Combining chain link sampling with a community-based participatory action study of people who smoke drugs in two cities in British Columbia, Canada

Background: The smoking of illicit drugs presents a serious social and economic burden in Canada. People who smoke drugs (PWSD) are at an increased risk of contracting multiple infections through risky drug practices. Peer-led harm reduction activities, and the resulting social networks that form around them, can potentially minimize the dangers associated with the smoking illicit drugs. Goal: The goal of this study is to examine the inuence of peer-led harm reduction initiatives on drug behaviour by comparing the attributes and social networks of PWSD in two British Columbia cities with different harm reduction programs. Methods: Using community-based participatory action research (CBPAR) and respondent- driven sampling (RDS), individuals with lived drug experiences were employed from communities in Abbotsford and Vancouver as ‘peer researchers’ to interview ten contacts from their social networks. Contacts completed a questionnaire about their harm reduction networks. Results: We found that PWSD residing in Abbotsford were more likely to report engaging in harm-promoting behaviours, such as sharing, reusing, or borrowing crack pipes. However, PWSD in the Downtown Eastside Community (DTES) of Vancouver were more likely to report engaging in harm-reducing activities, such as being trained in naloxone use and CPR. These differences are likely linked to the greater availability of harm reduction programs in DTES compared to Abbotsford.


Introduction
Illicit drug use has been the focus of much concern in recent years. In Canada, almost one million individuals over the age of 15 reported using at least one type of illegal substance within the previous year (Health Canada, 2018) excluding cannabis. Sustained and chronic use of drugs has been linked to multiple adverse health events resulting in death and disability.
In Canada, several harm reduction approaches have been adopted to address the dangers surrounding the consumption of drugs. In contrast to traditional drug cessation programs, harm reduction programs aim to mitigate harmful drug-related behaviours without requiring people who use drugs (PWUID) to abstain from drugs (Beirness et al., 2008). Effective programs often employ peers (individuals with lived drug experiences) to design, implement, and deliver activities (Broadhead et al., 1998).
Harm reduction measures can have different forms. The two most popular are the needle exchange programs (NEP) and supervised consumption facilities (SCF) (Beirness et al., 2008). While these programs have been found to alleviate the harms associated with intravenous drug use, less attention and fewer resources have been allocated to people who smoke drugs (PWSD) (Boyd et. al., 2008;Bungay et al., 2010;Haydon and Fischer, 2005). PWSD who engage in unsafe drug practices are at increased risk of contracting bloodborne and pulmonary infections (Edlin et al., 1994). DeBeck et al. found that the sharing and reuse of drug paraphernalia is a signi cant mechanism by which pathogens are transmitted between smokers. Speci cally, HIV and other bloodborne pathogens are conveyed to and from pipes via oral wounds and sores (DeBeck et al., 2009). In Vancouver, Canada, a survey of PWSD found that approximately half of surveyed participants reported sharing crack pipes within the previous six months (Ti et al., 2012).
The Downtown Eastside Community (DTES) of Vancouver has a large population of the drug users (Culbert & McMartin, 2015) and the largest proportion of the city's drug arrests (Linden, Mar, Werker, Jang, & Krausz, 2013). Consequently, it has been the focus of multiple harm reduction programs. Vancouver's Downtown East Side (DTES) community has functioned as a focal point for Canadian drug research for many years and has received a great deal of resources. The total cost of the services and programs available to this fteen-square-block community has been estimated at more than $1M per day. In 2013, 260 social and non-pro t agencies were operating in the community, totalling $360M per year (Culbert & McMartin, 2015). At the forefront of these operations is the Vancouver Area Network of Drug Users (VANDU), an organization comprised of individuals with lived drug experiences who elect board to represent them. Peer-led activities offered by VANDU include supervised drug consumption facilities, and needles and crack pipes distribution (Kerr et al., 2006).
An initiative similar to VANDU has been undertaken in rural Abbotsford, British Columbia. However, in contrast to the DTES, peer-led harm reduction activities in Abbotsford are informal and limited in scope.
Abbotsford has fewer resources available to manage the city's drug use and its related harms.
In this study, we wish to establish the feasibility of combining community based participatory research with respondent driven sampling, and compare the experience of peer-driven harm reduction services for PWSDs in the rural Abbotsford and urban Vancouver communities. We hypothesize that the differences in the concentrations of harm reduction programs in the two cities will be re ected in the level of harm promoting and reducing behaviours of PWSDs. In addition, networks of PWUIDs have been shown to play a signi cant role in in uencing risky drug behaviours (De, Cox, Boivin, Platt, & Jolly, 2007), which we hope to elucidate. Through the mapping of social networks, we will determine whether there is an association between the availability of peer-led harm reduction programs and the size of the networks of PWSD.
Results of this study will allow us to generate a pro le of the individuals residing in these communities, their harm reduction needs, and potential gaps in existing community programs.

Methods
This study was done in collaboration with the University of British Columbia and the drug using communities of Abbotsford and the DTES in Vancouver, British Columbia. Researchers used a community-based participatory action research (CBPAR) approach that engages the community in the development, implementation, and dissemination of research.
Participants were recruited using respondent-driven sampling (RDS) (Heckathorn, 1997). Eight and seven peer researchers, de ned as community members with lived drug experiences, were recruited from Abbotsford and Vancouver, respectively. Using RDS, each peer recruited ten current or former PWSDs from their established networks, who then each provided information on ten contacts and the relationships between them. A detailed description of the study methodology can be found in a published protocol . CBPAR is challenging in it itself, and in this study, participants not only guided the direction of the research, but also underwent ethics and informed consent training, crafted and administered the questionnaire, entered data and commented on the draft papers. Additionally, network questionnaires can be very time consuming as the numbers of network members nominated multiples the number of questions asked about each one, and respondents curtail their responses accordingly. In most studies in Canada, respondents are asked to nominate three friends who may then contact study personnel and if they consent, are recruited. Here, each peer researcher interviewed 10 friends (alters) and asked them questions about 10 of their friends, resulting in many questions and long interviews.
After consultations with two community groups above, the research proposal was approved, along with the questionnaire, and oral consent. Study stuff passed out advertisements for peer researchers were in Abbotsford, and posters were distributed by community members in Vancouver, both of which emphasised hiring community members. Eight and seven people who smoke drugs with lived drug experiences were recruited from Abbotsford and Vancouver, respectively, and paid $20.00 per hour, similar to research assistants at the university. These peer researchers were asked to recruit 10 "contacts or friends" in their networks, who used illegal drugs mainly through smoking; were 19 years of age or older, and to whom they felt comfortable administering the questionnaire . Besides providing information on themselves, each of the 10 friends then provided proxy information on ten of their "friends or contacts" and the relationships between them, for a total of about 1,500 people. Recruitment ended at the rst wave of contacts or friends, as this was a pilot to demonstrate feasibility of using CBPAR and RDS rather than to provide generalisable results. Each peer researcher completed 10 hours of training in applied ethics. Ethics approval for this research protocol and consent form was obtained from the University of British Columbia Certi cate H16-01580 and from the University of Ottawa H-05-18-741.
The questionnaire which each peer administered to each of their 10 friends was divided into two parts.
The rst included information from each friend, on housing and place of residence; age, gender, ethnic origins, drug smoking behaviours, including sharing equipment and frequency and type of drug smoked, medical conditions, overdosing; experience of violence, frequency of health care provider visits, mental health, injuries to the mouth as a result of smoking drugs, and drug smoking equipment. The second part of the questionnaire asked each of the 10 friends to list 10 of their friends whom they knew smoked drugs. For each one, questions were asked on; demographic and housing information, length of time that they have known the person, medical conditions, how close they were to the person, and smoking and injecting behaviours. Next, they were asked to select a from a list of roles or actions which one they considered the most valuable. For example, their friend had "taught me to x my pipe or dope", or had "administered naloxone when I overdosed". Last, they were asked to ll in a grid of which friends knew each other, and how close they were, close, somewhat close, not very close.
Because participants were permitted to use aliases in lieu of legal names, egocentric network sizes were determined using a hierarchical cross-network matching algorithm. The rst set of 10 friends of a peer researcher was added to the list of additional 10 their friends reported on and locations and sexes were compared. The social networks of peer research participants were joined together using a hierarchical cross-network matching mechanism. Successful matches were then matched on at least three of the following variables: age (within ten-year range), drug(s) of choice, current drug use status, and years of drug use (within a ve-year range) (Bouchard, Hashimi, Tsai, Lampkin, & Jozaghi, 2018) and were considered to be the same individual. Physical and mental health status, and routes of drug administration were used to verify matches and to resolve discrepancies. Numbers of people within a network has been reported to in uence injection equipment sharing behaviour, such that those in larger networks shared with more people than those in smaller networks (De et al., 2007). We analysed network size in order to determine whether this holds true for PWSD.
All variables with missing data exceeding 10%, were divided into two categories. We de ned missing data as incomplete, unclear, or 'don't know' responses, with the latter added due to the extremely low number of participants selecting this option. The association between missing data and the remaining variables was tested using Chi-squares or Fisher's exact tests. Bivariate analyses comparing participants in Abbotsford and the DTES were conducted using the same two tests with pairwise deletion for missing values.
Statistical analysis was conducted using SAS Software Version 9.4 and network analyses were conducted on UCINET 5.1. Networks were visualized using Organizational Risk Analyzer.

Results
Eight and seven peer researchers in Abbotsford and Vancouver recruited 79 and 70 friends and contacts, (alters) who reported on 739 and 498 friends, respectively (Figs. 1 and 2) The training and retainment of peer researchers has already been described; 8 of 10 completed 10 hours of ethics training from Abbottsford, and 6 of 7 from DTES . Although one participant in Abbotsford did not complete the questionnaire, all other peer researchers recruited 10 participants each for a total of 149 alters; one of the highest recruitment ever rates reported; 100% has been reported previously, but with a much lower sample size of 28 (Liao & Lai, 2017). Because there was substantial demographic and behavioural data missing, we compared the records with responses to those without in order to clarify possible biases (Table 1).
As above in our study aims, we compared to Abbotsford and DTES participants, and a greater proportion of participants from DTES were male (61% vs. 37%); self-identi ed as First Nations (58% vs. 28%); and reported living in supported housing (50% vs. 13%). In Abbotsford, more participants reported living with friends and family (16 % vs. 3%), (Table 2). There was no statistically signi cant difference in the proportion of self-reported medical conditions between participants in Abbotsford and DTES.

Discussion
The training and retention of peer researchers was laudable, and the success of community peer involvement is re ected in the number of participants recruited and the completion of questions about each of 10 additional alters. Table 1 shows the clusters of questions which people tended not to answer, many of which were concerning numbers of events. For example, answers to three consecutive questions about frequency of violence experienced by PWSD, from police, dealers and other PWSD decreased substantially from the rst to the third question. These are simple corrections which can be made in future research. To our knowledge this is the rst study combining CBPAR and RDS in people who use drugs, and we believe that the high response rate of 149 participants is a re ection of the value placed on personal relationships within the community of PWSD. The high number of participants referred by initial peer researchers has demonstrated beyond a doubt that the usual three recruits will be feasible, (Aglipay, Wylie, & Jolly, 2015) resulting in the optimal several waves of recruiters and respondents (Abdesselam K, Verdery, Pelude, Momoli, & Jolly, 2019).
The disparity in allocated resources is re ected in the socio-demographic and behavioural attributes of the participants and their listed contacts. Participants in Abbotsford were more likely to report engaging in harmful drug behaviours such as sharing, lending, or borrowing pipes and smoking in public areas, and less likely to report harm reducing activities, such as training on use of naloxone and CPR, carrying naloxone, and acquiring pipes from outreach organizations.
Fifty percent of DTES participants reported residing in a supported living environment, including single room occupancy (SRO) hotels and aboriginal housing. Participants in Abbotsford, however, reported a statistically higher percentage living with friends or families. Likewise, contacts of Abbotsford participants were more likely to be homeless or living in private housing while contacts of DTES participants were more likely to live in a supported living environment. This is consistent with previous research on the housing trends of drug users in DTES, where single occupancy housing was highly accessible. A 2015 survey of SRO hotels in DTES revealed a vacancy rate of only 4% among the 4,379 and 9,645 private and non-pro t SRO units in the community (City of Vancouver, 2015). Similarly, Shannon et al. reported that 70% of recruited DTES residents reported residing in SRO hotels and aboriginal housing (Shannon, Ishida, Lai, & Tyndall, 2006).
The comparatively fewer harm reduction services in Abbotsford is likely a factor in the greater prevalence of participants' contacts with HCV and mental illness. There was a signi cantly greater percentage of contacts (7.51%) in Abbotsford who reported being HCV positive compared to DTES contacts. The higher rates of infection are consistent with higher rates of reported pipe sharing and oral blisters, cuts, or sores among participants in Abbotsford. All of these are known to facilitate the transmission of bloodborne infections. The reasons for the difference in HCV infection rate, which was signi cant only among contacts and not participants, may be due to the small sample size, under-reporting of events due to social desirability bias, or unknown serostatus.
Over 77% of participants and 80% of contacts in Abbotsford reported consuming methamphetamine, whereas DTES participants and their contacts indicated crack as their drug of choice. Abbotsford participants were more likely to be female and Caucasian, consistent with previous studies where, relative to cocaine users, users of methamphetamine are more likely to be female and Caucasian (Rawson et al., 2000;Simon et al., 2001). The greater use of methamphetamine may be a consequence of Abbotsford's proximity to the United States border. Additionally, a greater number of participants in Abbotsford were female because recruitment posters were distributed in Warm Zone-Women's Resource Society, whereas VANDU caters to both female and male users of illicit drugs.
Although participants in Abbotsford had larger mean network sizes, this was statistically insigni cant. This suggests that urbanization has no in uence on the network size of people who smoke illicit drugs. Previous research on network structure has identi ed multiple individual-level factors that are associated with network size, including age, gender, and education level (Hill & Dunbar, 2003). Research into the network composition of urban versus rural social networks found urban residents to generally be socially isolated and rural residents to be socially connected and highly involved (Beggs, Haines, & Hurlbert, 1996;Fisher, 1982). However, Hooghe and Botterman found that among residents in Belgium, the population density and size of a region had no relationship to the quantity and degree of social association between residents (Hooghe and Botterman, 2012). Furthermore, users of illicit drugs are more likely to belong to a low socioeconomic class, and poverty has been associated with an increased sense of cohesion that may negate the in uence of urbanization (Belle, 1983). In addition, it is likely that because drug use is considered undesirable in North American society, all PWSD become part of networks of similar size and density, and marginalised from main stream society.
We used respondent-driven sampling (RDS) in partnership with people who smoke street drugs to better reach people who are usually marginalised by health care workers and researchers. This method assumes that a representative sample of the population is obtained following approximately six waves of chain-referrals, at which equilibrium is achieved (Heckathorn, 1997). However, this study consisted of a single wave of chain-referrals through initial key informants, to demonstrate feasibility of the communitybased participatory approach in conjunction with respondent driven sampling. Hence the study population cannot be assumed to be representative of the DTES and Abbotsford drug smoking community.
Nevertheless, the accuracy of participants' recollection of their contacts' behavioural and demographic characteristics can be assumed to be fairly accurate. Romney and Weller demonstrated that individuals who frequently interact with each other are a reliable source of information (Romney and Weller, 1984). Barrera and Arnold reported a high correlation (r=0.88) between test and retest reporting of social network members (Arnold, 1994;Barrera, 1980). Hammer observed a recall rate of 79% for contacts seen more than once a week (Hammer, 1984). Participants in Sudman's 39-person study were able to recall 92% of close contacts (Sudman, 1988), and Brewer found that injecting drug users remembered 78% of their drug using partners (Brewer et al., 2002). However, because responses were self-reported, they may have been subject to the social desirability bias.

Conclusion
There are key differences in the demographic and behavioural traits between PWSD in Abbotsford and DTES, which indicate the greater emphasis on harm reduction in DTES relative to Abbotsford. Recruited participants in DTES were found to engage in more harm reducing behaviours than Abbotsford participants, where harm reduction initiatives are limited. We recommend that closer attention be paid to Abbotsford and other rural regions across Canada that have traditionally suffered from lack of effective harm reduction programs.

Declarations
Declarations of interest; None Ethical approvals; included under Methods; end of paragraph 2 Funding; This research did not receive any speci c grant from funding agencies in the public, commercial or not-for-pro t sectors. Dr. Jozaghi's salary was provided by a post-doctoral fellowship from the Canadian Institutes of Health Research.

Consent for publication; All authors give consent for publication
Availability of data; The data may be available under request to Dr. Ehsan Jozaghi.  Figure 1 Social network of eight Abbotsford participants, 79 recruits, and their 739 friends. Dots represent individuals and lines between them relationships, including recruitment referrals into the study.

Figure 2
Social network of seven DTES participants, 70 recruits, and their 498 friends. Dots represent individuals and lines between them relationships, including recruitment referrals into the study.