Supervised injection facility use and exposure to violence among a cohort of people who inject drugs: A gender-based analysis

Abstract

Background:

Supervised injection facilities (SIFs) have been established in many settings, in part to reduce risks associated with injecting in public, including violence. However, the relationship between SIF use and experiencing violence has not yet been thoroughly evaluated. We sought to longitudinally examine the gender-specific relationship between SIF use and exposure to violence among people who inject drugs (PWID) in a Canadian setting.

Methods:

Data were drawn from two prospective cohort studies of PWID in Vancouver, Canada, between December 2005 and December 2016. Semi-annually, participants completed questionnaires that elicited data concerning sociodemographic characteristics, behavioural patterns, violent encounters and health service utilization. We used multivariable generalized estimating equations (GEE) to estimate the independent association between exclusively injecting drugs at a SIF and experiencing physical or sexual violence among men and women PWID, respectively.

Results:

Of 1930 PWID followed for a median of four years, 679 (35.2%) were women and the median age was 41 years at baseline. In total, 353 (52.0%) women and 694 (55.5%) men reported experiencing at least one incident of violence during follow-up. In multivariable analyses, exclusive SIF use was associated with decreased odds of experiencing violence among men after adjusting for potential confounders (Adjusted Odds Ratio [AOR] = 0.64; 95% confidence interval [CI]: 0.46 – 0.89). Exclusive SIF use was not significantly associated with experiencing violence among women in adjusted analyses (AOR = 0.97; 95% CI: 0.57 – 1.66).

Conclusion:

In light of the recent expansion of SIFs in Canada, our finding of a protective association between exclusive SIF use and exposure to violence among men is encouraging. The fact that we did not observe a significant association between SIF use and experiencing violence among women highlights the need for social-structural interventions that are more responsive to the specific needs of women PWID in relation to violence prevention.

INTRODUCTION

Exposure to physical and sexual violence among drug-using populations remains a public health problem in diverse settings worldwide (Degenhardt & Hall, 2012). Studies have documented particularly heightened rates of exposure to violence among marginalized drug-using populations, including people who inject drugs (PWID) (Hayashi et al., 2013; Kennedy et al., 2017a; Kutsa et al., 2016; Lorvick et al., 2014; Williams et al., 2018). For example, a community-based study of women who use drugs in San Francisco found that 26% had experienced partner violence and 28% had experienced non-partner violence in the previous six months (Lorvick et al., 2014). Similarly, another study found that 22% of a community-recruited cohort of PWID in Vancouver, Canada, had experienced physical or sexual violence in the previous six months (Richardson et al., 2015). In addition to physical injury, such violent encounters may have further adverse consequences for the health of PWID (Afifi et al., 2009; Braitstein et al., 2003; Hedtke et al., 2008). For example, exposure to violence has been associated with an increased likelihood of mental health concerns, including post-traumatic stress disorder and mood and anxiety disorders (Afifi et al., 2009; Hedtke et al., 2008). Moreover, PWID who have experienced violence have been found to be more likely to engage in drug use practices that increase risk of overdose and infectious disease transmission (Braitstein et al., 2003), and are also known to avoid health services near locations where they have experienced violence (McNeil et al., 2014a).

Over the past decade, there has been growing research interest in the role of contextual determinants of exposure to violence among PWID (Boyd et al., 2018a; Fairbairn et al., 2008; Kennedy et al., 2017a; Lorvick et al., 2014; Marshall et al., 2008; McNeil et al., 2014a; Richardson et al., 2015). Accompanying this shift has been the increasing application of Rhodes’ risk environment framework as a heuristic for guiding such investigations (Rhodes, 2002; Rhodes et al., 2012). Although first applied in studies of HIV risk (Rhodes et al., 2012), Rhodes’ risk environment framework has been extended to conceptualize how individual experiences of interpersonal violence among PWID are shaped by the interplay between various social, structural and environmental forces operating at the macro-, meso- and micro-levels of environmental influence (Boyd et al., 2018a; Fairbairn et al., 2008; Kennedy et al., 2017a; Lorvick et al., 2014; Marshall et al., 2008; McNeil et al., 2014a). For example, studies have illustrated how street-based drug scenes (i.e., inner-city areas characterized by high concentrations of people who use drugs and drug market activity) are key risk environments that contribute to the production of violence among PWID, with many violent encounters related to drug purchasing, drug dealing and sex work within these environments (Erickson, 2001; McNeil et al., 2014a; Richardson et al., 2015; Romero-Daza et al., 2003; Shannon et al., 2008; Small et al., 2007). Further, the preparation and injection of drugs in particular public spaces within street-based drug scenes (e.g., alleyways, public washrooms) has been found to increase susceptibility to violence, including assault by police officers as well as robbery and assault by strangers and acquaintances (Boyd et al., 2018a; McNeil et al., 2014a; Small et al., 2007). In addition, studies have demonstrated how social-structural exposures that disproportionately affect urban drug-using populations, including housing instability and street-based law enforcement, contribute to violence among PWID (Aitken et al., 2002; Boyd et al., 2018a; Kennedy et al., 2017a; Marshall et al., 2008; Werb et al., 2011).

Rhodes’ risk environment framework is also useful for conceptualizing how gendered structures and social relations may shape susceptibility to violence and exploitation within street-based drug scenes (Boyd et al., 2018a; Bungay et al., 2010; Epele, 2002; Fairbairn et al., 2008; Lorvick et al., 2014; McNeil et al., 2014a). For example, qualitative and ethnographic studies have highlighted how women are often rendered vulnerable to violence as a result of their subordination within street-based drug economy hierarchies (Boyd et al., 2018b; Maher & Hudson, 2007; McNeil et al., 2014a). Violence against women is also common in injection drug-using partnerships, with physical and emotional abuse often ensuing in conflicts over control of income and resources generated by women (Bourgois et al., 2004; Fairbairn et al., 2008; McNeil et al., 2014a). While less research attention has focused on exposure to violence among men, quantitative studies of PWID have demonstrated that correlates of received violence, as well as types and perpetrators of assault, may differ between men and women (Kutsa et al., 2016; Marshall et al., 2008). For example, men have been found to experience higher levels of assault from police officers and strangers than women (Kutsa et al., 2016; Marshall et al., 2008).

Increased recognition of the role of contextual and gendered forces in shaping vulnerability to violence among PWID has drawn attention to the need for interventions that seek to address social, structural and environmental determinants of risk (Kerr et al., 2007; McNeil & Small, 2014; Rhodes et al., 2006). Among such interventions are supervised injection facilities (SIFs), which have been implemented in many settings worldwide in response to health and social harms related to injection drug use, including street-based drug scene violence (European Monitoring Centre for Drugs and Drug Addiction, 2018; Kennedy et al. 2017b). SIFs provide secure and hygienic environments in which individuals can inject pre-obtained illicit drugs under the supervision of trained staff, as well as access sterile injecting equipment, receive emergency care in the event of overdose, and be connected with other on-site and external services (European Monitoring Centre for Drugs and Drug Addiction, 2018).

Since 2003, North America’s first sanctioned SIF, Insite, has been operating in the Downtown Eastside (DTES) neighbourhood of Vancouver, Canada, a 40- to 50-block inner-city area that is characterized by a highly visible street-based drug scene and disproportionately high rates of poverty, unemployment and criminalization (Kerr et al., 2017). Until 2016, Insite remained the only sanctioned SIF in the continent. However, at least thirty-nine additional federally-sanctioned SIFs have since been established in communities across Canada as part of the response to the ongoing overdose crisis (Health Canada, 2020). Additionally, more than thirty temporary low-threshold SIFs, termed overdose prevention sites, have since been provincially approved and implemented in Canada (Kerr et al., 2017; Irvine et al., 2019; Perkel, 2019). In total, three SIFs and six overdose prevention sites were operating in Vancouver as of January 2020 (Vancouver Coastal Health, 2020).

Numerous studies undertaken in Vancouver and internationally have demonstrated the positive health and social impacts of SIFs (Kennedy et al., 2017b; McNeil & Small, 2014; Potier et al., 2014). Of note, SIFs have been found to reduce the risk of drug-related harms, including overdose death and infectious disease transmission (Bravo et al., 2009; Irvine et al., 2019; Kerr et al., 2005; Kinnard et al., 2014; Marshall et al., 2011), and to facilitate uptake of addiction treatment among PWID (DeBeck et al., 2011; Gaddis et al., 2017; Wood et al., 2007; Wood et al., 2006a). The establishment of such facilities has also been shown to contribute to improvements in public order with no observed increases in local drug-related crime rates (Donnelly & Mahoney, 2013; Wood et al., 2004; Wood et al., 2006b). However, despite providing safer alternative environments to the open street-based drug scene for the consumption of drugs, the potential impact of SIF use on exposure to violence has not been thoroughly evaluated.

Several qualitative and ethnographic studies undertaken in Vancouver have provided evidence indicating that SIFs may offer protection from some forms of violence among certain subpopulations of PWID, including women and people who require help injecting (Boyd et al., 2018a; Fairbairn et al., 2008; McNeil et al., 2014b; Small et al., 2012a). For example, studies of sanctioned mixed-gender SIFs in this setting have illustrated how these facilities may serve to temporarily protect women PWID from violence during drug preparation and consumption, and might also help to mitigate exploitation and inter-partner violence by enabling women to exercise greater control over resources and the injecting process (Boyd et al., 2018a; Fairbairn et al., 2008; Small et al., 2012a). However, recent ethnographic work has also described how, despite operating as gender-neutral programmes, these facilities remain male-dominated environments in which women PWID are routinely subjected to harassment from men, which may deter women’s access (Boyd et al., 2018a). As such, questions remain about the potential role of SIFs in providing protection from violence among both women and men PWID. We therefore undertook the present study to longitudinally examine the gender-specific relationship between exclusively injecting at the Insite SIF and exposure to physical or sexual violence among a community-recruited cohort of PWID in Vancouver. This study may provide useful information to guide the development and implementation of evidence-based violence prevention strategies that are aligned with the specific needs of men and women PWID.

METHODS

The Vancouver Injection Drug User Study (VIDUS) and AIDS Care Cohort to evaluate Exposure to Survival Services (ACCESS) are open community-recruited prospective cohort studies of people who use drugs in Vancouver that began enrolment using community-based methods, including street outreach and self-referral, in May 1996. The two cohorts have been described in detail previously (Strathdee et al., 1998; Wood et al., 2001). In brief, VIDUS is a cohort of HIV-negative adults who have injected illicit drugs at least once in the month preceding enrolment. ACCESS is a cohort of HIV-positive adults who have used illicit drugs other than or in addition to cannabis (which was illegal in Canada during the study period) in the month preceding enrolment. VIDUS participants who seroconvert to HIV during follow-up are transferred into ACCESS. At baseline and semi-annually thereafter, participants provide blood samples for HIV/HCV serological testing and HIV clinical monitoring, as appropriate, and complete an interviewer-administered questionnaire that elicits information concerning sociodemographic characteristics, behavioural patterns, social-structural exposures and health service utilization. Participants receive a $40 CAD honorarium at each study visit. The cohorts have received ethical approval from the Providence Health Care/University of British Columbia Research Ethics Board.

The present analyses were restricted to participants who completed at least one study interview between December 1, 2005 and November 30, 2016, the time period during which all variables of interest were available. We further restricted to observations during the study period in which these participants reported having injected drugs in the previous six months, as only active injectors were eligible to use the SIF. The primary outcome for this analysis was response to the question: “Have you been attacked, assaulted (including sexual assault) or suffered any kind of violence in the last six months?” (December 2005 to June 2014) (yes vs. no). Post June 2014, this was defined in response to the questions “In the last six months, have you been physically attacked or suffered any kind of physical violence, including torture or punishment related to a drug debt?” and “In the last six months, have you been forced to have sex or perform a sexual act against your will, or experienced any kind of sexual assault?” (yes to either question vs. no to both questions). The primary explanatory variable of interest was exclusive SIF use, defined in response to the question, “In the last six months, what proportion of injections did you do at Insite?” Consistent with our past work (Lloyd-Smith et al., 2008; Wood et al., 2005a), responses were classified as all vs. most, some, a few or none. Given that SIFs have been found to engage PWID who are more likely to inject drugs in public and other potentially unsafe settings (Bravo et al. 2009; Kennedy et al. 2019; Kimber et al., 2003; Wood et al., 2005b; Wood et al., 2006c), this measure of SIF use was employed in effort to rule out the possibility that estimates of the association between SIF use and violence may be biased due to SIF users being more likely to experience violence while consuming drugs in such settings.

To estimate the independent association between exclusive SIF use and exposure to violence, we assessed a range of sociodemographic, social-structural and behavioural variables as potential confounders. The selection of these variables was informed by Rhodes’ risk environment framework (Rhodes et al., 2012), and previous studies investigating SIF use and/or violence among drug-using populations (Boyd et al., 2018a; Fairbairn et al., 2008; Jessell et al., 2017; Lorvick et al., 2014; Marshall et al., 2008; McNeil et al., 2014b; Richardson et al., 2015; Wenzel et al., 2001; Wenzel et al., 2004). These variables included: age (per year older); ancestry (white vs. non-white); relationship status (legally married/common law/regular partner vs. other); childhood emotional abuse (moderate/severe vs. low/none); calendar year of interview (per year increase); and use of injection heroin; injection cocaine; injection crystal methamphetamine; and non-injection crack cocaine (all ≥daily vs. <daily). Other variables considered included: DTES residence; employment (regular job/temporary job/self-employed); homelessness; HIV seropositivity; binge injection; requiring help injecting; heavy alcohol use (National Institute on Alcohol Abuse and Alcoholism, 2016); enrolment in methadone maintenance therapy (MMT); sex work involvement; drug dealing; and incarceration (all yes vs. no). Consistent with previous studies (Kennedy et al., 2017a; Lake et al., 2015), childhood emotional abuse was defined as a score of ≥13 (moderate/severe) vs. <13 (low/none) on the emotional abuse subscale of the Childhood Trauma Questionnaire (Bernstein & Fink, 1998). All variables were treated as time-updated based on each semi-annual follow-up visit and refer to the six-month period preceding the interview date unless otherwise indicated.

As a first step, we examined characteristics associated with recent exposure to violence at baseline, stratified by self-reported current gender identity (women (trans inclusive) vs. men (trans inclusive)). Participants with non-binary gender identities (n = 4; 0.2%) were included in the ‘women’ category given that we had insufficient statistical power to conduct meaningful analyses among this small group. While we recognize the limitations of such classification (Bauer et al., 2017; Bauer et al., 2009; Fraser, 2018), we believe that this was the most appropriate approach given that a women-only (trans-inclusive) SIF that also accommodates individuals with non-binary gender identities recently began operating in Vancouver (Boyd et al., 2018a) and our findings may be useful in informing the development of this service. Continuous variables were compared using the Wilcoxon rank-sum test. Categorical variables were compared using Pearson’s chi-squared test or Fisher’s exact test when expected cell counts were less than or equal to five. Next, we used time-updated generalized estimating equations (GEE) with a logit-link function to estimate unadjusted odds ratios (OR) for the association between past-six-month exposure to violence and each explanatory variable, stratified by gender. We used GEE for the analysis of correlated data given that the factors potentially associated with experiencing violence during follow-up were time-dependent measures. GEE analyses allowed for consideration of factors associated with experiencing violence over the full duration of the study period, with standard errors adjusted for multiple observations per individual using an exchangeable correlation structure (Lee et al., 2007). Therefore, data from every participant study visit were considered in this analysis. This approach allowed for examination of associations between the explanatory variables and six-month periods when violence did and did not occur, both within and between individuals. This method has been used successfully in previous cohort studies of violence among PWID (Kennedy et al., 2017a; Lake et al., 2015; Marshall et al., 2008; Richardson et al., 2015).

To estimate the gender-specific independent association between exclusive SIF use and exposure to violence, we constructed two multivariable models using an a-priori variable selection process described previously (Maldonado & Greenland, 1993). First, we fit a full model, for women and men, respectively, that included exclusive SIF use and all secondary explanatory variables that were significant at the level of p < 0.10 in bivariable GEE analyses. We then fit a series of reduced models that excluded each secondary explanatory variable. We compared the value of the coefficient associated with SIF use in each model to the full model and removed the secondary variable corresponding to the smallest relative change in the coefficient from further consideration. We continued this process until the minimum change in the value of the coefficient exceeded 5%. We have previously used this approach to estimate the independent relationship between a primary exposure variable and violence (Kennedy et al., 2017a; Lake et al., 2015).

Finally, as a subanalysis, we descriptively assessed characteristics of violent incidents experienced by participants, stratified by gender. Specifically, we analyzed responses to the following questions that were asked of participants who reported recent (i.e., past-six-month) exposure to violence: “Who has attacked you?” and “What type of attack was it?” Response options for the former question included: stranger; dealer; police; husband/wife; boyfriend/girlfriend; partner; sex work client; sex worker; friend; regular sex partner; casual sex partner; security guard; acquaintance; don’t know; and other (specify). Response options for the latter question included: beating; sexual assault/rape; attacked with weapons; strangled; attacked or threatened with a gun; robbery; and other (specify). Participants could provide more than one response for each of these questions. These questions were only included in the study questionnaire from December 2005 to December 2014, and thus this subanalysis was restricted to responses from participants’ first report of recent exposure to violence during this time period. We compared characteristics of violent encounters between women and men using Pearson’s chi-squared test or Fisher’s exact test when expected cell counts were less than or equal to five. We conducted all analyses with SAS version 9.4 (SAS Institute, Cary, NC). All p-values are two-sided.

RESULTS

Between December 2005 and December 2016, 2188 VIDUS and ACCESS participants completed at least one study visit. Of these, 1930 reported past-six-month injection drug use in at least one study interview during this period and were therefore included in the present study. At baseline, the median age of the study sample was 41 years (interquartile range [IQR]: 34 – 48), 679 (35.2%) were women, and 491 (25.4%) reported having experienced violence in the previous six months. Participants contributed a total of 9291 person-years of observation, with a median follow-up duration of 44 months (IQR: 12–106) per participant among women and 52 months (IQR: 12-105) per participant among men. Of the total 14,351 observations included in the study, 2451 (17.1%) included at least one reported incident of exposure to violence in the previous six-month period, including 782 (16.2%) among women and 1669 (17.8%) among men. In total, 353 (52.0%) women and 694 (55.5%) men reported at least one six-month period in which they experienced violence during follow-up, while 104 (15.3%) women and 197 (15.8%) men reported at least one six-month period of exclusive SIF use. Tables 1 and 2 present the baseline characteristics of the study sample, stratified by recent exposure to violence, among women and men, respectively.

TABLE 1:

BASELINE CHARACTERISTICS OF 679 WOMEN WHO INJECT DRUGS IN VANCOUVER, CANADA, STRATIFIED BY RECENT EXPOSURE TO VIOLENCE^* (2005-2016).

Characteristic	Yes n (%) n = 162	No n (%) n = 501	Odds Ratio (95% CI)‡
Age (per year older)
Median (IQR)†	38 (31–44)	38 (31–45)	0.99 (0.97 – 1.01)
Ancestry
White	90 (55.9)	242 (48.4)	1.35 (0.95 – 1.92)
Non-white	71 (44.1)	258 (51.6)
Downtown Eastside residence*
Yes	135 (83.3)	368 (73.5)	1.81 (1.14 – 2.86)
No	27 (16.7)	133 (26.6)
Employment*
Yes	25 (15.4)	62 (12.4)	1.29 (0.78 – 2.14)
No	137 (84.6)	439 (87.6)
In a relationship*
Yes	61 (37.9)	181 (36.6)	1.06 (0.73 – 1.53)
No	100 (62.1)	314 (63.4)
Homeless*
Yes	78 (48.2)	166 (33.3)	1.86 (1.30 – 2.66)
No	84 (51.9)	332 (66.7)
HIV seropositive*
Yes	45 (27.8)	210 (41.9)	0.53 (0.36 – 0.79)
No	117 (72.2)	291 (58.1)
Heroin injection*
≥Daily	77 (47.5)	177 (35.3)	1.66 (1.16 – 2.37)
<Daily	85 (52.5)	324 (64.7)
Cocaine injection*
≥Daily	17 (10.5)	59 (11.8)	0.88 (0.50 – 1.56)
<Daily	145 (89.5)	442 (88.2)
Crystal methamphetamine injection*
≥Daily	21 (13.0)	25 (5.0)	2.86 (1.55 – 5.26)
<Daily	140 (87.0)	476 (95.0)
Crack cocaine use*
≥Daily	85 (52.5)	231 (46.1)	1.29 (0.91 – 1.84)
<Daily	77 (47.5)	270 (53.9)
Binge injecting*
Yes	56 (34.8)	139 (28.0)	1.37 (0.94 – 2.01)
No	105 (65.2)	358 (72.0)
Require help injecting*
Yes	76 (46.9)	154 (31.1)	1.96 (1.37 – 2.82)
No	86 (53.1)	342 (69.0)
Exclusive supervised injection facility use*
Yes	4 (2.5)	16 (3.2)	0.77 (0.25 – 2.34)
No	157 (97.5)	484 (96.8)
Heavy alcohol use*
Yes	29 (18.1)	76 (15.2)	1.24 (0.77 – 1.98)
No	131 (81.9)	425 (84.8)
Enrolment in MMT*^
Yes	79 (48.8)	262 (52.3)	0.87 (0.61 – 1.24)
No	83 (51.2)	239 (47.7)
Sex work involvement*
Yes	82 (51.6)	199 (40.0)	1.60 (1.12 – 2.29)
No	77 (48.4)	299 (60.0)
Drug dealing*
Yes	68 (42.2)	151 (30.2)	1.69 (1.17 – 2.44)
No	93 (58.0)	349 (69.8)
Incarceration*
Yes	25 (15.4)	71 (14.2)	1.10 (0.67 – 1.81)
No	137 (84.6)	429 (85.8)
Childhood emotional abuse
Yes	96 (68.1)	244 (55.5)	1.71 (1.15 – 2.56)
No	45 (31.9)	196 (44.6)

^‡CI = confidence interval.

^*Refers to activities or experiences in the 6-month period prior to a baseline interview.

^†IQR = interquartile range.

^{^}MMT = methadone maintenance therapy.

Note: not all cells add up to n = 679 due to missing values.

Table 2:

Baseline characteristics of 1251 men who inject drugs in Vancouver, Canada, stratified by recent exposure to violence^* (2005-2016).

Characteristic	Yes n (%) n = 323	No n (%) n = 921	Odds Ratio (95% CI)‡
Age (per year older)
Median (IQR)†	42 (33–47)	44 (36–50)	0.97 (0.95 – 0.98)
Ancestry
White	238 (73.7)	662 (72.0)	1.09 (0.82 – 1.45)
Non-white	85 (26.3)	257 (28.0)
Downtown Eastside residence*
Yes	250 (77.4)	619 (67.2)	1.67 (1.24 – 2.24)
No	73 (22.6)	302 (32.8)
Employment*
Yes	77 (23.9)	262 (28.5)	0.79 (0.59 – 1.06)
No	245 (76.1)	659 (71.6)
In a relationship*
Yes	91 (28.4)	205 (22.5)	1.37 (1.03 – 1.83)
No	229 (71.6)	707 (77.5)
Homeless*
Yes	154 (47.8)	327 (35.6)	1.66 (1.28 – 2.15)
No	168 (52.2)	592 (64.4)
HIV seropositive*
Yes	111 (34.4)	374 (40.6)	0.77 (0.59 – 1.00)
No	212 (65.5)	547 (59.4)
Heroin injection*
≥Daily	97 (30.0)	233 (25.3)	1.27 (0.96 – 1.68)
<Daily	226 (70.0)	688 (74.7)
Cocaine injection*
≥Daily	22 (6.8)	99 (10.8)	0.61 (0.38 – 0.99)
<Daily	300 (93.2)	822 (89.3)
Crystal methamphetamine injection*
≥Daily	26 (8.1)	51 (5.5)	1.49 (0.91 – 2.44)
<Daily	297 (92.0)	869 (94.5)
Crack cocaine use*
≥Daily	123 (38.2)	286 (31.1)	1.37 (1.05 – 1.78)
<Daily	199 (61.8)	633 (68.9)
Binge injecting*
Yes	97 (30.1)	241 (26.4)	1.20 (0.91 – 1.59)
No	225 (69.9)	672 (73.6)
Require help injecting*
Yes	102 (31.7)	211 (23.2)	1.54 (1.16 – 2.04)
No	220 (68.3)	700 (76.8)
Exclusive supervised injection facility use*
Yes	11 (3.4)	52 (5.7)	0.58 (0.30 – 1.13)
No	312 (96.6)	860 (94.3)
Heavy alcohol use*
Yes	47 (14.6)	85 (9.2)	1.67 (1.14 – 2.45)
No	276 (85.5)	835 (90.8)
Enrolment in MMT*^
Yes	127 (39.6)	368 (40.0)	0.98 (0.76 – 1.28)
No	194 (60.4)	553 (60.0)
Sex work involvement*
Yes	16 (5.0)	37 (4.0)	1.24 (0.68 – 2.27)
No	307 (95.0)	882 (96.0)
Drug dealing*
Yes	166 (51.4)	288 (31.3)	2.32 (1.79 – 3.01)
No	157 (48.6)	632 (68.7)
Incarceration*
Yes	85 (26.3)	150 (16.3)	1.83 (1.35 – 2.48)
No	238 (73.7)	768 (83.7)
Childhood emotional abuse
Yes	154 (51.9)	358 (42.1)	1.48 (1.14 – 1.93)
No	143 (48.2)	492 (57.9)

^‡CI = confidence interval.

^*Refers to activities or experiences in the 6-month period prior to a baseline interview.

^†IQR = interquartile range.

^{^}MMT = methadone maintenance therapy.

Note: not all cells add up to n=1251 due to missing values.

Table 3 presents the results of the crude and adjusted longitudinal estimates of the odds of experiencing violence in the previous six months, stratified by gender. As shown, in unadjusted analyses, exclusive SIF use was associated with decreased odds of past-six-month exposure to violence among men (OR = 0.68; 95% confidence interval [95% CI]: 0.51 – 0.92, p = 0.014). Exclusive SIF use was not significantly associated with past-six-month exposure to violence among women in unadjusted analyses (OR = 0.82; 95% CI: 0.53 – 1.29, p = 0.398).

Table 3.

Bivariable and multivariable generalized estimating equation analyses of factors associated with experiencing violence, stratified by gender, among 1930 people who inject drugs in Vancouver, Canada (2005-2016).

	Women (n = 679)		Men (n = 1251)
Characteristic	Unadjusted OR (95% CI)	Adjusted OR (95% CI)	Unadjusted OR (95% CI)	Adjusted OR (95% CI)
Exclusive supervised injection facility use‡ (yes vs. no)	0.82 (0.53 – 1.29)	0.97 (0.57 – 1.66)	0.68 (0.51 – 0.92)*†	0.64 (0.46 – 0.89)*
Age (per year older)	0.98 (0.96 – 0.99)*†	1.01 (1.00 – 1.03)	0.95 (0.94 – 0.96)*†	0.97 (0.96 – 0.98)*
Ancestry (white vs. non-white)	1.17 (0.93 – 1.47)		1.18 (0.98 – 1.42)†
Downtown Eastside residence‡ (yes vs. no)	1.41 (1.14 – 1.75)*†	1.10 (0.86 – 1.41)	1.50 (1.30 – 1.73)*†	1.27 (1.10 – 1.47)*
Employment‡ (yes vs. no)	1.13 (0.88 – 1.44)		0.88 (0.77 – 1.00)*†
In a relationship‡ (yes vs. no)	1.06 (0.89 – 1.25)		1.15 (0.99 – 1.33)†
Homeless‡ (yes vs. no)	1.98 (1.67 – 2.36)*†	1.73 (1.40 – 2.13)*	1.87 (1.66 – 2.10)*†	1.29 (1.14 – 1.48)*
HIV seropositive (yes vs. no)	0.63 (0.50 – 0.79)*†	0.65 (0.50 – 0.85)*	0.76 (0.64 – 0.89)*†
Heroin injection‡ (≥ daily vs. <daily)	1.49 (1.25 – 1.77)*†	1.14 (0.93 – 1.41)	1.26 (1.09 – 1.45)*†
Cocaine injection‡ (≥ daily vs. <daily)	1.20 (0.96 – 1.51)		0.91 (0.75 – 1.11)
Crystal methamphetamine injection‡ (≥ daily vs. <daily)	1.45 (1.07 – 1.97)*†	1.47 (1.03 – 2.11)*	1.56 (1.28 – 1.92)*†
Crack cocaine use‡ (≥ daily vs. <daily)	1.48 (1.25 – 1.74)*†	1.10 (0.88 – 1.39)	1.61 (1.40 – 1.85)*†
Binge injection‡ (yes vs. no)	1.31 (1.11 – 1.54)*†	1.31 (1.09 – 1.59)*	1.07 (0.95 – 1.19)
Require help injecting‡ (yes vs. no)	2.01 (1.68 – 2.40)*†	1.72 (1.41 – 2.10)*	1.49 (1.28 – 1.73)*†
Heavy alcohol use‡ (yes vs. no)	1.29 (1.03 – 1.62)*†	1.21 (0.93 – 1.57)	1.59 (1.33 – 1.90)*†	1.66 (1.38 – 1.99)*
Enrolment in MMT‡ (yes vs. no)	0.69 (0.57 – 0.82)*†	0.86 (0.70 – 1.07)	0.87 (0.76 – 0.99)*†
Sex work involvement‡ (yes vs. no)	1.62 (1.36 – 1.93)*†	1.34 (1.09 – 1.63)*	1.65 (1.14 – 2.39)*†
Drug dealing‡ (yes vs. no)	1.58 (1.34 – 1.86)*†	1.26 (1.03 – 1.54)*	2.32 (2.05 – 2.62)*†	1.86 (1.63 – 2.12)*
Incarceration‡ (yes vs. no)	1.76 (1.38 – 2.25)*†	1.34 (1.00 – 1.78)*	2.28 (1.95 – 2.66)*†	1.57 (1.33 – 1.85)*
Childhood emotional abuse (yes vs. no)	1.99 (1.55 – 2.59)*†	2.07 (1.59 – 2.70)*	1.51 (1.28 – 1.80)*†
Calendar year of interview (per year increase)	0.94 (0.92 – 0.96)*†	0.96 (0.93 – 0.99)*	0.90 (0.89 – 0.92)*†	0.95 (0.92 – 0.97)*

Note: OR = odds ratio; CI = confidence interval.

^‡Refers to the 6-month period prior to the interview.

^*p < 0.05.

^†Considered for inclusion in multivariable model because p < 0.10 in bivariable generalized estimating equation analyses.

In the final multivariable model, exclusive SIF use remained independently associated with reduced odds of past-six-month exposure to violence among men after adjustment for age, DTES residency, homelessness, heavy alcohol use, drug dealing, incarceration and calendar year of interview (adjusted odds ratio [AOR] = 0.64; 95% CI: 0.46 – 0.89, p = 0.009). Among women, exclusive SIF use was not significantly associated with past-six-month exposure to violence after adjustment for age, DTES residency, homelessness, relationship status, HIV seropositivity, ≥daily heroin injection, ≥daily crystal methamphetamine injection, ≥daily crack cocaine use, enrolment in MMT, requiring help injecting, heavy alcohol use, sex work involvement, drug dealing, incarceration, childhood emotional abuse and calendar year of interview (AOR = 0.97; 95% CI: 0.57 – 1.66, p = 0.914).

Table 4 presents the characteristics of the first-reported incidents of violence among the 922 participants who experienced violence at least once between December 2005 and December 2014, stratified by gender. As shown, strangers and acquaintances were the most commonly reported perpetrators of violence among both men and women. Women were significantly more likely than men to report that perpetrators were current or former partners (17.4% vs. 4.6%), sex partners (2.4% vs. 0.2%), sex work clients (5.6% vs. 0.2%), and sex workers (2.4% vs. 0.2%). Men were more likely than women to report that perpetrators were strangers (45.4% vs. 33.0%), police officers (15.1% vs. 6.6%), and drug dealers (11.7% vs. 6.3%) (all p < 0.05). The most commonly reported types of violence experienced by both men and women were beatings and assaults with weapons. Men were more likely than women to report having been assaulted with weapons (29.2% vs. 17.4%), while women were more likely to report having been sexually assaulted (8.3% vs. 1.0%), and strangled (3.1% vs. 0.8%) (all p < 0.05).

Table 4.

Characteristics of first-reported violent incidents experienced by 922 people who inject drugs in Vancouver, Canada, between 2005 and 2014, stratified by gender.

	Number of participants (%)*†
Characteristic	Women (n = 288)	Men (n = 634)	p - value
Perpetrator of violence
Stranger	95 (33.0)	288 (45.4)	<0.001
Acquaintance	60 (20.8)	127 (20.0)	0.779
Police	19 (6.6)	96 (15.1)	<0.001
Current or former partner‡	50 (17.4)	29 (4.6)	<0.001
Drug dealer	18 (6.3)	74 (11.7)	0.011
Sex partner£	7 (2.4)	1 (0.2)	<0.001
Sex worker	7 (2.4)	1 (0.2)	<0.001
Sex work client	16 (5.6)	1 (0.2)	<0.001
Friend	20 (6.9)	34 (5.4)	0.343
Family	5 (1.7)	5 (0.8)	0.198
Unknown	8 (2.8)	17 (2.7)	0.933
Other	15 (5.2)	44 (6.9)	0.319
Type of violence
Beating	219 (76.0)	491 (77.4)	0.639
Robbery	26 (9.0)	79 (12.5)	0.128
Attacked with gun	2 (0.7)	5 (0.8)	0.879
Attacked with other weapons	50 (17.4)	185 (29.2)	<0.001
Sexual assault	24 (8.3)	6 (1.0)	<0.001
Strangled	9 (3.1)	5 (0.8)	0.007
Other	20 (6.9)	27 (4.3)	0.086

^*Refers to total number of study participants at their first reported incidents of received violence between December 2005 and December 2014.

^†Total percentages may exceed 100 as participants could select multiple response options.

^‡Refers to the pre-defined response categories of ‘husband/wife’, ‘boyfriend/girlfriend’ and ‘partner’ as well as ‘other (specify)’ responses referring to a current or former partner.

^£Refers to the pre-defined response categories of ‘regular sex partner’ or ‘casual sex partner.’

DISCUSSION

We found that exposure to violence was common among a community-recruited cohort of 1930 PWID in Vancouver, Canada, with 52% of women and 56% of men reporting having experienced at least one incident of physical or sexual violence over a median follow-up duration of four years. In longitudinal multivariable analyses adjusted for a range of sociodemographic, drug use, and social-structural factors, exclusive use of the Insite SIF for injections was independently associated with decreased odds of experiencing violence among men. However, we did not observe a significant association between exclusive SIF use and exposure to violence among women in bivariable or multivariable analyses. Strangers and acquaintances were the most commonly reported perpetrators of violence among both women and men. However, women were more likely than men to experience violence from intimate partners and sex workers and clients, while men were more likely than women to experience violence from strangers, police officers and drug dealers. Beatings and attacks with weapons were the most common forms of violence reported by both men and women. However, attacks with weapons were more prevalent among men than women, while sexual assault and strangulation were more prevalent among women than men.

The present study is the first, to our knowledge, to quantitatively examine the association between SIF use and received violence among a community-recruited prospective cohort of PWID. Our finding that exclusive SIF use was protective against exposure to violence among men but not women is perhaps understandable given that we also found that men were more likely than women to experience violence from strangers and police officers, with more than 60% of violent incidents among men involving such perpetrators compared to approximately 40% among women. As predatory and police violence often occur when PWID are preparing and consuming drugs in public settings within the local street-based drug scene (Boyd et al., 2018a; Fairbairn et al., 2008; McNeil et al., 2014b; Small et al., 2007), the provision of an alternative, off-street, monitored environment for the consumption of drugs likely provided men with protection from such violent encounters. In light of the recent expansion of SIFs in Vancouver and elsewhere in Canada (Health Canada, 2020), these findings are encouraging in highlighting how such facilities may reshape the social and environmental contexts of injection drug use to mitigate risk of violence among men PWID. As such, these findings build on past studies demonstrating the various health and social benefits of SIFs (Kennedy et al., 2017b; McNeil & Small, 2014; Potier et al., 2014) and provide further evidence to support the inclusion of these facilities within the continuum of services for this population.

Although past qualitative research undertaken in this setting has described how SIFs may offer women protection from violence around the time of injection (Boyd et al., 2018a; Fairbairn et al., 2008; McNeil et al., 2014b), we found that exclusive SIF use was not significantly associated with reduced exposure to violence overall among women. The underlying explanations for this finding cannot be determined with certainty based on the analyses presented herein. However, our measurement of our primary exposure as exclusive SIF use for injections (as opposed to a lower level of service use) allows us to exclude the possibility that this finding is simply explained by the confounding influence of SIFs tending to engage PWID who are more likely to inject drugs in public or other settings in which they are susceptible to violence (Bravo et al. 2009; Kennedy et al. 2019; Kimber et al., 2003; Wood et al., 2005b; Wood et al., 2006c). Instead, this finding is likely largely explained by gender dynamics of violence among PWID. Specifically, violence associated with the consumption of drugs may account for a relatively smaller proportion of violent encounters among women compared to men given the pervasiveness of violence in other contexts of their lives (Boyd et al., 2018a; Lorvick et al., 2014; Marshall et al., 2008). Indeed, although strangers and acquaintances were the most common perpetrators of violence among both men and women in the present study, past research has found that women often experience violence from such individuals in situations when they are not preparing or injecting drugs (Boyd et al., 2018b; Epele, 2002; Maher & Hudson, 2007; McNeil et al., 2014a). For example, studies have documented how gendered power relations operating within street-based drug scenes contribute to the relegation of women to low-level or marginal roles in illicit drug economies that increase their vulnerability to robbery and assault (Boyd et al., 2018b; Maher & Hudson, 2007; McNeil et al., 2014a). Other studies have described how women may be subjected to violence and exploitation by men, particularly men occupying higher positions within street-based drug scenes, because women are perceived as having more opportunities than men to obtain resources through activities such as sex work (Fairbairn et al., 2008; Maher & Hudson, 2007; McNeil et al., 2014a). It is also noteworthy that, in the present study, women were more likely than men to experience violence from sex workers and clients, and sex work involvement was positively associated with exposure to violence among women, as has been documented in previous studies (Lorvick et al., 2014; Marshall et al., 2008; Shannon et al., 2008). Further, consistent with existing studies of general populations (World Health Organization, 2014), current and former partners were much more likely to be perpetrators of violence among women than men in the current study, and past research of PWID has demonstrated that violent encounters with these individuals often occur outside of injection-related contexts (Braitstein et al., 2003; Fairbairn et al., 2008; McNeil et al., 2014a). As SIFs specifically aim to offer protection against violence during drug preparation and consumption, such forms of violence are likely beyond the scope of violence that SIFs would be expected to address.

Our finding of a lack of a protective association between exclusive SIF use and exposure to violence among women might also be partially explained by gender relations operating within SIF services. Specifically, consistent with the gender breakdown of the local PWID population, men account for a much larger proportion of Insite clients than women (Wood et al. 2006c; Kennedy et al. 2019). Moreover, past research has described how existing social patterns of violence towards women by men may extend into mixed-gender SIF service settings (Boyd et al. 2018a). Such dynamics may function to structure SIFs as male-dominated environments in which women commonly experience harassment from men (Boyd et al., 2018a). Although SIF staff have been found to intervene to stop harassment from escalating to violence (Boyd et al., 2018a), it is possible that violent encounters could subsequently ensue once women exit this service (Fairbairn et al., 2008). Further, such risks could be exacerbated by the fact that the Insite SIF is located in a geographical area in the DTES characterized by particularly high levels of violence (Boyd et al., 2018a). Indeed, previous studies have found that women often avoid injecting at Insite due to the threat of violence in the surrounding neighbourhood of the facility (McNeil et al., 2014a; Shannon et al., 2008). Thus, women who inject exclusively at Insite might have an elevated risk of violence due to their greater exposure time within this specific area of the local drug scene, thereby possibly offsetting potential safety benefits stemming from injecting within the SIF.

Our findings underscore the need for social-structural interventions and supports that are more responsive to the specific needs of women PWID in relation to violence prevention. First, given that women have been found to experience barriers in accessing local mixed-gender SIFs and harassment from men within these facilities (Boyd et al., 2018a; McNeil et al., 2014a), the implementation of women-only SIFs may help to foster service engagement and better support the health and safety of women PWID (Boyd et al., 2018a; Schäeffer et al., 2014). The integration of anti-violence programming within these services may further extend the impact of such programming in mitigating violence, including violence that occurs outside of injection-related contexts. The recent establishment of a women-only SIF in Vancouver, which also offers violence prevention resources and counselling, provides opportunities for further investigation of these issues (Boyd et al., 2018a). Additionally, future research should seek to examine if the recent scale-up of SIFs in the city has helped to increase service utilization and reduce instances of violence and other harms among PWID, including among women who avoid the area surrounding Insite due to the potential for violence.

Our findings also corroborate with past research indicating that requiring help with injecting increases the risk of exposure to violence (Marshall et al., 2008; McNeil et al., 2014b; Small et al., 2012b), and thus reinforce the need for policy reforms that allow SIFs to accommodate assisted injection in effort to enhance access and reduce harms among this highly-vulnerable subpopulation of PWID that is disproportionately comprised of women (Boyd et al., 2018a; Gagnon, 2017; Kolla et al., 2020; McNeil et al., 2014a). Until recently, federal regulations had prohibited assisted injection within all federally-sanctioned SIFs in Canada, despite evidence demonstrating that accommodating this practice within regulated settings may help to mitigate the risk of violence and other serious harms among people who require help injecting (Boyd et al., 2018a; Gagnon, 2017; Kolla et al., 2020; McNeil et al., 2014b). However, since July 2018, Health Canada has approved twenty federally-sanctioned SIFs across the country to allow peer-assisted injecting on a trial basis (Health Canada, 2020; Lupick, 2018), and thus the continued evaluation of such practices may provide critical evidence to inform the potential scale-up and optimization of assisted injection services within SIFs (Boyd et al., 2018a).

Finally, as SIFs alone are unlikely to provide sufficient protection from violence among women, interventions that seek to alter social-structural determinants of risk in the broader risk environment of women PWID are needed to better promote their safety and well-being (Lorvick et al., 2014; Marshall et al., 2008; McNeil & Small, 2014). In particular, given that homelessness and sex work involvement were associated with an increased likelihood of experiencing violence among women in the present study and previous studies (Lorvick et al., 2014; Marshall et al., 2008; McNeil et al., 2014a; Shannon et al., 2008), efforts to increase access to social-structural supports, including social housing and regulated sex work environments, should be included as part of broader strategies to minimize exposure to violence among women PWID in this setting (Marshall et al., 2008; McNeil & Small, 2014). Additionally, these findings support calls for legal reforms, including decriminalization of sex work, to enhance occupational health and safety among sex working populations (Platt et al., 2018).

This study has several limitations. First, our findings are based on a non-random sample of PWID and thus may not be generalizable to PWID in Vancouver or elsewhere. Second, our reliance on self-reported data may have resulted in reporting biases, including socially desirable reporting. Third, the study design limited our ability to determine a temporal relationship between explanatory variables and the outcome. As such, we cannot exclude the possibility that the observed association between SIF use and reduced exposure to violence among men is partially explained by those who have experienced violence being subsequently less likely to use the SIF. For example, given that Insite is located in an area characterized by high levels of violence (McNeil et al., 2014a), some men who have experienced violence near the facility might avoid the area of the SIF after experiencing such violent encounters. Fourth, although we assessed a range of potential confounders, our findings could be influenced by selection effects and residual confounding.

In summary, we found that more than half of women and men in a cohort of PWID experienced violence over a median of four years of follow-up. Exclusive SIF use was independently associated with decreased odds of exposure to violence among men, but was not significantly associated with this outcome among women. These findings suggest that the recent scale-up of SIFs in this setting may afford opportunities to reduce exposure to violence among men. However, further efforts are needed to tailor SIFs to the needs of women and to increase access to other social-structural supports in order to promote more comprehensive violence prevention among women PWID in this setting.