Longitudinal, population-based analysis of burn injuries across Canada pre-recreational and post-recreational cannabis legalisation Authors: Sarah Paleczny, Saifedine T Rjaibi, Olli Saarela, Peter Selby, Sergio Rueda, Adam Levitan, Edward Wang, Chris Brogly, Michael D Cusimano Published: 2026 ABSTRACT AbstractIntroductionRecreational cannabis legalisation (RCL) in Canada was implemented in October 2018 following a federal announcement in August 2015, with commercialisation occurring in a staggered, province-dependent manner. Cannabis may increase burn risk through hydrocarbon burns from butane hash oil extraction, inhalation injuries, psychomotor impairment, increasing the risk of flame-related accidents and illicit production practices. How burn-related injuries have changed over time around these policy shifts at a population level remains unclear.MethodsWe conducted an interrupted time series analysis of monthly burn-related hospitalisation rates (2006–2024) across Canada (all provinces and territories except Quebec) and burn-related emergency department (ED) visit rates (2010–2024) in Ontario and Alberta using population-based administrative data. Burns were identified using the Canadian adaptation of the International Classification of Diseases, 10th Revision (ICD-10-CA) codes X00–X19. Intervention periods reflected the RCL announcement (August 2015), enactment (October 2018), commercialisation (private retail expansion; province-dependent) and COVID-19 lockdowns as an effect modifier. Analyses were stratified by age, sex and urban/rural status and by commercialisation strategy groups. Generalised least squares segmented regression was used to estimate level and trend changes across periods.ResultsThere were 33 346 hospitalisations and 314 962 ED visits for burn-related injuries. No national change in burn injury rates was observed post-RCL when all subgroups were combined. Statistically significant increases in hospitalisation rates were observed among children aged 0–3 in Ontario immediately by 100.7% post-RCL (level β=1.4156; 95% CI 0.4666 to 2.3647; p=0.0039) and among urban residents in early-commercialising provinces over time by 0.6% per month following the RCL announcement period, corresponding to a 12.1% increase over the period (trend β=0.003; 95% CI 0.0005 to 0.0056; p=0.0201). National hospitalisation rates among males increased immediately by 20.6% post-COVID-19 lockdowns compared with the pre-RCL announcement period (level β=0.1561; 95% CI 0.0470 to 0.2652; p=0.0055). ED visits showed few statistically significant changes across periods.ConclusionsRCL-related policy milestones (including the announcement and implementation periods) coincided with increases in burn-related hospitalisations among young children, males and urban residents. More consistent changes in hospitalisations than ED visits suggest shifts concentrated in more severe cases, highlighting the importance of targeted public health strategies and ongoing surveillance during evolving cannabis policy environments. WHAT IS ALREADY KNOWN ON THIS TOPIC WHAT THIS STUDY ADDS Using national hospitalisation data and large provincial emergency department data, we identified periods in which burn-related hospitalisations increased among young children, males and urban residents in temporal relation to key RCL policy milestones, while overall national rates showed no clear change when all subgroups were combined. Using national hospitalisation data and large provincial emergency department data, we identified periods in which burn-related hospitalisations increased among young children, males and urban residents in temporal relation to key RCL policy milestones, while overall national rates showed no clear change when all subgroups were combined. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY These findings support targeted injury-prevention strategies (eg, safer storage messaging and prevention of unsafe extraction practices) and motivate improved surveillance to better understand mechanisms and monitor burn injury patterns as cannabis policy and access continue to evolve. These findings support targeted injury-prevention strategies (eg, safer storage messaging and prevention of unsafe extraction practices) and motivate improved surveillance to better understand mechanisms and monitor burn injury patterns as cannabis policy and access continue to evolve. Introduction In August 2015, the Canadian federal government announced plans to legalise recreational cannabis, with implementation set for October 2018.1 The policy aimed to protect public health and safety, provide adults with safer, regulated access to cannabis, reduce the burden on the criminal justice system and potentially decrease opioid use through substitution.2 Following the federal announcement of recreational cannabis legalisation (RCL), cannabis use increased across Canada, coinciding with rapid growth of the illicit and unregulated market during the pre-legalisation period, as demand expanded ahead of regulated retail supply.1 Self-reported usage from the Canadian Cannabis Survey and Canadian Tobacco, Alcohol and Drugs Survey reported an approximately 10% increase in use alongside the announcement, implementation and commercialisation of RCL.34 Recreational cannabis commercialisation, defined as the opening of licensed private retail stores, further amplified this access, with store numbers growing from 182 in the fourth quarter of 2018 to 3332 by the first quarter of 2023, and product diversity expanding from dried flower only to edibles, concentrates, topicals and high-potency extracts.5 Since its enactment, a growing body of research has investigated cannabis legalisation’s effects on public health, and burn-related injuries have emerged as one area of concern.6–8 Documented cannabis-related burn injuries include inhalation injuries and chemical pneumonitis from butane hash oil production (‘dabs’), as well as hydrocarbon burns from informal or home-based extraction processes.8–11 Burns from cannabis oil production can result in severe, life-altering injuries, often requiring multiple surgeries, prolonged hospitalisation and intensive healthcare resources.11–13 Separately, controlled laboratory studies show that tetrahydrocannabinol (THC) causes dose-dependent psychomotor impairment, including slowed reaction time, impaired divided attention and reduced hand-eye coordination.1014 Although direct observational evidence linking cannabis use to burns during routine kitchen food preparation remains limited, these experimentally demonstrated impairments provide a biologically plausible mechanism by which cannabis could increase the risk of flame or thermal (eg, stovetop) injury during food preparation. While studies have described the epidemiology of burns generally in smaller groups, such as paediatric hospitalisations and burns during COVID-19, as well as burns specifically from illegal cannabis oil manufacturing prior to legalisation, there are no national-level analyses of burn trends across the period of RCL.131516 However, existing literature has important limitations. Two single-centre reports from the USA provided local perspectives and found increases in hydrocarbon burns following cannabis legalisation on a small scale.911 Another study using the North American burn registry analysis analysed patients older than 12 years who received a urine drug screen pre-legalisation, and found that the proportion testing cannabis-positive rose from 6.0% in 2002 to 27.0% in 2011, with cannabis-positive patients having larger burns, more intensive care unit days and a greater number of operations compared with cannabis-negative patients.12 In Canada, one study has examined cannabis and burns, conducted at a single centre in Calgary, prior to legalisation (from 2012 to 2014). The authors reported that 7.5% of burn admissions were linked to illegal cannabis oil production, with those linked to cannabis having greater burn surface area (24% vs 9%) and increased need for grafting and hospitalisation.13 Population-level evaluations of burn injuries following cannabis legalisation have remained elusive, representing a critical gap given the national scope of this policy change in Canada. This study examined burn injury trends across Canada spanning key RCL policy milestones and commercialisation. National trends were assessed over 18 years, offering timely insight into a potentially overlooked consequence of cannabis reform. Methods Data sources This study used individual-level, monthly population-based data from the Canadian Institute for Health Information (CIHI). Hospitalisations related to burn injuries were extracted from the Discharge Abstract Database (DAD) for all provinces and territories except Quebec, as this data was not available.17 Emergency department (ED) visits were extracted from the CIHI National Ambulatory Care Reporting System (NACRS).18 Both DAD and NACRS are established data sources frequently used in national Canadian research.19–21 ED data submission to NACRS is mandatory only in Ontario, Alberta and the Yukon, and coverage in these jurisdictions is near-complete, representing over 21 million individuals.18 DAD data were available from April 2006 to February 2024, and NACRS data from April 2010 to February 2024. To protect privacy, cells with fewer than five non-zero observations were suppressed. All data were accessed through a secure research environment and aggregated prior to analysis. Each record in DAD and NACRS includes at least one diagnosis coded using the Canadian adaptation of the International Classification of Diseases, 10th Revision (ICD-10-CA). Burn-related injuries were identified using a set of predefined ICD-10-CA codes: X00–X19. These codes encompass exposure to smoke, fire and flames (X00–X09), such as ignition of highly flammable materials (X04), other specified smoke and fire exposures (X08) and unspecified smoke inhalation (X09), as well as contact with heat and hot substances (X10–X19), including contact with or inhalation of hot air and gases (X14), capturing both external burn mechanisms and inhalation-related burn injuries, consistent with prior epidemiological research. However, the ICD-10-CA dataset used in this study does not permit isolation of inhalation injuries as a distinct mechanistic category, as codes capturing inhalation-related injuries also encompass external thermal mechanisms; for example, codes for steam and hot vapour exposure include both inhalation injuries and external scalds, with no subdivision by route of injury. ICD-10-CA codes have been used in previous studies examining multiple conditions and have moderate to good reliability and validity.22 Study design We employed an interrupted time series (ITS) design to investigate monthly rates of burn-related hospitalisations and ED visits, evaluating the potential impacts of multiple policy-related interventions. ITS is a well-established quasi-experimental method widely used to evaluate the population-level trends in health outcomes surrounding cannabis policy changes, such as examining emergency department visits, paediatric poisonings, and traffic injuries across North America.202324 The four primary intervention periods analysed included the federal announcement of RCL, its enactment, the onset of legal retail sales (commercialisation) and the COVID-19 pandemic, which was included as a potential modifier of commercialisation impacts (figure 1). Commercialisation was defined by the date private retail cannabis outlets began operating in each province or territory, with these dates collected through correspondence with government agencies and licensing authorities (online supplemental appendix 1 and 2). Interventions (RCL announcement, RCL enactment, commercialisation and COVID-19) assessed in the analyses and timeline of the dates they encompassed. COVID-19 lockdowns began in March 2020 in Canada. Group 1: Newfoundland and Labrador; Nova Scotia; Manitoba; Saskatchewan; Alberta; British Columbia (commercialisation October 2018). Group 2: Ontario (commercialisation March 2020). Group 3: Prince Edward Island (no commercialisation); New Brunswick (commercialisation close to the study end so not examined exclusively due to low counts). Group 4: Northwest Territories; Yukon; Nunavut (data provided grouped together by CIHI due to low counts). Data for Quebec was not available. CIHI, Canadian Institute for Health Information; NB, New Brunswick; NT, Northwest Territories; NU, Nunavut; RCC, recreational cannabis commercialisation; RCL, recreational cannabis legalisation; YT, Yukon. Jurisdictions (provinces) were grouped into four commercialisation categories based on timing and rollout strategy (figure 1): group 1 (commercialisation coinciding with RCL enactment), group 2 (commercialisation coinciding with COVID-19) and group 3 (no or delayed commercialisation near study endpoint). Territories exhibited varied commercialisation but were aggregated into a separate group by CIHI due to their smaller population sizes (group 4). The demographic groups assessed within the study population were stratified by sex (male/female), age (0–3, 4–18, 19–39 and 40+ years) and geographical location (urban/rural). Urban or rural status was determined based on patients’ postal codes. Postal codes located within census metropolitan areas and census agglomerations were classified as urban, while those outside of these areas were classified as rural, in accordance with Statistics Canada’s guidelines.25 All territories were categorised as rural based on CIHI data classifications. Each post-intervention period was compared with the immediately preceding phase, except for the post-COVID-19-lockdown period, which was compared with the pre-RCL announcement baseline due to pronounced service disruptions during COVID-19 lockdowns. The time periods assessed included: Post-announcement of RCL (August 2015–September 2018) versus pre-announcement (April 2006–July 2015). Post-announcement of RCL (August 2015–September 2018) versus pre-announcement (April 2006–July 2015). Post-RCL implementation (October 2018–February 2020) versus post-announcement (August 2015–September 2018). Post-RCL implementation (October 2018–February 2020) versus post-announcement (August 2015–September 2018). Post-commercialisation versus pre-commercialisation (dates varied by group; see figure 1). Post-commercialisation versus pre-commercialisation (dates varied by group; see figure 1). COVID-19 period (March 2020–May 2021) versus pre-COVID-19 (October 2018–February 2020). COVID-19 period (March 2020–May 2021) versus pre-COVID-19 (October 2018–February 2020). Post-COVID-19 lockdowns (June 2021–March 2024) versus pre-RCL announcement period (April 2006–July 2015). Post-COVID-19 lockdowns (June 2021–March 2024) versus pre-RCL announcement period (April 2006–July 2015). Statistical analysis ITS analyses were conducted using generalised least squares (GLS) regression, which accounts for autocorrelation and underlying trends in the estimation of policy impacts over time. Monthly rates of burn-related ED visits (April 2010–March 2024; 168 months) and hospitalisations (April 2006–March 2024; 216 months) were calculated per 100 000 population nationally and for commercialisation groups 1 and 2. Groups 3 and 4 were not analysed outside of the national analysis due to low counts (<5 cases per month) leading to mandated data suppression. Numerators reflected the monthly count of hospitalisations or ED visits, and denominators were monthly stratum-specific population estimates interpolated from Statistics Canada census data.25 Segmented regression was used to estimate changes in both intercept (immediate level change) and slope (trend over time) after each intervention. Per cent changes were calculated for statistically significant findings, relative to the model-fitted rate at the intervention time point, estimated from pre-intervention trends. The post-COVID-19 period was compared with the pre-RCL baseline to reduce confounding due to the abrupt drop in non-COVID-19 healthcare use during lockdown. A p value cut-off of 0.05 was used to assess statistical significance. To account for multiple comparisons across intervention periods and subgroup strata, Holm-Bonferroni corrections were applied to all hypothesis tests. Model assumptions, including linearity and normality of residuals, were verified visually. To evaluate and adjust for autocorrelation, preliminary ordinary least squares regressions were used, and where needed, autoregressive and/or moving average terms were incorporated into the final GLS models.26 The ITS design provides a built-in control by comparing observed post-intervention data to a counterfactual trend based on the pre-intervention period, but to assess robustness, sensitivity analyses were conducted. This involved re-estimating models with the intervention date artificially shifted 8 months earlier, a period during which no policy change occurred, to evaluate whether observed trends might be driven by background trends or unmeasured confounding. All analyses were performed in RStudio (V.2023.12.1+402; R V.4.3.3) using the nlme and car packages.27 Results We analysed a total of 33 346 burn-related hospitalisations representing over 450 million person-years and 314 962 burn-related ED visits representing over 250 million person-years. Total cases of hospitalisations and ED visits for burns in each study group are presented in table 1. Table 1Total counts of hospitalisations and ED visits for burn injuries across Canada from April 2006 to March 2024 pre-RCL and post-RCL and COVID-19PatientsHospitalisationsTotal cases pre-RCL(Apr 2006–Sep 2018)Total cases post-RCL(Oct 2018–Feb 2020)Total cases during COVID-19(Mar 2020–May 2021)Total cases post-COVID-19(Jun 2021–Mar 2024)Total cases Total provinces and territories Total Canada22 01427682504606033 346Males all ages14 26217901596394321 591Females all ages7752978908211711 755 Ages 0–332014893728304892 Ages 4–18*23392142555473355 Ages 19–39476052246713057054 Ages 40+11 70415441410338018 038 Urban15 44320971834447923 853 Rural600859257612048380 Group 1 (NL, NS, MB, SK, AB, BC) hospitalisations Males all ages7705929802214511 581Females all ages403743344310565969 Ages 0–313861671242601937 Ages 4–18*129383971921665 Ages 19–3928003122768064194 Ages 40+617278373919039597 Urban7356939835220911 339 Rural41403863627485636 Group 2 (ON) hospitalisations Males all ages591978072916959123Females all ages339451243410175357 Ages 0–316773122355432767 Ages 4–18*758841313061279 Ages 19–3917571861794642586 Ages 40+496368860613797636 Urban76601109956221411 939 Rural14941641763992233Total cases pre-RCL(ON/AB: Apr 2010–Sep 2018)Total cases post-RCL(Oct 2018–Feb 2020)Total cases during-COVID-19(Mar 2020–May 2021)Total cases post-COVID-19(Jun 2021–Mar 2024)Total cases ON ED visits Males all ages72 76711 497887321 778114 915Females all ages69 38011 622812519 818108 945Ages 0–3 males10 46418261316305016 656Ages 0–3 females70711266910205111 298Ages 4–18 males989514201012285115 178Ages 4–18 females10 39515671033279315 788Ages 19–39 males25 65537582943734939 705Ages 19–39 females24 03940602854658137 534Ages 40+ males26 75344873602852843 370Ages 40+ females27 87548363328839344 432 Urban114 51418 94313 88133 914181 252 Rural26 24940412913703940 242 AB ED visits Males all ages31 70548073929923349 674Females all ages25 99343153264785641 428Ages 0–3 males491177762512177530Ages 0–3 females35195354028425298Ages 4–18 males484466355113267384Ages 4–18 females447961548712356816Ages 19–39 males12 56717421456326219 027Ages 19–39 females979216231255302215 692Ages 40+ males938316251297342815 733Ages 40+ females820314831120275713 563 Urban34 5185972474911 27356 512 Rural22 60830092392551733 526*Hospitalisations of patients aged 4–18 were omitted from statistical analyses due to low monthly counts and risk of re-identification. Minor discrepancies between totals and stratified counts reflect small cell suppression (n<5) for privacy protection and missing/unspecified substrata (eg, sex, age, urban/rural). Patient groups are based on commercialisation implementation policies which are summarised in figure 1.AB, Alberta; BC, British Columbia; ED, emergency department; MB, Manitoba; NL, Newfoundland and Labrador; NS, Nova Scotia; NT, Northwest Territories; ON, Ontario; RCL, recreational cannabis legalisation; SK, Saskatchewan; Total counts of hospitalisations and ED visits for burn injuries across Canada from April 2006 to March 2024 pre-RCL and post-RCL and COVID-19 Hospitalisations of patients aged 4–18 were omitted from statistical analyses due to low monthly counts and risk of re-identification. Minor discrepancies between totals and stratified counts reflect small cell suppression (n<5) for privacy protection and missing/unspecified substrata (eg, sex, age, urban/rural). Patient groups are based on commercialisation implementation policies which are summarised in figure 1. AB, Alberta; BC, British Columbia; ED, emergency department; MB, Manitoba; NL, Newfoundland and Labrador; NS, Nova Scotia; NT, Northwest Territories; ON, Ontario; RCL, recreational cannabis legalisation; SK, Saskatchewan; Hospitalisations Following the announcement of RCL, rates of hospitalisation for burns significantly increased immediately in ages 0–3 in group 2 (β=0.6111; 95% CI 0.0072 to 1.2151, p=0.0487), representing an immediate level increase of 28.3%. Urban areas in group 1 also showed a significant increase over time (β=0.003; 95% CI 0.0005 to 0.0056, p = 0.0201) (figure 2; online supplemental appendix 3), with the monthly trend increasing by 0.6% per month, and rates reaching 12.1% higher than expected by the end of the announcement period. Following RCL implementation, rates of hospitalisation for burns in ages 0–3 in Ontario (group 2) increased immediately by 1.4156 monthly burns per 100 000 (95% CI 0.4666 to 2.3647, p=0.0039), representing an immediate level increase of 100.7% (figure 3; online supplemental appendix 3). Rate of hospitalisations for burns per 100 000 population in group 1 urban areas. Solid coloured lines=trend observed; dashed coloured lines=predicted counterfactual trend without intervention. The post-COVID-19 period was compared directly to the pre-RCL announcement period. RCL, recreational cannabis legalisation. Rate of hospitalisations for burns per 100 000 population in group 2 ages 0–3. Solid coloured lines=trend observed; dashed coloured lines=predicted counterfactual trend without intervention. The post-COVID-19 period was compared directly to the pre-RCL announcement period. RCL, recreational cannabis legalisation. Our analysis did not find statistically significant evidence of a change in rates of hospitalisations for burns during COVID-19 lockdowns. After COVID-19 lockdowns ended (compared with pre-RCL announcement), rates of hospitalisations for burns across Canada (in all groups combined) significantly increased immediately (β=0.0935; 95% CI 0.0257 to 0.1614, p=0.0075), representing an immediate level increase of 18.8%, and over time (β=0.0026; 95% CI 0.0003 to 0.0049, p=0.0287), representing a monthly trend increase of 0.4% per month (online supplemental appendix 4). From subgroup stratifications, these increases post-COVID-19 lockdowns were observed in males and urban areas across Canada. Among all groups combined, males increased immediately by 20.6% (β=0.1561; 95% CI 0.0470 to 0.2652, p=0.0055), and urban areas increased immediately by 18.1% (level β=0.0928; 95% CI 0.0232 to 0.1624, p=0.0097). In group 1, males increased immediately by 25.6% (level β=0.0068; 95% CI 0.0012 to 0.0123, p=0.0173), and urban areas increased immediately by 31.6% (level β=0.0035; 95% CI 0.0011 to 0.0059, p=0.0229). Additionally, there were increases in ages 0–3 in group 1 (β=0.6683; 95% CI 0.0717 to 1.2650, p=0.0293), representing an immediate level increase of 41.7%, ages 19–39 in group 1 (β=0.0107; 95% CI 0.0057 to 0.0158, p<0.0001), with the monthly trend increasing by 2.1% per month, and ages 40+ in group 1 (β=0.2264; 95% CI 0.1260 to 0.3267, p<0.0001), representing an immediate level increase of 34.7% (online supplemental appendix 4). Furthermore, following the end of COVID-19 lockdowns, there were also significant immediate level increases among ages 0–3 in group 2 by 37.6% (β=0.9134; 95% CI 0.0427 to 1.7841, p=0.0411) (online supplemental appendix 4). ED visits Following the announcement of RCL, in Ontario, there was a statistically significant increase in ED visits for burns over time in males aged 4–18, with the monthly trend increasing by 0.4% per month (β=0.0427; 95% CI 0.0029 to 0.0825, p=0.0371) (online supplemental appendix 3), such that by the end of the RCL announcement intervention period, rates were 118.8% higher than expected based on the pre-intervention trend. During COVID-19 lockdowns, most groups showed statistically significant decreases in rates of ED visits for burns in the total Ontario population (β=−0.9188; 95% CI −1.3367 to −0.5009, p<0.0001), representing an immediate level decrease of 23.8%, and the total Alberta population (β=−1.6114; 95% CI −2.7094 to −0.5134, p=0.0046), representing an immediate level decrease of 16.6% (online supplemental appendix 3). Following the end of COVID-19 lockdowns, in Ontario and Alberta, most groups had either no significant change or significant decreases immediately and over time (Alberta β=−0.0215; 95% CI −0.0615 to −0.0185, p=0.2390; Ontario β=−0.0123; 95% CI: −0.0257 to 0.0011, p=0.0750) (online supplemental appendix 4). However, in Alberta, there was a statistically significant increase in rates of ED visits for burns in males aged 4–18 over time following the end of COVID-19 lockdowns (β=0.1130; 95% CI 0.0411 to 0.1849, p=0.0025), with the monthly trend increasing by 1.0% per month, when compared with the pre-RCL announcement period (online supplemental appendix 4). Sensitivity analyses When the intervention dates chosen for RCL were shifted 8 months prior to the intervention, there were no significant changes associated with the interventions, supporting the robustness of the model in assessing the interventions of interest (online supplemental appendix 5). Additionally, the comparison of the data as a Poisson generalised linear model with a log-link compared with a GLS model showed a similar fit (online supplemental appendix 6). Discussion Main findings Our analyses identified statistically significant increases in burn-related hospitalisations, and more limited changes in ED visits, concentrated within specific subgroups and time periods rather than at the population level, reflecting temporal associations with policy milestones. The most consistent hospitalisation increases were observed among young children (ages 0–3), males and urban populations, with timing and magnitude varying by cannabis commercialisation strategy. In particular, urban areas in early-commercialising provinces (group 1) demonstrated increasing hospitalisation trends following the federal announcement period, while immediate increases among children aged 0–3 were observed in Ontario after both the announcement and implementation periods. Following the end of COVID-19 lockdowns, compared with the pre-announcement baseline, burn-related hospitalisation rates increased nationally and across multiple subgroup strata, most notably among males and urban residents, with additional age-specific increases observed within early-commercialising provinces. In contrast, ED visit rates showed fewer and less consistent changes over time. The primary ED signals were limited to increases among males aged 4–18 in Ontario following the announcement period and in Alberta during the post-COVID period, while most other ED strata demonstrated stable or declining trends. Taken together, these findings suggest that while overall burn-related healthcare utilisation remained broadly stable at the national level, periods of overlapping policy and societal change were accompanied by time-specific shifts in hospitalisation patterns for burns among certain demographic and geographical subgroups. To account for multiple comparisons across intervention periods and strata, Holm-Bonferroni corrections were applied; key subgroup findings, including those among young children, urban populations and males, remained statistically significant after adjustment, indicating that the most consistent signals were robust to correction. Interpretation in context of existing literature Increased cannabis access and paediatric burn trends The increase in paediatric (aged 0–3 years) burn-related hospitalisations in Ontario observed after the announcement and implementation of RCL may be associated with increased exposure to hot surfaces or scalding liquids in contexts of caregiver impairment or reduced supervision. Cannabis usage can cause caregiver impairment, such as reduced attentiveness or coordination, leading to these injuries in toddlers who may be exposed to hot surfaces, spills or indirectly injured through caregiver impairment or distraction. There was a large increase in cannabis usage following the government’s announcement of RCL, which may be partially attributed to the sharp rise in illicit and non-regulated cannabis availability during this period, along with a reduced perceived risk of cannabis once it was planned to be legalised, which coincided with higher rates of cannabis use across Canada.34 Cannabis oil extraction and urban burn injuries In urban areas of group 1 provinces, the sharp rise in burn-related hospitalisations following the RCL announcement may be attributable in part to a surge in illicit cannabis oil extraction activities. Between 2015 and 2018, cities such as Calgary, Edmonton and Toronto saw dramatic increases in butane hash oil ‘labs’ and unregulated cannabis storefronts.13 These operations circumvented safety regulations, and often used volatile solvents and improvised tools such as butane torches in residential settings with no ventilation.13 This led to numerous cases of severe burn injuries from fires and explosions. For example, during a 2-year period preceding legalisation, 12 of 161 burn unit admissions in Calgary were directly attributed to cannabis oil extraction and involved severe burns affecting an average of 24% of the total body surface area and resulted in prolonged hospital stays.13 These patterns align with earlier case series from California that also found larger burn sizes from amateur THC extraction using volatile butane.28 These makeshift labs represent a major urban public safety threat, particularly during transitional legalisation periods when demand outpaces regulated supply. The surge in urban burn-related hospitalisations post-announcement may reflect efforts by untrained individuals to meet anticipated market demand through unsafe production.2829 Incidents like the Toronto 2016 explosion and reports from Edmonton Police highlight how these hazardous practices often occurred in densely populated residential areas, compounding the risk to both producers and bystanders.2930 Together, these incidents illustrate how gaps in regulatory oversight during the process of RCL may have contributed to conditions for unregulated extraction practices that increased risk of burn injuries. Sex differences in burn injury risk Moreover, the sex-based trends observed in our study align closely with previous literature. In both Canada and the USA, two small-scale case series and one pre-RCL American National Burn Repository study have reported that the majority of cannabis extraction burn victims are young (75% under age 41) and predominantly male (83%).121328 Similarly, Stanton et al (2025) found that cannabis-positive burn patients were significantly more likely to be male and younger, mirroring the demographic profile identified in our analysis.31 Notably, we observed significant post-RCL announcement and post-COVID-19 lockdown increases in burn-related hospitalisation rates among males across Canada, indicating that these disparities may have intensified alongside legalisation. This male predominance likely reflects underlying sex-based differences in exposure to hazardous extraction practices. Prior research has shown that men are more likely to occupy direct, risk-intensive roles in illicit substance production and handling,32 and that risk-taking behaviours are more strongly aligned with traditional masculinity norms.33 These findings highlight the importance of considering behavioural predispositions and risk profiles in interpreting population-level injury trends. This pattern may also signal a resurgence in unregulated or high-risk cannabis behaviours following the pandemic, when enforcement and oversight may have lagged and personal production increased. Thus, jurisdictions seeking to implement RCL policies should anticipate and mitigate high-risk extraction practices to prevent avoidable burn injuries. Injury severity and healthcare utilisation Finally, the greater proportion of significant results observed in hospitalisations compared with ED visits reinforces the severity of these injuries. Hospitalisation visits represent more serious burns, as patients were admitted, likely with injuries that covered a larger percentage of total body surface area, involved multiple layers of skin, and necessitated surgical interventions such as grafting. This aligns with previous research in Calgary and across the USA, which reported that cannabis-attributable burn injuries were more likely to affect a larger total body surface area (24% vs 9%), required extensive surgical management (skin grafting in 75% of cases) and resulted in long lengths of stay in burn units (mean 32 days).121328 These findings emphasise both the substantial burden these injuries place on individuals and the healthcare system and the need for proactive public health measures to prevent further harm. Future directions Our findings point to several key areas for future research and intervention. To protect public health and safety in light of emerging RCL policies, surveillance systems should systematically capture cannabis-related injury mechanisms, including whether cannabis use or exposure contributed to burns. Future studies using data sources with greater mechanistic granularity, such as burn registry or chart-level data, could also distinguish inhalation injuries from external burns, a distinction with direct relevance for prevention, as these mechanisms warrant different public health strategies. Alongside this, studies should continue to track cannabis usage and resulting injuries over time to monitor the development of the impacts of these policy changes on public health. This would require integrating toxicology screening, caregiver interviews and possibly fire department data into injury databases. Furthermore, targeted paediatric and urban-focused injury prevention campaigns are urgently needed. These should emphasise safe storage, the dangers of home extraction and illegal practices and the risks of impaired caregiving. It is also vital that other countries considering RCL implementation integrate targeted educational campaigns for these specific groups if they move forward with legalisation for preventative measures. Limitations This study has some limitations. Full ED data are only available for Ontario and Alberta, which may under-represent national trends. However, the data are population-based, from two diverse provinces, and represent over 21 million people. Data were restricted to documented ED visits and hospitalisations, so our study could not account for burns that did not result in ambulatory/hospital care, or deaths occurring where the patient never went to the ED or hospital. Additionally, burn encounters could not be attributed to cannabis use at the individual level. Cannabis-specific ICD-10-CA codes (eg, F12.x) were not used to define or stratify the groups, as these codes are assigned only when cannabis use is disclosed by the patient and documented by the clinician, and cannabis use is substantially under-disclosed in healthcare settings.3435 Critically, this under-coding is differential across the study period, as patients had stronger incentives to conceal use prior to legalisation; stratifying on these codes would therefore have introduced a time-varying ascertainment bias, confounding policy effects with disclosure effects.3435 Such codes also would not capture mechanistically relevant exposures, including extraction-related injuries, where cannabis usage would usually not be coded. Furthermore, we were unable to account for potential confounding from concurrent tobacco use, such as burns caused by smoking cigarettes, which may accompany cannabis use; however, this is likely a minor consideration given the steady decline in Canadian tobacco smoking rates over the study period.36 Also, as vape devices are increasingly used to vaporise cannabis, the potential for device-related thermal injury represents a plausible cannabis-attributable mechanism.37 Lastly, given the large number of analyses conducted, there was a potential for chance findings. To account for this, Holm-Bonferroni corrections were applied, and the results remained statistically significant after adjustment. Conclusion In this national analysis, we observed time-specific increases in serious burn injuries requiring hospitalisation around the announcement and rollout periods of RCL, with the most pronounced changes among toddlers, males and urban residents. Fewer and less consistent increases were observed for ED visits, suggesting that changes in burn-related healthcare utilisation during these periods were more concentrated among hospitalised cases, which typically reflect more severe injuries. Together, these findings extend existing evidence by characterising national and subgroup-specific patterns in burn injuries across key policy and societal transition periods. As cannabis policies continue to evolve, targeted injury-prevention and harm-reduction efforts, particularly those focused on child safety and high-risk contexts, remain important to mitigate severe burn injuries in the setting of expanding access. Supplementary material