Abstract
Background: With the recent legalization of recreational cannabis in Canada, cannabis-impaired driving is an important public safety concern. Our aim was to examine the association between recreational cannabis legalization and fatal motor vehicle collisions using data from the United States, which present a timely natural experiment of cannabis legalization.
Methods: We conducted an ecologic study using the number of fatal motor vehicle collisions and the associated number of deaths for US jurisdictions with legalized recreational cannabis (2007–2018) retrieved from the US Fatality Analysis Reporting System. We examined jurisdiction-specific rates of fatal motor vehicle collisions and associated deaths before and after recreational cannabis legalization using Poisson regression and meta-analyzed estimates across jurisdictions using DerSimonian and Laird random-effects models.
Results: After adjustment for calendar year, legalization was associated with increases in rates of fatal motor vehicle collisions (incidence rate ratio [IRR] 1.15, 95% confidence interval [CI] 1.06–1.26) and associated deaths (IRR 1.16, 95% CI 1.06–1.27). Differences between the first 12 months after legalization relative to subsequent months were inconclusive for rates of fatal motor vehicle collisions (IRR 0.92, 95% CI 0.84–1.02) and associated deaths (IRR 0.92, 95% CI 0.84–1.01).
Interpretation: Recreational cannabis legalization in the US was associated with a relative increased risk of fatal motor vehicle collisions of 15% and a relative increase in associated deaths of 16%, with no conclusive difference between the first and subsequent years after legalization. These findings raise concern that there could be a similar increase in fatal motor vehicle collisions and associated deaths in Canada following recreational cannabis legalization.
Canada legalized recreational cannabis nationally on Oct. 17, 2018. By 2019, more than 5.1 million Canadians (16.8% of the population aged ≥ 15 yr) reported cannabis use in the previous 3 months (compared to 14.9% before legalization).1 Although legalization better aligned the law with Canadian values and practices, it raised public health concerns regarding impaired driving.2 Of people who reported any cannabis use in the previous 3 months, 13.2% reported driving within 2 hours of consuming cannabis; this proportion increased to 28.8% among daily or almost daily cannabis users.1 The relation between cannabis use and impaired driving is complicated by varied methods of consumption (e.g., smoking, vaping, edibles), individual factors (e.g., metabolism, frequency of use) and time since consumption, as well as challenges in the reliable assessment of cannabis impairment.3 However, cannabis is known to affect psychomotor skills necessary for driving, including critical tracking, divided attention and reaction time.4
Despite this knowledge, few studies have examined the impact of large-scale public policies that increased access to cannabis on impaired driving and related outcomes. Given the varied approaches taken to cannabis regulation by different jurisdictions in the United States,5 data from the US present a timely natural experiment to assess the impact of recreational cannabis legalization. Our objective was to examine the association between recreational cannabis legalization and the rate of fatal motor vehicle collisions in the US to inform impaired-driving policy and public health prevention measures in Canada.
Methods
Design and setting
We performed an ecologic study to examine the association between recreational cannabis legalization and fatal motor vehicle collisions and associated deaths in the US between 2007 and 2018. Legalization was in effect in 11 jurisdictions (10 states and the District of Columbia) before 2019 (Table 1).29
Data sources
We retrieved the number of fatal motor vehicle collisions and of deaths due to motor vehicle collisions (as each collision could result in more than 1 fatality) by month for US jurisdictions with legalized recreational cannabis from the US Fatality Analysis Reporting System (FARS).30 The FARS database includes data on collisions that occur on public roadways and result in at least 1 death within 30 days. We selected 2007 as the start of our study period to include at least 5 years before the first year in which legalized recreational cannabis was in effect in any jurisdiction (2012). Although recreational cannabis legalization came into effect in other states after 2019, we did not include those jurisdictions because the most recent year of FARS data available at the time of analysis (September 2020) was 2018.
The FARS database is the only source of national-level statistics on police-reported fatalities in the US, with data collected through collaborative agreements between the National Highway Traffic Safety Administration and each state.31 Data collection forms for FARS are completed by personnel trained by the National Highway Traffic Safety Administration using multiple source documents (e.g., police, medical and administrative reports) and entered into a standardized, validated database.31,32
Exposure definitions
We defined exposure using a time-dependent approach with 2 categories: legalized recreational cannabis and no legalized recreational cannabis (reference group). We defined legalization for each jurisdiction as the date on which recreational cannabis legalization was in effect, rounded up or down to the nearest full month. Each jurisdiction contributed person-time of observation (determined with the use of the annual population of each jurisdiction for each year between 2007 and 2018)30 and data on fatal motor vehicle collisions to both categories (legalized and nonlegalized) in the analysis.
Statistical analysis
We estimated the rates of fatal motor vehicle collisions and associated deaths within each jurisdiction before and after recreational cannabis legalization. Rates were estimated per 100 000 person-years for years during which recreational cannabis was legal and not legal for each included jurisdiction. We then determined crude and adjusted incidence rate ratios (IRRs) and 95% confidence intervals (CIs) for each jurisdiction using a generalized linear model with a log link and Poisson distribution by means of the proc genmod procedure in SAS (version 9.4) (SAS Institute) for the associations between recreational cannabis legalization and fatal motor vehicle collisions and fatalities due to motor vehicle collisions. We adjusted models for calendar year, which was modelled as a continuous variable, to account for underlying time trends (e.g., baseline trends in substance use and driving, impaired-driving policies). We then meta-analyzed jurisdiction-specific estimates using DerSimonian and Laird random-effects models with inverse variance weighting. We conducted the meta-analysis in R (R Foundation for Statistical Computing) using the meta package.
We performed 2 secondary analyses. In the first, we defined exposure using the date when commercial cannabis dispensaries opened (rounded up or down to the nearest full month) instead of the date that legalization came into effect. Commercial dispensaries were open in 7 states before 2019 (Alaska, California, Colorado, Massachusetts, Nevada, Oregon and Washington). The remaining jurisdictions did not plan to allow dispensaries or opened (or planned to open) dispensaries after the study period.
In the other secondary analysis, we subcategorized data from the months during which cannabis was legal (or the months during which commercial dispensaries were opened) into the first 12 months and subsequent months to assess whether any observed increased risk was short term. We then compared rates from the first 12 months and subsequent months to each other and to those from nonlegalized periods.
In sensitivity analyses, we used a fixed-effects rather than a random-effects model. We also examined temporal trends in the incidence of fatal motor vehicle collisions and associated deaths before legalization to visually assess whether these were log-linear. We assessed colinearity between calendar year and legalization and open dispensaries using the variance inflation factor.
Ethics approval
Ethics approval was not required for this research as it used publicly available data.
Results
A total of 73 982 fatal motor vehicle collisions resulting in 80 402 deaths occurred between 2007 and 2018 in the 11 US jurisdictions that had recreational cannabis legalization in effect before 2019 (Table 2). Of these collisions, 17 116, resulting in 18 580 deaths, occurred during periods when recreational cannabis was legal (Table 2; Appendix 1, Supplemental Tables S1–S4, available at www.cmajopen.ca/content/9/1/E233/suppl/DC1).
After adjustment for calendar year, legalization was associated with increased rates of fatal motor vehicle collisions (IRR 1.15, 95% CI 1.06–1.26) (Figure 1A) and deaths from motor vehicle collisions (IRR 1.16, 95% CI 1.06–1.27) (Figure 1B). When we used the date commercial dispensaries opened rather than the date legalization came into effect in the analyses, the findings were similar for the risk of motor vehicle collisions (IRR 1.18, 95% CI 1.06–1.32) (Figure 2A) and associated deaths (IRR 1.18, 95% CI 1.06–1.32) (Figure 2B).
After adjustment for calendar year, the difference in the first 12 months versus subsequent months after legalization was inconclusive for rates of fatal motor vehicle collisions (IRR 0.92, 95% CI 0.84–1.02) and deaths from motor vehicle collisions (IRR 0.92, 95% CI 0.84–1.01) (Table 3). When we used the date dispensaries opened rather than the date legalization came into effect in the analyses, we likewise found an inconclusive difference in rates of fatal motor vehicle collisions (IRR 0.88, 95% CI 0.76–1.02) and associated deaths (IRR 0.88, 95% CI 0.76–1.01) in the first 12 months versus subsequent months (Table 3).
Sensitivity analysis
Sensitivity analyses that used fixed-effects models showed results similar to those of the random-effects models for both rates of fatal motor vehicle collisions and deaths associated with recreational cannabis legalization (Figure 1) and the opening of dispensaries (Figure 2). Temporal trends in the incidence of fatal motor vehicle collisions and associated deaths before legalization were roughly log-linear, although mild departures were present in some jurisdictions. The variance inflation factor between calendar year and legalization for fatal collisions ranged from 1.02 in Michigan to 4.03 in Colorado and Washington (Appendix 1, Supplemental Table S5). The IRR for calendar year and fatal motor vehicle collisions ranged from 0.90 (95% CI 0.85–0.96) in the District of Columbia to 1.003 (95% CI 0.99–1.01) in Michigan, with a similar range for associated deaths. Similar findings were obtained for the variance inflation factor for opening of dispensaries.
Interpretation
We found that recreational cannabis legalization in US jurisdictions was associated with a relative increased risk of fatal motor vehicle collisions of 15% and a relative increase in associated deaths of 16%. From the FARS database, in 2018, an estimated 30 270 motor vehicle fatalities occurred in jurisdictions without legalized recreational cannabis. Our results suggest that legalization on a national scale could result in an additional 4843 motor vehicle fatalities per year in the US.
Previous literature concerning the association between cannabis legalization and impaired driving is limited. A small number of observational studies suggest that medical or recreational cannabis legalization may increase the proportion of drivers with detectable levels of cannabis in analyses of blood or urine; however, these observations may be confounded by increases in enforcement activities and in cannabis testing and reporting following legalization.33–36 Likewise, an increase in detectable levels of cannabis (e.g., owing to increased use following legalization) would not necessarily correspond directly to increases in impaired driving.3 Another study showed an increase in self-reported driving under the influence of cannabis following legalization.37 However, people may be more likely to report cannabis use and temporally associated driving when cannabis use is legal.
Several previous studies in which ecologic approaches were used did not rely on the measurement and reporting of the presence of cannabis or data on self-reported impaired driving. Aydelotte and colleagues38 used a difference-in-differences approach to compare changes in fatal motor vehicle collisions associated with legalization, using FARS data (2007–2017) from Colorado and Washington and 9 control states. Although their results were not conclusive, their analysis suggested a potential increase in fatal collisions of 1.2 per billion vehicle miles travelled (95% CI −0.6 to 2.1) in Colorado and Washington in the 5 years after recreational cannabis legalization. When they used the date of opening of commercial dispensaries, legalization was associated with an increase in fatal collisions of 1.8 per billion vehicle miles travelled (95% CI 0.4–3.7). Santaella-Tenoirio and colleagues39 examined similar data (2005–2017) from Colorado and Washington, which they compared to synthetic reference groups (created from a pool of comparison states whose fatality rates best resembled those of the exposed states before legalization). They found an increase in motor vehicle fatalities in Colorado (1.5/billion vehicle miles travelled, p = 0.047) but not in Washington (0.08/billion vehicle miles travelled, p = 0.7) following legalization of recreational cannabis retail sales. They hypothesized that higher retail density, cannabis use and cannabis tourism in Colorado relative to Washington may have contributed to the observed increase in fatalities from collisions in Colorado.
Two additional ecologic studies included data from other states with recreational cannabis legalization. Kamer and colleagues40 used a difference-in-differences approach to compare motor vehicle fatalities in the FARS database (2008–2018) in Colorado, Washington, Oregon and Alaska to those in the 20 states without legalized recreational or medical cannabis as of the beginning of 2018. After adjusting for covariates, they found an increase in motor vehicle fatalities of 2.1 per billion vehicle miles travelled (95% CI 1.3–3.0) following the opening of recreational cannabis retail stores. Lane and colleagues41 used data from the Centers for Disease Control and Prevention’s Wide-ranging Online Data for Epidemiologic Research Web application and RoadSafetyBC to examine changes in traffic fatalities between 2009 and 2016 in 3 states with recreational cannabis legalization (Colorado, Washington and Oregon), using neighbouring jurisdictions without legalization as comparators. They conducted an interrupted time-series analysis of traffic fatality rates using generalized least squares regression models adjusted for seasonality and autocorrelation to test the association between legalization and traffic fatalities. They found a pooled step increase of 1.08 traffic fatalities per million residents (p < 0.001), which suggested an increase in traffic fatalities less than 1 year after legalization, followed by a reduction in trend of 0.06 per month (p < 0.001).
Our analysis extends the findings of these previous studies by including an additional 7 jurisdictions with recreational cannabis legalization through 2018. Overall, our results support previous literature suggesting that recreational cannabis legalization is associated with increased impaired driving. Although our findings suggest that there may be some variability between jurisdictions in increased risk, IRRs for 10 of the 11 jurisdictions were greater than 1.0 (range 1.01–1.92) for fatal motor vehicle collisions following legalization. Michigan was the only jurisdiction with a point estimate that suggested a lower risk after legalization (IRR 0.80); however, recreational cannabis was legal for only 1 month in that state before 2019. Among jurisdictions with recreational cannabis dispensaries, all point estimates suggested increased rates of fatal motor vehicle collisions after the opening of dispensaries (IRR range 1.02–1.36).
In contrast to the findings of Lane and colleagues,41 our findings suggest that the increase in fatal motor vehicle collisions may be sustained over time. We found no evidence of a transient increase in rates of fatal collisions and associated deaths in the first 12 months following legalization or the opening of commercial dispensaries compared to subsequent months. However, our estimates suggest that there may be lower rates of fatal collisions and associated deaths in the first 12 months after legalization relative to subsequent months. This may be due to relatively limited cannabis availability in many jurisdictions during the first years after legalization and opening of dispensaries (e.g., fewer stores open, product shortages owing to high demand), which might deter new cannabis users in particular.
In Canada, 1922 motor vehicle fatalities were reported in 2018;42 a relative increase of 16% would correspond to 308 additional deaths yearly. However, the national approach taken to recreational cannabis legalization may mitigate increases in impaired driving in Canada. For example, Bill C-46 (enacted in 2018) established national per se driving limits for tetrahydrocannabinol (THC), with corresponding minimum penalties for exceeding these limits.43 Although these limits are controversial owing to a lack of direct correlation between THC levels and driving impairment,44,45 public awareness of the establishment of legal limits alone has the potential to decrease substance-impaired driving.46,47 Of the 11 US jurisdictions included in our analyses, 7 did not have a threshold limit for THC.6 Overall, our analysis suggests that Canada should remain vigilant regarding the potential for increases in cannabis-impaired driving.
Limitations
Our study was observational, and there were several factors that varied between jurisdictions and were likely to influence fatal motor vehicle collisions (e.g., substance use, population density, speed limits). Our jurisdiction-specific analyses were self-controlled and adjusted for calendar year to account for temporal trends; however, confounding remains possible. Although we found increased rates of fatal motor vehicle collisions and associated deaths following legalization, there was an amplification of the effect after we controlled for calendar year.
In contrast to other investigators, we did not select neighbouring or matched control jurisdictions as comparators; rather, we used the prelegalization period (minimum of 5 yr) for each jurisdiction to serve as the comparator for the period following legalization. Natural trends in the incidence of fatal motor vehicle collisions and associated deaths were roughly log-linear before legalization in most jurisdictions. Although mild departures from linearity were present in some jurisdictions, to facilitate analyses across jurisdictions, we selected the modelling approach that was most appropriate for the most jurisdictions. Given the growing number of US jurisdictions with legalized cannabis, this approach minimizes the effect of spillover or contamination (i.e., residents of neighbouring states without legalization consuming cannabis purchased in states with legalized cannabis) on the analyses of association.
Conclusion
Our findings suggest that there is the potential for an important increase in fatal motor vehicle collisions following the legalization of recreational cannabis in Canada. Although differences between the US and Canada may mitigate potential increases in impaired driving, the observed 15% relative increase in fatal motor vehicle collisions and 16% relative increase in associated deaths could result in 308 additional road fatalities annually in Canada.
Acknowledgement
The authors thank Hanan Abramovici PhD for his helpful suggestions.
Footnotes
Competing interests: None declared.
This article has been peer reviewed.
Contributors: Kristian Filion and Mark Eisenberg conceived of the study and supervised it. Pauline Reynier and Josselin Cabaussel analyzed the data. Sarah Windle drafted the manuscript. All of the authors contributed to the study design and data interpretation, approved the final version to be published and agreed to be accountable for all aspects of the work.
Funding: This study was funded by grant SHI-155407 from the Canadian Institutes of Health Research. Sarah Windle is supported by a Maysie MacSporran Graduate Studentship, an Alma Mater Fellowship and a Graduate Excellence Award from McGill University. Kristian Filion is supported by a Senior Research Scholar award from the Fonds de recherche du Québec – Santé (FRQS) and a William Dawson Scholar award from McGill University. Brett Thombs is supported by an FRQS research salary award.
Data sharing: Data used in this study were accessed from the publicly available United States Fatality Analysis Reporting System database. These data are available at https://www-fars.nhtsa.dot.gov/Main/index.aspx.
Supplemental information: For reviewer comments and the original submission of this manuscript, please see www.cmajopen.ca/content/9/1/E233/suppl/DC1.
This is an Open Access article distributed in accordance with the terms of the Creative Commons Attribution (CC BY-NC-ND 4.0) licence, which permits use, distribution and reproduction in any medium, provided that the original publication is properly cited, the use is noncommercial (i.e., research or educational use), and no modifications or adaptations are made. See: https://creativecommons.org/licenses/by-nc-nd/4.0/
References
- © 2021 Joule Inc. or its licensors