Skip to main content
SearchLoginLogin or Signup

Typologies of Canadian young adults who drive after cannabis use: A two-step cluster analysis

Published onApr 28, 2022
Typologies of Canadian young adults who drive after cannabis use: A two-step cluster analysis


Young adults that drive after cannabis use (DACU) may not share all the same characteristics. This study aimed to identify typologies of Canadians who engage in DACU. 910 cannabis users with a driver’s license (17-35 years old) who have engaged in DACU completed an online questionnaire. Two-step cluster analysis identified four subgroups, based on driving-related behaviors, cannabis use and related problems, and psychological distress. Complementary comparative analysis among the identified subgroups was performed as external validation. The identified subgroups were: 1) frequent cannabis users who regularly DACU; 2) individuals with generalized deviance with diverse risky road behaviors and high levels of psychological distress; 3) alcohol and drug-impaired drivers who were also heavy frequent drinkers; 4) well-adjusted youths with mild depressive-anxious symptoms. Individuals who engaged in DACU were not a homogenous group. When required, prevention and treatment need to be tailored according to the different profiles.

Keywords: Automobile Driving, Canada, Cannabis, Driving Under the Influence, Risk-Taking, Typology


In 2020, among the 27% of Canadians who reported cannabis use in the past 12 months, 22% declared to have driven within two hours after using cannabis, a decrease from 26% in 2019 (Health Canada, 2020). Driving after cannabis use (DACU) could be a risky behavior, considering that delta-9-tetrahydrocannabinol (THC), the main psychoactive molecule in cannabis, can alter motor coordination, short-term memory, shared attention, concentration, reaction time, and time perception (Bondallaz, Favret, Chtioui, Fornari, Maeder et al., 2016; Capler, Bilsker, Van Pelt & MacPherson, 2017; Doroudgar, Mae Chuang, Bohnert, Canedo, Burrowes et al., 2018; Hartman & Huestis, 2013; Mikulskaya & Martin, 2018a, 2018b; Rogeberg & Elvik, 2016). Cognitive and psychomotor impairments could explain associations between recent cannabis use and a greater likelihood of collisions (Asbridge, Hayden, Cartwright, 2012; Bédard, Dubois & Weaver, 2007). Although DACU may not lead systematically to accidents and fatalities, 4,407 Canadians were physically injured and 75 died in a road accident involving cannabis in 2012 (Wettlaufer, Florica, Asbridge, Beirness, Brubacher et al., 2017).

Many sociodemographic and psychological characteristics have been associated with higher probabilities of engaging in DACU: age under 35 years old (Domingo-Salvany, Herrero, Fernandez, Perez, del Real et al., 2017; Voas, Lacey, Jones, Scherer et Compton., 2013), male gender (Arterberry, Treloar, Smith, Martens, Pedersen et al., 2013; Voas et al., 2013), high frequency of cannabis use (Arterberry et al., 2013; Arterberry, Treloar & McCarthy, 2017; Berg, Daniel, Vu, Li, Martin et al., 2018; Borodovsky, Marsch, Scherer, Grucza, Hasin et al., 2020; Cuttler, Sexon & Mischley, 2018; Matthews, Bruno, Dietze, Butler & Burns, 2014; Sukhawathanakul, Thompson, Brubacher & Leadbeater, 2019; Whitehill, Rodriguez-Monguio, Doucette & Flom, 2019), presence of cannabis-related problems or disorders (Choi, DiNitto & Marti, 2019; Cook, Shank, Bruno, Turner & Mann, 2017; Le Strat, Dubertret & Le Foll, 2015; Scherer, Voas & Furr-Holden, 2013), participation in other risky behaviors, e.g., driving under the influence of alcohol (DUI-A), speeding, and erratic driving (Bédard et al., 2007; Bingham, Shope & Zhu, 2008), and peer approval of engaging in DACU (Aston, Merrill, McCarthy & Metrik, 2016; Ward, Schell, Kelly-Baker, Otto & Finley, 2018). However, these risk factors may not be shared by all individuals who have DACU, as they constitute a heterogeneous population. Identification of subgroups is clinically relevant for tailoring awareness, prevention, and treatment programs to obtain the best outcomes.

As such, these aforementioned studies used variable-centered approaches, where associations of parameters were examined and assumed to be eventually representative of the population of interest, notably through analyses such as ANOVA, factor analysis, regression, and structural equation modeling (Howard & Hoffman, 2018). Instead, a person-centered approach recognizes heterogeneity in a given population and thus allows identifying meaningful and/or useful subpopulations of similar individuals (also known as typologies, clusters, or subgroups), within a sample according to a set of chosen variables, using cluster analyses, e.g., K-means, two-step cluster analysis, Ward’s method, and latent class analysis. (Howard & Hoffman, 2018; Tan, Steinbach, Karpatne & Kumar, 2019). Cluster analyses, therefore, are a collection of exploratory data-driven methods. Over the past 40 years, many studies identified diverse typologies of individuals who had been driving while impaired (DWI) on any substance, with some subgroups having been repeatedly found in the past decade. A first subgroup often labeled as “well-adjusted” or “normal”, presented normative scores in diverse personality features and low/mild psychiatric, criminal, or substance-related problems (Donovan & Marlatt, 1982; Moore, 1994; Nelson, Shoov, LaBrie & Shaffer, 2019; Nolan, Johnson & Pincus, 1994; Okamura, Kosuge, Kihira & Fujita, 2014; Saltstone & Poudrier, 1989; Scherer, Beck, et al., 2021; Scherer, Nochajski, Romano, Manning, Romosz et al., 2021; Shim, Wang & Bahk, 2016; Snowden, Nelson & Campbell, 1986; Steer, Fine & Scholes, 1979; Wells-Parker, Cosby & Landrum, 1986). A second group was mainly characterized by high levels of neuroticism, psychological distress, or psychiatric symptoms (Donovan & Marlatt, 1982; Hubicka, Källmén, Hiltunen & Bergman, 2010; Moore, 1994; Nelson et al., 2019; Roma, Mazza, Ferracuti, Cinti, Ferracuti et al., 2019; Saltstone & Poudrier, 1989; Scherer, Nochajski, et al., 2021; Shim et al., 2016; Steer et al., 1979). A third group was defined by frequent/heavy substance use or by substance-related problems (Ball, Jaffe, Crouse-Artus, Rounsaville & O’Malley, 2000; Hubicka et al., 2010; Okamura et al., 2014; Saltstone & Poudrier, 1989; Scherer, Beck, Taylor, Romosz, Voas et al., 2021; Scherer, Nochajski, et al., 2021; Snowden et al., 1986; Steer et al., 1979; Wells-Parker et al., 1986). A fourth group presented a general tendency towards deviant road behaviors, as expressed through speeding, driving-related aggression, poor driving history, risky driving, repeated DWI offenses; they were sometimes characterized as being aggressive, hostile, resentful towards authority, and non conforming (Donovan & Marlatt, 1982; Moore, 1994; Nelson et al., 2019; Roma et al., 2019; Saltstone & Poudrier, 1989; Scherer, Nochajski, et al., 2021; Shim et al., 2016; Wells-Parker et al., 1986; Wieczorek & Miller, 1992). Finally, some studies reported a subgroup with a severe presentation across all measured characteristics: criminality, mental health, substance use, and related problems, etc. (Okamura et al., 2014; Steer et al., 1979; Wells-Parker et al., 1986; Wieczorek & Miller, 1992).

All these aforementioned studies focused on alcohol or general substance use; none have examined cannabis specifically. In Canada, young drivers tend to hold unfavorable and negative views towards DUI-A, while perceiving lower risk levels associated with DACU (Goodman, Leos-Toro, Hammond, 2020). Different sets of risk factors were associated with DWI depending on the detected psychoactive substance in toxicological tests (Scherer, Harrell & Romano, 2015). Therefore, observations from studies on DUI-A may not be directly transposable to DACU. Furthermore, all of these studies were conducted among samples constituted of suspected or convicted DWI offenders, the vast majority of them being identified as males. As such, knowledge is scarce concerning typologies of cannabis users who have engaged in DACU, including females and those who have managed to avoid being caught. This study aimed at identifying subgroups of individuals who DACU based on their road-related behaviors, their cannabis use and associated problems, and their level of psychological distress. It also compared them according to previously identified demographic and psychosocial risk factors. Based on previous literature, it was expected to find the following subgroups: 1) individuals mainly defined by chronic or problematic cannabis use, 2) individuals that present diverse deviant road behaviors, 3) individuals with generalized and elevated problems related to cannabis use, mental health, and driving; 4) individuals with high psychological distress, and 5) “well-adjusted” individuals with low scores on all indicators.


2.1.1 | Study setting

This study was geographically limited to Canada, including its 10 provinces and three territories. In Paragraph 320.14 of the Criminal Code of Canada, an offence of “operation while impaired” occurs when someone, while operating or having care or control of conveyance (e.g., motor vehicle), has a blood alcohol concentration (BAC) equals or exceeding 80 mg of alcohol per 100 ml of blood, has a drug concentration equaling or exceeding the concentration prescribed by regulation, is impaired to any degree by alcohol and/or a drug, or fail or refuse to comply with a demand to perform physical coordination tests, to submit to an evaluation to determine drug impairment or to provide bodily fluids for alcohol/drug detection. As such, law enforcement can demand a driver with reasonable suspicion of impaired driving to complete a standardized field sobriety test. If the driver fails, the police officer can bring in a drug recognition expert for a detailed 12-step assessment or ask for a blood sample testing. They can also test the driver’s saliva for THC on the roadside using an approved device (Drager DrugTest 5000). For cannabis, there are three legal thresholds: 2 ng (but less than 5ng) of THC per 1 ml of blood (lower level offence), 5 ng/ml or more (serious level of offence), and 2.5nl/mL combined with a BAC of 50 mg/mL. Penalties vary according to alcohol or drug concentration, whether it is a first or a repeated offence and if the impaired driving caused bodily harm or death to another person. Additionally, each Canadian province can apply supplemental laws and can enforce stricter penalties.

2.1.2 | Study population

Participants were recruited through Facebook paid advertising between August 2018 and March 2019 to fill out an online questionnaire on LimeSurvey. The four inclusion criteria were: 1) age between 17 and 35 years old; 2) residing in Canada; 3) cannabis use in the past 12 months; 4) to possess a regular or a probationary driver’s license. The paid Facebook advertising presented the inclusion criteria and an infographic picture of a car with cannabis leaves in the car front headlights (as presented in Figure 1). Facebook thus targeted individuals with age and location data in their profile that matched our inclusion criteria. When participants clicked on the ad, they were immediately redirected on the survey’s website. The survey was available in both French and English. The survey was in French in a default mode, but the participant had the option to choose the appropriate language before starting the questionnaire.

[insert Figure 1 about here]

Consent was obtained for 2,270 participants. Cases were excluded if they had entire missing datasets (n=544) or if they were deemed as careless respondents (n=99), i.e., that they provided inconsistent or improbable answers throughout the questionnaire. For this present study, 717 individuals who had not engaged in DACU in the past 12 months were also excluded. The final analyzed sample was composed of 910 participants. This study was approved by the local institutional review board of an integrated university health and social services center (Centre intégré universitaire de santé et de services sociaux du Centre-Sud-de-l’Île-de-Montréal; approbation number DIS-1819-54). It was performed following the ethical standards outlined in the 1964 Declaration of Helsinki and its later amendments.

2.2 | Measures

2.2.1 | Clustering variables

Clustering variables were chosen based on theory, parsimony, and validity, i.e., the silhouette measure of cohesion and separation of clusters (Rousseeuw, 1987). Considering the exploratory nature of cluster analyses, testing different sets of variables was recommended to determine which clustering algorithm would produce the best solution (Kantardzic, 2019). As such, the following clustering variables were selected based on 1) the literature on DWI typologies as well as their meaningfulness for potential psychosocial interventions on modifiable cognitive and behavioral characteristics, and 2) their contribution in yielding the statistical solution with the most acceptable silhouette of cohesion and separation. In other words, these clustering variables had to allow sufficient differentiation between clusters, not be highly correlated between them, and present a reasonable ratio to the sample size (Sarstedt & Mooi, 2019). For this study, clustering variables were: 1) speeding; 2) DUI-A; 3) dangerous driving (a. negative emotions; b. aggressiveness; c. risking driving); 4) frequency of cannabis use; 5) cannabis use problems (CUDIT score); 6) psychological distress (a. anxiety; b. depression; c. irritability; d. cognitive problems). | Risky driving behaviors

Speeding was assessed through a 7-item Likert-type question “Do you speed?”, ranging from “never” to “always”. DUI-A was defined by the number of times in the last 12 months they drove within an hour of drinking alcohol above the legal limit. This variable was then dichotomized (yes/no) for the cluster analysis. The original version of the Dula Dangerous Driving Index (DDDI) assessed the frequency of risky and aggressive driving (Dula & Ballard, 2003). This 29 Likert-type questionnaire (1=never to 5=always) had good internal consistency (α=.88 to .92) and was composed of three subscales: aggressiveness (7 items; e.g., “I verbally insult drivers”; α=.74 to .83), risky driving (12 items; e.g., “I illegally pass a vehicle moving too slowly”; α=.76 to .86), and negative emotions (11 items; “I drive while I'm angry”; α=.79 to .80). For each subscale, the score was computed with the sum of respective items. A higher score denoted higher levels of dangerous driving. | Cannabis use variables

The frequency of cannabis use in the past 12 months was coded into three categories: 1) occasionally (i.e., a few times per month or less), 2) weekly or almost every week and 3) daily or almost every day. The Cannabis Use Disorders Identification Test (CUDIT), a 10-item Likert-type questionnaire measured cannabis-related problems according to DSM-IV criteria (Adamson & Sellman, 2003). The answers of the first eight items ranged from 0 ("never") to 4 ("almost every day") and the last two were dichotomous (yes/no) with a respective value of 4 and 0. The total score was computed with the sum of the items, a score above a threshold of 8 indicating the presence of cannabis-related problems (Thake & Davis, 2011). Internal consistency was considered as good (α=.84). | Psychological distress

The short version of the psychological distress scale of the Quebec Health Survey containing 14 items (IDPESQ-14) assessed the level of psychological distress defined by symptoms related to four subscales: depression, anxiety, irritability, and cognitive problems (Boyer, Préville, Légaré & Valois, 1993). This questionnaire was derived from the Psychiatric Symptom Index and had a Cronbach’s alpha of .89, indicating a good internal consistency (Ilfeld, 1976). Participants declared how often they experience each symptom on a scale of 1 (“never”) to 4 (“really often”) The mean of related items was used to compute subscales.

2.2.2 | External validation variables

The identified clusters were compared on a set of diverse variables not included in the clustering algorithm to validate them as distinct subgroups. This external validation also allowed further characterization or profiling of the subgroups in terms of sociodemographic factors, clinical features, behaviors and attitudes, etc. (Sarstedt & Mooi, 2019). They were the variables that could not have been included in the clustering algorithm because they did not allowed generating a fair to good silhouette of cohesion and separation (Rousseeuw, 1987). | Sociodemographic variables

Participants self-declared their age and their gender to which they identified, i.e., female, male, cultural gender (two-spirit, agowka, etc.), or other gender identities (gender-fluid, non-binary, etc.). Other sociodemographic variables included occupation (studying exclusively, working exclusively, both studying and working, other [unemployed, on disability leave, a stay-at-home parent, etc.]), and education level attainment (high school or less, college or vocational training, and university). | Alcohol use

Participants were asked about the frequency of their alcohol consumption. Possible responses ranged from never to twice or more per day. Those who responded at least two to three times per week were classified as “regular”, one drink per week or less as “occasional”, and no alcohol use as “never”. According to the Canadian low-risk alcohol drinking guidelines, binge drinking in the past 30 days was defined as the number of episodes when participants had drunk four (for women) or five (for men) standard drinks or more on the same occasion (Health Canada, 2021). | Impulsivity traits

The short version of the Urgency, Premeditation, Perseverance, and Sensation Seeking Impulsive Behaviour Scale, a 20-Likert type item questionnaire, assessed different facets of impulsivity (Billieux, Rochat, Ceschi, Carré, Offerlin-Meyer et al., 2012). The answers ranged from 1 (“strongly agree”) to 4 (“strongly disagree”). Five sub-scores were obtained through the sum of corresponding items: negative urgency (e.g., “When I'm upset, I often act without thinking”; α=.78), positive urgency (e.g., “When I'm really excited I tend not to think about the consequences of my actions”; α=.70), premeditation (e.g., “Usually I think carefully before I do anything”; α=.79), perseverance (e.g., “I generally prefer to see things through to the end”; α=.84), and sensation seeking (e.g., “I sometimes like to do things that are a little bit scary”; α=.83). A high subscore implied a higher level of impulsivity. | DACU frequency

Participants declared how many times in the last 12 months they drove within two hours after cannabis use. Four categories have been derived from it: once, two to three times, four to six times, and seven times and more. | Attitudes towards cannabis

The Marijuana Effect Expectancies Questionnaire (MEEQ) measured to which extent a person expected diverse effects to occur when using cannabis (Schafer & Brown, 1991). The short version used in this study had a good internal consistency (α=.84) and was made up of six subscales computed as the mean of relevant items (Guillem, Notides, Vorspan, Debray, Nieto et al., 2011). The first subscale, Cognitive and Behavioural Impairment, referred to difficulties concentrating or understanding others, expressing oneself clearly, and slow thinking and actions (e.g., “Cannabis slows down my thoughts and actions”). Individuals who endorsed the Relaxation and Tension Reduction subscale expected cannabis to help them unwind, make them less tense, relieve their anxiety, make them calm, and get a sense of relaxation (e.g., “I find a feeling of relaxation while consuming cannabis.”). Social and Sexual Facilitation defined increased talkativeness and sociability, as well as feeling more romantic or sexually attracted to others (e.g., “I am more sociable when I use cannabis.”). Perceptual and Cognitive Enhancement was related to increased creativity and imagination, different perceptions of music, and a better focus on one thing (e.g., “I become more creative or imaginative on cannabis.”). Individuals who had a high score on Global Negative Effects expected to feel down, lose control, become careless, or be angry and possibly violent after using cannabis (e.g., “Cannabis causes me to lose control and become careless.”). The Craving and Physical Effects subscale referred to hunger and craving for snacks or other things (e.g., “Consuming cannabis makes me hungry.”) | Attitudes towards cannabis and driving

Eight Likert-type items developed by Davis, Allen, Duke, Nonnemaker, Bradfield et al. (2016) assessed the perceptions related to DACU. Possible answers ranged from 1 (“strongly disagreeing”) to 5 (“strongly agreeing”). Two sub-constructs: knowledge of legal consequences (e.g., “I could lose my license if I drove high”) and risk perception (e.g., “I can drive safely under the influence of cannabis”) were computed as the mean of corresponding items. They had respectively a moderate (α=.63) and good (α=.91) internal consistency. | Motives for DACU

This questionnaire was adapted for DACU by the research team from a scale of motives for driving after drug use (Bonar, Arterberry, Davis, Cunningham, Blow et al., 2018). Participants declared how frequently each motive explained why they engaged in DACU. The 26 Likert-type questions had four anchors ranging from 1 (“never”) to 4=(“always”). Seven subscales were derived as the mean of relevant items: low impact of cannabis on driving (α=.79; e.g., “You consider that it causes little or no harm to your driving”), lack of alternative transportation (α=.86; e.g., “There was no more public transportation”), low/no risk associated with DACU (α=.85; e.g., “You consider having few chances of getting caught”), being a designated driver (α=.82; e.g., “Others want you to drive them somewhere”), perception of non-intoxication (α=.46; e.g., “You do not feel high”), fate/destiny (α=.65; e.g., “In any case, it is destiny that decides”), and peer influence (α=.71; e.g., “Your friends also do it”). | Descriptive and injunctive norms

Descriptive norms were determined by the participant’s estimation of the number of friends that have engaged in DACU. Answers ranged from “none” to “all or almost all of them”. The degree of friends’ approval of DACU defined injunctive norms. Participants declared how likely most of their friends would approve of them engaging in DACU. Answers ranged from 1 (“they would definitely disapprove”) to 7 (“they would definitely approve”).

2.3 | Analyses

All analyses were performed using IBM SPSS 26. A two-step cluster analysis was first performed to identify subgroups based on the following clustering variables: speeding, DUI-A, dangerous driving cognitions and behaviors (all three DDDI subscales), frequency of cannabis use, CUDIT total score, and psychological distress (depression, anxiety, irritability, cognitive problems). These sets of variables were chosen as they were the most commonly found in previous studies on DWI typologies as presented in the Introduction. The two-step cluster analysis combined two methods. First, the pre-clustering step separated groups by constructing a clustering feature tree with distance measure using Euclidean distance and log-likelihood method (Arminger, Clogg & Sobel, 1995). Second, the optimal subgroup model was chosen with a probabilistic approach, similar to latent class analysis, using Schwarz's Bayesian Criterion (BIC) as the clustering criterion (Gelbard, Goldman & Spiegler, 2007; Kent, Jensen & Kongsted, 2014). Latent class analysis is a model-based approach that uses parametric probability distribution and assumes the existence of an unobserved latent outcome variable that underlies differences among groups, while two-step cluster analysis also relies on an approach of distance/similarities between observations (Benassi, Garofalo, Ambrosini, Sant’Angelo, Raggini et al., 2020; Cunha, Amaral, Jacinto, Sousa-Pinto & Fonseca, 2021). Both analyses can produce convergent similar clustering solutions on the same dataset, but two-step cluster analysis tended to be easier to use and interpret than latent class analysis (Benassi et al., 2020; Kent et al., 2014). This latter analysis has several desirable features compared to more traditional cluster analysis techniques (K-means, Ward’s method, etc.), such as allowing categorical and continuous variables simultaneously, analyzing outliers, determining the number of clusters based on a statistical measure of fit, and being able to handle large datasets, i.e < 500 (Benassi et al., 2020; Gelbard et al., 2007; Kent et al., 2014).

The assumption of multicollinearity was not violated: none of these variables had a correlation coefficient ≥ .9 (Pallant, 2007). The first attempt of cluster analysis produced a two-group model with a “fair” average silhouette coefficient. This initial model was discarded as the software had yielded one group with high levels of problems on all indicators and another very low scores; these two groups remained difficult to interpret as heterogeneity was possibly still present. A three-cluster model was then yielded with a “fair” average silhouette with distinguishable subgroups. To determine if hidden heterogeneity was still present, a four-cluster solution was then tested, with an average silhouette coefficient that remained as “fair”. The four-cluster solution was chosen as it offered a stronger model in terms of statistics and interpretability, as it was more grounded on previous literature, than the three-cluster solution. Of note, a five-cluster solution produced a “weak” average silhouette coefficient, suggesting poor separation and cohesion between these clusters. As a reminder, two-step cluster analysis is an exploratory technique where different clustering algorithms need to be tested to yield the most statistically and theoretically optimal solution.

External validation of this four-cluster solution was executed through bivariate comparative analyses (χ2 or ANOVA) using variables that were not used for clustering. These included the socio-demographic variables, attitudes regarding DACU, and impulsivity. When external variables were statistically significant (p<.05), posthoc tests were performed to identify these differences. Specifically, the Bonferonni posthoc test was used for continuous variables, and standardized Pearson residuals were used for categorical variables. For the latter, a result was considered significant when the standardized residual was higher than 2 or lower than −2 (Everitt & Skrondal, 2010).


3.1 | Sample description

The final sample consisted of 910 Canadian young adults with a valid driver's license who all reported having DACU at least once in the last 12 months before taking the online survey. Among them, 40.1% (n=365) declared to be female, 54.0% (n=491) considered themselves as males, and 5.9% (n=54) identified themselves as being non-binary, gender-fluid, bispiritual, or belonging to another gender identity. They were 25.5 ± 5.4 years old on average. Less than half resided in an urban area (40.8%; n=371). About two-thirds of the sample were exclusively workers (66.7%; n=606); 17.6% (n=160) were exclusively students, and 12.3% (n=112) were both working and studying. The majority (63.5%; n=578) had a high school level or less. Half of the sample (50.7%; n=461) had DACU seven times or more in the past 12 months. The majority of participants (77.4%) filled out their questionnaire in English; the rest completed it in French. Participants were mainly from Ontario (27.8%) and Quebec (27.2%). For the other respondents, 10.5% lived in Alberta, 10.4% from British-Columbia, 8.2% from Nova Scotia, 4.2% from Saskatchewan, 4.1% from New-Brunswick, 3.3% from Newfoundland-Labrador, 2.4% from Manitoba, and 1.9% of Prince-Edward-Island. There was no participant from either of the three territories.

3.2 | Cluster analysis

Comparison analyses showed significant differences between the four groups for all clustering variables, as shown in Table 1.

[insert Table 1 about here]

The first subgroup was constituted of 398 individuals (43.7% of the sample). They were more likely to have either never or very rarely been speeding on the road (27.9%) and less likely to have done it either often, very often, or always (18.1%). This subgroup, as well the fourth one, had the lowest scores for negative emotions on the road, road aggressiveness, and risky driving behaviors. None of them had DUI-A and all of them declared using cannabis on a daily or almost daily basis. Their level of cannabis-related problems was significantly lower than the second subgroup, significantly higher than the fourth one, but equivalent to the third one. They had significantly lower scores for anxiety and depression symptoms than the second and fourth subgroups. Their scores for irritability and cognitive problems were significantly lower than the second subgroup. They were labeled as “high frequent cannabis users”.

The second subgroup, composed of 236 individuals (25.9% of the entire sample), had high scores or values for all clustering variables. Therefore, they were labeled as “individuals with generalized deviance”. Compared to the other subgroups, they were significantly the most likely to have been speeding either often, very often, or always (51.3%) and to present diverse deviant road behaviors and attitudes. The majority (78.4%) had DUI-A. In fact, 8.5% did it once, 25.4% two to three times, 29.7% four to six times, and 14.8% seven times or more in the past 12 months. Although half used cannabis either on a weekly or occasional basis, they had the highest score for cannabis related-problems. They also significantly presented the highest level of psychological distress for all related variables.

The third subgroup (n=145; 15.9% of the sample) distinguished itself from the other subgroups, notably that all of them had also DUI-A. As such, 29.0% did once, 40.0% two to three times, 17.9% four to six times, and 13.1% seven times or more in the past year. They did not stand out in terms of speeding behavior. They had significantly lower levels of negative emotions on the road than the second subgroup, but similar to the first and fourth subgroups. Their aggressiveness level on the road was also significantly lower than the second subgroup, but higher than the first subgroup. They presented lower risky road behaviors than the second subgroup, but significantly higher than the other two subgroups. More than half used cannabis on a daily or almost daily basis, and their cannabis-related problems score was significantly lower than the second subgroup, higher than the fourth one, and equivalent to the first one. Similar to the first subgroup, they had the lowest scores for all variables related to psychological distress. Therefore, they were labeled as “alcohol and drug-impaired drivers”.

The fourth subgroup, composed of 131 young adults (14.1%), was more likely to speed sometimes on the road. Almost none had DUI-A; a single individual had done it only once in the past 12 months. They did not differ from the first and third subgroups on the DDDI scores, except for risky driving which was significantly lower than the third subgroup. None were daily or almost daily cannabis users and they had the lowest score for cannabis-related problems. They had the second-highest scores for anxiety and depression, but they were similar to the first and third subgroups for irritability and cognitive problems. They were labeled as “well-adjusted youths with mild depressive-anxious symptoms”.

3.3 | External validation

The four subgroups were then compared on other external variables to further validate the generated model, as presented in Table 2. In terms of sociodemographic characteristics, the “well-adjusted youths with mild depressive-anxious symptoms” (subgroup 4) were significantly younger than the “high frequent cannabis users” (subgroup 1) and the “individuals with generalized deviance” (subgroup 2). The “young adults with generalized deviance” (subgroup 2) were predominantly males and workers exclusively, while the “well-adjusted youths with mild depressive-anxious symptoms” (subgroup 4) had a higher proportion of females and they were most likely to be studying as their main occupation. The “high frequent cannabis users” (subgroup 1) were more likely to have only a high school level or less, while a higher proportion of “well-adjusted youths with mild depressive-anxious symptoms” (subgroup 4) had a university degree.

[insert Table 2 about here]

Regarding alcohol use behavior, the “high frequent cannabis users” (subgroup 1) were more likely to be occasional or no alcohol drinkers (84.0%), and to have seldom binge drinking, i.e., once or never in the past month (71.4%). The majority of the “individuals with generalized deviance” (subgroup 2) had been drinking regularly and three-quarters had been binge drinking at least twice in the past month. Regarding “alcohol and drug-impaired drivers” (subgroup 3), they were all alcohol drinkers, with 41.0% being regular drinkers. Compared to the first and fourth subgroups, they were more numerous to be drinking regularly and binge drinking between three to five times in the past month.

The “individuals with generalized deviance” (subgroup 2) had the highest scores for all impulsivity subscales, except for sensation seeking, compared to the three other subgroups. The “alcohol and drug-impaired drivers” (subgroup 3) had significantly higher scores for lack of premeditation, positive urgency, and lack of perseverance than the “high frequent cannabis users” (subgroup 1). The latter had lower scores for positive and negative urgency and lack of perseverance compared to the “well-adjusted youths with mild depressive-anxious symptoms” (subgroup 4).

The “high frequent cannabis users” (subgroup 1) were less likely to expect cannabis to create cognitive disturbances and to generate negative effects, and more likely to declare that cannabis relaxes and reduces tension than the three other subgroups. They are also more likely to expect cannabis to amplify sensorial perceptions compared to “individuals with generalized deviance” (subgroup 2). This latter group was also more likely to endorse that cannabis causes cognitive disturbances than subgroup 3 (“alcohol and drug-impaired drivers”) and that cannabis is associated with negative effects than subgroups 3 and 4 (“well-adjusted youths with mild depressive-anxious symptoms”).

About opinions regarding DACU, the “high frequent cannabis users” (subgroup 1) were the least likely to perceive DACU as a risky behavior compared to the three other subgroups. The “well-adjusted youths with mild depressive-anxious symptoms” (subgroup 4) were more likely to be aware of legal consequences associated with DACU compared to the “high frequent cannabis users” (subgroup 1) and the “individuals with generalized deviance” (subgroup 2), the first subgroup being significantly more knowledgeable concerning legal repercussions than the second subgroup.

The “individuals with generalized deviance” (subgroup 2) were most likely to have engaged in DACU because they perceived cannabis as having low or no impact on driving than the three other subgroups. The “well-adjusted youths with mild depressive-anxious symptoms” (subgroup 4) were less likely to use this reason to explain why they DACU compared to subgroups 1 and 3. Lack of alternative transport, being the designated driver, perception of not being intoxicated, and accepting fate were significantly more likely to be important reasons to DACU for “individuals with generalized deviance” (subgroup 2) than for the three other subgroups. Subgroup 2 was the most likely to declare that they DACU because it was a low-risk behavior and that their peers influence them into doing so, followed by subgroup 3 (“alcohol and drug-impaired drivers”). Subgroups 1 and 4 were significantly less likely than subgroup 3 to endorse these two motives.

While the “high frequent cannabis users” (subgroup 1) were higher in proportion to think that their peers will definitely approve that they DACU and to believe that all or almost of them were doing it, the “well-adjusted youths with mild depressive-anxious symptoms” (subgroup 4) were more likely to declare that most of the peers will definitely disapprove DACU, considering that none or only a few of them were doing it themselves. Concerning the “individuals with generalized deviance” (subgroup 2), this behavior was somewhere in the midrange for peer approval and the number of friends who DACU.


This study was the first to identify typologies of cannabis users who DACU based on driving-related behaviors and attitudes, cannabis use and related problems, and psychological distress. Four distinct clusters emerged from the analyses, demonstrating the heterogeneity among individuals who DACU, and the need to adopt diverse and targeted strategies to prevent DACU, either on an individual level or a societal standpoint. Although the results partially converged with our hypotheses and most previous studies concerning identified typologies of DWI individuals, a few divergences had been observed. Before discussing specific differences and similarities of each cluster compared to previous literature, some overall points had to be considered to explain some global divergences. Considering that each study had not used the same set of clustering variables, some variations were expected to be found either in the number of clusters or in the main shared characteristics within one subgroup (Wells-Parker, Anderson, Pang & Timken, 1993). However, when using similar concepts to form clusters, some types of grouping were indeed repetitively found. Other divergences may also be explained by the fact that this study examined individuals who DACU, regardless if they have been arrested or not, while all previous studies focused on alcohol and were conducted among convicted offenders. Differences in sample characteristics may account for the absence of a fifth subgroup of individuals that presented diverse deviant road behaviors in our study. Also, they may have been merged with the generalized deviance subgroup.

Daily or almost daily cannabis users formed the first cluster. Previous typology studies on alcohol-impaired drivers had also identified subgroups constituted mainly of frequent and/or problematic substance use (Ball et al., 2000; Hubicka et al., 2010; Okamura et al., 2014; Saltstone & Poudrier, 1989; Scherer, Beck, et al., 2021; Scherer, Nochajski, et al., 2021; Snowden et al., 1986; Steer et al., 1979; Wells-Parker et al., 1986). In this study, this typology was characterized mainly by frequent cannabis use and more positive attitudes towards cannabis and DACU. They were also more likely to believe that the vast majority, if not all of their friends, will approve of DACU and that this behavior was frequent among their peers. Otherwise, they did not present other risky behaviors and they had low scores regarding psychological distress or impulsiveness. This suggests that frequent cannabis users do not systematically present high levels of associated problems. Asbridge, Duff, Marsh & Erickson (2014) highlighted that assessments of cannabis related-problems relied disproportionally on cannabis use frequency rather than on actual harms; these two distinct concepts were often conflated together as if they were synonymous. Frequent cannabis use in itself, with low or no associated problems, should therefore not be viewed as the unique reason to provide clinical intervention. Nonetheless, considering their frequent cannabis use, DACU became unavoidable if they needed to drive around for different purposes (e.g., do their groceries, go to work or school, etc.). DACU has been reported previously by daily users as an unintentional behavior because it was an integral part of their everyday life (Watson, Mann, Wickens & Brands, 2019). This might largely explain also why the frequency of DACU was high within this subgroup. In North America where “car culture” is still dominant, suburban and rural communities are spread out across a vast territory, requiring most individuals to rely on their car if they need to go somewhere (Filion, 2014). Nonetheless, there remains a potential risk of road accidents associated with DACU (Asbridge et al., 2012; Bédard et al., 2007). From a macroscopic perspective, policymakers should consider implementing transportation strategies that would permit these frequent cannabis users to move around conveniently while reducing road accidents. Convenience has been highlighted as a key reason by some daily cannabis users on why they DACU; for some, current transportation alternatives were viewed as time-consuming or needing too much planning (Watson et al., 2019). For densely populated areas, efficient, extremely frequent, reliable, and around-the-clock public transportation should be developed and maintained for citizens to perceive it as a truly viable alternative transit option. Another initiative, more adapted for distant and rural areas, would be to offer a safe-ride-home service. In some Canadian provinces, Red Nose Operation is a free service provided during the Christmas holiday season to accompany drunk drivers back home safely and may have decreased the number of road accidents, considering that 69,029 Canadians reached for such a safe ride in 2019 (Fell, Scolese, Achoki, Burks, Goldberg et al., 2020). Local initiatives in Canada are available all year long, but they tend to charge a fee comparable to taxis (e.g., Safe Ride Home in Vancouver, Tolerance Zero 8 in Québec, etc.) and they are not as well-known as Red Nose Operation. Otherwise, some frequent cannabis users may be forced to drive to their friend’s place to be able to legally consume cannabis, since many Canadian municipalities have prohibited cannabis use in public spaces or rented dwellings (Gourdet, Gagnon, Moscetti & Obradovic, 2021). On another note, they tended to downplay the negative cannabis effect expectations and perceived low risk associated with DACU, while being less knowledgeable concerning the legal repercussions of such behavior. Sensitization campaigns need to address trivialization attitudes towards cannabis, as DACU is not risk-free, but they should avoid patronizing individuals who frequently use cannabis or dramatizing the issue.

The second identified cluster was characterized by individuals who presented diverse dangerous road behaviors and who manifested high levels of problems regarding psychological distress and cannabis use. Previous studies had also reported such a similar subgroup with high scores on all indicators (Okamura et al., 2014; Steer et al., 1979; Wells-Parker et al., 1986; Wieczorek & Miller, 1992). This typology corresponded to the Problem Behavior Theory developed by Jessor, which posits that some individuals tend to deviate from desirable or conventional social norms as a result of different risk factors that were not attenuated by protective factors (Jessor, 2016; Jessor & Jessor, 1977). Risk factors included models for engaging in risk behaviors, opportunities and contexts favorable for general deviance, as well as personal psychological and biological vulnerabilities. Protection factors encompassed positive models for prosocial behavior, social controls, and experiences of health-enhancing behaviors. While the Problem Behavior Theory has been used to explain DUI-A and DACU among adolescents and young adults (Jessor, 1987; Shope & Bingham, 2002), this study showed that this theoretical model seemed to apply only to a subgroup that represented one-quarter of drivers who engaged in DACU. If clinical intervention is required, a holistic assessment that tackles diverse problems, from risky behaviors to mental health symptoms, needs to be considered for individuals with this profile. Repressive deterrence measures by themselves will not be efficient, if DACU is considered only as criminal behavior, and the psychosocial difficulties that underlie it are not tackled.

The third identified cluster was characterized by individuals who drive after using any substance, notably alcohol. In some regards, they appeared to be similar to the first group of frequent cannabis users, but they were principally alcohol drinkers rather than cannabis users. They distinguished themselves from the first cluster by their frequent alcohol use, as well as their high tendency for binge drinking. Also, compared to the first subgroup of frequent cannabis users, their cannabis use was less frequent, and their opinions and attitudes towards cannabis effects and DACU were less downplayed. They were also the second-highest impulsive group, but they did not manifest psychological distress symptoms, nor did they adopt other risky road behaviors and attitudes. Their profile suggested that they may constitute a subgroup that thrived at being highly intoxicated, probably inclined for partying hard. Like for the first group, initiatives should be geared towards providing easily accessible transportation options that will prevent them from driving themselves. Furthermore, binge drinking has been associated with DUI-A (Hingson, Zhi & White, 2017; Minaker, Bonham, Elton-Marshall, Leos-Toro et al., 2017; Sloan, Chepke & Davis, 2014). In this profile, impulsiveness seemed to underlie their characteristics, albeit not as intense as in the general deviance typology. Any program aiming at reducing DACU for this subgroup should be able to provide harm reduction strategies that are easy, convenient, quick to apply, and adapted to situations where decisions can be taken impulsively in the heat of the moment.

Like in previous studies, the fourth identified cluster was characterized by individuals who did not present any “extreme” behavior, often labeled as “well-adjusted” (Donovan & Marlatt, 1982; Moore, 1994; Nelson et al., 2019; Nolan et al., 1994; Okamura et al., 2014; Saltstone & Poudrier, 1989; Scherer, Beck, et al., 2021; Scherer, Nochajski, et al., 2021; Shim et al., 2016; Snowden et al., 1986; Steer et al., 1979; Wells-Parker et al., 1986). In this analysis, they seemed to be non-frequent cannabis users and occasional drinkers who had less frequently engaged in DACU than the rest of the sample. They did not exhibit any other problematic road-related behaviors or attitudes, but they had the second-highest scores in terms of self-reported anxiety and depression symptomatology. However, this observation may be partly explained by the over-representation of individuals who identify themselves as females and students with a university education level. Worldwide, females in the general population tend to report higher levels of psychological distress than men; this gender difference has been often attributed to social, economic, and cultural inequities (Drapeau, Beaulieu-Prévost, Marchand, Boyer, Préville et al., 2010; Viertiö, Kiviruusu, Piirtola, Kaprio, Korhonen et al., 2021). Previous studies had reported that high psychological distress was also prevalent and persistent among university students, partly due to academic stress and financial instability (Sharp & Theiler, 2018). The results did not allow determining why they had engaged in DACU on a few occasions, but none of the assessed motives appeared to specifically characterize them. This highlights that DACU is not an “all-or-nothing” behavior: for the same individual, risk assessment related to the decision of driving or not can vary from one event to another, based on context and state of mind (Hakamies-Blomqvist, 2006).

4.1 | Limitations

This study presented some limitations. As the data was self-reported, participants may had either minimized or exaggerated their answers or refused to disclose sensitive information. However, self-reported criminal behaviors and substance use have moderate to strong correlations with official records, suggesting that they constitute valid data (Johnston & O’Malley, 1985; Thornberry & Krohn, 2002). As the participant completed alone and anonymously a computerized questionnaire, the impact of social desirability has been minimized (Richman et al., 1999). Memory bias may had occurred, considering that some questions referred to events that happened in the past 12 months. As the participant was the sole informant providing the information, data could not have been cross-checked with other sources. Additionally, the cross-sectional design of this study did not allow inferring any causal effects. Furthermore, this study included a convenience sample of individuals who were motivated to fill out a 20-minute questionnaire and who had Internet access. Another limitation concerned age inclusion. Adults older than 35 years old were not recruited for feasibility reasons, as the sample size was limited by the allotted budget. Removing the age limit of 35 years old would have reduced considerably the sample size of the group of individuals who declared engaging in DACU, as this behavior is mainly associated with younger age. Therefore, this study’s results may not be applicable to older adults. Their inclusion may have allowed for other typologies to emerge. Previous studies had highlighted that some older adults who engage in DACU with specific characteristics, although they are less numerous in proportion to do it (Choi et al., 2019). It would therefore be recommended to also recruit older adults in a future research project that allows for a more substantial sample size capacity due to methodological consideration (e.g. budget). Moreover, DACU was conceptualized in this study as driving after using any quantity of cannabis, which can range from being not intoxicated to being very “high”. However, our definition, i.e., driving within two hours after cannabis consumption, was more likely to include events where the person was intoxicated. Cannabis intoxication and induced impairment typically last within approximately three to six hours, but there is a wide interindividual variation and from one consumption to another (Fischer, Russell, Sabioni, van den Brink, Le Foll et al., 2017). Many different factors will moderate the level of cannabis induced impairment after use, such as product quantity, product quality (such as THC content, etc.), tolerance, body weight, lung capacity, metabolic rate, mental state, etc. Finally, clusters were determined by the selection of variables entered into the model, which can hinder the replicability of results. As such, previous studies did present variations in the number of clusters or the main shared characteristics within one subgroup among different studies (Wells-Parker et al., 1993). However, when using similar concepts to form clusters, some types of grouping were indeed repetitively found, both in this study and previous ones.


Although previous DWI typology studies allowed identifying clusters of arrested or convicted DUI-A offenders, knowledge remained scarce concerning specifically DACU and drivers who have not necessarily been apprehended. As cannabis use is becoming relatively more socially accepted as some Western countries have either already legalized this substance or are considered to do so, a better understanding of DACU is necessary to better prevent potential risks related to this behavior. This study’s results highlighted the need to acknowledge heterogeneity among individuals who DACU. Preventing DACU is more likely to be effective if programs target more specifically the specific set of risk factors associated with drivers who share similar characteristics. Furthermore, when assessing referred individuals into impaired-driving programs, these different profiles should be considered to provide adapted and tailored services.


Adamson, S. J., & Sellman, J. D. (2003). A prototype screening instrument for cannabis use disorder: The Cannabis Use Disorders Identification Test (CUDIT) in an alcohol-dependent clinical sample. Drug and Alcohol Review, 22(3), 309–315.

Arminger, G., Clogg, C. C., & Sobel, M. E. (1995). Handbook of Statistical Modeling for the Social and Behavioral Sciences. Plenum Press.

Arterberry, B. J., Treloar, H., & McCarthy, D. M. (2017). Empirical profiles of alcohol and marijuana use, drugged driving, and risk perceptions. Journal of Studies on Alcohol and Drugs, 78(6), 889–898.

Arterberry, B. J., Treloar, H. R., Smith, A. E., Martens, M. P., Pedersen, S., & McCarthy, D. M. (2013). Marijuana use, driving, and related cognitions. Psychology of Addictive Behaviors : Journal of the Society of Psychologists in Addictive Behaviors, 27(3), 854–860.

Asbridge, M., Duff, C., Marsh, D. C., & Erickson, P. G. (2014). Problems with the identification of ‘problematic’ cannabis use: Examining the issues of frequency, quantity, and drug use environment. European Addiction Research, 20(5), 254–267.

Asbridge, M., Hayden, J. A., & Cartwright, J. L. (2012). Acute cannabis consumption and motor vehicle collision risk: Systematic review of observational studies and meta-analysis. The British Medical Journal, 344.

Aston, E. R., Merrill, J. E., McCarthy, D. M., & Metrik, J. (2016). Risk factors for driving after and during marijuana use. Journal of Studies on Alcohol and Drugs, 77(2), 309–316.

Ball, S. A., Jaffe, A. J., Crouse-Artus, M. S., Rounsaville, B. J., & O’Malley, S. S. (2000). Multidimensional subtypes and treatment outcome in first-time DWI offenders. Addictive Behaviors, 25(2), 167–181.

Bédard, M., Dubois, S., & Weaver, B. (2007). The impact of cannabis on driving. Canadian Journal of Public Health, 98(1), 6–11.

Benassi, M., Garofalo, S., Ambrosini, F., Sant’Angelo, R. P., Raggini, R., De Paoli, G., Ravani, C., Giovagnoli, S., Orsoni, M., & Piraccini, G. (2020). Using Two-Step Cluster Analysis and Latent Class Cluster Analysis to Classify the Cognitive Heterogeneity of Cross-Diagnostic Psychiatric Inpatients. Frontiers in Psychology, 11, 1085.

Berg, C. J., Daniel, C. N., Vu, M., Li, J., Martin, K., & Le, L. (2018). Marijuana use and driving under the influence among young adults: A socioecological perspective on risk factors. Substance Use & Misuse, 53(3), 370–380.

Billieux, J., Rochat, L., Ceschi, G., Carré, A., Offerlin-Meyer, I., Defeldre, A.-C., Khazaal, Y., Besche-Richard, C., & Van der Linden, M. (2012). Validation of a short French version of the UPPS-P Impulsive Behavior Scale. Comprehensive Psychiatry, 53(5), 609–615.

Bingham, C. R., Shope, J. T., & Zhu, J. (2008). Substance-involved driving: Predicting driving after using alcohol, marijuana, and other drugs. Traffic Injury Prevention, 9(6), 515–526.

Bonar, E. E., Arterberry, B. J., Davis, A. K., Cunningham, R. M., Blow, F. C., Collins, R. L., & Walton, M. A. (2018). Prevalence and motives for drugged driving among emerging adults presenting to an emergency department. Addictive Behaviors, 78, 80–84.

Bondallaz, P., Favrat, B., Chtioui, H., Fornari, E., Maeder, P., & Giroud, C. (2016). Cannabis and its effects on driving skills. Forensic Science International, 268, 92–102.

Borodovsky, J. T., Marsch, L. A., Scherer, E. A., Grucza, R. A., Hasin, D. S., & Budney, A. J. (2020). Perceived safety of cannabis intoxication predicts frequency of driving while intoxicated. Preventive Medicine, 131, 105956.

Boyer, R., Préville, M., Légaré, G., & Valois, P. (1993). La détresse psychologique dans la population du Québec non institutionnalisée: Résultats normatifs de l’Enquête Santé Québec. The Canadian Journal of Psychiatry, 38(5), 339–343.

Capler, R., Bilsker, D., Van Pelt, K., & MacPherson, D. (2017). Cannabis use and driving: Evidence review (pp. 1–66). Canadian Drug Policy Coalition.

Choi, N. G., DiNitto, D. M., & Marti, C. N. (2019). Older adults driving under the influence: Associations with marijuana use, marijuana use disorder, and risk perceptions. Journal of Applied Gerontology, 38(12), 1687–1707.

Cook, S., Shank, D., Bruno, T., Turner, N. E., & Mann, R. E. (2017). Self-reported driving under the influence of alcohol and cannabis among Ontario students: Associations with graduated licensing, risk taking, and substance abuse. Traffic Injury Prevention, 18(5), 449–455.

Cunha, F., Amaral, R., Jacinto, T., Sousa-Pinto, B., & Fonseca, J. A. (2021). A systematic review of asthma phenotypes derived by data-driven methods. Diagnostics, 11(4), 644.

Cuttler, C., Sexton, M., & Mischley, L. K. (2018). Driving under the influence of cannabis: An examination of driving beliefs and practices of medical and recreational cannabis users across the United States. Cannabis, 1(2), 1–13.

Davis, K. C., Allen, J., Duke, J., Nonnemaker, J., Bradfield, B., Farrelly, M. C., Shafer, P., & Novak, S. (2016). Correlates of marijuana drugged driving and openness to driving while high: Evidence from Colorado and Washington. PLOS ONE, 11(1), e0146853.

Domingo-Salvany, A., Herrero, M. J., Fernandez, B., Perez, J., del Real, P., González-Luque, J. C., & de la Torre, R. (2017). Prevalence of psychoactive substances, alcohol and illicit drugs, in Spanish drivers: A roadside study in 2015. Forensic Science International, 278, 253–259.

Donovan, D. M., & Marlatt, G. A. (1982). Personality subtypes among driving-while-intoxicated offenders: Relationship to drinking behavior and driving risk. Journal of Consulting and Clinical Psychology, 50(2), 241.

Doroudgar, S., Mae Chuang, H., Bohnert, K., Canedo, J., Burrowes, S., & Perry, P. J. (2018). Effects of chronic marijuana use on driving performance. Traffic Injury Prevention, 19(7), 680–686.

Drapeau, A., Beaulieu-Prévost, D., Marchand, A., Boyer, R., Préville, M., & Kairouz, S. (2010). A life-course and time perspective on the construct validity of psychological distress in women and men. Measurement invariance of the K6 across gender. BMC Medical Research Methodology, 10, 68.

Dula, C. S., & Ballard, M. E. (2003). Development and evaluation of a measure of dangerous, aggressive, negative emotional, and risky driving. Journal of Applied Social Psychology, 33(2), 263–282.

Everitt, B. S., & Skrondal, A. (2010). The Cambridge Dictionary of Statistics. Statistics and probability (4th edition). Cambridge University Press.

Fell, J. C., Scolese, J., Achoki, T., Burks, C., Goldberg, A., & DeJong, W. (2020). The effectiveness of alternative transportation programs in reducing impaired driving: A literature review and synthesis. Journal of Safety Research, 75, 128–139.

Filion, P. (2014). The Creation and Perpetuation of an Automobile-Oriented Urban Form: Dispersed Suburbanism in North America. In The Organization of Transport (1st Edition, pp. 173–194). Routledge.

Fischer, B., Russell, C., Sabioni, P., Van Den Brink, W., Le Foll, B., Hall, W., ... & Room, R. (2017). Lower-risk cannabis use guidelines: a comprehensive update of evidence and recommendations. American Journal of Public Health, 107(8), e1-e12.

Gelbard, R., Goldman, O., & Spiegler, I. (2007). Investigating diversity of clustering methods: An empirical comparison. Data & Knowledge Engineering, 63(1), 155–166.

Goodman, S. E., Leos-Toro, C., & Hammond, D. (2020). Risk perceptions of cannabis- vs. Alcohol-impaired driving among Canadian young people. Drugs: Education, Prevention and Policy, 27(3), 205–212.

Gourdet, C., Gagnon, F., Moscetti, C., & Obradovic, I. (2021). Regulating private and public places of non-medical cannabis consumption in North America: Public health and public safety issues. Journal of Canadian Studies, 55(2), 279–306.

Guillem, E., Notides, C., Vorspan, F., Debray, M., Nieto, I., Leroux, M., & Lépine, J.-P. (2011). Cannabis expectancies in substance misusers: French validation of the Marijuana Effect Expectancy Questionnaire. The American Journal on Addictions, 20(6), 543–554.

Hakamies-Blomqvist, L. (2006). Are there safe and unsafe drivers? Transportation Research Part F: Traffic Psychology and Behaviour, 9(5), 347–352.

Hartman, R. L., & Huestis, M. A. (2013). Cannabis effects on driving skills. Clinical Chemistry, 59(3), 478–492.

Health Canada. (2020, December 21). Canadian Cannabis Survey 2020—Summary [Surveys].

Health Canada. (2021, July 5). Low-risk alcohol drinking guidelines [Education and awareness].

Hingson, R. W., Zha, W., & White, A. M. (2017). Drinking beyond the binge threshold: Predictors, consequences, and changes in the US. American Journal of Preventive Medicine, 52(6), 717–727.

Howard, M. C., & Hoffman, M. E. (2018). Variable-centered, person-centered, and person-specific approaches: Where theory meets the method. Organizational Research Methods, 21(4), 846–876.

Hubicka, B., Källmén, H., Hiltunen, A., & Bergman, H. (2010). Personality traits and mental health of severe drunk drivers in Sweden. Social Psychiatry and Psychiatric Epidemiology, 45(7), 723–731.

Ilfeld, F. W. (1976). Further validation of a Psychiatric Symptom Index in a normal population. Psychological Reports, 39(3_suppl), 1215–1228.

Jessor, R. (1987). Risky driving and adolescent problem behavior: An extension of problem-behavior theory. Alcohol, Drugs & Driving, 3(3–4), 1–11.

Jessor, R. (2016). Problem Behavior Theory over the Years. In R. Jessor (Ed.), The Origins and Development of Problem Behavior Theory: The Collected Works of Richard Jessor (pp. 15–42). Springer International Publishing.

Jessor, R., & Jessor, S. L. (1977). Problem behavior and psychosocial development: A longitudinal study of youth. New York: Academic Press.

Kantardzic, M. (2019). Cluster Analysis. In Data Mining (pp. 295–334). John Wiley & Sons, Ltd.

Kent, P., Jensen, R. K., & Kongsted, A. (2014). A comparison of three clustering methods for finding subgroups in MRI, SMS or clinical data: SPSS TwoStep Cluster analysis, Latent Gold and SNOB. BMC Medical Research Methodology, 14(1), 113.

Le Strat, Y., Dubertret, C., & Le Foll, B. (2015). Impact of age at onset of cannabis use on cannabis dependence and driving under the influence in the United States. Accident Analysis & Prevention, 76, 1–5.

Matthews, A. J., Bruno, R., Dietze, P., Butler, K., & Burns, L. (2014). Driving under the influence among frequent ecstasy consumers in Australia: Trends over time and the role of risk perceptions. Drug and Alcohol Dependence, 144, 218–224.

Mikulskaya, E., & Martin, F. (2018a). Visual attention to motion stimuli and its neural correlates in cannabis users. European Journal of Neuroscience, 47(3), 269–276.

Mikulskaya, E., & Martin, F. H. (2018b). Contrast sensitivity and motion discrimination in cannabis users. Psychopharmacology, 235(8), 2459–2469.

Minaker, L. M., Bonham, A., Elton-Marshall, T., Leos-Toro, C., Wild, T. C., & Hammond, D. (2017). Under the influence: Examination of prevalence and correlates of alcohol and marijuana consumption in relation to youth driving and passenger behaviours in Canada. A cross-sectional study. Canadian Medical Association Journal Open, 5(2), E386–E394.

Moore, R. H. (1994). Underage female DUI offenders: Personality characteristics, psychosocial stressors, alcohol and other drug use, and driving-risk. Psychological Reports, 74(2), 435–445.

Nelson, S. E., Shoov, E., LaBrie, R. A., & Shaffer, H. J. (2019). Externalizing and self-medicating: Heterogeneity among repeat DUI offenders. Drug and Alcohol Dependence, 194, 88–96.

Nolan, Y., Johnson, J., & Pincus, A. (1994). Personality and drunk driving: Identification of DUI types using the Hogan Personality Inventory. Psychological Assessment, 6, 33–40.

Okamura, K., Kosuge, R., Kihira, M., & Fujita, G. (2014). Typology of driving-under-the-influence (DUI) offenders revisited: Inclusion of DUI-specific attitudes. Addictive Behaviors, 39(12), 1779–1783.

Pallant, J. (2007). SPSS survival manual: A step by step guide to data analysis using SPSS for Windows (3rd edition). Open University Press.

Rogeberg, O., & Elvik, R. (2016). The effects of cannabis intoxication on motor vehicle collision revisited and revised. Addiction, 111(8), 1348–1359.

Roma, P., Mazza, C., Ferracuti, G., Cinti, M. E., Ferracuti, S., & Burla, F. (2019). Drinking and driving relapse: Data from BAC and MMPI-2. PloS One, 14(1), e0209116.

Rousseeuw, P. J. (1987). Silhouettes: A graphical aid to the interpretation and validation of cluster analysis. Journal of Computational and Applied Mathematics, 20, 53–65.

Saltstone, R., & Poudrier, L. M. (1989). Suggested treatment interventions for impaired driving offenders based upon research with impaired driver subtypes. Alcoholism Treatment Quarterly, 6(3–4), 129–141.

Sarstedt, M., & Mooi, E. (2019). Cluster Analysis. In M. Sarstedt & E. Mooi (Eds.), A Concise Guide to Market Research: The Process, Data, and Methods Using IBM SPSS Statistics (pp. 301–354). Springer.

Schafer, J., & Brown, S. A. (1991). Marijuana and cocaine effect expectancies and drug use patterns. Journal of Consulting and Clinical Psychology, 59(4), 558–565.

Scherer, M., Beck, K., Taylor, E. P., Romosz, A., Voas, R., & Romano, E. (2021). A latent class analysis of DUI offender motivation and awareness as predictors of performance while on alcohol ignition interlocks. Journal of Substance Use, 26(3), 250–255.

Scherer, M., Harrell, P., & Romano, E. (2015). Marijuana and other substance use among motor vehicle operators: A latent class analysis. Journal of Studies on Alcohol and Drugs, 76(6), 916–923.

Scherer, M., Nochajski, T. H., Romano, E., Manning, A. R., Romosz, A., Tippetts, S., Taylor, E., Voas, R., & Paul, R. (2021). Typologies of drivers convicted of driving under the influence of alcohol as predictors of alcohol ignition interlock performance. Alcoholism Treatment Quarterly, 39(1), 96–109.

Scherer, M., Voas, R. B., & Furr-Holden, D. (2013). Marijuana as a predictor of concurrent substance use among motor vehicle operators. Journal of Psychoactive Drugs, 45(3), 211–217.

Sharp, J., & Theiler, S. (2018). A review of psychological distress among university students: Pervasiveness, implications and potential points of intervention. International Journal for the Advancement of Counselling, 40(3), 193–212.

Shim, I. H., Wang, H.-R., & Bahk, W.-M. (2016). Typical MMPI-2 profiles of multiple-DWI individuals. International Journal of Mental Health and Addiction, 14(2), 149–153.

Shope, J. T., & Bingham, C. R. (2002). Drinking-driving as a component of problem driving and problem behavior in young adults. Journal of Studies on Alcohol, 63(1), 24–33.

Sloan, F. A., Chepke, L. M., & Davis, D. V. (2014). Addiction, drinking behavior, and driving under the influence. Substance Use & Misuse, 49(6), 661–676.

Snowden, L. R., Nelson, L. S., & Campbell, D. (1986). An empirical typology of problem drinkers from the Michigan Alcoholism Screening Test. Addictive Behaviors, 11(1), 37–48.

Steer, R. A., Fine, E. W., & Scoles, P. E. (1979). Classification of men arrested for driving while intoxicated, and treatment implications. A cluster-analytic study. Journal of Studies on Alcohol, 40(3), 222–229.

Sukhawathanakul, P., Thompson, K., Brubacher, J., & Leadbeater, B. (2019). Marijuana trajectories and associations with driving risk behaviors in Canadian youth. Traffic Injury Prevention, 20(5), 472–477.

Tan, P.-N., Steinbach, M., Karpatne, A., & Kumar, V. (2019). Cluster Analysis: Basic Concepts and Algorithms. In Introduction to Data Mining (Second Edition, pp. 525–612). Pearson Education.

Thake, J., & Davis, C. G. (2011). Assessing problematic cannabis use. Addiction Research & Theory, 19(5), 448–458.

Viertiö, S., Kiviruusu, O., Piirtola, M., Kaprio, J., Korhonen, T., Marttunen, M., & Suvisaari, J. (2021). Factors contributing to psychological distress in the working population, with a special reference to gender difference. BMC Public Health, 21(1), 611.

Voas, R. B., Lacey, J. H., Jones, K., Scherer, M., & Compton, R. (2013). Drinking drivers and drug use on weekend nights in the United States. Drug and Alcohol Dependence, 130(0), 215–221.

Ward, N. J., Schell, W., Kelley-Baker, T., Otto, J., & Finley, K. (2018). Developing a theoretical foundation to change road user behavior and improve traffic safety: Driving under the influence of cannabis (DUIC). Traffic Injury Prevention, 19(4), 358–363.

Watson, T. M., Mann, R. E., Wickens, C. M., & Brands, B. (2019). “Just a Habit”: Driving Under the Influence of Cannabis as Ordinary, Convenient, and Controllable Experiences According to Drivers in a Remedial Program. Journal of Drug Issues, 49(3), 531–544.

Wells-Parker, E., Anderson, B., Pang, M., & Timken, D. (1993). An examination of cluster-based classification schemes for DUI offenders. Journal of Studies on Alcohol, 54(2), 209–218.

Wells-Parker, E., Cosby, P. J., & Landrum, J. W. (1986). A typology for drinking driving offenders: Methods for classification and policy implications. Accident Analysis & Prevention, 18(6), 443–453.

Wettlaufer, A., Florica, R. O., Asbridge, M., Beirness, D., Brubacher, J., Callaghan, R., Fischer, B., Gmel, G., Imtiaz, S., Mann, R. E., McKiernan, A., & Rehm, J. (2017). Estimating the harms and costs of cannabis-attributable collisions in the Canadian provinces. Drug and Alcohol Dependence, 173, 185–190.

Whitehill, J. M., Rodriguez-Monguio, R., Doucette, M., & Flom, E. (2019). Driving and riding under the influence of recent marijuana use: Risk factors among a racially diverse sample of young adults. Journal of Ethnicity in Substance Abuse, 18(4), 594–612.

Wieczorek, W. F., & Miller, B. A. (1992). Preliminary typology designed for treatment matching of driving-while-intoxicated offenders. Journal of Consulting and Clinical Psychology, 60(5), 757–765.


TABLE 1. Cluster analysis

Group 1: High frequent cannabis users

n=398 (43.7%)

Group 2: Individuals with generalized deviance

n=236 (25.9%)

Group 3: Alcohol and drug-impaired drivers

n=145 (15.9%)

Group 4: Well-adjusted youths with mild depressive- anxious symptoms

n=131 (14.1%)




n (%) / M (S.D.)

n (%) /
M (S.D.)

n (%) /
M (S.D.)

n (%) /
M (S.D.)

n (%) / M (S.D.)




38 (9.5%)*

4 (1.7%)‡

3 (2.1%)‡

8 (6.1%)

53 (5.8%)

Very rarely

73 (18.3%)*

12 (5.1%)‡

20 (13.8%)

22 (16.8%)

127 (14.0%)


76 (19.1%)

37 (15.7%)

31 (21.4%)

16 (12.2%)

160 (17.6%)


139 (34.9%)

62 (26.3%)‡

59 (40.7%)

58 (44.3%)*

318 (34.9%)


53 (13.3%)

79 (33.5%)*

22 (15.2%)

16 (12.2%)‡

170 (18.7%)

Very often

18 (4.5%)

25 (10.6%)*

10 (6.9%)

11 (8.4%)

64 (7.0%)


1 (0.3%)

17 (7.2%)*

0 (0.0%)

0 (0.0%)

18 (2.0%)




0 (0.0%)‡

185 (78.4%)*

145 (100.0%)*

1 (0.8%)‡

331 (36.4%)


398 (100.0%)*

51 (21.6%)‡

0 (0.0%)‡

130 (99.2%)*

579 (63.6%)

Frequency of cannabis use


Daily / almost daily

398 (100.0%)*

112 (47.5%)‡

82 (56.6%)‡

0 (0.0%)‡

592 (65.1%)

Weekly / almost weekly

0 (0.0%)‡

121 (51.3%)*

57 (39.3%)

116 (88.5%)*

294 (32.3%)


0 (0.0%)‡

3 (1.3%)

6 (4.1%)

15 (11.5%)*

24 (2.6%)

CUDIT score

11.36 (6.08)a

19.00 (6.79)b

10.93 (5.98)a

6.10 (4.74)c

12.52 (7.40)



Negative emotions

18.25 (5.38)a

29,00 (5.02)b

18.84 (4.97)a

18.63 (5.72)a

21.19 (7.01)



10.42 (3.57)a

21.35 (4.90)b

11.81 (3.80)c

10.82 (3.32)a,c

13.53 (6.11)


Risky driving

20.22 (5.12)a

38.30 (8.22)b

23.68 (5.39)c

19.57 (5.38) a

25.37 (9.90)


Psychological distress


1.86 (0.70)a

2.73 (0.67)b

1.77 (0.60)a

2.07 (0.81)c

2.10 (0.79)



1.95 (0.75)a

2.70 (0.62)b

1.82 (0.53)a

2.14 (0.79)c

2.15 (0.77)



1.71 (0.57)a

2.74 (0.57)b

1.67 (0.50)a

1.74 (0.54)a

1.97 (0.72)


Cognitive problems

1.70 (0.66)a

2.63 (0.71)b

1.72 (0.60)a

1.86 (0.78)a

1.97 (0.79)


Note. χ2 comparisons are made for each row reporting percentages. When a significant difference is observed in a row, the standardized Pearson residual is used to identify the cells representing a proportion different from the expected one. A standardized residual over 2 or less than -2 is statistically significant. This is denoted with an asterisk (*) when proportions are higher than expected and with a double dagger (‡) when proportions are lower than expected. For continuous variables, means in the row sharing subscripts (a, b or c) are not significantly different from each other.

* p < .05; ** p < .01; *** p < .001

TABLE 2. External validation variables

Group 1: High frequent cannabis users

n=398 (43.7%)

Group 2: Individuals with generalized deviance

n=236 (25.9%)

Group 3: Alcohol and drug-impaired drivers

n=145 (15.9%)

Group 4: Well-adjusted youths with mild depressive- anxious symptoms

n=131 (14.1%)




n (%) / M (S.D.)

n (%) / M (S.D.)

n (%) / M (S.D.)

n (%) / M (S.D.)

n (%) / M (S.D.)

Sociodemographic characteristics


25.70 (5.65)a

26.25 (4.72)a

25.13 (5.19)a,c

23.96 (5.53)b,c

25.5 (5.37)


Gender identity



166 (41.7%)

69 (29.2%)

53 (36.6%)

77 (58.8%)*

365 (40.1%)


202 (50.8%)

153 (64.8%)*

87 (60.0%)

49 (37.4%)

491 (54.0%)


30 (7.5%)

14 (5.9%)

5 (3.4%)

5 (3.8%)

54 (5.9%)



Studying only

65 (16.4%)

39 (16.5%)

20 (13.8%)

36 (27.5%)*

160 (17.6%)

Working only

263 (66.4%)

179 (75.8%)*

101 (69.7%)

63 (48.1%)

606 (66.7%)

Studying and working

51 (12.9%)

14 (5.9%)


26 (19.8%)*

112 (12.3%)


17 (4.3%)

4 (1.7%)

3 (2.1%)

6 (4.6%)

30 (3.3%)

Education attainment


High school or less

277 (69.6%)*

146 (61.9%)

80 (55.2%)

75 (57.3%)

578 (63.5%)

College or vocational training

72 (18.1%)

59 (25.0%)

36 (24.8%)

22 (16.8%)

189 (20.8%)

University (bachelor, master's, Ph.D.)

49 (12.3%)

31 (13.1%)

29 (20.0%)

34 (26.0%)*

143 (15.7%)

Alcohol use

Alcohol drinking frequency



24 (6.4%)*

0 (0.0%)

0 (0.0%)

3 (2.5%)

27 (3.3%)


291 (77.6%)*

74 (37.9%)

82 (59.0%)

87 (73.7%)*

534 (64.6%)


60 (16.0%)


57 (41.0%)*

28 (23.7%)

266 (32.2%)

Binge drinking



164 (43.9%)*

22 (11.4%)


44 (37.3%)

249 (30.3%)


103 (27.5%)*

26 (13.5%)

31 (22.5%)

25 (21.2%)

185 (22.5%)



48 (24.9%)*

28 (20.3%)

24 (20.3%)

151 (18.3%)

Three to five times

35 (9.4%)

63 (32.6%)*

44 (31.9%)*

18 (15.3%)

160 (19.4%)

Six times or more


34 (17.6%)*

16 (11.6%)

7 (5.9%)

78 (9.5%)

Driving after cannabis use



31 (7.8%)

26 (11.0%)

16 (11.0%)

44 (33.6%)*

117 (12.9%)

Two to three times

61 (15.3%)

66 (28.0%)*

28 (19.3%)

38 (29.0%)*

193 (21.2%)

Four to six times

41 (10.3%)

51 (21.6%)*

22 (15.2%)

25 (19.1%)

139 (15.3%)

Seven times and more

265 (66.6%)*

93 (39.4%)

79 (54.5%)

24 (18.3%)

461 (50.7%)


Lack of premeditation

7.08 (2.59)a

9.17 (2.15)b

7.95 (2.31)c

7.62 (2.49)a,c

7.84 (2.57)


Positive urgency

8.84 (2.61)a

11.27 (1.97)b

9.61 (2.29)c

9.55 (2.65)c

9.70 (2.61)


Sensation seeking

10.49 (2.70)a

11.06 (2.18)b,c

10.73 (2.41)a,c

10.54 (2.53)a,c

10.69 (2.51)


Negative urgency

8.09 (3.00)a

11.33 (2.10)b

8.27 (2.79)a,c

9.16 (3.04)c

9.11 (3.08)


Lack of perseverance

7.00 (2.54)a

9.00 (2.08)b

7.76 (2.48)c

7.77 (2.57)c

7.75 (2.55)


Expectations of cannabis effects

Cognitive disturbances

1.97 (0.98)a

3.00 (0.90)b

2.59 (1.01)c

2.77 (1.08)b,c

2.45 (1.08)


Relaxation and tension reduction

4.54 (0.60)a

3.59 (0.92)b

4.05 (0.87)c

4.06 (0.83)c

4.15 (0.87)


Social facilitation

3.00 (0.91)a

3.09 (0.66)a

2.93 (0.89)a

3.17 (0.88)a

3.04 (0.85)


Perceptual amplification

3.61 (0.90)a

3.32 (0.80)b

3.43 (0.92)a,b

3.55 (0.92)a,b

3.50 (0.89)


Negative effects

1.33 (0.55)a

2.76 (0.93)b

1.73 (0.85)c

1.59 (0.69)c

1.80 (0.94)


Craving and physical effects

3.56 (1.03)a

3.42 (0.81)a

3.55 (0.86)a

3.67 (0.97)a

3.54 (0.94)


Attitudes and opinions towards DACU

DACU is a risky behaviour

2.07 (0.98)a

2.57 (0.82)b

2.48 (1.04)b

2.64 (1.03)b

2.35 (0.99)


DACU is associated with legal consequences

3.62 (1.27)a

3.34 (0.97)b

3.65 (1.14)a,b,c

3.97 (1.02)c

3.60 (1.16)


Motives for DACU

Low impact

2.29 (0.74)a

2.58 (0.58)b

2.25 (0.69)a

1.95 (0.75)c

2.31 (0.72)


Lack of alternative transport

2.05 (0.78)a

2.65 (0.56)b

2.19 (0.65)a

1.99 (0.72)a

2.22 (0.74)


Low risk

1.71 (0.70)a

2.49 (0.60)b

1.95 (0.64)c

1.71 (0.63)a

1.95 (0.73)


Designated driver

1.73 (0.76)a

2.50 (0.60)b

1.79 (0.61)a

1.70 (0.75)a

1.93 (0.77)


Perception intoxication

2.02 (0.67) a

2.60 (0.65)b

2.11 (0.64)a

1.92 (0.66)a

2.17 (0.71)



1.32 (0.58)a

2.35 (0.77)b

1.41 (0.61)a

1.31 (0.52)a

1.60 (0.77)


Peer influence

1.37 (0.50)a

2.30 (0.67)b

1.53 (0.49)c

1.35 (0.46)a

1.63 (0.68)


Peer influence

Peer approval


1 = will definitely disapprove


9 (3.8%)

8 (5.5%)

14 (10.7%)*

58 (6.4%)


19 (4.8%)

15 (6.4%)

8 (5.5%)

24 (18.3%)*

66 (7.3%)


40 (10.1%)

51 (21.6%)*

24 (16.6%)

28 (21.4%)

143 (15.7%)


83 (20.9%)

63 (26.7%)

36 (24.8%)

30 (22.9%)

212 (23.3%)


72 (18.1%)

55 (23.3%)

29 (20.0%)

26 (19.8%)

182 (20.0%)


64 (16.1%)*

19 (8.1%)

19 (13.1%)

5 (3.8%)

107 (11.8%)

7 = will definitely approve

93 (23.4%)*

24 (10.2%)

21 (14.5%)

4 (3.1%)

142 (15.6%)

Number of friends who DACU



31 (7.8%)

6 (2.5%)

8 (5.5%)

19 (14.5%)*

64 (7.0%)

Some of them

156 (39.2%)

123 (52.1%)*

63 (43.4%)

76 (58.0%)*

418 (45.9%)

Many of them

105 (26.4%)

54 (22.9%)

41 (28.3%)

26 (19.8%)

226 (24.8%)

All or almost all of them

53 (22.5%)

33 (22.8%)

10 (7.6%)

202 (22.2%)

Note. χ2 comparisons are made for each row reporting percentages. When a significant difference is observed in a row, the standardized Pearson residual is used to identify the cells representing a proportion different from the expected one. A standardized residual over 2 or less than -2 is statistically significant. This is denoted with an asterisk (*) when proportions are higher than expected and with a double dagger (‡) when proportions are lower than expected. For continuous variables, means in the row sharing subscripts (a, b or c) are not significantly different from each other.

* p < .05; ** p < .01; *** p < .001


No comments here
Why not start the discussion?