polysubstance use as any use of 2 or more of the following drugs in the last 12 months (n = 15 863): prescription opioids, prescription benzodiazepines, prescription stimulants, prescription antidepressants, cannabis, methamphetamine, cocaine powder, psilocybin or mushrooms, other psychedelics (including ketamine, lysergic acid diethylamide [LSD], 3,4-methylenedioxymethamphetamine [MDMA], mescaline, or phencyclidine [PCP]), or illicit opioids (including heroin, fentanyl, or fentanyl analogues). Prescription benzodiazepines and antidepressants were not commonly included in other studies, but drugs that affect the central nervous system can have abuse potential16 and therefore we included them to be more comprehensive of dispensing in the general population. The recreational drugs included the most common used in the general population; inclusion of rarer drugs likely would not facilitate further delineation of behavioral profiles. We did not include alcohol and tobacco in the definition per previous research suggestions to understand more nuanced profiles.14 We included concomitant drug use (simultaneous) as an additional behavior, defined by report of each drug listed here being used at the same time as alcohol, cannabis, recreational drugs, or other psychoactive prescriptions.
In addition, we collected information on demographics and other risk factors (e.g., drug treatment history, mental health diagnoses). The survey included the Drug Abuse Screening Test (DAST-10), a validated self-administered instrument measuring severity of problematic drug use found to discriminate well (scores ≥ 3) for patients with drug use disorders with or without diagnoses.17,18 We used this measure as a proxy for likely SUD because the instrument has several beneficial characteristics: it can discriminate SUD without diagnoses, it is agnostic to which drugs are used, it can be used to compare people using different drugs,19 and it intentionally excludes alcohol use disorder.
Statistical Analysis
We conducted a latent class analysis (LCA) of drug use behaviors exhibited in the last 12 months to identify subgroups among the population who reported polysubstance use. LCA takes a person-centered approach, examining patterns in the data and characterizing individuals into mutually exclusive groups.12 The contribution to the models of both the drug used (individual drug indicators) and situational behaviors (medical use vs NMU, concomitant use) defined patterns of polysubstance use.
We classified the LCA model indicators as follows: (1) prescription drugs were treated as a 3-level categorical variable with NMU taking priority, then medical use, then no use; (2) cannabis and recreational drugs were treated as dichotomous use versus no use; and (3) concomitant use was summarized across any reported drug (if any drugs were used with alcohol then an indicator for concomitant alcohol use was included). Sample size considerations are evolving, but studies have posed that 300 or more observations per class is desirable.20,21 Our sample size was therefore well powered to detect up to 5 latent classes.22
No a priori number of latent classes were assumed, and we determined model selection by evaluating model separation through goodness of fit tests using Bayesian Information Criterion in addition to qualitative model interpretability.12,21 We reported other model fit and diagnostics based on best practices: Akaike Information Criterion, G2 statistic, log-likelihood, smallest class size, entropy, and average posterior probabilities.21 We incorporated the individual-level design weights from NMURx into the class estimation through the pseudo-maximum likelihood approach.23 The chosen model was interpreted qualitatively based on the class probabilities and item-response probabilities.12 We calculated posterior probabilities based on Bayes’ Theorem and assigned individuals to the class with the highest probability.24 We calculated proportions of demographics and other risk factors among the latent classes. Based on individual best class assignment, we used the weighted entire NMURx data to calculate prevalence of no drug use, single drug use, and each of the polysubstance use classes.
Finally, we were interested in a stratified analysis comparing LCA profiles among those with (n = 3473) and without (n = 12 390) likely SUD based on the DAST-10 (defined by scores ≥ 3 and < 3, respectively). We calculated measurement invariance across groups using a likelihood ratio test.23 We estimated item-response probabilities independently within each SUD group. Given the cross-sectional nature of the survey, this approach was taken to generate independent latent profiles by SUD.
Given the online survey nature, respondents were required to answer all questions so that there were no item-level missing data. We conducted all analyses in SAS version 9.4 (Cary, NC).
RESULTS
Among adults in the United States in 2022, the prevalence of polysubstance use of 2 or more of the most common prescription and recreational drugs based on our definition was 20.9% (95% confidence interval [CI] = 20.5%, 21.3%), whereas the prevalence of use of only 1 drug was 26.5% (95% CI = 26.1%, 27.0%) and the prevalence of no drug use was 52.6% (95% CI = 52.1%, 53.1%). Compared to all adults, those using multiple drugs had higher proportions who self-reported being female, of White race, having annual household incomes of less than $50 000, and being current health care professionals or students (Table 1). Adults using multiple drugs also had higher proportions with a history of the mental health diagnoses, experience of acute or chronic pain, and use of alcohol or tobacco. Additionally, 17.6% of adults using multiple drugs had a likely SUD compared to 4.1% among those who used only a single drug.
TABLE 1—
Weighted Characteristics of Total Adult Population vs Polysubstance Use Population: United States, 2022
| Characteristics | Total Adult Population | Polysubstance Use |
| Unweighted no. | 59 041 | 15 800 |
| Estimated weighted no. | 259 008 600 | 54 044 600 |
| Age, y, mean (95% CI) | 47.6 (47.4, 47.7) | 44.3 (43.9, 44.6) |
| Male, % (95% CI) | 51.2 (50.7, 51.7) | 43.9 (42.9, 44.9) |
| Self-reported race,a % (95% CI) | ||
| White | 79.1 (78.6, 79.5) | 83.3 (82.4, 84.1) |
| Black or African American | 11.5 (11.1, 11.8) | 9.9 (9.2, 10.5) |
| Any other race | 12.6 (12.2, 12.9) | 10.7 (10.0, 11.4) |
| Hispanic/Latino, % (95% CI) | 11.5 (11.2, 11.9) | 12.5 (11.8, 13.2) |
| Household annual income, $, % (95% CI) | ||
| < 50 000 | 44.7 (44.2, 45.2) | 48.6 (47.6, 49.7) |
| 50 000 to < 75 000 | 35.2 (34.8, 35.7) | 33.8 (32.8, 34.8) |
| ≥ 75 000 | 20.1 (19.7, 20.5) | 17.5 (16.7, 18.3) |
| Highest education, % (95% CI) | ||
| High school or less | 23.2 (22.8, 23.6) | 23.8 (22.9, 24.7) |
| Some college or associate’s degree | 31.3 (30.9, 31.8) | 35.9 (34.9, 36.9) |
| Bachelor’s degree or trade school | 30.5 (30.0, 31.0) | 27.9 (26.9, 28.8) |
| Graduate degree | 14.9 (14.6, 15.3) | 12.5 (11.8, 13.1) |
| Current health care professional, % (95% CI) | 5.4 (5.1, 5.6) | 7.3 (6.7, 7.8) |
| Current student, % (95% CI) | 7.2 (6.8, 7.5) | 9.1 (8.4, 9.8) |
| History of mental health diagnoses, % (95% CI) | ||
| Any anxiety disorder | 23.5 (23.0, 23.9) | 52.8 (51.7, 53.8) |
| Major depressive disorder | 9.2 (8.9, 9.4) | 23.9 (23.0, 24.8) |
| ADD/ADHD | 6.9 (6.7, 7.2) | 17.5 (16.7, 18.3) |
| PTSD | 6.2 (6.0, 6.4) | 16.3 (15.6, 17.0) |
| Any acute or chronic pain in last 12 mo, % (95% CI) | 38.3 (37.8, 38.8) | 59.1 (58.1, 60.2) |
| Likely SUD, DAST-10 Score 3–10, % (95% CI) | 5.3 (5.1, 5.5) | 17.6 (16.8, 18.4) |
| Drug use in last 12 mo, % (95% CI) | ||
| Prescription opioids | 21.4 (21.0, 21.8) | 63.9 (62.9, 64.9) |
| Prescription antidepressants | 18.1 (17.8, 18.5) | 57.8 (56.8, 58.9) |
| Prescription benzodiazepines | 10.1 (9.9, 10.4) | 39.7 (38.7, 40.7) |
| Prescription stimulants | 5.2 (5.0, 5.4) | 21.3 (20.4, 22.1) |
| Cannabis | 21.4 (21.0, 21.8) | 56.2 (55.1, 57.2) |
| Other psychedelicsb | 1.9 (1.8, 2.1) | 8.7 (8.1, 9.2) |
| Methamphetamine | 1.8 (1.7, 1.9) | 8.1 (7.6, 8.6) |
| Psilocybin or mushrooms | 1.6 (1.5, 1.7) | 7.5 (6.9, 8.1) |
| Cocaine powder | 1.3 (1.2, 1.4) | 6.0 (5.5, 6.5) |
| Illicit opioids | 1.1 (1.0, 1.1) | 4.8 (4.4, 5.2) |
| Alcohol | 46.3 (45.8, 46.8) | 52.7 (51.6, 53.7) |
| Tobacco smoker | 12.0 (11.8, 12.2) | 23.1 (22.5, 23.8) |
Note. ADD/ADHD = attention deficit disorder/attention deficit hyperactivity disorder; CI = confidence interval; DAST-10 = drug abuse screening test; PTSD = posttraumatic stress disorder; SUD = substance use disorder. All estimates shown are weighted except the unweighted number to show sample size.
Respondents may self-select multiple races, estimates may not sum to 100; any other race may include American Indian/Alaska Native, Asian, or Native Hawaiian/Pacific Islander.
Other psychedelics included any use of: ketamine, LSD, MDMA, mescaline, or PCP.
Among the adults using multiple drugs, the most common behaviors included prescription opioid and antidepressant use (63.9% and 57.8%, respectively), followed by cannabis use (56.2%), then the remaining prescription drug classes (39.7% for benzodiazepines and 21.3% for stimulants), whereas all recreational drugs accounted for 8.7% or less (Table 1).
Latent Class Findings
The 4-latent-class model resulted in the best fit by Bayesian Information Criterion (Table A, which includes all diagnostics; available as a supplement to the online version of this article at http://www.ajph.org); models with more than 4 classes did not converge on a single solution. This 4-class model had an entropy of 0.83; entropy values of 0.8 or higher are acceptable, and it indicates how well the model defines the classes.21 The average posterior class probability was 0.91, indicating that there was high certainty across individuals for class assignment, where 0.9 or higher is ideal.21 After examination of the model inputs and the prevalence estimates, the following 4 classes were described and subjectively named (Figure 1, Table 2):
FIGURE 1—
Model Parameters for 4 Polysubstance Use Latent Classes, (a) Medically Guided Polysubstance Use, (b) Self-Guided Polysubstance Use, (c) Principle Cannabis Use Variety, and (d) Indiscriminate Coexposures: United States, 2022
Note. NMU = nonmedical use. Four observed latent classes emerged among those who used 2 or more psychoactive drugs in the last 12 months in the United States in 2022 (standard errors of estimates in supplemental material). Medically guided polysubstance use (prevalence = 11.5%) and self-guided polysubstance use (3.4%) were primarily defined by medical or NMU of prescription drugs, respectively. Principal cannabis use variety (4.0%) had high probability of cannabis use with various prescription and recreational drugs concomitantly used. Indiscriminate coexposures (2.1%) indicated concomitant drug use with indiscriminate drug preference. The interpretation of the first bar would be that an individual belonging to the medically guided polysubstance use class had a 0.6 probability of using opioids medically in the last 12 months (standard errors shown in Table B, available as a supplement to the online version of this article at http://www.ajph.org).
TABLE 2—
Weighted Characteristics Among 4 Polysubstance Use Latent Classes: United States, 2022
| Measure | Medically Guided Polysubstance Use | Self-Guided Polysubstance Use | Principal Cannabis Use Variety | Indiscriminate Coexposures |
| Unweighted no. | 7 862 | 2 719 | 3 104 | 2 115 |
| Estimated weighted no. | 29 718 900 | 8 696 200 | 10 268 000 | 5 361 500 |
| Demographics and risk factors | ||||
| Age, y, mean (95% CI) | 48.0 (47.4, 48.5) | 44.3 (43.4, 45.1) | 37.8 (37.1, 38.5) | 36.2 (35.5, 37.0) |
| Male, % (95% CI) | 36.2 (34.8, 37.6) | 46.6 (44.1, 49.2) | 52.8 (50.4, 55.3) | 65.0 (62.2, 67.8) |
| Self-reported race,a % (95% CI) | ||||
| White | 86.5 (85.4, 87.5) | 82.9 (80.9, 85.0) | 77.1 (74.9, 79.3) | 78.1 (75.2, 81.1) |
| Black or African American | 8.5 (7.7, 9.4) | 8.7 (7.3, 10.1) | 13.6 (11.8, 15.4) | 11.9 (9.6, 14.1) |
| Any other race | 9.1 (8.3, 10.0) | 10.3 (8.6, 12.0) | 14.3 (12.4, 16.2) | 13.5 (11.0, 16.0) |
| Hispanic/Latino, % (95% CI) | 9.2 (8.3, 10.0) | 14.7 (12.8, 16.6) | 16.2 (14.2, 18.1) | 20.5 (17.9, 23.2) |
| Household annual income, $, % (95% CI) | ||||
| < 50 000 | 47.8 (46.3, 49.2) | 45.5 (43.0, 48.0) | 55.5 (53.0, 58.0) | 45.6 (42.5, 48.7) |
| 50 000 to < 75 000 | 34.3 (32.9, 35.6) | 35.8 (33.3, 38.2) | 29.7 (27.4, 32.0) | 36.2 (33.3, 39.2) |
| ≥ 75 000 | 18.0 (16.9, 19.1) | 18.7 (16.7, 20.7) | 14.8 (12.9, 16.6) | 18.2 (16.0, 20.4) |
| Highest education, % (95% CI) | ||||
| High school or less | 20.8 (19.6, 21.9) | 23.7 (21.5, 25.9) | 31.8 (29.5, 34.1) | 25.2 (22.6, 27.8) |
| Some college or associate’s degree | 38.3 (36.9, 39.7) | 30.3 (28.0, 32.6) | 36.0 (33.7, 38.4) | 31.3 (28.3, 34.2) |
| Bachelor’s degree or trade school | 28.1 (26.8, 29.4) | 31.5 (29.1, 33.9) | 23.9 (21.7, 26.0) | 28.6 (25.8, 31.4) |
| Graduate degree | 12.8 (11.9, 13.8) | 14.5 (12.7, 16.2) | 8.3 (7.0, 9.7) | 15.0 (12.9, 17.0) |
| Current health care professional, % (95% CI) | 6.4 (5.7, 7.1) | 7.3 (6.0, 8.6) | 6.0 (4.9, 7.2) | 14.4 (12.1, 16.7) |
| Current student, % (95% CI) | 7.7 (6.7, 8.6) | 7.8 (6.2, 9.3) | 9.8 (8.2, 11.5) | 17.7 (15.2, 20.2) |
| History of mental health diagnosis, % (95% CI) | ||||
| Any anxiety disorder | 55.8 (54.4, 57.2) | 49.3 (46.8, 51.9) | 49.7 (47.2, 52.2) | 47.3 (44.2, 50.4) |
| Major depressive disorder | 26.7 (25.4, 27.9) | 21.1 (19.0, 23.1) | 20.2 (18.3, 22.2) | 19.7 (17.3, 22.1) |
| ADD/ADHD | 17.3 (16.2, 18.4) | 13.3 (11.5, 15.0) | 19.2 (17.2, 21.2) | 22.3 (19.8, 24.8) |
| PTSD | 16.3 (15.3, 17.3) | 13.3 (11.7, 14.9) | 17.6 (15.8, 19.4) | 18.3 (16.1, 20.5) |
| Any acute or chronic pain in last 12 mo, % (95% CI) | 64.9 (63.5, 66.4) | 51.2 (48.6, 53.7) | 50.3 (47.8, 52.8) | 56.7 (53.6, 59.8) |
| Substance use factors | ||||
| Likely SUD, DAST-10 Score 3-10, % (95% CI) | 6.1 (5.4, 6.7) | 14.5 (12.8, 16.2) | 31.9 (29.6, 34.2) | 58.9 (55.9, 62.0) |
| Received substance use treatment in last year, % (95% CI) | 2.1 (1.7, 2.5) | 5.1 (4.1, 6.1) | 8.1 (6.8, 9.3) | 23.2 (20.6, 25.8) |
| Concomitant use when using prescription drugs, % (95% CI) | ||||
| With alcohol | Rareb | 23.8 (21.6, 26.0) | 27.3 (25.0, 29.6) | 65.4 (62.5, 68.2) |
| With cannabis | Rare | 2.9 (2.1, 3.6) | 47.8 (45.3, 50.3) | 68.7 (65.9, 71.6) |
| With other prescription drugs | Rare | 44.8 (42.3, 47.3) | 12.2 (10.6, 13.8) | 79.2 (76.7, 81.6) |
| With recreational drugs | Rare | 3.7 (2.7, 4.6) | 8.1 (6.8, 9.4) | 66.7 (63.8, 69.6) |
| Concomitant use when using recreational drugs, % (95% CI) | ||||
| With alcohol | Rare | 2.3 (1.6, 2.9) | 24.4 (22.2, 26.5) | 44.7 (41.7, 47.8) |
| With cannabis | Rare | Rare | 34.9 (32.5, 37.2) | 46.0 (42.9, 49.1) |
| With other prescription drugs | Rare | 3.6 (2.6, 4.5) | 3.5 (2.8, 4.3) | 50.6 (47.5, 53.7) |
| With recreational drugs | Rare | 0.3 (0.1, 0.5) | 7.6 (6.3, 8.8) | 39.8 (36.8, 42.8) |
| Prescription drug sourcing from own health care professional, % (95% CI) | ||||
| Among those using opioids | 96.0 (95.3, 96.7) | 74.1 (71.7, 76.5) | 68.8 (65.5, 72.1) | 59.8 (56.5, 63.1) |
| Among those using antidepressants | 97.9 (97.4, 98.4) | 78.0 (75.5, 80.5) | 80.8 (77.5, 84.2) | 62.8 (58.8, 66.9) |
| Among those using benzodiazepines | 97.7 (97.1, 98.4) | 70.6 (67.7, 73.6) | 64.0 (59.3, 68.7) | 53.1 (49.2, 57.1) |
| Among those using stimulants | 98.3 (97.3, 99.3) | 60.3 (55.4, 65.2) | 58.2 (51.9, 64.5) | 49.7 (44.7, 54.7) |
Note. ADD/ADHD = attention deficit disorder/attention deficit hyperactivity disorder; CI = confidence interval; DAST-10 = drug abuse screening test; PTSD = posttraumatic stress disorder; SUD = substance use disorder. All estimates shown are weighted except the unweighted number to show sample size.
Respondents may self-select multiple races, estimates may not sum to 100; any other race may include American Indian/Alaska Native, Asian, or Native Hawaiian/Pacific Islander.
Sparse data cells (< 5 respondents) were suppressed.
-
1.
Medically guided polysubstance use (Figure 1a): national prevalence 11.5% (95% CI = 11.2%, 11.8%). This class was defined by the highest endorsement of medical use of multiple prescribed drugs classes (mean = 1.8 types) sourced from their own health care providers. There was 52.4% probability of cannabis use but negligible concomitant drug use. Among adults representing this class, in the last year 6.1% were estimated to have an SUD while 2.1% received substance use treatment.
-
2.
Self-guided polysubstance use (Figure 1b): national prevalence 3.4% (95% CI = 3.2%, 3.5%). This class was defined primarily by NMU of multiple prescription drug classes (mean = 2.3 types) with some concomitant use with alcohol or other prescriptions. These prescriptions were likely to be sourced from providers, but approximately 25% sourced exclusively outside health care providers (data not shown). Among adults representing this class, in the last year 14.5% were estimated to have an SUD while 5.1% received substance use treatment.
-
3.
Principal cannabis use variety (Figure 1c): national prevalence 4.0% (95% CI = 3.8%, 4.2%): This class was chiefly defined by high probability of cannabis use at 95.3% with smaller, but approximately equal, probabilities across NMU of prescriptions and recreational drugs. These drugs were used concomitantly with cannabis and alcohol. When we considered each drug type included in this analysis, there were on average 2.0 other drugs used besides cannabis. Among adults representing this class, in the last year 31.9% were estimated to have an SUD and only 8.1% received substance use treatment.
-
4.
Indiscriminate coexposures (Figure 1d): national prevalence 2.1% (95% CI = 1.9%, 2.2%). This class was defined by concomitant use of cannabis, alcohol, prescriptions, and recreational drugs with patterns indicating indiscriminate drug preferences. When we considered each drug included in this analysis, there were on average 4.6 drugs used in the last 12 months. When prescription drugs were used, they were commonly sourced outside of health care providers. Among adults representing this class, in the last year 58.9% were estimated to have an SUD and only 23.2% received substance use treatment.
Table 2 shows the demographic characteristics and other risk factors for members of the 4 classes. Those belonging to the medically guided polysubstance use and self-guided polysubstance use classes generally were older and there were slightly more females than males. The principal cannabis use variety and indiscriminate coexposures classes had the highest proportions of self-reported minority groups compared to other classes without noticeable differences in income or education.
Sensitivity Analysis by Substance Use Disorder
The measurement invariance likelihood ratio test indicated that there were differences in the measurement model across the SUD groups (P < .001). When modeled independently, 4 analogous latent classes were observed among those with and without likely SUD by subjective interpretation, albeit in different proportions (Figure 2). The difference in class membership probabilities was likely driving the measurement invariance findings more than item-response probabilities. The group without a likely SUD comprised 82.4% of the sample, so latent class proportions were similar to the proportions of each class in the overall analysis. The medically guided polysubstance use class was 6% larger proportionally compared to the overall findings, the self-guided polysubstance use class was similar in size, while the other 2 classes were smaller. The indiscriminate coexposures class in this group was indicated with an asterisk because the item-response probabilities for recreational drugs were slightly lower than the overall findings with high probabilities for NMU of prescription drugs (online Figure A). The group with likely SUD constituted 17.6% of the sample and had almost equal proportions of the 4 latent classes, with principal cannabis use variety being largest at 34.5% (Figure 2). The self-guided polysubstance use class was also indicated with an asterisk because the item-response probabilities for recreational drugs and concomitant use were slightly higher than the main findings (Figure A).
FIGURE 2—
Proportion of 4 Polysubstance Use Latent Classes Comparing Those With or Without Substance Use Disorders: United States, 2022
Note. DAST-10 = drug abuse screen test; SUD = substance use disorder. Those adults without likely SUD constituted 82.4% of the polysubstance sample, while those with likely SUD were 17.6% of sample. An analogous 4 latent classes were observed in both SUD groups as observed in the main analysis, albeit in different proportions, indicating that SUD could arise in someone belonging to any of these classes. aIndicates analogous classes upon interpretation but different contributions in latent class item-response probabilities (provided in Table B and Figure A, available as a supplement to the online version of this article at http://www.ajph.org).
DISCUSSION
This study, which utilized a general population survey among adults in the United States, found that 21% used 2 or more psychoactive drugs in 2022. This research elucidated the landscape of polysubstance use by using methods to reduce the dimensionality of this multifaceted behavior. Four unique behavioral classes were found that crossed multiple drug classes and drug types (prescription or recreational), and they were partially defined by patterns of concomitant drug use.
The most prevalent profile contained individuals with fewer markers of SUD and was categorized primarily by prescription drugs used as intended, sourced by their own providers. Although this group is unsurprising in the context of common behaviors, it is not commonly discussed in the context of polysubstance use. This group equates to about 11.5% of all adults, and it had other health concerns such as pain history and high proportion of comorbidities. Additionally, about half of this group reported also using cannabis throughout a similar timeframe. This medically guided polysubstance use group is a new contribution to the polysubstance use literature, although it is consistent with prior studies linking cannabis use with increased use of psychoactive prescriptions.25 More research examining this association and the interplay of mental health disorders is warranted, because these use patterns of multiple drugs, even under the direction of a health care provider, could represent individuals attempting to treat complex comorbidities. This group reports interactions with health care providers, where identification, education, or prevention strategies could be helpful to address underlying mental health concerns or prevent potential prescription drug interactions.
The second largest latent class found, the principal cannabis use variety, constitutes 4.0% of all adults. Cannabis was the primary drug, but there are about equal proportions of people using other prescription drugs nonmedically, recreational drugs, or both. Of note, while there is variety in the other drugs of choice, these other drugs in the profile are commonly used concomitantly with cannabis or alcohol. Approximately one third of this group has a likely SUD, but with the variety in the other contributing drug, education tactics or prevention strategies would need to be more customized to the individual. With this being the second largest profile and as more states continue to legalize cannabis, there might be opportunities to explore intervention points through other environments outside health care providers. This group reported only modest interaction with health care, so new places to add “connect to care” opportunities may be required, such as dispensaries or cannabis retail outlets.
The smallest 2 latent classes found are the ones that are already discussed in the literature but are often grouped into a single, low-prevalence class that combines both recreational drug use with NMU of prescription drugs, often called “the polysubstance use class.”14,26 This work separates it into 2 distinct profiles. First, the self-guided polysubstance use class (3.4%) consists of those primarily nonmedically using prescription medications and mixing with other prescriptions or alcohol. This group has also been identified in other studies, particularly around prescription opioid or stimulants use27,28 and around use of alcohol with sedatives.29 Although this class still has an opportunity for health care provider intervention, it could be a group that would benefit from different screening and early referral; 14.5% had markers for a SUD. Many in the class could benefit from prevention of potential drug interactions or education around advancement to SUDs.
Finally, the indiscriminate coexposures class is consistent with other studies and is the most often cited polysubstance group, despite it having the lowest prevalence (2.1%). It is typically associated with high risk for acute advanced health outcomes like overdose and has major clinical and public health implications. We observed a large discrepancy between those with a likely SUD (58.9%) and those reporting getting treatment (23.2%) who could benefit from psychosocial or pharmacological treatment.30 A better understanding of treatment access or barriers to access among this group would be beneficial. Pharmacological treatments typically address single substances, and new treatments that address psychostimulant use or concomitant use should be explored given that there are currently none approved by the Food and Drug Administration.31 There is likely a subset of this group who do not believe they are at risk for acute health events, despite commonly sourcing their drugs from illicit sources and mixing drugs concomitantly.32
After stratifying the analysis by those with likely SUD, 4 similar classes appeared with differing prevalence, demonstrating that SUD is not a defining feature of the classes. This is a shift in perspective that is necessary. Most literature describes those with SUDs as being more aligned with specific high-risk profiles, which was shown here, but this work shows that an SUD could arise in someone belonging to any of these classes. Therefore, there is a need to take a more personalized approach where identification into an individual’s total drug use behaviors is quantified and their health care approaches them more holistically. “Polysubstance use” lacks diagnostic specificity, and these individuals can be clinically complex cases with ranging profiles of drug use histories and high rates of comorbidities or mental health disorders.26,33,34 Approaches to identifying these individuals should be expanded to capture a wide range of drugs and use behaviors. Given that an SUD can cause a spectrum of difficulties in an individual’s life, an SUD may manifest differently across these profiles, which would be important to explore in the future.
Strengths and Limitations
One strength of this large population study was the person-centric analysis approach, which helps elucidate behavior profiles not described in the context of polysubstance use. Within our classes, we found parallels to polysubstance use features seen in other studies, giving strength to the novel findings of this study. The weighted analysis accounts for both health and demographic metrics to make more generalized drug use prevalence estimates. Limitations to this study include possible drug recall bias within the last year. Online formats have been shown to be less prone to social desirability bias, which is important for accuracy of self-report.35 Also, the latent class approach can be sensitive to inclusion criteria, although we found that profiles were defined across drug classes and behaviors; it did not appear to be sensitive to inclusion of any 1 drug. Additionally, class interpretation can be subject to the naming fallacy based on qualitative interpretation of the probabilistic assignment. It is important to note that the landscape of drug use is rapidly evolving in the United States. These observed profiles could change over time, such as with the introduction of new substances or changes in use patterns (e.g., use of xylazine, changes in prescribing guidelines or decriminalization of substances).
Public Health Implications
About 1 in 5 US adults have used multiple psychoactive drugs in the last year across 4 distinct profiles. Current diagnostic criteria for identifying polysubstance use lack the ability to identify the extensive combinations of drug use patterns.26,33 No one-size-fits-all prevention approach is likely to succeed given the clear existence of subpopulations with very different behaviors and histories.
Both clinical practice and public health should adapt the current approaches. First, for more effective clinical care, the education of providers around these profiles and correlated behaviors would be beneficial. Knowledge of use behaviors obtained from current screening tools could prompt future questioning into correlated harmful behaviors in a nonaccusatory manner. Second, improved screening tools could help determine appropriate clinical risk stratification, which dictates education, future screening, or next steps to reduce adverse health effects. Third, an SUD can manifest in different ways, and understanding which behaviors are leading to problematic areas of a patient’s life would improve overall patient care, both physical and mental. Not everyone in these profiles interacts regularly with health care providers, so public health professionals should consider alternative access points to interact with individuals in their own communities to educate about risks, encourage safe use patterns, or implement programs to eliminate use patterns, depending on individuals’ goals and health risks.
ACKNOWLEDGEMENTS
Research reported in this publication was supported by the National Institute on Drug Abuse of the National Institutes of Health (award no. R36DA057413).
This work was presented at the 2023 annual meeting of the American Public Health Association.
Note. The contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Institutes of Health.
CONFLICTS OF INTEREST
The authors have no conflicts of interest to report.
HUMAN PARTICIPANT PROTECTION
Secondary data research application for this analysis was approved by the Colorado Multiple Institutional Review Board (COMIRB) on March 7, 2023 (#23-0373). The NMURx Program study was given a certificate of exemption by COMIRB on July 5, 2016 (#16-0922).
See also Xu et al., p. 646.
REFERENCES
- 1.Connor JP, Gullo MJ, White A, Kelly AB. Polysubstance use: diagnostic challenges, patterns of use and health. Curr Opin Psychiatry. 2014;27(4):269–275. https://doi.org10.1097/YCO.0000000000000069 [DOI] [PubMed] [Google Scholar]
- 2.Crummy EA, O’Neal TJ, Baskin BM, Ferguson SM. One is not enough: understanding and modeling polysubstance use. Front Neurosci. 2020;14:569. 10.3389/fnins.2020.00569 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Salom CL, Betts KS, Williams GM, Najman JM, Alati R. Predictors of comorbid polysubstance use and mental health disorders in young adults—a latent class analysis. Addiction. 2016;111(1):156–164. 10.1111/add.13058 [DOI] [PubMed] [Google Scholar]
- 4.Key Substance Use and Mental Health Indicators in the United States: Results From the 2020 National Survey on Drug Use and Health. Rockville, MD: Substance Abuse and Mental Health Services Administration; 2021. HHS Publication No. PEP21-07-01-003, NSDUH Series H-56. https://www.samhsa.gov/data/sites/default/files/reports/rpt35325/NSDUHFFRPDFWHTMLFiles2020/2020NSDUHFFR1PDFW102121.pdf [Google Scholar]
- 5.Lorvick J, Browne EN, Lambdin BH, Comfort M. Polydrug use patterns, risk behavior and unmet healthcare need in a community-based sample of women who use cocaine, heroin or methamphetamine. Addict Behav. 2018;85:94–99. 10.1016/j.addbeh.2018.05.013 [DOI] [PubMed] [Google Scholar]
- 6.Gilmore D, Zorland J, Akin J, Johnson JA, Emshoff JG, Kuperminc GP. Mortality risk in a sample of emergency department patients who use cocaine with alcohol and/or cannabis. Subst Abus. 2018;39(3):266–270. 10.1080/08897077.2017.1389799 [DOI] [PubMed] [Google Scholar]
- 7.Mattson CL, Tanz LJ, Quinn K, Kariisa M, Patel P, Davis NL. Trends and geographic patterns in drug and synthetic opioid overdose deaths—United States, 2013–2019. MMWR Morb Mortal Wkly Rep. 2021;70(6):202–207. 10.15585/mmwr.mm7006a4 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Black JC, Rockhill KM, Dart RC, Iwanicki J. Clustering patterns in polysubstance mortality in the United States in 2017: a multiple correspondence analysis of death certificate data. Ann Epidemiol. 2023;77:119–126. 10.1016/j.annepidem.2022.03.011 [DOI] [PubMed] [Google Scholar]
- 9.Ciccarone D. The rise of illicit fentanyls, stimulants and the fourth wave of the opioid overdose crisis. Curr Opin Psychiatry. 2021;34(4):344–350. 10.1097/YCO.0000000000000717 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Jenkins RA. The fourth wave of the US opioid epidemic and its implications for the rural US: a federal perspective. Prev Med. 2021;152(pt 2):106541. 10.1016/j.ypmed.2021.106541 [DOI] [PubMed] [Google Scholar]
- 11.Townsend T, Kline D, Rivera-Aguirre A, et al. Racial/ethnic and geographic trends in combined stimulant/opioid overdoses, 2007–2019. Am J Epidemiol. 2022;191(4):599–612. 10.1093/aje/kwab290 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.Collins LM, Lanza ST. Latent Class and Latent Transition Analysis: With Applications in the Social, Behavioral, and Health Sciences. Hoboken, NJ: John Wiley and Sons Inc; 2010. [Google Scholar]
- 13.Tomczyk S, Isensee B, Hanewinkel R. Latent classes of polysubstance use among adolescents—a systematic review. Drug Alcohol Depend. 2016;160: 12–29. 10.1016/j.drugalcdep.2015.11.035 [DOI] [PubMed] [Google Scholar]
- 14.de Jonge MC, Bukman AJ, van Leeuwen L, Onrust SA, Kleinjan M. Latent classes of substance use in young adults—a systematic review. Subst Use Misuse. 2022;57(5):769–785. 10.1080/10826084.2022.2040029 [DOI] [PubMed] [Google Scholar]
- 15.Black JC, Rockhill K, Forber A, et al. An online survey for pharmacoepidemiological investigation (Survey of Non-Medical Use of Prescription Drugs Program): validation study. J Med Internet Res. 2019; 21(10):e15830. 10.2196/15830 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16.Dart RC. Monitoring risk: post marketing surveillance and signal detection. Drug Alcohol Depend. 2009;105(suppl 1):S26–S32. 10.1016/j.drugalcdep.2009.08.011 [DOI] [PubMed] [Google Scholar]
- 17.Skinner HA. The drug abuse screening test. Addict Behav. 1982;7(4):363–371. 10.1016/0306-4603(82)90005-3 [DOI] [PubMed] [Google Scholar]
- 18.Bohn M, Babor T, Kranzler H. Validity of the Drug Abuse Screening Test (DAST-10) in inpatient substance abusers. Probl Drug Depend. 1991;119: 233–235. [Google Scholar]
- 19.Janulis P, Luo J, Tang X, Schalet BD. Can severity of substance use be measured across drug classes? Estimating differential item functioning by drug class in two general measures of substance use severity. Drug Alcohol Depend. 2023;250:110877. 10.1016/j.drugalcdep.2023.110877 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Nylund-Gibson K, Choi AY. Ten frequently asked questions about latent class analysis. Transl Issues Psychol Sci. 2018;4(4):440–461. 10.1037/met0000486 [DOI] [Google Scholar]
- 21.Weller BE, Bowen NK, Faubert SJ. Latent class analysis: a guide to best practice. J Black Psychol. 2020;46(4):287–311. 10.1177/0095798420930932 [DOI] [Google Scholar]
- 22.Dziak JJ, Lanza ST, Tan X. Effect size, statistical power and sample size requirements for the bootstrap likelihood ratio test in latent class analysis. Struct Equ Modeling. 2014;21(4):534–552. 10.1080/10705511.2014.919819 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Lanza ST, Collins LM, Lemmon DR, Schafer JL. PROC LCA: a SAS procedure for latent class analysis. Struct Equ Modeling. 2007;14(4):671–694. 10.1080/10705510701575602 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.Bakk Z, Kuha J. Relating latent class membership to external variables: an overview. Br J Math Stat Psychol. 2021;74(2):340–362. 10.1111/bmsp.12227 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Rognli EB, Bramness JG, von Soest T. Cannabis use in early adulthood is prospectively associated with prescriptions of antipsychotics, mood stabilizers, and antidepressants. Acta Psychiatr Scand. 2020;141(2):149–156. 10.1111/acps.13104 [DOI] [PubMed] [Google Scholar]
- 26.Connor JP, Leung J, Chan GCK, Stjepanović D. Seeking order in patterns of polysubstance use. Curr Opin Psychiatry. 2023;36(4):263–268. 10.1097/YCO.0000000000000881 [DOI] [PubMed] [Google Scholar]
- 27.Chan GCK, Butterworth P, Becker D, et al. Longitudinal patterns of amphetamine use from adolescence to adulthood: a latent class analysis of a 20-year prospective study of Australians. Drug Alcohol Depend. 2019;194:121–127. 10.1016/j.drugalcdep.2018.08.042 [DOI] [PubMed] [Google Scholar]
- 28.Rockhill KM, Olson R, Dart RC, Iwanicki JL, Black JC. Differing behaviors around adult nonmedical use of prescription stimulants and opioids: latent class analysis. J Med Internet Res. 2023;25:e46742. https://doi.org10.2196/46742 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.Quek LH, Chan GC, White A, et al. Concurrent and simultaneous polydrug use: latent class analysis of an Australian nationally representative sample of young adults. Front Public Health. 2013;1:61. 10.3389/fpubh.2013.00061 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30.Ray LA, Meredith LR, Kiluk BD, Walthers J, Carroll KM, Magill M. Combined pharmacotherapy and cognitive behavioral therapy for adults with alcohol or substance use disorders: a systematic review and meta-analysis. JAMA Netw Open. 2020;3(6):e208279. 10.1001/jamanetworkopen.2020.8279 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Lee JC, Janda KD. Development of effective therapeutics for polysubstance use disorders. Curr Opin Chem Biol. 2022;66:102105. 10.1016/j.cbpa.2021.102105 [DOI] [PubMed] [Google Scholar]
- 32.Carroll JJ, Marshall BDL, Rich JD, Green TC. Exposure to fentanyl-contaminated heroin and overdose risk among illicit opioid users in Rhode Island: a mixed methods study. Int J Drug Policy. 2017;46:136–145. 10.1016/j.drugpo.2017.05.023 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Shearer RD, Bart G, Reznikoff C. Rethinking the use of “polysubstance” to describe complex substance use patterns. J Gen Intern Med. 2022;37(12):3174–3175. 10.1007/s11606-022-07424-5 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 34.Ellis JD, Rabinowitz JA, Ware OD, Wells J, Dunn KE, Huhn AS. Patterns of polysubstance use and clinical comorbidity among persons seeking substance use treatment: an observational study. J Subst Use Addict Treat. 2023;146:208932. 10.1016/j.josat.2022.208932 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 35.van Gelder MM, Bretveld RW, Roeleveld N. Web-based questionnaires: the future in epidemiology? Am J Epidemiol. 2010;172(11): 1292–1298. 10.1093/aje/kwq291 [DOI] [PubMed] [Google Scholar]