This originally appeared at https://pubmed.ncbi.nlm.nih.gov/36908044/
Authors: Madison Walker 1 2 , Olli Saarela 3 , Robert Mann 4 , Melissa Carpino 1 , Michael D Cusimano 1 2 3
PMID: 36908044 DOI: 10.1111/add.16188
Abstract
Aims
To measure the impact of Canada’s recreational cannabis legalization (RCL) in October 2018 and the subsequent impact of the coronavirus disease 2019 (COVID-19) lockdowns from March 2020 on rates of emergency department (ED) visits and hospitalizations for traffic injury.
Design
An interrupted time series analysis of rates of ED visits and hospitalizations in Canada recorded in population-based databases from January/April 2010 to March 2021.
Setting
ED visits in Ontario and Alberta and hospitalizations in Ontario, Alberta, British Columbia, the Prairies (Manitoba and Saskatchewan) and the Maritimes (Nova Scotia, New Brunswick, Newfoundland and Prince Edward Island).
Participants
Monthly counts of presentations to the ED or hospital for motor vehicle injury or pedestrian/cyclist injury, used to calculate monthly rates per 100 000 population.
Measurements
An occurrence of one or more International Statistical Classification of Diseases and Related Health Problems, 10th Revision, Canada (ICD-10-CA) code for motor vehicle injury (V20–V29, V40–V79, V30–V39 and V86) and pedestrian/cyclist injury (V01–V09 and V10–V19) within the National Ambulatory Care Reporting System and Discharge Abstract Database.
Findings
There were no statistically significant changes in rates of ED visits and hospitalizations for motor vehicle or pedestrian/cyclist injury after RCL after accounting for multiple testing. After COVID-19, there was an immediate decrease in the rate of ED visits for motor vehicle injury that was statistically significant only in Ontario (level change β = −16.07 in Ontario, 95% CI = −20.55 to −11.60, P = 0.000; β = −10.34 in Alberta, 95% CI = −17.80 to −2.89, P = 0.008; α of 0.004) and no changes in rates of hospitalizations.
Conclusions
Canada’s recreational cannabis legalization did not notably impact motor vehicle and pedestrian/cyclist injury. The rate of emergency department visits for motor vehicle injury decreased immediately after COVID-19 lockdowns, resulting in rates below post-recreational cannabis legalization levels in the year after COVID-19.
INTRODUCTION
Road traffic injury is one of the top five global causes of death for those ages 5 to 29 [1] and one of the top three leading causes of injury-related hospitalization in Canada [2]. Driving under the influence of drugs is a particular public health concern for road traffic safety, especially because of the recent nationwide enactment of recreational cannabis legalization (RCL) in Canada in October 2018, which has increased access to cannabis products. Because of the pharmacodynamic properties of cannabis, its use is associated with impaired motor function, increased reaction time and risky decision making [3]. As such, there is anincrease in the risk of serious injury or death from motor vehicle incidents when driving under the influence of cannabis [4–6]. Other jurisdictions have implemented RCL, including Uruguay nationwide in 2013 and Colorado and Washington State in the United States (US) in 2012 followed by other states. The impact of RCL on traffic fatality has been the primary focus of injury-related studies. Four studies found a significant increase in traffic fatalities after RCL in the US and Uruguay [7–10], two found an increase in traffic fatalities after RCL only in certain states [11, 12] and three did not report any significant changes in traffic fatality in US states after RCL [13, 14]. Only one study has been done in Canada, which found no change in emergency department (ED) visits for non-fatal traffic injury after RCL [15].
Traffic injury has also been impacted by the coronavirus disease 2019 (COVID-19) pandemic, as lockdowns and restrictions that began in March 2020 have greatly reduced the number of vehicles on the road [16]. Although a decrease in all injury-related ED visits was reported after March 2020, the largest decrease (between 48% and 50% worldwide) was seen for traffic-related injuries [16, 17]. A transient decrease in pedestrian deaths was seen in Toronto, Ontario, after COVID-19 [18]; however, non-fatal traffic injury in relation to COVID-19 is less studied in Canada. Since the pandemic started, there has been an increase in hospitalizations because of substance use and a self-reported increase in substance use among Canadians, with 40% reporting increased cannabis use [19]. Therefore, it is important to consider the effects of the COVID-19 pandemic when looking at the effects of RCL in Canada over time. This study will assess the impact of RCL in Canada on rates of ED visits and hospitalizations for motor vehicle and pedestrian/cyclist incidents and subsequently assess the impact of COVID-19 lockdowns after March 2020 on these outcomes. Currently, no studies on RCL and injury have also assessed the effect of COVID-19 in a legal cannabis market. This is also the first study to report the impact of RCL on rates of hospitalizations for traffic injury worldwide and expands on the current literature in Canada by assessing the impact of RCL on pedestrian/cyclist injury in addition to motor vehicle injury.
Our main objectives are to:
- (1) Assess the impact of RCL on rates of ED visits and hospitalizations for (i) motor vehicle injury and (ii) pedestrian/cyclist injury in Canadian provinces.
- (2) Assess the impact of COVID-19 lockdowns on rates of ED visits and hospitalizations for (i) motor vehicle injury and (ii) pedestrian/cyclist injury within a legal cannabis landscape in Canadian provinces.
METHODS
Data sources
Population-based aggregate data from the Canadian Institute for Health Information (CIHI) was used. ED visits were measured using CIHI’s National Ambulatory Care Reporting System (NACRS) for Ontario beginning January 2010 (2021 Ontario population 14 223 942) and Alberta beginning April 2010 (2021 Alberta population 4 262 635) [20]. Reporting to NACRS is mandated in these two provinces, making it a representative database for this outcome [21]. Data from NACRS encompasses additional measures associated with ED visits including day surgeries, clinics, diagnostic imaging and other ambulatory care. To maximize our ability to capture all relevant cases per month, all measures from NACRS were included in the monthly counts and collectively referred to as ED visits.
Hospitalizations were measured using CIHI’s Discharge Abstract Database (DAD), with mandatory reporting across Canada (beginning January 2010) except in Québec as their data is not submitted to CIHI [22]. The included provinces were Ontario, Alberta, British Columbia (2021 British Columbia population 5 000 879) [20], Manitoba, Saskatchewan, Nova Scotia, New Brunswick, Newfoundland and Prince Edward Island (PEI). Certain provinces were grouped based on similar demographics, population and culture; namely, the Prairies (Manitoba and Saskatchewan) representing 2 474 658 Canadians (based on 2021 population) and the Maritimes (Nova Scotia, New Brunswick, Newfoundland and PEI) representing 2 409 874 Canadians (based on 2021 population) [20]. Data from DAD encompasses additional measures associated with hospitalizations including acute inpatient care, inpatient rehabilitation/general rehabilitation, inpatient complex continuing care/chronic care facility, long-term care (24-hour nursing care), inpatient psychiatry (mental health), nursing stations, special rehabilitation facility, inpatient palliative care, day surgery and correctional facilities. To maximize our ability to capture all relevant cases per month, all measures from DAD were included in the monthly counts, collectively referred to as hospitalizations.
For NACRS ED visit data, the pre-RCL period was January 2010 to September 2018 for Ontario (n = 105 months) and April 2010 to September 2018 for Alberta (n = 102 months). The pre-RCL period for DAD hospitalization data was January 2010 to September 2018 (n = 105 months). For all data, the post-RCL period was October 2018 to February 2020 (n = 17 months), and the post-COVID period was March 2020 to March 2021 (n = 13 months).
Outcomes
The outcomes of interest were derived from International Statistical Classification of Diseases and Related Health Problems, 10th Revision, Canada (ICD-10-CA) codes: (1) motor vehicle incidents (V20–V29, V40–V79, V30–V39 and V86) and (2) pedestrian/cyclist incidents (V01–V09 and V10–V19). Occurrence of the outcomes in any diagnostic position was used (i.e. not confined to primary diagnosis). Additionally, one medical record could contribute to more than one outcome count (i.e. a motor vehicle incident and a pedestrian/cyclist incident) if there are multiple outcomes of interest at different diagnostic positions. Given the difference in frequency and characteristics of presentations, rates of ED visits and hospitalizations were analysed separately for each outcome.
Statistical analysis
Suppressed data
Data were provided to the researchers in aggregate form with monthly counts of any outcome between one and five suppressed to protect the privacy of individuals. For outcomes with suppressed data, a sensitivity analysis was performed by imputing a value of one or five. If there was no difference in the interpretation of results, a value of five was imputed for that outcome [23]. This method was used instead of multiple imputationbecause the values were within a known range (one to five) [24]. If there was no difference in the interpretation when imputing the minimum and maximum values, then there would also be no difference in interpretation when using multiple imputations, which would likely impute values in the middle of the range. Provinces with large amounts of suppressed data were excluded for the given outcome (Supporting information Appendix S1).
For ED visits, there were no months with suppressed data. For hospitalizations for motor vehicle injury, only PEI had suppressed data. There was suppressed data for hospitalizations for pedestrian/cyclist injury in all Prairie provinces. Based on sensitivity analyses for these outcomes, there was no difference in the interpretation of results when a value of one or five was imputed for months with suppressed data. As such, a value of five was imputed for these outcomes. Because of a high amount of suppressed data, hospitalizations for pedestrian/cyclist injury in the Maritime provinces were excluded.
Modelling
An interrupted time series with segmented linear regression was used to examine the effect of RCL enactment and the COVID-19 lockdowns on monthly rates of ED visits and hospitalizations, calculated per 100 000 population [25]. Before fitting the statistical model, all outcomes of interest were plotted as a time series to visually inspect for any evidence of non-linearity, outliers or other apparent population-level changes between 2010 to 2021 that would require adjustment. Because of the long pre-RCL period, visual inspection of linear trends was possible.
To fit the models, generalized least-squares (GLS) regression was used, which extends ordinary least-squares by allowing for correlations between errors [26].
A sensitivity analysis was performed to compare a GLS model assuming piecewise linear time trends to a generalized linear model (GLM) with Poisson distribution assuming log-linear time trends to confirm that the GLS model appropriately fit this data. GLS and GLM models were fitted for ED visits for motor vehicle injury in Ontario and plotted to visually assess the model fits (Supporting information Appendix S2).
For month t, the GLS model is given by the following equation:
ε
The trend change represents the change in slope (a gradual change over time), whereas the level change represents the change in y-intercept (an immediate change after the intervention). To account for seasonality, an indicator variable was created for each month. April was chosen as a reference month and variables for the other 11 months were added to the model. To remove the seasonal component and adjust to the reference month when plotting, a variable was created that consisted of the offset for the 11 monthly terms.
To account autoregressive (AR) and/or moving average (MA) correlation, the autocorrelation function (ACF) and partial ACF plots for each outcome were visually inspected to determine whether AR and/or MA correlation structures needed to be added to the final model. Based on the ACF and partial ACF plots, rates of hospitalization for motor vehicle injury in Ontario, British Columbia and the Prairies did not require AR or MA correlation terms. All other models included an adjustment for AR correlation.
As a qualitative control, an interrupted time series analysis was performed on the rate of ED visits and hospitalizations for appendicitis (ICD-10-CA K35.80) per 100 000 population (Supporting information Appendix S3), as these cases were not expected to be affected by RCL [27]. No statistically significant changeafter RCL and a level decrease immediately after COVID-19 was expected for this outcome, which would indicate that the interrupted time series analysis is appropriate to detect notable changes in the outcomes of interest.
An additional analysis was performed to assess the nationwide effect of RCL and COVID-19 on ED visits and hospitalizations for motor vehicle injury and pedestrian/cyclist injury (n = 4 models) allowing for clustered data by including the province as a grouping variable in the correlation structure of the GLS model. This model provides an estimate of the average level (immediate) and trend (gradual) changes while accounting for within-province correlation. For ED visits, the combined analysis included Ontario and Alberta; the pre-RCL period for Ontario was shortened to 102 months (beginning April 2010 instead of January 2010) for the purposes of a combined analysis. For hospitalizations, the combined analysis included Ontario, Alberta, British Columbia and the Prairies. The Maritime provinces were excluded for both motor vehicle and pedestrian/cyclist traffic injury for the purpose of consistency across outcomes within the nationwide analysis. The province-level and nationwide average rates per 100 000 were plotted over time for descriptive purposes (Supporting information Appendix S4).
Bonferroni correction was used to conservatively interpret the results for each outcome. Therefore, the α value with Bonferroni correction for the province-level analyses was 0.004 (α value = 0.05/13 statistical models). The α value for the additional nationwide analysis using Bonferroni correction was 0.013 (α value = 0.05/4 statistical models).
All statistical analyses were performed with RStudio using the packages nlme, car and stats [28–30]. The analysis was not pre-registered and the results should be considered exploratory.
Ethics approval
This study obtained ethics approval from the Research Ethics Board at Unity Health Toronto (REB 20-330).
RESULTS
Modelling
After visual inspection, there was no evidence of non-linearity in the data as determined by the similar model fit of the Poisson versus GLS model (Supporting information Appendix S2). There was no change in the rate of ED visits or hospitalizations for appendicitis after RCL (Supporting information Appendix S3). Based on these results, an interrupted time series analysis using a using a GLS model was deemed suitable.
Injury characteristics
Traffic injury presentations as a total number and proportion are reported in Table 1. Before COVID-19 lockdowns, the proportion of ED visits was between 1.01% and 1.21% for motor vehicle injury and between 0.28% and 0.40% for pedestrian/cyclist injury. The highest proportion of hospitalizations for motor vehicle traffic injury was in the Maritimes (>3%) (Table 1). TABLE 1. The number and proportion (%) of ED visits and hospitalizations for traffic injury (motor vehicle and pedestrian/cyclist) in each province, reported for descriptive purposes.
| Visit type | Province | Type of traffic injury | Total no. of traffic injury visits (% of all visits) | ||
|---|---|---|---|---|---|
| Pre-RCL | Post-RCL | Post-COVID-19 | |||
| ED visits | Ontario | Motor vehicle | 586 097 (1.05) | 101 658 (1.04) | 50 546 (0.85) |
| Pedestrian/cyclist | 222 164 (0.40) | 29 094 (0.30) | 26 963 (0.45) | ||
| Alberta | Motor vehicle | 230 910 (1.21) | 33 219 (1.01) | 20 984 (1.05) | |
| Pedestrian/cyclist | 69 839 (0.37) | 9302 (0.28) | 10 240 (0.51) | ||
| Hospitalizations | Ontario | Motor vehicle | 41 609 (0.50) | 6554 (0.46) | 4612 (0.47) |
| Pedestrian/cyclist | 18 226 (0.22) | 2663 (0.19) | 2220 (0.23) | ||
| Alberta | Motor vehicle | 25 849 (0.90) | 3176 (0.66) | 2430 (0.73) | |
| Pedestrian/cyclist | 7322 (0.26) | 1046 (0.22) | 952 (0.29) | ||
| British Columbia | Motor vehicle | 29 451 (0.42) | 4341 (0.34) | 3324 (0.38) | |
| Pedestrian/cyclist | 18 004 (0.26) | 3046 (0.24) | 2695 (0.31) | ||
| Prairies | Motor vehicle | 16 044 (0.42) | 2133 (0.33) | 1699 (0.41) | |
| Pedestrian/cyclist | 4376 (0.11) | 637 (0.10) | 546 (0.13) | ||
| Maritimes | Motor vehicle | 112 953 (3.42) | 16 204 (3.17) | 12 065 (3.58) | |
- Note: The denominator for the proportion is the total number of ED visits or hospitalizations for any reason in the given period. Prairies = Manitoba and Saskatchewan. Maritimes = Nova Scotia, New Brunswick, Newfoundland and Labrador and Prince Edward Island (PEI).
- Abbreviations: COVID-19, coronavirus disease 2019; ED, emergency department; RCL, recreational cannabis legalization.
The effect of RCL
The results of all interrupted time series models are shown in Table 2. There were no statistically significant changes in the rate of ED visits for motor vehicle or pedestrian/cyclist injury after RCL (Table 2 and Figures 1 and 2). TABLE 2. The impact of RCL and COVID-19 on ED visits and hospitalizations for motor vehicle and pedestrian/cyclist injury in Canada, reported as level (immediate) and trend (gradual) changes based on interrupted time series models.
| Visit type | Province | Type of traffic injury | Effect of RCL | Effect of COVID-19 | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Level change (95% CI) | P value | Trend change (95% CI) | P value | Level change (95% CI) | P value | Trend change (95% CI) | P value | |||
| ED visits | Ontario | Motor vehicle | 0.45 (−2.95 to 3.86) | 0.795 | −0.08 (−0.39 to 0.23) | 0.606 | −16.07 (−20.55 to −11.60) | <0.001* | 0.37 (−0.20 to 0.95) | 0.207 |
| Pedestrian/cyclist | −1.58 (−4.07 to 0.91) | 0.216 | 0.10 (−0.13 to 0.33) | 0.381 | −0.39 (−3.52 to 2.75) | 0.809 | −0.03 (−0.46 to 0.40) | 0.901 | ||
| Alberta | Motor vehicle | 1.03 (−4.92 to 6.99) | 0.734 | 0.01 (−0.53 to 0.56) | 0.968 | −10.34 (−17.80 to −2.89) | 0.008 | 0.36 (−0.67 to 1.39) | 0.495 | |
| Pedestrian/cyclist | 0.17 (−3.62 to 3.96) | 0.929 | −0.02 (−0.37 to 0.33) | 0.919 | 2.05 (−2.67 to 6.77) | 0.396 | 0.01 (−0.65 to 0.67) | 0.971 | ||
| Hospitalizations | Ontario | Motor vehicle | 0.019 (−0.19 to 0.23) | 0.859 | −0.06 (−0.03 to 0.004) | 0.124 | −0.28 (−0.58 to 0.02) | 0.067 | 0.03 (−0.004 to 0.07) | 0.086 |
| Pedestrian/cyclist | −0.14 (−0.33 to 0.04) | 0.138 | 0.005 (−0.01 to 0.02) | 0.597 | 0.06 (−0.19 to 0.31) | 0.649 | −0.01 (−0.04 to 0.03) | 0.668 | ||
| Alberta | Motor vehicle | 0.06 (−0.69 to 0.80) | 0.879 | 0.002 (−0.07 to 0.07) | 0.949 | −0.40 (−1.31 to 0.52) | 0.395 | 0.06 (−0.06 to 0.18) | 0.353 | |
| Pedestrian/cyclist | 0.05 (−0.34 to 0.44) | 0.793 | −0.01 (−0.05 to 0.02) | 0.515 | 0.48 (−0.04 to 1.00) | 0.073 | −0.03 (−0.09 to 0.04) | 0.419 | ||
| British Columbia | Motor vehicle | 0.22 (−0.31 to 0.76) | 0.416 | −0.05 (−0.10 to −0.002) | 0.042 | −0.01 (−0.77 to 0.74) | 0.971 | 0.08 (−0.01 to 0.17) | 0.077 | |
| Pedestrian/cyclist | 0.42 (−0.20 to 1.03) | 0.185 | −0.02 (−0.07 to 0.04) | 0.559 | 0.37 (−0.44 to 1.18) | 0.375 | 0.02 (−0.08 to 0.12) | 0.699 | ||
| Prairies | Motor vehicle | −0.27 (−1.02 to 0.49) | 0.488 | 0.01 (−0.06 to 0.08) | 0.784 | −0.07 (−1.13 to 0.98) | 0.894 | 0.04 (−0.09 to 0.16) | 0.576 | |
| Pedestrian/cyclist | −0.22 (−0.64 to 0.19) | 0.296 | 0.01 (−0.02 0.05) | 0.461 | 0.09 (−0.46 to 0.64) | 0.754 | −0.03 (−0.10 to 0.04) | 0.377 | ||
| Maritimes | Motor vehicle | −0.44 (−0.93 to 0.05) | 0.082 | 0.02 (−0.03 to 0.07) | 0.376 | −0.26 (−1.24 to 0.71) | 0.597 | −0.02 (−0.11 to 0.08) | 0.708 | |
- Note: Level change = immediate (step) change, trend change = gradual (slope) change. Prairies = Manitoba and Saskatchewan. Maritimes = Nova Scotia, New Brunswick, Newfoundland and Labrador and Prince Edward Island (PEI). The interrupted time series analysis used population-based monthly rates of each outcome. The post-RCL period began in October 2018 and the post-COVID period began in March 2020.
- Abbreviations: CI, confidence interval; COVID-19, coronavirus disease 2019; ED, emergency department; RCL, recreational cannabis legalization.
- * P < 0.004 (P value threshold with Bonferroni correction for 13 statistical models).
Similarly, there were no statistically significant immediate or gradual changes in the rate of hospitalizations for motor vehicle or pedestrian/cyclist injury after RCL in Ontario, Alberta, British Columbia, the Prairies and the Maritimes (Table 2 and Figures 3 and 4).
Visual interpretation of the plotted data did not provide evidence for any population-level changes after RCL in any jurisdiction (Figures 1–4).
Similar findings were seen in the nationwide analysis, with no statistically significant changes in the rate of ED visits for motor vehicle or pedestrian/cyclist injury after RCL (Supporting information Appendix S4).
The effect of COVID-19 relative to RCL
There was a statistically significant immediate decrease in the rate of ED visits for motor vehicle injury after COVID-19 lockdowns began in March 2020 in Ontario (level change β = −16.07, 95% CI = −20.55 to −11.60, P < 0.001). A similar effect was observed for the rate of ED visits for motor vehicle injury in Alberta after COVID-19 (level change β = −10.34, 95% CI = −17.80 to −2.89, P = 0.008) (Table 2 and Figure 1). With Bonferroni correction for conservative interpretation, this finding was no longer statistically significant. In the 10 years before COVID-19, rates of ED visits for motor vehicle injury were between 27.7 and 52.2 per 100 000 population in Ontario and 35.1 and 78.9 per 100 000 population in Alberta. In April 2020, the rates decreased to 11.1 and 19.9 ED visits per 100 000 population in Ontario and Alberta, respectively. There were no statistically significant immediate or gradual changes in rates of ED visits for pedestrian/cyclist injury (Table 2 and Figure 2).
For rates of hospitalizations in all provinces, there were no statistically significant changes in the rate of motor vehicle and pedestrian/cyclist injury after COVID-19 lockdowns began (Table 2 and Figures 3 and 4).
The results of the nationwide analysis confirmed a statistically significant immediate decrease in the rate of ED visits for motor vehicle injury after RCL in Canada (Ontario and Alberta, level change β = −13.19, 95% CI = −3.27 to 5.18, P < 0.001). As with the province-specific models, no changes in ED visits for pedestrian/cyclist injury or hospitalizations for either type of injury were seen with the nationwide analysis.
DISCUSSION
This study used population-based administrative databases, representative of over 36 million people, to assess the effects of RCL and COVID-19 on rates of ED visits and hospitalizations for motor vehicle and pedestrian/cyclist injury in Canada using an interrupted time series analysis. There is no clear evidence that RCL had any effect on rates of ED visits and hospitalizations for either motor vehicle or pedestrian/cyclist injury across Canada. However, after COVID-19 lockdowns began, there was an immediate drop in the rate of ED visits for motor vehicle injury relative to the post-RCL period.
The results of our study are consistent with the 2021 Canadian study by Callaghan et al., which found no change in rates of ED visits for drivers involved in traffic injury in Ontario and Alberta [15]. Our study expanded on this by looking at all traffic-related injuries, whether to drivers, passengers, cyclists or pedestrians. As well, our study assessed rates of hospitalizations, which encompasses injuries of greater severity than ED visits. Although some US studies found increased rates of traffic fatalities after RCL, the different outcome measure (fatalities compared to ED visits and hospitalizations) could explain why our study did not match these findings. In addition, some studies in the US found a change in traffic fatality in certain RCL states only [11, 12], suggesting that there may be differences within each individual jurisdiction that enacts RCL, includingbetween state level RCL (in the US) and nationwide RCL (in Canada). Jurisdiction-specific factors such as cannabis retail store restrictions and driving laws in US states versus Canada are other important considerationsthat may influence the impact of RCL on traffic outcomes in different studies. Like our study, two of these US studies, as well as the 2021 Canadian study and the study from Uruguay, used an interrupted time series analysis [9, 11, 12, 31, 15]. Therefore, our study complements other studies that have looked solely at fatalities, which is the most severe traffic outcome. It should be noted that our data includes fatalities, but only among patients who made it to the hospital.
Our findings on the effects of RCL may be associated with existing public health measures in Canada since RCL implementation, which could have contributed to heightened awareness of the risks of cannabis use for new and previous users. Although we did not find that RCL impacted traffic injury, this is an important finding as it may indicate that Canada’s public health efforts are effectively preventing cannabis-related traffic injury—highlighting the importance of continuing to promote safe cannabis use within a legal cannabis landscape. For instance, those who increased cannabis use after RCL did so with appropriate prevention measures for traffic injury such as not driving under the influence of cannabis. Public health education and awareness measures implemented in Canada include evidence-based information tools as well as advertising and marketing campaigns (e.g. on social media, television and print) focusing on the health and safety risks of cannabis use, travel restrictions and strict penalties for impaired driving, with a focus on children and youth [32]. Our results, therefore, inform what happened in Canada from a public health perspective.
The decrease in rates of ED visits for motor vehicle injury immediately after COVID-19 lockdowns began is likely because of public health restrictions such as the closure of work, school and other activities, which reduced the need for driving and ultimately decreased the risk exposure for motor vehicle and other traffic injury. Future jurisdictions that implement RCL should note that rates of ED visits for motor vehicle injury were still below baseline levels 1 year after COVID-19 lockdowns began, which will affect interpretation of results when assessing traffic injury at this stage. The rate of hospitalizations is intrinsically lower than the rate of ED visits for motor vehicle injury, as hospitalizations represent more severe cases of injury. This may reduce the ability to detect statistically significant changes in rates of hospitalizations after COVID-19, whereas drastic reductions in the rate of ED visits were observed.
A major strength of this study is that our analysis was able to control for factors associated with longitudinal data such as seasonality, secular trends and autocorrelation, which increases the confidence of determining the effect of the intervention [33, 34]. Our study also used databases that are representative of the entire Canadian population due to reporting of health information to CIHI being required for the included provinces. Canada’s nationwide legalization of cannabis provide a strong basis for analysis of health effects at a population level. However, because these results are based on aggregate population data, this study does not attempt to explain the impact on individualsor differences between demographic groups, which were not explored. The assumption of linear trends can be a limitation of using a GLS segmented regression model; however, we addressed this as we did not observe evidence against linearity in the data through the sensitivity analyses and visual inspection. Additionally, the short post-RCL period, because of COVID-19 lockdowns occurring only 17 months after RCL implementation, may preclude the ability to detect trend changes that might have occurred over a longer period of time. We acknowledge the inherent difficulty in assessing the impact of RCL over time given the large global impact of COVID-19 on the health care system. Another option to account for the effects of COVID-19 would have been to include COVID-19 as a covariate as opposed to a separate intervention; however, separating the two effects might be difficult because of multicollinearity.
Another limitation of this study is that the databases used do not provide information on cannabis use at the time of injury. Nonetheless, it is important to note that our population-based study on rates of total traffic injury can detect changes in both reported and unreported cannabis-related incidents and is not confounded by subjective measurements of cannabis use.
Our study provided an overview of the effects of RCL on overall rates of ED visits and hospitalizations, as well as COVID-19 lockdowns within a legal cannabis landscape, which is a helpful baseline for future studies that should consider cannabis-related traffic injuries as well. Future studies should also consider how RCL has affected rates of traffic injury for specific at-risk populations, such as young males [35–37]. As additional years of data become available, further analysis is required to assess rates of traffic injury over time. Although COVID-19 restrictions have loosened, the pandemic has also had a harmful impact on mental health and substance use that could impact traffic injury over time [19, 38, 39], requiring continued surveillance within a legal cannabis landscape.
This novel population-level study is important for understanding the impact of RCL on traffic injury in the Canadian context, which has implications for public health measures regarding the prevention of traffic incidents and safe cannabis use. For instance, continued education and awareness about the risks of driving under the influence of cannabis that Canada has implemented since RCL will be important to prevent injury in the future. These results will also inform future policy changes for jurisdictions globally that are considering implementing RCL after COVID-19 and must understand all potential health implications.
AUTHOR CONTRIBUTIONS
Madison Walker: Formal analysis (lead), funding acquisition (lead), visualization (lead), writing—original draft preparation (lead), investigation (equal), methodology (equal), writing—review and editing (equal), conceptualization (supporting). Olli Saarela: Methodology (equal), formal analysis (supporting), conceptualization (supporting), writing—review and editing (supporting). Robert Mann: Methodology (equal), conceptualization (supporting), writing—review and editing (supporting). Melissa Carpino: Investigation (equal), writing—review and editing (equal), writing—original draft preparation (supporting), methodology (supporting), visualization (supporting). Michael D. Cusimano: Conceptualization (lead), supervision (lead), methodology (equal), funding acquisition (supporting), writing—review and editing (supporting).
DECLARATION OF INTERESTS
All authors have completed the ICMJE uniform disclosure form at http://www.icmje.org/disclosure-of-interest/ and declare: M.D.C. is a practicing neurosurgeon who cares for individuals as a result of cannabis usage and has received a research grant from the Public Health Agency of Canada who may be interested in this work as well the Canadian Institutes of Health Research for a similar study (on which O.S. and R.M. were co-applicants); he and M.W. have received a research grant from the Toronto Cannabis and Cannabinoid Research Consortium who might have an interest in the submitted work; M.W. also received support from the Ontario Graduate Scholarship—Masters; O.S. received a research grant from the Natural Sciences and Engineering Research Council of Canada who might also have an interest in the submitted work; no other relationships or activities that could appear to have influenced the submitted work.
