Prevalence and Correlates of Cannabis Use and Cannabis Use Disorder Among U.S. Veterans: Results From the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC-III)

The legal status of cannabis has changed dramatically in the United States. Although cannabis remains illegal at the federal level, 36 states and the District of Columbia have legalized cannabis for medical use, 17 states and the District of Columbia have legalized adult recreational use (1), and adults are increasingly likely to perceive cannabis use as harmless (2). Concomitantly, the prevalence of adult nonmedical cannabis use and cannabis use disorder has increased in general population and clinical samples (3–6).

Many conditions approved for medical cannabis use are common among veterans (e.g., posttraumatic stress disorder [PTSD]), and veteran groups have advocated for legalization (1, 7–9). However, little is known about the prevalence and correlates of cannabis use among U.S. veterans. Studies of Veterans Health Administration (VHA) patients showed that 11%–14% used cannabis (10, 11). However, many U.S. veterans do not utilize VHA care (12). Only one study examined cannabis use in a nationally representative sample that included veterans without regard to VHA enrollment (13), showing that 9% of veterans reported past-year cannabis use. Other studies examining trends, recreational versus medical cannabis use, or cannabis use disorder among veterans are older (4) or are limited to particular subsamples (4, 13–15).

Given growing legalization and acceptability of cannabis use, nationally representative data are needed to understand cannabis use among veterans, including frequent use patterns (e.g., daily use) and cannabis use disorder, and to identify high-risk subgroups (3, 13, 16). In civilian samples, frequent cannabis use and cannabis use disorder are associated with other substance and psychiatric disorders—for example, PTSD and anxiety and mood disorders (17–19)—and may worsen their outcomes (20–22). Such disorders are overrepresented among veterans (23, 24). Additionally, since civilian studies show higher rates of adult cannabis use in states with legalized cannabis than in other states (16, 25–28), understanding the risk for cannabis use among veterans in states with and without medical marijuana laws is critical to understanding public health effects of legalization among veterans. Only one such study exists, a VHA medical record study suggesting that in 2009, ICD-9-CM cannabis use disorder was more common among veterans in VHA care in states with medical marijuana laws (4). More recent, representative information is needed.

Data from the 2012–2013 National Epidemiologic Survey on Alcohol and Related Conditions–III (NESARC-III) can be used to examine the prevalence and correlates of cannabis use and use disorder among U.S. veterans prior to recreational cannabis sales (1), providing important baseline data for understanding veterans’ cannabis use and use disorder prior to the additional influence of recreational marijuana laws. NESARC-III data further afford the opportunity to examine an array of sociodemographic and psychiatric correlates of use, including DSM-5 psychiatric and substance use disorders. We analyzed NESARC-III data to assess the following among veteran respondents: 1) the prevalence of past-12-month and lifetime cannabis use, frequent use, and cannabis use disorder; 2) the demographic, substance use disorder, and psychiatric disorder correlates of these outcomes; and 3) differences in prevalence of cannabis use and use disorder between veterans in states with and without medical marijuana laws.

METHODS

Samples and Procedures

The NESARC-III target population included U.S. noninstitutionalized civilians, at least 18 years of age, living in households and group quarters, including group homes and workers’ dormitories (29, 30). Respondents were selected using multistage probability sampling, including primary (counties or groups of contiguous counties), secondary (Census-defined blocks), and tertiary sampling units (households within secondary sampling units), with oversampling of Black, Asian, and Hispanic respondents. Interviews were conducted from April 2012 to June 2013. Data were adjusted for nonresponse and weighted to represent the U.S. population (31). Weighting adjustments adequately compensated for nonresponse (30). All respondents gave informed consent and received $90 for participation. The institutional review boards of the National Institute on Alcohol Abuse and Alcoholism and Westat approved survey procedures. The overall response rate was 60.1%, comparable to other contemporaneous U.S. national surveys (32, 33). The NESARC-III sample included 36,309 participants. The present study used data from the 3,119 respondents (8.6% of sample) who reported that they had ever served on active duty in the U.S. Armed Forces, Military Reserves, or National Guard (excluding training only, including activation) and were no longer on active duty at the time of data collection.

Assessments

Substance use and use disorders.

The Alcohol Use Disorder and Associated Disabilities Interview Schedule–5 (AUDADIS-5) was the assessment measure of conditions in the past-12-month and lifetime (ever) time frames (34), which were not mutually exclusive. Nonmedical cannabis use was defined as use without a prescription or other than prescribed (e.g., to get high) (16). Medical use was assessed by asking whether or not respondents had received prescriptions for or used “medical cannabis” ever and in the past 12 months (35). Any nonmedical cannabis use was defined as at least one use within the past 12 months. Daily or near daily nonmedical use was defined as using cannabis five to seven times per week on average over the past 12 months. Consistent with AUDADIS procedures, only those endorsing nonmedical cannabis use were assessed for 12-month and lifetime cannabis use disorder, which required at least two of 11 criteria within the past 12 months. DSM-5 12-month and lifetime opioid (related to nonmedical use of prescription opioids), alcohol, tobacco, and other drug use disorders were assessed and defined similarly. Test-retest reliability of 12-month and lifetime cannabis use was found to be substantial (kappa values, 0.78 and 0.77, respectively) in a general population sample (36). Test-retest reliabilities of 12-month and lifetime cannabis use disorder (kappa values, 0.41 and 0.41, respectively) and respective criteria scales (intraclass correlation coefficients, 0.70 and 0.71, respectively) were fair to substantial in a subsample of NESARC-III participants (N=1,006) (34). Procedural validity was assessed through blind clinician reappraisal using the semistructured, clinician-administered Psychiatric Research Interview for Substance and Mental Disorders, DSM-5 version (PRISM-5) (37) in a separate NESARC-III subsample (N=712). Concordance between the AUDADIS-5 and the PRISM-5 was moderate for cannabis use disorder (kappa values, 0.60 and 0.51) and substantial for its dimensional criteria scale (intraclass correlations, 0.79 and 0.78) (38). Concordance between the AUDADIS-5 and the PRISM-5 for alcohol and tobacco use disorders and corresponding criteria scales was fair to substantial (kappa values, 0.36–0.66; intraclass correlations, 0.68–0.91) (38).

DSM-5 psychiatric disorders assessed using the AUDADIS-5 included mood disorders (primary major depressive disorder, dysthymia, bipolar I and bipolar II disorders) and anxiety disorders (panic, agoraphobia, social and specific phobias, generalized anxiety disorder). Primary mood and anxiety diagnoses excluded substance-induced and medically induced disorders. PTSD and schizotypal, borderline, and antisocial personality disorders were also assessed. Test-retest reliability and validity of the diagnoses were fair to substantial (kappa values, 0.40–0.87) in the NESARC-III reliability subsample (38, 39).

Sociodemographic characteristics.

Sociodemographic characteristics included sex, age, race/ethnicity, education, marital status, employment, family income, urbanicity, region, military service era, and residing in a state with medical marijuana laws by 2012. Medical marijuana law status was based on evaluations by legal and economic experts, as described elsewhere (40).

Statistical Analysis

Weighted prevalence estimates of nonmedical and medical cannabis use and cannabis use disorder were calculated for the full veteran sample. Adjusted prevalence estimates and adjusted odds ratios from multivariable logistic regressions were used to test the associations between sociodemographic characteristics and any nonmedical cannabis use (i.e., with or without medical use) and cannabis use disorder, adjusting for all other variables. Similar logistic regressions were used to test associations between nonmedical cannabis use and cannabis use disorder and each of the substance use and psychiatric disorders examined, adjusting for sociodemographic characteristics. Only 12 veterans reported medical cannabis use without nonmedical use. Cannabis use disorder was not assessed in these participants, so the correlates of their cannabis use were not examined. Finally, logistic regression models that were run on the full veteran sample included state medical marijuana laws to examine whether any past-12-month nonmedical, medical and nonmedical, or nonmedical only use or use disorder differed by state legalization status after adjusting for demographic covariates and were rerun after adjusting for additional state characteristics, including percent male, White, <30 years of age, and ≥25 years of age without a high school diploma. Analyses were performed using SUDAAN (41), which adjusts standard errors for complex survey designs using Taylor series linearization. Odds ratios were considered statistically significant when 95% confidence intervals excluded 1.00.

RESULTS

Demographic Characteristics and Prevalence of Cannabis Use and Cannabis Use Disorder

Veteran respondents were primarily White (79.5%), male (90.2%), ≥45 years of age (81.2%), and married or cohabitating (67.6%) (Table 1). Many were retired, in school, or disabled (53.1%). Most resided in urban areas (75.0%); 29.6% resided in states with medical marijuana laws by 2012.

The prevalence of 12-month cannabis use and cannabis use disorder was 7.3% and 1.8%, respectively (Table 2). Lifetime use and cannabis use disorder prevalence was 32.5% and 5.7%, respectively. Among past-12-month users, 15.2% endorsed any medical use and 96.0% reported nonmedical use. Among past-12-month users, 84.9% endorsed nonmedical use only, 4.0% endorsed medical use only, and 11.2% endorsed both medical and nonmedical use. The prevalence of 12-month cannabis use disorder among past-12-month users was 24.4%. Among those endorsing lifetime use, nearly all veterans endorsed nonmedical use only (95.5%).

Correlates of Past-12-Month Nonmedical Cannabis Use and Cannabis Use Disorder

Table 3 lists the prevalences of 12-month nonmedical cannabis use and cannabis use disorder by sociodemographic characteristics and adjusted odds ratios indicating associations between each characteristic and cannabis use and use disorder. Men had higher odds of cannabis use than women (adjusted odds ratio=2.2, 95% CI=1.2, 3.9), and those in the 18- to 29-year age range (adjusted odds ratio=9.0, 95% CI=4.1, 19.9), the 30- to 44-year age range (adjusted odds ratio=8.8, 95% CI=4.7, 16.3), and 45- to 64-year age range (adjusted odds ratio=4.9, 95% CI=3.0, 8.2) had higher odds of use than those ≥65 years of age. Compared with married veterans, those who were widowed, separated, or divorced (adjusted odds ratio=2.3, 95% CI=1.5, 3.4) or never married (adjusted odds ratio=2.7, 95% CI=1.7, 4.3) had higher odds of cannabis use. Compared with those with incomes ≥$70,000, those in the income categories ≤$19,999 and $20,000–$34,999 had higher odds of cannabis use (adjusted odds ratios, 2.4 [95% CI=1.3, 4.4] and 1.8 [95% CI=1.0, 3.3], respectively).

Similarly, men had higher odds of cannabis use disorder than women (adjusted odds ratio=10.6, 95% CI=2.7, 42.1). Those in the 30- to 34-year age group (adjusted odds ratio=2.8, 95% CI=1.3, 6.0) had greater odds of cannabis use disorder than those in the 45- to 64-year age group; prevalence among those ≥65 years of age was too sparse for comparison. Compared with married veterans, widowed, separated, or divorced veterans had greater odds of cannabis use disorder (adjusted odds ratio=2.8, 95% CI=1.5, 5.0), and those with incomes in the ≤$19,999 category had higher odds of cannabis use disorder than those in the ≥$70,000 category (adjusted odds ratio=5.0, 95% CI=1.4, 18.2).

Table 4 lists the prevalences of 12-month nonmedical cannabis use and cannabis use disorder by psychiatric and other substance other disorders and adjusted odds ratios indicating associations between these disorders and cannabis use and use disorder. All substance use and psychiatric disorders were associated with cannabis use (adjusted odds ratios, 2.0–12.6) except 12-month PTSD. All substance use and psychiatric disorders were associated with cannabis use disorder (adjusted odds ratios, 2.1–5.9) except past-12-month PTSD, opioid use disorder, and other drug use disorders.

Correlates of Lifetime Nonmedical Cannabis Use and Cannabis Use Disorder

Associations between lifetime cannabis use and cannabis use disorder and sociodemographic correlates were similar to 12-month associations (see Table S1 in the online supplement). The only difference was related to income: participants in all lower income categories had higher odds of lifetime use disorder than those in the ≥$70,000 category (adjusted odds ratios, 2.1–2.5).

Associations with lifetime nonmedical cannabis use and cannabis use disorder and substance use and other psychiatric disorders were also similar to 12-month findings, although associations with PTSD and opioid use disorder varied slightly (see Table S2 in the online supplement). Lifetime cannabis use was associated with all substance use disorders and psychiatric disorders examined, including PTSD (adjusted odds ratios, 1.5–9.0). Lifetime cannabis use disorder was associated with all disorders examined, including lifetime PTSD and opioid use disorder (adjusted odds ratios, 2.8–7.6).

Past-12-month Cannabis Use and Cannabis Use Disorder by State Medical Marijuana Law Status

In states with legalized medical marijuana, the prevalence and odds of any cannabis use, daily or near daily use, nonmedical use, medical use, medical and nonmedical use, and use disorder were significantly higher than in other states (adjusted odds ratios, 1.6–4.4) (Table 5).

DISCUSSION

Among veteran respondents in the NESARC-III survey, over 7% had used cannabis during the previous 12 months, and 1.8% met criteria for cannabis use disorder. Corresponding lifetime prevalence estimates were 32.5% and 5.7%, respectively. Approximately 1 in 4 (24.4%) who used nonmedical cannabis in the past 12 months met criteria for 12-month cannabis use disorder, and more than 1 in 6 (17.4%) who reported lifetime nonmedical use met criteria for lifetime cannabis use disorder. Sociodemographic correlates of nonmedical use and cannabis use disorder included being younger, being male, being unmarried, and having a lower income. Cannabis use and cannabis use disorder were associated with most psychiatric and substance use disorders examined. Finally, veterans in states with legalized medical cannabis were more likely than those in other states to use cannabis nonmedically and to have cannabis use disorder.

Two previous studies addressing cannabis use or cannabis use disorder among veterans in VHA care with mental health needs (10, 11) reported higher prevalence of cannabis use than we found in the present study, perhaps owing to the association of cannabis use with psychiatric disorders (10, 11). A study utilizing 2009 VHA-wide electronic medical record data (4) found a prevalence of past-year cannabis use disorder among VHA-enrolled veterans ∼70% lower than we found using the 2012–2013 NESARC-III data. This may reflect national increases in cannabis use disorder (3), or it may be because VHA providers do not routinely assess cannabis use disorder, underestimating prevalence in the medical record. Veterans in VHA care tend to be of lower socioeconomic status, a risk factor for cannabis use disorder, and many are diagnosed with substance use and psychiatric disorders, characteristics associated with nonmedical cannabis use and use disorder in the present study and in previous research (17, 42–46). Given VHA patient characteristics and national increases in cannabis use disorder (3, 6), cannabis use and use disorders may be more common among VHA patients now than in 2009. Because the VHA serves approximately 6 million veterans a year (47), understanding VHA clinical and treatment needs related to cannabis use is of considerable national importance. Therefore, updated, comprehensive studies of cannabis use and use disorder prevalence in VHA patients are warranted.

In the only other study of cannabis use among veterans in a nationally representative survey (13), the prevalence of 12-month cannabis use in 2014 was slightly higher (9%) than in our study (7.3%). Interviewer administration of the NESARC-III surveys may have contributed to lower rates compared with the previous survey, which was self-administered. However, the previous study did not examine lifetime prevalence or DSM-5 substance use and psychiatric disorders, all of which are important in understanding the distribution of cannabis use among U.S. veterans. Results from the present study are more easily compared with findings from the full NESARC-III sample, in which the prevalence of cannabis use and use disorder was also higher (16) than in our veteran subsample. Study findings provide an important benchmark of veteran status relative to nonveterans prior to implementation of recreational marijuana laws. Whether such trends continue with increasing legalization will be an important focus for future study.

Identified demographic correlates of cannabis use and cannabis use disorder were similar to those found in previous veteran and civilian studies (13, 17), as were associations between substance use and other psychiatric disorders and cannabis use and use disorder (13, 17, 19, 42). Importantly, our study extends earlier findings by examining DSM-5 diagnoses. Taken together, these findings add to evidence suggesting that certain individuals are consistently at increased risk for cannabis use and use disorder, while highlighting that veterans share risk factors with the general population. The demographic composition of the veteran population (i.e., predominantly male, with a bimodal age distribution including older veterans from earlier wars and younger veterans of recent conflicts) (48), and findings indicating that veterans experience disproportionately high levels of substance use and psychiatric disorders (46, 49) underscore the need for careful screening, assessment, and treatment of cannabis use disorder in clinical settings serving veterans, as well as ongoing monitoring of cannabis use and related functional and symptom outcomes. Screening may be particularly important among veterans with substance use and psychiatric disorders, such as PTSD, as cannabis use during treatment of these disorders is associated with worse outcomes (21, 22).

PTSD is a condition of high relevance to veterans, with 9%–13% of veterans meeting current criteria (23, 50). In this study, lifetime cannabis use and lifetime use disorder were significantly associated with lifetime PTSD. While past-12-month cannabis use and use disorder were not significantly associated with PTSD in a sociodemographically adjusted model, associations were in the same direction, with near-significant confidence intervals. Among 2001–2002 NESARC survey respondents, lifetime PTSD was significantly associated with cannabis use disorder and marginally associated with lifetime cannabis use (51). In the full NESARC-III sample, PTSD was significantly associated with past-12-month and lifetime cannabis use disorder (17). Among veterans with cannabis use disorder enrolled in the VHA, PTSD is the most common psychiatric comorbidity (4). Together, these findings show a relationship between PTSD and cannabis use as well as cannabis use disorder, with more consistent associations with cannabis use disorder. Despite a lack of evidence that plant cannabis is an effective treatment for PTSD, 25 of 36 states with medical marijuana laws include PTSD as an approved condition for medical use (1, 8, 20), and a study of post–9/11 veterans showed that PTSD was one of the leading reasons for seeking medical cannabis (14). Given the relevance of PTSD to veterans, further efforts are needed to increase our understanding of the relationship between PTSD and cannabis use and use disorder, including the sequence of their occurrence and whether the association between PTSD and cannabis use differs by reasons for use (e.g., medical-only versus medical and recreational), which was not examined in this cross-sectional study.

The current U.S. opioid crisis has also focused attention on the role of cannabis in increasing or decreasing opioid use and adverse outcomes (40, 52). Our findings suggest that veterans engaging in nonmedical cannabis use have increased odds of opioid use disorder, findings that are consistent with a large study of nonveterans showing that cannabis use predicted higher risk of nonmedical opioid use and use disorder among those with moderate to severe pain than among those without pain (52). Taken together, these studies add to evidence that advocating cannabis legalization to remediate the opioid crisis is premature.

Similar to past civilian and veteran studies (4, 16, 25, 28), we found higher odds of cannabis use and use disorder in states with medical marijuana laws, including odds almost five times higher for daily or near daily cannabis use. Because frequent use increases risk for negative consequences (e.g., cannabis use disorder, respiratory symptoms, intoxication-related injury, psychotic symptoms) (16, 17, 53–55), this finding is of particular concern. Further, at the time of NESARC-III data collection, 16 states had medical marijuana laws, but none had enacted recreational marijuana laws. Currently, 17 states and the District of Columbia (all of which previously had medical marijuana laws) have recreational laws. Because recreational laws further increase adult nonmedical cannabis use, frequent use, and cannabis use disorders (27), continued study of how cannabis use and cannabis use disorder prevalence change among veterans as legalization expands will be important, as will efforts to examine other conditions potentially affected by marijuana laws, such as opioid and psychiatric medication use. Overall, our study results highlight the importance of careful clinical screening, assessment, and monitoring of veterans as additional states pass permissive marijuana laws and continued consideration of policies guiding veterans and providers, particularly within the VHA, a federal system that currently defines cannabis as a Schedule I controlled substance.

Our study has some limitations. NESARC-III relied on self-report, which is subject to bias. Data were cross-sectional; thus, causality cannot be determined. Assessment of medical cannabis use relied on a single question, and it did not differentiate between use recommended by a physician or for self-determined reasons and thus may have been interpreted differently in states with and without medical marijuana laws. Cannabis use disorder was not assessed among those reporting medical use only, but should be in future surveys. AUDADIS-5 interviewers were not clinicians, although a validation substudy utilizing clinicians revealed nearly identical past-year prevalence (39). The relationship of our study findings to veterans treated in the VHA is unknown, as NESARC-III did not assess whether respondents were receiving VHA care. Additionally, in this large exploratory study, corrections for multiple testing were not made because of the paucity of data examining prevalence and correlates of cannabis use and use disorder among veterans. Finally, our findings may underestimate cannabis use among veterans overall if many were excluded from the survey because of incarceration (56), institutionalization, or homelessness.

In summary, our findings provide important information to inform future studies of veterans and may help guide clinical services. A clear risk for cannabis use disorder was evident among veterans reporting nonmedical cannabis use, with up to 37% meeting criteria for a cannabis use disorder in their lifetime. The odds of cannabis use and use disorder were also higher among vulnerable veteran subgroups, including those of lower socioeconomic status and those with psychiatric and substance use disorders. Our results indicate that the odds of use, daily or near daily use, and use disorder are even higher among veterans in states with medical marijuana laws. Such findings highlight the importance of continued research examining the impact of changing marijuana laws on our nation’s veterans. These findings are also important to communicate to policy makers, to health care professionals who may need to consider cannabis screening and intervention services when caring for veterans, and to veterans themselves so they can be well-informed about potential risks and benefits of cannabis use.