Cannabis Use Associations with Adverse Psychosocial Functioning among North American College Students

Abstract

Introduction:

Adolescent-onset cannabis use (CU) (before age 18) is associated with multiple adverse psychosocial outcomes, but rates of CU peak between the ages of 18 and 22, coinciding with college matriculation. Whether CU among college-enrolled young adults is associated with similar psychosocial outcomes is poorly understood. In the present study, we examined relationships between CU and multiple psychosocial outcomes in North American college students.

Methods:

Data for this report come from N = 40,250 North American college students ages 18-to-25 years (mean age = 20.7 years, 69% female, 66% Caucasian) who participated in the Healthy Minds Study (HMS) 2016–17. HMS is a web-based annual survey querying multiple mental health, substance use, and psychosocial variables in representative student populations from 53 universities across North America. Student respondents were stratified in two groups based upon their self-report of past 30-day CU and compared on psychosocial variables.

Results:

Approximately 20% (n = 8,327) of student respondents reported past 30-day CU. After adjusting for socio-demographics, knowledge of campus services, and use of other drugs, the odds of depression (aOR = 1.3), suicidal thoughts and behaviors (aORs ~1.4–1.7), anxiety (aOR = 1.2), eating disorders (aOR = 1.2), and violence victimization (aOR = 1.4) were all higher for CU students. Additionally, CU students had higher rates of other drug use and lower rates of perceived supportive relationships.

Conclusion:

Our results indicated that CU is common among North American college students and associated with adverse psychosocial consequences across multiple domains. Based upon these findings, colleges should consider expanding educational, prevention, and early-intervention programs for students who use cannabis.

Introduction

Broad cannabis-related legislative changes over the past 25 years have made the use of cannabis for medical and recreational purposes legal in Canada and in many states in the United States (U.S.), leading to major societal shifts in attitudes and use patterns in North Americans (Hammond et al., 2020). The legal cannabis market in the U.S., worth $33.1 billion in 2021, is growing rapidly, with a projected increase to $84 billion by 2028 (Grandview Research, 2021). Perceptions that cannabis use (CU) is harmful to one’s physical and mental health have decreased among North Americans across age groups (Azofeifa et al., 2016). Additionally, cannabis remains the most commonly used federally-illicit drug in the U.S., with 7.5% (19.8 million) of individuals over the age of 12 years reporting past year use in 2013 and increasing CU rates among adults (Azofeifa et al., 2016). Considering shifting legislation, perceptions, and use patterns, it is increasingly important to identify groups of individuals at elevated risk for CU-related adverse health outcomes.

Prevalence rates of CU in ‘college age’ young adults are higher than in other age groups and have increased since 2009, reaching historically high levels in recent years (Schulenberg et al., 2020; University of Michigan, 2018). In 2019, 43% of full-time college students aged 19–22 reported CU at least once in the last year, with 26% reporting CU in the past 30 days (Schulenberg et al., 2020). In addition, CU patterns and modes of administration have also changed among young adults over the past 20 years (Hall & Degenhardt, 2007; UNODC, 2006). College students represent a distinct subgroup of young people undergoing a critical period of identity development (Evans et al., 2009). Additionally, the typical years of college education coincide with a stage of continuous maturation of brain regions involved in decision-making and emotion regulation (Gogtay et al., 2004), making them vulnerable to environmental insults from drug use (Hammond et al., 2014). The convergence of developmental vulnerabilities and high prevalence of CU during college makes this period an optimal time for intervention (Halladay et al., 2019).

There is compelling evidence showing associations between CU and poor social and health outcomes in adolescents and adults (Gorey et al., 2019). These associations include higher rates of depression (Horwood et al., 2012), anxiety disorders (Kedzior & Laeber, 2014), suicidality (Moore et al., 2007; Gobbi et al., 2019), and eating disorders (Krug et al., 2008). Other psycho-social outcomes associated with CU include victimization (Hyman & Sinha, 2009). Additionally, regular use of cannabis is associated with increased risk of initiating and using other drugs (Kandel, 2002) and developing substance use disorders in adulthood (Blanco et al., 2016). This association is present even in longitudinal studies adjusting for confounders (Hall & Lynskey, 2005). While there is growing evidence of poor physical and mental health outcomes associated with chronic recreational use of cannabis during adolescence, especially with use of high delta-9-tetrahydrocannabinol (Δ−9-THC) potency chemotypes, fewer studies have focused on college students (Davis et al., 2016; Dutra et al., 2018).

In the present study, we sought to characterize behavioral and psychosocial correlates of past 30-day CU among North American college students using population-representative data from over 40,000 undergraduates and graduate students who participated in the Healthy Minds Study.

Prior research suggests college students have changing CU risk trajectories, with those most at risk for long-term negative health outcomes not necessarily being heavy cannabis users or use cannabis at all at the beginning of college (Dennhardt & Murphy, 2013). Additionally, a report of life-time use included participants who had only used once in their lifetime, as well as those with distant use, both of which types of use would be less likely to impact current health outcomes. Based on this information about varying trajectories in CU among college students, we explored a more temporally close use (within the past 30 days) of cannabis. We hypothesized that CU in the past 30 days would be associated with poorer psychosocial and behavioral outcomes in college students. Further, we hypothesized that these CU-psychosocial associations would be seen across different psychosocial domains suggesting a general or ‘non-specific’ relationship between CU and poorer psychosocial outcomes.

Methods

Study procedures and participants

Data used in the present analysis are from the 2016–2017 Healthy Minds Study (HMS), a population-level, web-based survey examining mental health, substance use, and related behaviors in undergraduate and graduate students in 53 universities across the United States and one in Canada (total data set N = 53,760 student respondents from 54 universities) (The Healthy Minds Network, 2022). The HMS survey data used in this report were collected during the 2016–2017 academic year with surveys administered from September to December 2016. At each HMS university, a random sample of 4,000 students were invited by email to participate, except at smaller universities (<4000 student body) where all students were invited to participate. The invitation email described the study and contained a URL web-link that students could use to gain access to the anonymous online survey, administered by Qualtrics. All students who were invited to participate in the survey were asked to consent before starting the survey. Students who were over the age of 18 years and were currently enrolled at participating universities were eligible to participate in the study. There were no other inclusion or exclusion criteria. Schools with graduate students typically included both undergraduate and graduate students randomly sampled from those respective universities. Participating students did not receive payment for their study participation. To incentivize participation, all students who were invited to participate were informed that they were entered into a national sweepstakes where they would be eligible to win one of several prizes totaling $2,000. Students who did not initially complete the survey were engaged with up to four follow-up reminders over the 30-day data collection period. For the 2016–2017 HMS survey, the response rate was 23% with 53,760 students agreeing to participate. To minimize the effects of non-response bias, sample probability weights were constructed by the HMS research group based upon estimates for each type of student using administrative data on gender, race/ethnicity, academic level, and grade-point average from the full student population. The HMS was approved by the Institutional Review Boards at University of Michigan (the main study coordinating center) and at all participating university sites. To additionally protect participants’ privacy, the HMS is covered by a Certificate of Confidentiality from the National Institutes of Health.

Participants included in the present analysis were 40,250 young adult students between the ages of 18 to 25 years with complete HMS survey data. Participants older than 25 years of age (n = 12,368), and those who refused to complete the survey (n = 1,142) were excluded. Participants who endorsed past 30-day CU (cannabis users) comprised 20.69% (n = 8,327) of all respondents.

Measures

Cannabis use assessment

The main variable of interest was self-reported past-30-day cannabis use, which was assessed using the question: “Over the past 30 days, have you used marijuana?” From the responses, a binary variable was constructed to examine cannabis users against non-users.

Psychosocial functioning and behavioral health assessments

Our analyses focused on eight binary measures of psychosocial functioning and mental/behavioral health, including depression; anxiety disorders; eating disorders; suicidal behaviors including ideation, planning, and attempts; and one ordinal variable measuring the supportiveness of social relationships.

Depression was measured using the Patient Health Questionnaire 9 (PHQ-9), which scores each of the 9 Diagnostic Statistical Manual IV (DSM-IV, APA) criteria for depression as “0” (not at all) to “3” (nearly every day). The total score can range from 0 to 27. We created a binary variable, and classified respondents as “depressed” and “not depressed” using a cut off score of ≥10 in the PHQ-9. The PHQ-9 score ≥10 has shown a sensitivity of 88% and a specificity of 88% for major depression. PHQ-9 scores of 10 represent moderate depression. Higher scores represent moderately severe (between 15 and 19) and severe depression (≥ 20). (Kroenke et al., 2001).

Suicidal ideation was assessed by asking the participants “in the past year, did you ever seriously think about attempting suicide?” (where 0 is “no” and 1 is “yes”). Participants responding “yes” to this question were asked about a suicide plan and suicide attempts with two questions: “In the past year, did you make a plan for attempting suicide?” (where 0 is “no” and 1 is “yes”); and “In the past year, did you attempt suicide?” (where 0 is “no” and 1 is “yes”).

Anxiety was measured using the Generalized Anxiety Disorder-7 (GAD-7), a 7-item questionnaire (Spitzer et al., 2006) reflecting the DSM-IV symptom criteria for GAD. It asks how often in the last two weeks the participant was bothered by each symptom. The responses are “not at all,” “several days,” “more than half the days,” and “nearly every day,” and they are scored 0, 1, 2, and 3, respectively. The internal consistency of the GAD-7 was excellent (Cronbach alpha = .92). Test-retest reliability was also good (intraclass correlation = .83), indicating good procedural validity. We created a binary variable using a cut off score of 10 in the GAD-7 for Generalized Anxiety Disorder. Cut off scores of 7–10 have shown a sensitivity of .83 and a specificity of 0.84 (Plummer et al., 2016).

Eating disorders were measured with the SCOFF (Luck et al., 2002), a brief tool designed to detect anorexia and bulimia nervosa. The SCOFF consists of 5 questions related to aspects of eating disorders, such as making oneself “sick because you feel uncomfortably full,” worrying about losing “control over how much you eat?,” having lost weight in the past 3 months, believing “to be fat when others say you are too thin?,” or saying that “food dominates your life.” Each item is answered as “yes” or “no.” Scores range from 0 to 5, with a score of ≥ 2 indicating a positive screen. The SCOFF has shown excellent validity and reliability in prior studies (Luck et al., 2002). Based on a cut off of 2, the sensitivity of the SCOFF is 53.7% and the specificity is 93.5% (Solmi et al., 2015).

Victim of violence was assessed with the question: “Over the past 12 months, have you experienced emotional, physical, or sexual abuse (either from someone you know or don’t know)?” Respondents answered either “no” or “yes.”

Quality of social relationships measured using one item on the Flourishing Scale (Diener et al., 2010). The item asks young adults to respond to the following statement “My social relationships are supportive and rewarding.” Responses were made on a 7-point scale with the following options: “strongly disagree” (1), “disagree” (2), “slightly disagree” (3), “mixed or neither agree nor disagree” (4), “slightly agree” (5), “agree” (6), “strongly agree” (7). The scale was reverse coded for analysis.

Covariates

We controlled for seven socio-demographic characteristics, including age; sex at birth (female, male and intersex); race (White, Black, Asian, Hispanic and Multi-racial or Other Races); sexual orientation (heterosexual, LGBTQ); U.S. citizenship (no, yes); relationship status (single, in a relationship, married, in a domestic partnership, engaged, divorced or separated, widowed/other), financial stress as measured by asking “How would you characterize your current financial situation right now?” (always stressful, often stressful, sometimes stressful, rarely stressful or never stressful).

We also controlled for five academic characteristics including whether the student was first generation to attend college in his or her family (no, yes), grade point average (A, B, C, D or below, no grade or unknown), year in school (1^st, 2^nd, 3^rd, 4^th, 5^th or more); location of university (Northeast USA, South USA, Midwest USA, West USA, Canada), and knowledge of campus services accessible to the student (no knowledge, some knowledge). Finally, we controlled for other substance use:

Binge alcohol consumption assessed using the National Institute on Alcohol Abuse and Alcoholism (NIAA) definition of binge drinking which states: “A ‘binge’ is a pattern of drinking alcohol that brings blood alcohol content to about 0.08 gram-percent or above. For the typical adult, this pattern of drinking alcohol corresponds to 5 or more drinks (male), or 4 or more drinks (female), in about two hours” (NIAAA, 2004). Consistent with Cranford et al. (2006) who adapted the NIAAA definition into a survey question, respondents were asked: “Over the past 2 wk, about how many times did you have 4 (female) or 5 (male) or more alcoholic drinks in a row?” Responses were made on a 6-point scale with the following options: none (1), once (2), twice (3), 3 to 5 times (4), 6 to 9 times (5), and 10 or more times (6). A drink was defined as a glass of wine, a bottle of beer or wine cooler, or a shot of liquor straight or in a mixed drink. A gender-specific measure of past-2-week binge drinking was constructed by recoding scores on the frequency of binge drinking item, so that participants who reported at least 1 binge drinking episode in the past two weeks were classified as “binge drinkers.”

Cigarette use was assessed with a single question that asked respondents: “Over the past 30 days, about how many cigarettes did you smoke per day?” Responses were made on a 5-point scale with the following options: zero cigarettes (1), less than one cigarette (2), one to five cigarettes (3), about one-half pack (4), one or more packs (5). A measure of past 30-day cigarette use was constructed by recoding scores, so that participants who reported any cigarette use in the past 30 days were classified as “smokers.”

Lastly, we used a single question that asked respondents: “Over the past 30 days, have you used any of the following drugs?” Respondents could select all that applied from the following list: cocaine (including crack, powder, or freebase), heroin, methamphetamines, stimulants (i.e, use of prescription stimulants such as Ritalin or Adderall without a prescription or more than the prescribed amount), and ecstasy. From these responses a binary variable for each drug was computed with “No” (0) or “Yes” (1).

Statistical analysis

For our analyses, we examined associations between past-30-day CU and several measures indexing different domains of psychosocial functioning and behavioral health among young adults. Prior to the main analysis, a multiple imputation procedure was used to address missing data which has been shown to be superior to traditional methods when data are missing due to random reasons unrelated to observed or non-observed variables (Cox et al., 2014). Using Rubin’s (2009) guidance, five imputations were completed and pooled for all the analyses conducted in this study.

We computed chi-square statistics and associated pvalue for the independent variables against CU. To test our assumptions regarding the association between CU, psychosocial functioning, and health, we examined bivariate relationships, reporting unadjusted odds ratios. We then assessed for potential multicollinearity among independent variables by calculating correlation coefficients, variance inflation factor (VIF) and tolerance statistics. These analyses revealed no potential issues with multicollinearity among independent variables. Correlation matrices only show bivariate relationships, and multicollinearity is fundamentally a multivariate issue. For this reason, variance inflation factor (VIF) and tolerance statistics are more reliable indicators of multicollinearity because they consider the multivariate relationships among the variables. A VIF of greater than 5 is considered to indicate high multicollinearity, indicating standard errors for the coefficients of these variables are inflated, and may lead to issues with significance testing. None of the independent variables had VIFs greater than 1.30, including age and year in academic program. Likewise, tolerance, a measure of the impact of one variable on all the other independent variables, had no calculated values below .80, where values less than 0.1 may indicate issues with multicollinearity. We also re-ran the analysis with all covariates except for years in program, and then again with all covariates except for age. The removal of these variables did not alter the study results. The re-run analysis excluding variables, along with the correlation matrix, and the calculation of collinearity statistics indicate that the potential for multicollinearity was addressed in this analysis.

Finally, we estimated logistic regression models for suicide ideation, suicide planning, suicide attempts, depression, anxiety, eating disorders, insomnia, and victim of violence. An ordinal regression model was estimated for quality of social relationships. Analyses were conducted using SPSS v.26 and R 4.1.2. Adjusted odds ratios (aOR) were presented from the logistic and ordinal regression models. The full models for each dependent variable, correlation matrix of independent variables, and calculations of VIF and tolerance statistics are included in the online supplemental materials.

Results

Socio-demographics and substance use behaviors in CU and Non-CU college students

The rates of past-30-day CU among the 40,250 college students was 20.7%. There were statistically significant differences in CU by age (older students reported slightly lower CU), sex at birth (23.8% male vs. 19.3% female users), and race (21.6% White, 16.6%, Black 10.1%, Asian 10.1%, Hispanic 24.6%, and Multi or Other 25.6% users). LGBTQ + students (30.8% users) used cannabis more frequently compared to heterosexual students (18.6% users). U.S. citizens (21.5% users) had a higher CU than nonresidents (8.8%). Those “in a relationship” (21.4% users), “widowed” (33.3% users) or “other” (34.2% users) reported higher CU than those who were single (20.6% users), married (12.1% users), or divorced/separated (19.3% users). College students whose financial situation is “always” (26.2% CU) or “often stressful” (23.2%) were more likely to be cannabis users than those who reported being in a “sometimes” (18.7% users), “rarely” (18.6% users), or “never stressful” (18.3%) financial situation (Table 1).

Table 1.

Socio-demographic and academic characteristics of the sample (N = 40,250) comparing cannabis users to non-users.

	Cannabis Non-Users (N = 31923)		Cannabis Users (N = 8327)
Sociodemographic and Academic Characteristics	N	% or $\overline{\text{X}}$	N	% or $\overline{\text{X}}$	χ2	p
Age (range 18 – 25)	31923	20.73	8326	20.65	45.53	.0006
Sex at Birth
Male	9462	29.64	2949	35.41
Female	22428	70.25	5374	64.54
Intersex	33	.10	4	.05	104.91	<.0001
Race
White	20938	65.59	5784	69.46
Black	1743	5.46	341	4.10
Asian	4056	12.71	455	5.46
Hispanic	1464	4.59	479	5.75
Multi or Other	3722	11.66	1268	15.23	434.12	<.0001
Sexual Orientation
Heterosexual	27197	85.20	6223	74.73
LGBTQ+	4726	14.80	2104	25.27	513.13	<.0001
U.S. Citizen (Ref = No)
Yes	29495	92.39	8093	97.19	24.5.92	<.0001
Relationship Status
Single	17422	54.58	4508	54.14
In a relationship	12737	39.90	3471	41.68
Married	1472	4.61	203	2.44
Divorced/Separated	25	.08	6	.07
Widowed	4	.01	2	.02
Other	263	.82	137	1.65	125.86	<.0001
Financial Situation
Always stressful	3686	11.55	1309	15.72
Often stressful	7709	24.15	2325	27.92
Sometimes stressful	11828	37.05	2714	32.59
Rarely stressful	6459	20.23	1477	17.74
Never stressful	2241	7.02	502	6.03	197.12	<.0001
First Generation Student (Ref = Yes)	28093	88.11	7494	90.01	23.56	<.0001
No
Grade Point Average
A	15934	49.91	3418	41.05
B	10934	34.25	3355	40.29
C	2286	7.16	854	10.25
D or below	155	.49	74	.89
No grade or unknown	2614	8.19	626	7.51	279.49	<.0001
Year in School
1st year	9638	30.19	2194	29.40
2nd year	8374	26.23	2144	26.13
3rd year	7227	22.64	2002	22.92
4th year	5518	17.29	1597	17.68
5th year or more	1166	3.65	390	3.87	71.00	<.0001
Location of University
Northeast USA	3016	9.45	775	9.30
South USA	7967	24.96	1858	22.31
Midwest USA	11414	35.75	2248	26.30
West USA	8462	26.51	3186	38.26
Canada	1064	3.33	260	3.12	484.46	<.0001
Services Knowledge (Ref = No Knowledge)
Some Knowledge	24938	78.11	6547	78.62	.98	.3205

χ² = Chi-square.

Additionally, first-generation students (20% reporting CU) were significantly less likely to report CU than those who had parents who previously attended college (21.0% reporting CU). Students who had an “A” grade point average (17.7% reporting CU) and those who reported having “no grade or unknown” (19.3% reporting CU) were also less likely to be use cannabis than those with lower grade point averages (“B” 23.5%; “C” 27.2%; “D or below” 32.3% reporting CU). The proportion of the student body who reported past 30-day CU significantly increased in each successive year at college (1^st year 18.5% users; 2^nd year 20.4% users; 3^rd year 21.7% users; 4^th year 22.4% users; 5^th year or more 25.1% users). Students who attended universities located in the Northeast (20.4% reporting CU), South (18.9% reporting CU), and Midwest (16.4% reporting CU) of the U.S. and the university in Canada (19.7% reporting CU) reported using less cannabis compared to students in universities in the West Coast (27.3% with CU) of the U.S. There were no differences between the cannabis and non-cannabis users related to knowledge of available mental health services at their university. Regarding non-cannabis substance use, cannabis using college students compared to non-CU college students had higher rates of substance use of all other drugs of abuse (see Table 2). For example, students reporting CU had higher rates of a binge drinking episode in the past 2 wk (87% vs. 56%, p < 0.0001) and of past 30-day combustible cigarette use (26% vs. 6%, p < 0.0001), non-medical prescription stimulant use (10% vs. <1%, p < 0.0001), cocaine use (9% vs. <1%, p < 0.0001), ecstasy use, heroin use, and methamphetamine use (all p’s <0.0001) compared to students who reported no CU.

Table 2.

Drug use characteristics of the sample (N = 40,250) comparing cannabis users to non-users.

	Cannabis Non-Users (N = 31923)		Cannabis Users (N = 8327)
Drug Use Characteristics	N	%	N	%	χ2
Binge drinking episode, past 2 wk	17528	54.91	7281	87.44	2955.82*
Combustible cigarette use, past	1904	5.96	2202	26.44	3023.79*
30-days
Cocaine use, past 30-days	136	.43	708	8.50	2098.42*
Ecstasy use, past 30-days	30	.09	233	2.80	743.95*
Heroin use, past 30-days	5	.02	22	.26	60.86*
Methamphetamine use, past 30-days	7	.02	41	.49	122.71*
Non-medical prescription stimulant	215	.67	825	9.91	2237.23*
use, past 30-days

χ² = Chi-square.

^*p <.0001.

Associations between psychosocial functioning, behavioral health, and CU in college students

Unadjusted and adjusted OR (aOR) indexing variation in psychosocial functioning and behavioral health measure results as a function of CU status are presented in Table 3. In unadjusted univariate analyses, respondents in the CU group reported higher rates of depression (OR = 1.7, pvalue < .0001), anxiety (OR = 1.4, pvalue < .0001), eating disorders (OR = 1.4, pvalue < .0001), and violence victimization (OR = 2.0, pvalue < .0001) compared to the non-CU group. Respondents in the non-CU group were more likely to report strongly agreeing or agreeing that they have supportive relationships (OR = 1.2, pvalue < .0001).

Table 3.

Univariate and multivariate associations between cannabis use and dependent variables representing mental, physical, and social well-being.

		Cannabis Non-Users (N = 31923)		Cannabis Users (N = 8327)
Psychosocial and Behavioral Measures	% Total	N	%	N	%	OR	OR 95% CI	AOR	AOR 95% CI
Depression	32.58	9609	30.10	3506	42.10	1.69*	1.61 − 1.78	1.32*	1.24 − 1.40
Suicide Ideation	11.79	3223	10.10	1523	18.29	1.99*	1.87 − 2.13	1.54*	1.42 − 1.66
Suicide Planned	4.81	1295	4.06	641	7.70	1.97*	1.79 − 2.18	1.39*	1.24 − 1.57
Suicide Attempted	1.11	268	.84	179	2.15	2.60*	2.14 − 3.14	1.66*	1.31 − 2.09
Victim of Violence	6.69	1794	5.62	900	10.81	2.04*	1.87 − 2.21	1.36*	1.23 − 1.50
Anxiety	28.33	8525	26.70	2876	34.54	1.45*	1.38 − 1.53	1.19*	1.12 − 1.27
Eating Disorder	24.68	7445	23.32	2489	29.89	1.40*	1.33 − 1.48	1.20*	1.12 − 1.27
Supportive Relationships
Strongly Agree	21.83	7123	22.32	1662	19.96
Agree	39.80	1282	40.16	3200	38.43
Slightly Agree	19.48	6087	19.07	1755	21.08
Neither agree/disagree	9.89	3124	9.79	856	10.28
Slightly Disagree	4.75	1470	4.60	443	5.32
Disagree	2.49	741	2.32	260	3.12
Strongly Disagree	1.76	557	1.74	151	1.81	1.17*	1.12 − 1.22	1.10*	1.05 − 1.16

^*Denotes a p value < .0001; OR = Odds Ratio; AOR = Adjusted Odds Ratio and controlled for age, sex-at-birth, race, sexual orientation, citizenship status, relationship status, current financial stress, first generation student status, grade point average, year in degree program, location of University, knowledge of campus services, binge alcohol use, tobacco use, cocaine use, ecstasy use, heroin use, methamphetamine use, and stimulant use.

In adjusted univariate analyses controlling for age, sex-at-birth, race, sexual orientation, citizenship status, relationship status, current financial stress, first generation student status, grade point average, year in degree program, location of university, knowledge of campus services, binge alcohol use, cigarette use, cocaine use, ecstasy use, heroin use, methamphetamine use, and stimulant use, higher rates of self-reported suicide ideation (aOR = 1.5, pvalue < .0001), suicide planned (aOR = 1.4, pvalue <.0001), and suicide attempted (aOR = 1.7, pvalue < .0001), depression (aOR = 1.3, pvalue < .0001), anxiety (aOR = 1.2, pvalue < .0001), eating disorder (aOR = 1.2, pvalue <.0001), and violence victimization (aOR = 1.4, pvalue <.0001) were again observed in the CU group compared to the non-CU group, and higher rates of agreeing or strongly agreeing that their relationships were supportive (aOR = 1.2, pvalue < .0001) was again observed in the non-CU group compared to the CU group. Results from the multivariate logistic regression (see supplemental data Table S1) are consistent with the above findings.

Discussion

Using data from the HMS 2016–17 surveys, the present study explored psychosocial and behavioral health correlates of CU among North American college students. The main objective of this study was to investigate associations between past 30-day CU and psychosocial functioning and behavioral health outcomes in North American college students. Our results show that college-attending young adults who reported past-30-day cannabis use had higher rates of depression, anxiety, suicidal thoughts and behaviors, eating disorders, and violence victimization compared to college-attending young adults who do not use cannabis. College students who recently used cannabis also had higher rates of alcohol and other drug use and lower rates of perceived supportive social relationships compared to their non-using counterparts. These associations between cannabis use and poor psychological functioning and health persisted when controlling for age, sex-at-birth, first generation college student status, grade point average, year in degree program, location of the University, use of other substances, or knowledge of services in the university. The implications of these findings are discussed below.

Recent use of cannabis was common among college students in our sample, with one out of five respondents reporting past 30-day CU, and is consistent with other nationally representative samples of college-enrolled young adults (Schulenberg et al., 2020; University of Michigan, 2018). Age differences in our sample were statistically significant and showed a pattern of increased CU rates from 18 years of age with a peak at age 22 and a later decrease in use, consistent with earlier studies (Mauro et al., 2018). Interestingly, and contrary to our age-related findings, our analysis by grade showed that the relative proportion of students reporting recent CU increased with every grade level. This could reflect a cohort effect, especially given the divergence of grade-level from age-related findings.

The magnitude of cannabis-mood and cannabis-anxiety relationships identified in the present analysis are consistent with prior studies in adolescents and adults showing that CU is associated with poorer course of depression (Degenhardt & Hall, 2012; Hall and Degenhardt, 2009; Degenhardt et al., 2012; Horwood et al., 2012; Kedzior & Laeber, 2014) and anxiety disorders (Kedzior and Laeber, 2014). Our findings regarding suicidal thoughts and behaviors also converge with other studies showing increased suicidality related to cannabis exposure during adolescence (Gobbi et al., 2019; Moore et al., 2007) and suggest that cannabis-suicide associations extend beyond adolescence into young adulthood and are present in college student populations. In unadjusted analyses, we identified a relationship between CU and increased likelihood of having an eating disorder that remained significant after controlling for all other co-variates. Higher rates of eating disorders in those who use cannabis are consistent with existing evidence showing that eating disorders and substance use disorders commonly co-occur (Carr, 2007; CASA, 2003; Holderness et al., 1994; Krug et al., 2008; Salbach-Andrae et al., 2008). Rates of drug and alcohol use among individuals with eating disorders are 50% compared to the 9% prevalence found in the general population (CASA, 2003). The rates of having an eating disorder are also higher among individuals with a substance use disorder (over 35%) than in the general population (1–3%) (CASA, 2003; Krug et al., 2008).

Our results showed that college students in the CU group were more likely to report past year violent victimization compared to students in the no CU group. This finding is consistent with prior studies showing associations between CU and victimization in young people (Hyman & Sinha, 2009). Our finding showing increased rates of alcohol and other substance use in students who used cannabis is consistent with previous research (Hall & Lynskey, 2005; Kandel, 2002). This increase in rates of alcohol and substance use in CU students could be the result of common predispositional factors or emerge as a consequence of psychosocial or biological differences related to regular CU. For example, students who use cannabis have more access to other illicit drugs and are more likely to develop social relationships with and be influenced by peers who drink and use other drugs (Miech et al., 2017; Neighbors et al., 2008). Additionally, growing evidence suggests that repeated use of cannabis during adolescence and young adulthood may prime the midbrain dopaminergic system and shift the hedonic set-point of regular cannabis users, increasing the likelihood of using other drugs such as tobacco, opioids, and cocaine in the future (Volkow et al., 2019). While our results identified associations between CU and negative health in college students, it is important to note that these associations are complex. For example, CU may be associated with some areas of perceived positive social relationships, like the perception of “contributing to other people’s happiness,” a characteristic of social flourishing (Diener et al., 2010) in U.S. college students (Vidal et al., 2022), suggesting potential age-stratified differences on the psychosocial buffering effects of positive social relationships on CU-health associations that warrants further investigation.

While not the focus of the study, we identified distinct demographic student-level and university-level characteristics that differed as a function of CU status in college students. Regarding academic success, students who had higher grade point averages were less likely to use cannabis. This converges with other studies showing a relationship between CU and poorer academic and vocational outcomes. Arria et al. (2015) examined college students longitudinally, finding that CU at baseline had negative effects on class attendance and GPA of college students. At the university-level, geographic location of a university was associated with CU in our analyses, with students attending universities located on the West Coast of the U.S. having higher rates of past 30-day CU compared to students attending universities in other regions of the U.S. Regional differences warrant further exploration as they may relate to cultural acceptance and/or cannabis legalization laws.

Given the changing legal status of cannabis throughout North America, it is important to understand the health effects of CU and to identify vulnerable subgroups. Our findings converge with other studies and provide growing support that college-aged young adults represent a population that is at elevated risk for negative outcomes related to CU (Hammond et al., 2020). Based upon our findings showing associations between college student CU and worse psychosocial functioning and behavioral health across different health domains, prevention and treatment interventions in the college setting may be key to reducing long-term poor psychosocial outcomes. Campus-wide educational approaches have shown little success in decreasing CU in college students, but brief counseling interventions, including motivational interviewing and skill-based interventions, as well as parent-based approaches, have shown promise. Research to develop intervention components tailored to students who use drugs, both cross-cutting and tailored to specific drug types, need to be explored (Dennhardt & Murphy, 2013). College students overestimating normative levels of cannabis use and believing that cannabis is relatively benign also increases the risk of drug use. Compiling information about the possible adverse outcomes related to CU in the context of brief interventions may be useful (Elliot et al., 2011). Finally, harm reduction approaches that shift the intervention focus away from abstinence and toward reducing the risk for negative health outcomes related to cannabis such as preventing students from driving while intoxicated on cannabis (McCambridge & Strang, 2004) and interventions focused on decreasing college student use of high-THC potency cannabis products may improve health outcomes in this population.

This study has a number of important limitations and several notable strengths. Limitations of this study include the cross-sectional design, which does not allow for conclusions about causality. While the outcome measures used are validated for mental health disorders, they consisted of self-reports, contributing to recall bias. Likewise, CU and use of other substances may have been underreported for social desirability purposes, and the reports of use over the past 30 days may not be indicative of chronic use. Furthermore, there was no characterization of frequency of CU. Additionally, CU groups were based on a single self-report item and there was a lack of biochemical verification of substance use. At the same time, the study also has several relevant strengths, including the size and representativeness of the sample with a college student population and breadth of constructs measured.

Conclusion

Results from our study indicate that one out of five North American college students report recent CU, and that past 30-day CU in this population is associated with increased rates of depression, anxiety, suicidal thoughts and behaviors, eating disorder, violence victimization, and perception of less supportive social relationships. Given these findings, additional research, prevention messaging, and targeted interventions in college student populations are warranted especially given the changing cannabis policy landscape that has expanded availability and use of cannabis among college-enrolled young adults in the US.