Study design and setting

We conducted a population-based, cross-sectional study of all accidental benzodiazepine-related toxicity deaths that occurred in Ontario, Canada between January 1, 2018 and June 30, 2022. Accidental benzodiazepine toxicity deaths were defined as those arising from an acute intoxication due to the direct contribution of a benzodiazepine, either alone or in combination with other drugs or substances, and which were confirmed as unintentional (i.e., an injury where death was not intended, foreseen or expected) by the investigating coroner. The study is reported as per the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [13] with a preliminary analysis of this data previously made available online [14,15].

Data sources

We obtained data from ICES (formerly known as the Institute for Clinical Evaluative Sciences), an independent, non-profit research institute with legal status allowing for the analysis of administrative data for healthcare system evaluation and improvement. To identify people who died of benzodiazepine-related toxicity, we used the Drug and Drug/Alcohol Related Death Database, which holds records from coroner’s investigations of all confirmed toxicity deaths where benzodiazepines, opioids, alcohol, and/or stimulants directly contributed to death in Ontario. The database captures post-mortem toxicology of the substances directly involved and contributing to death, as well as sociodemographic information and details about the manner and circumstances surrounding death as determined by the investigating coroner, including the location of the acute toxicity incident, the number and type of substances contributing to death, and whether a bystander was present.

To enumerate the population of Ontario and ascertain the sociodemographic characteristics of the study population, we used the Registered Persons Database, a registry of all Ontario residents eligible for the universal and publicly funded Ontario Health Insurance Plan. We used the Narcotics Monitoring System, which contains all prescriptions for controlled substances dispensed from community pharmacies in Ontario regardless of the payer, to capture prior dispensed prescriptions for benzodiazepines and opioids. We identified healthcare encounters in the emergency department and in inpatient settings using the Canadian Institute for Health Information’s (CIHI) National Ambulatory Care Reporting System and Discharge Abstract Database, respectively, and we ascertained visits to outpatient care using the Ontario Health Insurance Plan Claims Database. These datasets were linked using unique encoded identifiers and analyzed at ICES. The use of the data in this project is authorized under section 45 of Ontario’s Personal Health Information Protection Act and does not require review by a Research Ethics Board. The study team did not have access to any identifying personal information and accessed the data for research purposes between September 29, 2023 and January 31, 2025.

Measures

We measured the monthly rate of benzodiazepine-related toxicity deaths to examine temporal trends over the study period. We summarized the characteristics of each person who died of benzodiazepine-related toxicity over the study period and the circumstances surrounding their death. This included age (0–24, 25–44, 45–64, 65 + years), sex, neighbourhood income quintile (using CIHI’s standard definition [16]), location of residence (urban or rural region; northern or southern region), setting of the toxicity incident (private residence, other, unknown), other substances (opioids, alcohol, or stimulants) directly contributing to death, and the number substances directly involved in the death (benzodiazepines alone or multiple substance). We also assessed the prevalence of dispensed prescriptions for benzodiazepines and opioids in the 30 days, one year, and five years prior to death. Finally, we summarized prior healthcare encounters for acute substance-related toxicities (due to benzodiazepines, opioids, alcohol, or stimulants) in the year before death, and for substance use disorder diagnoses (any substance, and benzodiazepines specifically) in the five years before death.

We stratified all characteristics and circumstances according to the type of the benzodiazepine (approved vs. not-approved for medical use) directly contributing to death, based on each benzodiazepines’ status in Canada (i.e., whether each benzodiazepine is legally available by prescription) [17]. Benzodiazepines not approved for medical use in Canada included: etizolam, flubromazolam, flualprazolam, and bromazolam, while benzodiazepines approved for medical use in Canada included: alprazolam, bromazepam, camazepam, chlordiazepoxide, clobazam, clonazepam, demoxepam, diazepam, flurazepam, hydroxyalprazolam, lorazepam, midazolam, nitrazepam, oxazepam, temazepam, and clorazepate. Deaths were attributed to non-approved benzodiazepines if post mortem toxicology identified these substances, either alone or in combination with an approved benzodiazepine. Deaths were attributed to an approved benzodiazepine if toxicology identified only benzodiazepines approved for medical use. See S1 Table for details on definitions, diagnosis codes and benzodiazepine types.

Statistical analysis

We calculated the monthly crude rate of benzodiazepine-related toxicity deaths (per 100,000 Ontario residents) across the study period. We summarized the characteristics and circumstances of people who died of benzodiazepine-related toxicity using descriptive statistics, reported overall and stratified by the type of benzodiazepine directly involved in death. We used chi-square tests and Fisher’s exact tests to compare characteristics and circumstances according to benzodiazepine type. All analyses were conducted at ICES using SAS Enterprise Guide 8.3 (SAS institute, Cary, North Carolina, USA), and used a type 1 error rate of 0.05 as the threshold for statistical significance.

Involvement of people with lived experience

The Ontario Drug Policy Research Network (ODPRN) regularly engages a Lived Experience Advisory Group, comprised of people with lived and living experience using opioids. Throughout the study, we consulted with this group, along with a separate group of five individuals with lived and living experience using a range of substances, including opioids, stimulants, benzodiazepines, and alcohol, about the study scope, methodology, and interpretation of the results. Throughout the conduct of this work, we met six times, where these individuals provided feedback on the study scope, methodology, contextualization, and reporting of results. All people with lived and living experience were compensated for their time spent on this study in line with standard best practices, [18] and were offered co-authorship (or when preferred by the individual, acknowledgment) in this manuscript.

Trends

Between January 2018 and June 2022, 782 Ontarians had an accidental death involving benzodiazepine toxicity. Over the first two-and-a-half years of the study period, the crude rate of benzodiazepine-related toxicity death remained relatively stable (from 0.08 per 100,000 in January 2018 [N = 12] to 0.05 per 100,000 in June 2020 [N = 8]), notwithstanding minor seasonal variations in rates observable as troughs in summer months and peaks during cooler months (Fig 1). However, in July 2020, rates of benzodiazepine-related toxicity death began to rise in Ontario, reaching a peak of 0.27 per 100,000 (N = 40) in December 2020. By October 2021, the rate of benzodiazepine-related toxicity deaths had declined to 0.04 deaths per 100,000 (N = 6) and remained relatively stable for the remainder of the study period (Fig 1).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Trends in the rate of benzodiazepine-involved toxicity deaths in Ontario, January 2018 to June 2022.

https://doi.org/10.1371/journal.pone.0355210.g001

Characteristics and circumstances

Over the study period, benzodiazepine-related toxicity deaths were evenly distributed by benzodiazepine type, with half (50.1%; N = 392) involving benzodiazepines approved for medical use, and the other half (49.9%; N = 390) involving benzodiazepines not approved for medical use (Table 1). Overall, the median age at death was 40 years (interquartile range [IQR] 30–53 years), although this was higher for deaths related to approved benzodiazepines (median age: 45 years [IQR: 33–56 years] vs. 37 years [IQR: 28–49 years] for deaths attributed to non-approved benzodiazepines; p < 0.001). Males accounted for more than two-thirds of deaths (70.5%; N = 551), and this predominance was more pronounced for deaths involving benzodiazepines not approved vs. approved for medical use (76.9% vs. 64.0%; p < 0.001). Overall, most benzodiazepine-related toxicity deaths occurred among people residing in urban (91.2%; N = 713) and southern (88.6%; N = 693) regions of the province. Notably, northern-residing individuals accounted for a larger proportion of deaths with benzodiazepines approved vs. not approved for medical use (14.8% vs. 7.9%; p = 0.003). In general, deaths were concentrated in lower-income neighbourhoods, with approximately 60% of decedents (N = 462; 59.1%) residing in communities in the two lowest quintiles of income (Table 1).

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Characteristics of Ontarians who died of benzodiazepine-involved toxicity, January 1, 2018 to June 30 2022.

https://doi.org/10.1371/journal.pone.0355210.t001

Three-quarters (75.4%; N = 590) of all benzodiazepine-related toxicity deaths occurred at a private residence, and this was even more common among those involving benzodiazepines approved for medical use (79.8% vs. 71.0%; p = 0.007; Table 2). Moreover, the vast majority of deaths (98.1%; N = 767) involved more than one substance. Opioids were the most common co-occurring substance, with direct involvement in 9 of 10 benzodiazepine-related toxicity deaths overall (91.2%; N = 713), more commonly among deaths involving benzodiazepines not approved for medical use (95.1% vs. 87.2%; p < 0.001). Stimulants were co-involved in approximately half of all benzodiazepine-related toxicity deaths (47.1%; N = 368); however, this prevalence was considerably higher among deaths involving benzodiazepines not approved for medical use (60.5% vs. 33.7%; p < 0.001). Alcohol was involved in 18.5% of all benzodiazepine-related toxicity deaths (N = 145), and this was more common among deaths involving approved benzodiazepines (24.5% vs. 12.6%; p < 0.001), in contrast to the pattern observed with stimulants. With regards to prescribing history, we observed a much higher prevalence of benzodiazepine dispensing among deaths attributed to benzodiazepines approved for medical use, in the 30 days (60.7% vs. 10.8%; p < 0.001), one year (70.4% vs. 20.8%; p < 0.001), and five years before death (80.1% vs. 36.9%; p < 0.001). A similar pattern was observed for a history of opioid prescribing in the 30 days (46.7% vs. 24.9%; p < 0.001), one year (66.8% vs. 46.9%; p < 0.001), and five years prior to death (85.5% vs. 73.3%; p < 0.001). Notably, one-fifth (22.6%; N = 177) of all Ontarians who died of benzodiazepine-related toxicity experienced a hospital-treated substance toxicity in the year prior to death. This history was more common among deaths involving benzodiazepines not approved for medical use (27.4% vs. 17.9%; p = 0.001), driven by a history of opioid toxicity in particular (25.1% vs. 13.8%; p < 0.001). Finally, in the five years preceding death, nearly two-thirds (64.1%; N = 501) of decedents had a healthcare encounter where a substance use disorder diagnosis was recorded, and this was similar across benzodiazepine type (Table 2).

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Circumstances surrounding and preceding death among people who died of benzodiazepine-involved toxicity, January 1, 2018 to June 30 2022.

https://doi.org/10.1371/journal.pone.0355210.t002

Limitations

A core strength of this study is that it is the first to examine trends and patterns in benzodiazepine toxicity deaths using a population-based database in Ontario. However, several limitations warrant discussion. First, in accordance with institutional privacy policy prohibiting the reporting of small cells (counts <5), we were unable to stratify trends in deaths by benzodiazepine type, limiting our ability to understand the degree to which changes in trends were driven by approved vs. non-approved benzodiazepines. Future research should examine how different benzodiazepines may affect toxicity risk, particularly in the context of exposure to multiple substances including fentanyl. Second, our analysis was restricted to deaths in which benzodiazepines directly contributed to the toxicity incident. However, the thresholds for differentiating between whether a benzodiazepine either directly contributed or was involved in death are not well established. Therefore, although our analysis may underestimate the rate of benzodiazepine-related mortality in Ontario, the specificity of our case definition helps to avoid misclassification. Third, information on indication of benzodiazepine use is not readily available in administrative data, and therefore we were not able to assess the reason for prescribing among people whose death arose from a benzodiazepine approved for medical use. Fourth, there is no validated definition of substance use disorders, and therefore we relied on diagnoses made during prior healthcare encounters and receipt of opioid agonist treatment in the prior five years. This definition may under-report people with substance use disorders who are disconnected from the healthcare system, although we anticipate this potential misclassification to be minimal. Finally, we used standard area-level definitions to define neighbourhood income quintile which may only moderately correlate with individual-level income data, and could underestimate the magnitude of sociodemographic differences [38].