Associations between Adolescent Marijuana Use, Driving after Marijuana Use and Recreational Retail Sale in Colorado, USA

Abstract Objective We investigated associations between the retail distribution of recreational marijuana in Colorado and (i) past 30-day marijuana use and (ii) driving after marijuana use (DAMU) among a representative sample of public high school students using four waves of data from a state surveillance system. Methods Past 30-day marijuana use was assessed among all sampled students (n = 85,336). DAMU was assessed among students 15 years or older who indicated driving (n = 47,518). Modified Poisson regression with robust variance estimates was used to estimate prevalence ratios (PR) comparing the pre-distribution (2013) and post-distribution (2015, 2017, 2019) periods for marijuana-related behaviors. Frequency of behavioral engagement was assessed using a multinomial approach. Results An estimated 20.3% of students engaged in past 30-day marijuana use and 10.5% of student drivers engaged in DAMU. Retail distribution of recreational marijuana was not significantly associated with the prevalence of any marijuana use or DAMU. However, it was associated with 1.16 (95% CI: 1.04–1.29) times the prevalence of using marijuana one or two times in the last 30 days, 1.27 (1.03, 1.55) times the prevalence of DAMU one time, and 0.82 (0.69, 0.98) times the prevalence of DAMU six or more times. No significant associations were observed for the remaining frequency categories. Conclusions Approximately 1 in 10 students who drive reported DAMU. Varying prevalence in the frequency of past 30-day marijuana use and DAMU was observed following the retail distribution of recreational marijuana in Colorado. Care should be taken to properly educate adolescent drivers regarding the dangers of DAMU.


Introduction
The state-level legalization of recreational marijuana in the United States began in 2012 with the passage of legislation in Colorado and Washington.Now, as more states continue to legalize marijuana for medical and recreational purposes, the issues of youth marijuana use and its associated behaviors, including driving after marijuana use (DAMU), are receiving increased attention.While the prevalence of marijuana use among US high school students is well knownapproximately 21% of high school students reported currently using marijuana in the 2019 administration of the Youth Risk Behavior Survey (Jones et al., 2020)-studies investigating the effects of policies legalizing recreational marijuana use on behaviors and attitudes relating to marijuana use among adolescents have observed mixed results.
Nationally, Anderson et al. (2021) estimated recreational marijuana legalization was associated with 6% lower odds of marijuana use among adolescents.In Colorado, adolescents' perceived ease of access to marijuana was higher following recreational legalization, through their use behaviors and perception of harm remained unchanged (Brooks-Russell et al., 2019;Harpin et al., 2018).Vigil et al. (2018) found no change in adolescent marijuana use following legalization, though they observed increases in calls to poison control centers and hospitalizations involving marijuana among this age group, similar to Wang et al. (2018).In Washington state, evidence of increased use among 8th and 10th graders following legalization was observed using data from the Monitoring the Future survey (Cerdá et al., 2017).Dilley et al. (2019) found the opposite in their analysis of data from Washington's Healthy Youth Survey: recreational marijuana legalization was associated with lower marijuana use among 8th and 10th graders.These mixed findings may have arisen due to differences in methods, or they may reflect heterogeneity in retail laws' effect by state or over time.Therefore, continued research should monitor impact of retail marijuana sales across different states and time periods.
Adolescent DAMU is also of concern.Motor vehicle crashes are a leading cause of death among this age group (Centers for Disease Control and Prevention, National Centers for Injury Prevention and Control, 2021).Further, reckless driving behaviors commonly exhibited by younger drivers, such as cellphone-related distraction and alcohol use, tend to result in a higher crash risk compared to older drivers (Bingham, 2014;Rudisill & Zhu, 2015).Therefore, it is important to quantify the relationship between permissive marijuana policies and DAMU, and in turn driver impairment.Unfortunately, there is a paucity of literature on this topic, particularly regarding younger drivers in states where recreational marijuana was legalized.
Among adults, however, studies of self-reported DAMU reveal little evidence of its association with recreational marijuana legalization (Benedetti et al., 2021a(Benedetti et al., , 2021b)).Lensch et al. (2020) found DAMU was more prevalent in states that legalized recreational sales.However, they also found that users in these states planned ahead or decided to not drive while impaired more frequently than their counterparts in states without legal recreational sales (Lensch et al., 2020).Studies leveraging crash data suggest that recreational marijuana legalization may be associated with higher prevalence of THC in drivers in fatal motor vehicle crashes (Couper & Peterson, 2014), though not with higher motor vehicle fatality rates (Aydelotte et al., 2017;Hansen et al., 2018) or motor vehicle-related injuries (Callaghan et al., 2021).
These studies highlight the relative uncertainty surrounding the relationship between recreational marijuana legalization, marijuana use, and DAMU.Additionally, little is known about the long-term impact of recreational marijuana policies, due to the recency in which they have been enacted (Hasin, 2018).Furthermore, studies investigating marijuana impaired driving among adolescents are relatively scarce, none directly evaluate associations between recreational legalization and DAMU.This study aims to address these research gaps by analyzing responses to four waves of data (2013-2019) from a state-based surveillance system in Colorado, one of the first two states to legalize recreational marijuana use.Utilizing these data, we aim to investigate the association between the retail distribution of recreational marijuana in Colorado and (i) past 30-day marijuana use and (ii) DAMU among public high school students.

Participants
This study utilized data from the 2013, 2015, 2017, and 2019 administrations of the Healthy Kids Colorado Survey (HKCS).The HKCS is a biennial, 2-stage cluster sample of public high school students in Colorado, United States.In the first stage, a sampling frame of all public high schools in the state was divided into strata based on Colorado's health statistics regions.Schools were then ordered by their enrollment size before being systematically sampled.Within selected schools, a course or period common across all students was chosen, then a simple random sample was used to select sessions in which all students were administered the HKCS.In most districts, parents were provided information about the survey and could opt-out their students, while a few districts required parents to provide active consent.
Each administration of the HKCS contained two versions, the High School Module A, and the High School Module B. Both versions share a common set of core questions along with a different set of supplementary questions.For example, the question regarding DAMU was included in Module A only for 2013-2017 before being included as a core question in 2019.

Marijuana-related behaviors
Past 30-day marijuana use was assessed using the question "During the past 30 days, how many times did you use marijuana?"with responses ranging from 0, 1 or 2, 3 to 9, 10 to 19, 20 to 39, and 40 or more times.DAMU was assessed using the question "During the past 30 days, how many times did you drive a car or other vehicle when you had been using marijuana?"with responses "I did not drive a car or other vehicle during the past 30 days, " 0, 1, 2 or 3, 4 or 5, and 6 or more times.Students who did not drive were not included in the 0 times category for DAMU.

Retail distribution of recreational marijuana
On November 6th, 2012 Colorado voters passed a ballot measure legalizing recreation marijuana for adults age 21 year older within the state, effective December 10th, 2012 (Task Force on the Implementation of Amendment 64, 2013).While its use was legal from this point on, recreational marijuana was not available for retail purchase until January 1st, 2014 (Cerdá et al., 2017;Livingston et al., 2017).Therefore, we utilized January 1st, 2014 to assign a policy variable to participating students, with those taking part in the 2013 HKCS categorized as pre-retail-distribution and those responding to the 2015, 2017, and 2019 cycles categorized as post-retail-distribution (henceforth pre-distribution and post-distribution respectively).

Driving status
Driving status was assessed through responses to two driving-related questions: the aforementioned DAMU question and "During the past 30 days, how many times did you drive a car or other vehicle when you had been drinking alcohol?" Students were categorized as non-drivers (responded "did not drive" to both questions) or drivers (any other response to both questions), while those with inconsistencies between the questions were excluded.

Demographics
Age was assessed through the question "How old are you?" with response options: "12 years old or younger, " "13 years old, " "14 years old, " "15 years old, " "16 years old, " "17 years old, " and "18 years old or older." Students aged 14 years or younger were collapsed into a single level due to small sample size.Sex was determined through the question "What is your sex?" and categorized as male and female.Race/ethnicity was assessed using the questions "Are you Hispanic or Latino?" and "What is your race?" with responses stratified into four categories: non-Hispanic White, non-Hispanic Black, Hispanic/Latino, and non-Hispanic Other.

Analytic population
Past 30-day marijuana use was assessed among the entire sample of students (n = 85,336).DAMU was assessed among students age 15 years or older who indicated driving in the last 30 days (n = 47,518), consistent with Colorado's minimum age for learner's permit licensure (Governors Highway Safety Administration, 2021).

Statistical analysis
Survey weighted procedures were used in conjunction with a finite population correction for sampled public high schools to account for the HKCS's complex design and produce representative estimates of the analytic populations at the state-level.Weighted prevalence estimates and their corresponding 95% confidence intervals (CIs) were estimated for survey year and demographics among all students and student drivers age 15+ overall, stratified by marijuana-related behaviors.Additionally, the unadjusted prevalence difference between policy periods was calculated for each outcome.
We estimated the association between recreational marijuana distribution in Colorado and marijuana-related behaviors using modified Poisson regression (Zou, 2004).As missing data was below acceptable levels (<10%), a complete case analysis was conducted.Associations were initially assessed for each outcome as dichotomized responses (1 or more time relative to 0 times).Specifically, the linear portion of our model took the following form: x where Y i is an indicator that respondent i self-reported the outcome, either past 30-day marijuana use or DAMU; PostLaw i is an indicator that the respondent took the survey after recreational marijuana distribution began; and x i is a vector encoding the respondent's sex, race/ethnicity, and age.
To further investigate the impact of Colorado's recreational marijuana distribution on the frequency of marijuana-related outcomes, we generalized the modified Poisson regression framework to a multinomial setting (Blizzard & Hosmer, 2007).Much like a generalized logit model, separate modified Poisson regression models were fit to estimate the adjusted prevalence of each non-zero level of the outcome variables relative to "0 times." As an illustrative example, let Y i ( ) 1 2 − equal one if respondent i reported marijuana use one or two times in the last month, and zero if they reported marijuana use zero times in the last month.Y i − , and Y i ( ) 4 0 + are defined similarly.Our multinomial regression approach proceeded by fitting the following models: Similar models were used to model DAMU, with outcome categories ranging from 1 time to 6 or more times.A multinomial regression approach was chosen over an ordinal approach as the proportional odds assumption was not met for the outcomes of interest.Stata 16.0 (StataCorp LLC., College Station, TX) was used for all analyses, with statistical significance set at an alpha level of 0.05.

Descriptive analysis
Within Colorado, 20.3% (95% CI: 19.6, 20.9) of public high school students engaged in past 30-day marijuana use at the time of their survey (Table 1).The yearly prevalence of past 30-day marijuana use remained relatively consistent over the study period (Figure 1).Relative to the pre-distribution period, past 30-day marijuana use was 1.0% (-0.3, 2.4) higher following the retail distribution of recreational marijuana, though this difference was not statistically significant.Students aged 16, 17, or 18 years or older had a similar prevalence of past 30-day marijuana use, ranging from 24% to 25% (Table 1).Past 30-day use was also similar among male and female students.Lastly, students who drive had a higher prevalence of past 30-day marijuana use (23.2%: 22.4, 24.0) compared to those who do not drive (13.9%: 12.9, 15.0).Demographics for past 30-day marijuana use among drivers age 15 year or older can be found in Supplementary Table 1.Among Colorado's public high school students 52.6% (51.5, 53.7) were drivers aged 15 years or older.Within this subpopulation of students, 10.5% (10.0, 11.0) engaged in DAMU, with yearly prevalence estimates collected in Figure 1.The prevalence of DAMU was 10.9% in the pre-distribution period and 10.3% in the post-distribution period (Table 2).When comparing these two periods, no statistically significant difference in prevalence was observed.Consistent with the overall sample, the prevalence of DAMU was lowest for 15-year-old drivers (6.0%) and highest for those age 18 years or older (15.7%).A higher prevalence of male students aged 15 years or older engaged in DAMU (12.3%) compared to female (8.5%).Finally, when stratified by race/ethnicity, the prevalence of DAMU ranged from 9.7% (non-Hispanic White students) to 12.2% (non-Hispanic, Other).

General engagement in marijuana-related behaviors
Among all high school students, retail distribution of recreational marijuana was not statistically significantly associated with a higher prevalence of past 30-day marijuana use.Students in the post-distribution period had 1.06 (0.99, 1.14) times the prevalence of past 30-day marijuana use compared to students in the pre-distribution period, controlling for age, sex, and race/ethnicity (Table 3).Similarly, no statistically significant association was observed between retail distribution and the prevalence of DAMU (aPR: 0.95; 0.85, 1.05).

Frequency of engagement in marijuana-related behaviors
When utilizing the full range of response options through multinomial regression, the adjusted prevalence of using marijuana 1 or 2 times in the last 30 days (relative to 0  times, excluding 3 or more times) following the retail distribution recreational marijuana in Colorado was 1.16 (1.04, 1.29) times that of the pre-distribution period.No such change was observed for higher frequencies of past 30-day marijuana use (Table 3).DAMU saw a similar increase; student drivers age 15+ had 1.27 (1.03, 1.55) times the prevalence of DAMU one time (relative to 0 times, excluding 2 or more times) following retail distribution.Higher frequencies of DAMU exhibited a lower prevalence following retail distribution (Table 4); the prevalence of driving 6 or more times (relative to 0 times, excluding 5 times or less) was 0.82 (0.69, 0.98) times lower after enactment of the law.

Discussion
This study leveraged a state surveillance system, over nearly a decade, to investigate associations between the retail distribution of recreational marijuana in Colorado, past 30-day marijuana use, and DAMU among adolescents.Using one wave of baseline data from before retail distribution, and three waves of data post-distribution, we found retail distribution was associated with increases in infrequent past 30-day marijuana use and DAMU and a statistically significant decrease in frequent DAMU.Although marijuana use remained illicit for high school aged students during the study period, we hypothesized the retail distribution of recreational marijuana for adults aged 21 and older, and the resulting normalization of the    behavior, as seen in previous samples of youth (Wong et al., 2020) and young adults (Koval et al., 2019;Wallace et al., 2020), would impact the overall prevalence of marijuana-related behaviors.We found little support for this hypothesis, as the prevalence of past 30-day marijuana use was relatively consistent across the study period.Our results were consistent with previous studies involving Colorado and other states with recreational marijuana programs that have either found no significant change in the prevalence of marijuana use (Brooks-Russell et al., 2019;Harpin et al., 2018), mixed findings by grade (Cerdá et al., 2017;Dilley et al., 2019), or variation in the direction of policy effects (Anderson et al., 2019;Dilley et al., 2019).
Previous studies and data have shown that this lack of an association is not mirrored among adults.The National Survey on Drug Use and Health identified an increase in past-month marijuana use among young adults (ages 18 to 25 years), from 19.1% in 2013 to 23.0% in 2019 (Substance Abuse & Mental Health Service Administration, 2019).Within Colorado, adult (18 years old and up) marijuana use increased from 13.6% in 2014 to 17.5% in 2018(Colorado Department of Public Health & Environment, 2021).This suggests retail distribution differentially affected adults as compared to adolescents, consistent with the target demographic of the legislation.That said, using a multinomial approach we identified a higher prevalence of infrequent marijuana use (one time in the past 30 days) following retail distribution.
When investigating DAMU, we observed an increase in the prevalence of infrequent DAMU (one or two times in the past 30 days) and a decrease in frequent DAMU (six or more times in the past 30 days).There are few studies regarding the effect of marijuana legalization on adolescent impaired driving, and this is the first we are aware observing these patterns.Less harmful perceptions of marijuana use are associated with DAMU (Cantor et al., 2021) and these perceptions are becoming more prevalent among high school students (Sarvet et al., 2018).It is possible that frequent engagers in DAMU had prior perceptions of marijuana use that were robust to more permissive marijuana policy, leading to no change or less engagement with the behaviors over time.Meanwhile, infrequent users may be more susceptible to changing social norms as a result of permissive marijuana policy, increasing their prevalence.Further research into relationship between marijuana policy and frequency of engagement in marijuana-related behaviors is necessary to better contextualize these results.
The prevalence of DAMU was relatively high among student drivers; with approximately 1 in 10 engaging in the behavior.As a comparison, the national prevalence of driving after alcohol use among student drivers is 5.4% (Centers for Disease Control and Prevention, 2021).The prevalence of DAMU is concerning because adolescent and young adults are disproportionately involved in fatal crashes (National Highway Traffic Safety Administration, 2020).In a 2019 study, Brubacher et al. (2019) noted inexperienced marijuana users may be more prone to its acute effects and could be less capable in compensating for these effects when driving.With the increase in both infrequent use and DAMU among adolescents following retail distribution, it is possible that marijuana-related motor vehicle crashes in the state, particularly among this age group, may disproportionately involve infrequent, inexperienced marijuana users.
Regardless of a state's marijuana policy status, more can be done to prevent adolescent impaired driving, from marijuana or any substance.First, there are policies that impact youth access to substances and impaired driving behaviors, beyond overall legalization status.These include policies affecting retailers such as required retailer trainings, compliance checks of retailers, and associated penalties of underage sales (Harding et al., 2016).Other policy levers known to impact youth substance use include increasing taxes, although there can be concerns that raising taxes too high could lead to unintended consequences for illegal market demand (Caulkins & Kilmer, 2016;Caulkins et al., 2012).
Also effective at preventing impaired driving are zero tolerance laws for those under 21 (Alderman and Johnston, 2018;Fell et al., 2016) and increased graduated drivers' license requirements and delays to full licensure (Chen et al., 2006).Prevention programming can also be implemented at schools to prevent impaired driving and have been found to be particularly effective at preventing riding with impaired drivers (Elder et al., 2005;Venkatraman et al., 2021).Brief interventions, often implemented by health care providers, have been found to prevent impaired driving related to alcohol use, likely mediated by reducing problematic substance use (Steinka-Fry et al., 2015).
Finally, parents have an important role to play in prevention by talking with their children about expectations for driving, closely monitoring their vehicle access and choices about passengers, and ensuring evidence-based policies such as graduate driver licensing laws are being followed (Zhu et al., 2013).The Substance Abuse and Mental Health Services Administration offers resources to parents such as creating family agreements related to substance use.Continued identification and implementation of effective strategies by policy makers, physicians, parents, and the community can help prevent fatalities from impaired driving (National Research Council (US) and Transportation Research Board, 2007).

Limitations
This study had several limitations.One such limitation was data availability.Although the baseline wave of data (2013) occurred before the retail distribution of recreational marijuana in 2014, medical marijuana has previously been sold in the state of Colorado.Therefore, secular trends in the perception of risk and permissiveness attitudes toward of marijuana use may have been in motion before the study period.Study data came from a repeated cross-sectional surveillance system.While useful for examining social trends, they do not indicate individual changes in behaviors over time.Moreover, the study design and data do not allow us to attribute observed associations directly to causal effects of the policy change.HKCS only includes data on Colorado students, therefore we were unable to incorporate a control state that could have strengthened any causal interpretations.Next, students were asked about having driven a car after using marijuana.It is not known if students were impaired or to what degree they experienced impairment, only that they reported DAMU.Additional limitations inherent to similar survey research apply, such as the potential for social desirability bias given the relatively sensitive and illicit nature of DAMU among adolescents.In addition, response rates for certain years were lower than desired, especially in 2017 (when seven schools were excluded due to administering a truncated version of the HKCS), potentially impacting generalizability.Finally, while these results are representative of Colorado's public high school students further research is needed to investigate potential policy effects nationwide.

Conclusions
We found evidence of associations of both the prevalence infrequent adolescent marijuana use and DAMU with the retail distribution of recreational marijuana in Colorado.Additionally, retail distribution of recreational marijuana was associated with a lower prevalence of high frequency DAMU among student drivers.Over the study period, one in ten high school age drivers engaged in DAMU, which is concerning given the high risk of motor vehicle-related injury and death arising from impaired driving among adolescents.Further action should be taken to properly communicate the risks of DAMU to teen drivers.

Figure 1 .
Figure 1. this figure collects the weighted prevalence of marijuana use in the last 30 days (past 30-day marijuana use) among all high school students and the weighted prevalence of driving after marijuana use among student drivers 15+ years of age.Prevalence estimates are presented along with their 95% confidence intervals for each survey year.
a among all students who drive, age 15 years or older.b comparing post-distribution period(2015, 2017, 2019)  to pre-distribution period (2013).

Table 1 .
Demographics of colorado high school students, by past 30-day marijuana use, Healthy Kids colorado Survey 2013 to 2019.
a column percentage may not sum to 100 due to rounding.b Row percentage.c absolute difference in prevalence, post-distribution vs. pre-distribution.

Table 2 .
Demographics of colorado high school student drivers (age 15+), by driving after marijuana use engagement, Healthy Kids colorado Survey 2013 to 2019.absolute difference in prevalence, post-distribution vs. pre-distribution.
a column percentage may not sum to 100 due to rounding.b Row percentage.c

Table 3 .
association between the expansion of recreational marijuana sales in colorado and past 30-day marijuana use, Healthy Kids colorado Survey 2013 to 2019.

Table 4 .
association between the expansion of recreational marijuana sales in colorado and driving after marijuana use, Healthy Kids colorado Survey 2013 to 2019.