This study utilized data from the National Survey on Drug Use and Health (NSDUH), collected by the Substance Abuse and Mental Health Services Administration from 2015 to 2022 (23). The data were collected via an annual, cross-sectional survey that asks non-institutionalized individuals aged 12 years and older about their experience with substance use and health status. To provide population estimates, the survey uses a stratified, multi-stage area probability sampling design. Details of the sampling design and methods used for the survey are provided elsewhere (24). Also, the survey does not collect any identifiable information. We focused on adults (aged 18 years and older) because adolescents have different substance use behaviors and outcomes from adults (25). For example, adolescents are more prone to engage in risky substance use behaviors, which can be influenced by social and developmental factors, such as peer substance use (26, 27). This study was exempt from IRB review because the NSDUH data are de-identified and publicly available, which means it does not constitute human subject research. As a result, participant consent was not required.

The main outcome variables of this study are prescription pain reliever (opioid) and/or cannabis use in the past year. The NSDUH survey included the following question: “In the past 12 months, which, if any, of these pain relievers have you used?” A “yes” response indicated that the individual had used a prescription opioid in the past year. We excluded individuals who reported misusing prescription opioids in terms of frequency, amount, and duration, as prior research indicates that nonmedical use is associated with distinct drug use patterns and a higher likelihood of polysubstance use, including cannabis (28–30). To ensure a more focused analysis, we limited our sample to individuals who reported medical use of opioids. Another question was, “How long has it been since you last used marijuana or hashish?” If the response was either “within the past 30 days” or “more than 30 days ago but within the past year,” we considered that as cannabis use within the past year. Using the responses to those two questions, we defined the following groups: (1) opioid-only (yes to opioid question, no to cannabis question), (2) cannabis-only (no to opioid question, yes to cannabis question), and (3) OC group (yes to both questions).

Demographic factors included age, gender, and race/ethnicity. Age was grouped into young (18–34 years old), middle-aged (35–64 years old), and older adults (65 + years old). Racial/ethnic groups were classified into non-Hispanic (NH) white people, NH black people, Hispanic people, and other minorities. Socioeconomic and geographic factors included family income (<$20,000, $20,000–$49,999, $50,000–$74,999, >$75,000), education completed (less than or higher than high school diploma/GED credential), health insurance type (private, Medicaid/Medicare, self-pay, other), employment status (unemployed, employed), and metropolitan statistical area (non-metro, small metro, large metro areas).

We considered the health status of participants. Overall self-reported health status (question: “Would you say your health in general is excellent, very good, good, fair, or poor?”) was categorized into four levels (Fair/Poor, Good, Very Good, Excellent). Mental health status was estimated using a serious psychological distress indicator based on the worst Kessler-6 total score in the past year (31). The survey year (2015–2022) was included as a covariate.

First, we used descriptive statistics to compare the characteristics of the opioid-only, cannabis-only, and OC groups. Continuous variables were compared using one-way ANOVA tests with the assumption of equal variances, and categorical variables were compared using chi-square tests (α = 0.05). Next, we used a weighted multinomial logistic regression model to evaluate factors associated with co-use, comparing the OC group to both the opioid-only and cannabis-only groups. To examine potential multicollinearity among categorical independent variables, we computed generalized variance inflation factor (GVIF), which extend the traditional VIF to handle multi-level categorical predictors (32). All adjusted GVIFs were less than 2, which indicates no significant multicollinearity. Model adequacy was assessed using a likelihood ratio test that compares the full multinomial logistic regression model to the intercept-only model. All analyses incorporated survey weights and accounted for the complex survey design of NSDUH, including clustering and stratification. Statistical analyses were conducted using the R software, version R 4.3.0 (33). Specifically, we constructed a complex survey design object using the final analysis weight (ANALWT_C), clustering (VEREP), and stratification variable (VESTR) from the NSDUH data, implemented via the svydesign function in survey package. A weighted multinomial logistic regression model was then fitted using svy_vglm function from svyVGAM package in R (32).

Our study estimated the prevalence of independent prescription opioid use, and cannabis use, and their co-use among non-institutionalized US adults between 2015 and 2022. While describing prevalence trends can serve useful descriptive epidemiologic purposes, we did not perform formal trend analyses, as comparisons of 2020 estimates with those of prior years are discouraged because of methodological changes in the NSDUH survey implemented during the COVID-19 pandemic.

When comparing the rates between 2015 and 2022, opioid use without cannabis decreased significantly by 36.1% (from 30.2 to 19.3%), while cannabis use without opioids more than doubled (from 7.1 to 15.6%). Co-use of opioids and cannabis (OC group) increased by 30.6% (from 4.9 to 6.4%). National statistics show similar trends in opioid and cannabis use, although those figures are not conditional on other substance use. For example, between 2011 and 2020, opioid prescriptions decreased by 44% (34), and overdose deaths involving prescription opioids did not increase (35). These improvements may reflect the impact of nationwide efforts to respond to the national opioid crisis (36–38). Other studies have shown rapidly increasing cannabis use, especially among younger adults, and cannabis use disorder over the last decade (39, 40). This trend may be attributed primarily to the widespread state legalization of medical and recreational cannabis (41).

Relatively little research has focused on distinguishing the characteristics of individuals who use opioids or cannabis only from those who use both substances, although many studies have examined the broader relationship between cannabis and opioid use. Much of existing literature has evaluated whether cannabis legalization or use influences opioid consumption, the risk of opioid use disorder, or treatment outcomes for opioid use disorder, often with mixed findings. For example, some studies have reported that opioid prescriptions, hospitalizations related to opioid misuse or dependence, and opioid overdose mortality rates were lower in states that legalized medical cannabis use (14, 42, 43). Other research indicated that greater availability of cannabis does not necessarily increase opioid use (44), and that individuals with pain may reduce or avoid opioid use by substituting medical cannabis (9, 45–47). Public health programs have also documented the use of medical cannabis as a harm reduction strategy (48–51).

The main objective of this study was to examine demographic and health-related characteristics of opioid-only and cannabis-only groups in comparison with the co-use group. Interestingly, several characteristics showed distinct patterns when contrasting opioid-only with co-use versus cannabis-only with co-use.

For example, compared with women, men had a greater risk of using cannabis only, relative to using OC, but a lower risk of using opioids only, relative to using OC. In other words, among individuals who use cannabis, men were more likely than women to use cannabis without opioids, whereas among individuals who use prescription opioids, men were more likely than women to combine opioids with cannabis. One possible explanation for this finding is the higher prevalence of recreational cannabis use among men compared with women. Men who use cannabis for recreational purposes may be more likely to co-use alcohol more often than prescription opioids. Prior studies have shown that the simultaneous use of alcohol with cannabis is the most frequent pattern of polysubstance use among individuals who use cannabis for recreational purposes (52, 53), and this co-use pattern is more common among men than women (54).

Similarly, different age groups showed contrasting patterns in their risks of co-use relative to single-substance use. The risk that individuals in the middle-aged and older adult groups would use OC, relative to their chance of using cannabis alone, was about 1.8 times higher than for the younger adult group. However, the two older age groups had a 2.2 to 7.37 times higher probability of using opioids only, relative to using OC, than the youngest age group. This may imply that older adults who use cannabis may continue to use opioids to address their complex health needs. Longitudinal studies that track opioid use among older adults who use cannabis will be helpful for understanding their substance use patterns and resultant health outcomes.

The probabilities of using a single substance versus OC also varied by type of metropolitan statistical area. Among residents of large-metro areas, the risk of using OC was higher, relative to the risk of using cannabis only, than for residents of non-metro areas. On the other hand, large-metro residents had a higher probability of using opioids only than of using OC, compared to non-metro residents. These findings may point to different pathways into substance use. In non-metro areas, where healthcare access is more limited and cannabis availability is lower (55, 56), individuals who use cannabis may be less likely to combine it with prescription opioids than those in large-metro areas. On the other hand, in large-metro areas, individuals who already use opioids may have easier access to prescription opioids and thus be less likely to complement their use with cannabis than those in non-metro areas. To better understand these pathways, more detailed longitudinal information on substance use experiences will be needed. In addition, these patterns may change if analyses incorporate both licit and illicit opioids, given geographical differences in types of opioid use (57, 58).

The findings on the characteristics of individuals who use opioids and/or cannabis have important implications for clinical care and public health. A deeper understanding of the traits associated with opioid and cannabis use or co-use can help healthcare professionals make more informed decisions about pain management, medication counseling, and follow-up care. Insights from this study may also support the development of predictive model that flag patterns of co-use across specific demographic and clinical profiles. When paired with data on other substance use behaviors and health outcomes, such tools can help identify individuals who may benefit from additional clinical attention, supportive services, or discussions about alternative therapies.

The observed variability in substance use patterns across large-metro and non-metro areas highlights the importance of geographically tailored public health efforts. In metropolitan areas, where co-use of opioids and cannabis is more common, public health campaigns that address the risks of concurrent use, particularly in regions where both substances are legally accessible, can help raise awareness and support informed decision-making. In addition, incorporating enhanced screening and patient-provider communication about polysubstance use during routine visits, especially in pain management and behavioral health settings, may help mitigate potential adverse outcomes. In non-metropolitan areas, expanding access to telemedicine for pain management and behavioral health can improve understanding the reasons individuals use opioids or cannabis and play a preventive role by offering earlier support and alternatives before other substances are introduced. Furthermore, community-based outreach programs can provide substance use education and stigma reduction tailored to rural populations, which may help reduce the risk of co-use and related harms.

Across regions, enhanced monitoring of polysubstance use, involving opioids and cannabis, can help detect early signals of emerging public health concerns and better identify populations at elevated risk. Building on the current study with temporal–spatial analyses of co-use patterns and associated health outcomes could further inform the geographic distribution of need. Such analyses can guide the strategic deployment of prevention and education initiatives, as well as the implementation of integrated care models in areas where co-use is more prevalent. This approach may support early intervention and promote safer, more coordinated management of opioid and cannabis use.

Our study has several limitations. First, the NSDUH survey data are based on self-reporting of drug use and health status, and our results may be subject to recall bias influenced by social desirability (59). Second, the survey asked about general cannabis use without details about illegal use or purposes of use (i.e., medical, leisure, or both). Substance use behaviors and risks would be different among those who use cannabis for medical purposes only and among those who use it for leisure purposes only. More detailed information about cannabis use would elucidate its increasing use trends and identify individuals at a high risk of misusing it, either by itself or with other substances. Lastly, since the survey does not include drug use among those who are institutionalized, homeless, or in the military, our study results are limited to U. S. civilian, non-institutionalized individuals.