The MIX cohort constitutes the largest Australian community-based PWID cohort to-date, and differs from other Australian PWID cohorts in several important ways.
Firstly, our cohort is recruited from the community, and includes a large sample of out-of-treatment PWID; just over one third of our participants were prescribed OST at recruitment, compared with 51%-63% of street-based PWID and NSP-attendees interviewed in recent Victorian drug trend monitoring studies [48–50]. As such, it does not possess the selection effects associated with recruitment from a particular place, such as treatment facilities. Although PWID who regularly attend primary care centres or pharmacies to obtain pharmacotherapy treatment may be easier to retain in longitudinal studies, PWID in-treatment tend to be different to those out-of-treatment, commonly being older and further progressed in their injecting careers [34, 40]. At the time of recruitment, the heroin market in Melbourne had been relatively depressed for some time [38, 51], and research suggests that this reduction in heroin supply was associated with both reduced heroin injection among current injectors and reduced initiation into injecting [42, 52]. This decreased the pool of newer, out-of-treatment PWID, preventing us from recruiting as large a sample of these users as hoped. Despite this, our cohort will still provide vital information about transitions into and out of drug treatment and the factors which motivate these decisions. Further, the inclusion of individuals both in and out of treatment will allow for assessments of a range of barriers to treatment as well as evaluations of the impact of treatment.
Participants in our cohort were recruited from three locations across Melbourne, where illicit drug markets and/or NSPs are located, with significant differences in socio-demographic and drug use patterns detected across sites. The Inner-West and Central areas are historically working-class and industrial; today, they include large public housing estates, and are home to significant Asian migrant populations, and more recently, refugee populations from the Horn of Africa [53–55]. Following a transition from predominantly private dealing, street-based heroin markets emerged in these areas in the mid-1990s and continue to remain active despite ongoing policing [38, 53, 56]. In contrast, the outer-urban recruitment site is home to a predominantly Anglo-Australian community, with manufacturing and construction the main industries [57, 58]; MIX study participants from this site displayed a preference for amphetamine and pharmaceutical opiate injection, presumably reflecting limited access to heroin due to geographic distance from active heroin markets. Differences in patterns of alcohol consumption were also recorded across research sites and may reflect a number of factors including neighbourhood liquor outlet density  and differing cultural attitudes towards alcohol consumption. The role of the geographical environment in drug use and associated risks and harms warrants further investigation, and will be examined in future.
Rather than focusing specifically on BBV incidence or drug treatment outcomes – the main focus of previous cohorts of Australian PWID [34–37] - our study collects data on a broad range of other health outcomes, including patterns of drug injection and injecting cessation, physical and mental health, and engagement with health services. Of particular interest is the fact that although 58% of participants reported attending a GP clinic in the past month, just 17% reported recent attendance at one of the five state-funded free PWID-specific PHC clinics, despite these clinics generally being located reasonably close to participants’ residences. Further analysis is required to explore the characteristics of clients attending these services and their presenting complaints, and to understand the ways in which patterns of health service utilisation are associated with factors such as recruitment site, service availability and patterns of drug use. The use of prospective data will also enable examination of longer-term drug use and other health outcomes among PWID attending these services.
While we used a combination of RDS, traditional snowballing and street outreach to ensure that a diverse sample of PWID were included in the study, there were few significant differences across recruitment arms. While not the focus of this paper, further analysis, including the calculation of RDS-weighted population prevalence estimates, will facilitate better understanding of the usefulness of this recruitment strategy.
Despite having worked in these field sites for a number of years [60, 61], and conducting formative research prior to study commencement (field-based observations and pilot interviews), the Melbourne drug market is dynamic, and unanticipated changes in both people accessing the market, and availability of different drug types did occur [62, 63]. In response, a number of changes to the eligibility criteria of the study, as well as study procedures were implemented.
Firstly, we relaxed our age restriction on eligibility, which resulted in the inclusion of 95 participants aged 30–31, and 38 participants aged over 31 in the study. As such our sample is slightly older than initially hoped, with a median age of 27.6 years, making them slightly younger than participants in the Victorian cohort recruited by Crofts et al. in the early 1990s , but older than cohorts recruited in Sydney and Melbourne in the mid-2000s [35, 36]. It has been noted that PWID in this jurisdiction are an ageing population; repeat cross-sectional surveys have indicated that the median age of NSP attendees in Victoria has increased significantly from 26 years in 1997 to 35 years in 2010 . Similar increases in mean ages have been observed among PWID survey participants in Victoria’s illicit drug trends monitoring system over the past ten years [48, 64]. This is likely to be due to the population of ageing PWID who initiated injecting in the 1980s and 1990s and continue to inject today, combined with decreasing numbers of young people initiating injection . The median year of injecting initiation among our sample, however, was 1999 (IQR: 1996–2003), with a median delay of one year to regular injecting drug use. Thus, while a proportion of participants initiated injecting during the latter years of the heroin ‘glut’ , there are few participants in our study for whom drug use was already entrenched during this period, with the majority commencing regular injecting in the setting of limited heroin availability.
Our study initially aimed to recruit both primary heroin and methamphetamine injectors, as most previous Australia cohorts have been comprised mostly of heroin injectors [36, 37, 65] and, despite reported recent increases in crystal methamphetamine use, relatively little was known about patterns of methamphetamine injection. By the time study recruitment commenced however, recent reports of crystal methamphetamine use had again decreased , meaning that only a small number of primary methamphetamine injectors met the study eligibility criteria. Nonetheless, prospective data collection will enable ongoing monitoring of trends in methamphetamine use, and provide opportunities to explore potential changes in drug use and health outcomes as participants transition between different patterns of primary heroin and methamphetamine use.
While other Australian PWID cohorts have been limited by short durations of follow-up, the MIX cohort will be followed up annually for a minimum of four years (with funding for further follow-up to be sought). At 12-months’ follow-up, the retention rate was 71%, comparable to similar international studies, which had follow-up rates from 68%-83% reported over durations ranging from three months to four years [27–30, 32]. Similar to other longitudinal studies of vulnerable populations, we found that the collection of detailed contact information at baseline, comprehensive follow-up procedures and an ongoing field presence that allowed researchers to build familiarity and trust with participants, were all integral in tracking respondents [67, 68]; participants who did not provide complete contact details at baseline were more likely to be lost to follow-up. Importantly, while attrition was associated with male gender, those lost to follow-up were otherwise similar to participants retained in the study, thus limiting the impact of attrition bias on our findings. The long duration of follow-up, combined with future data linkage through administrative data (e.g. the Medicare system) beyond the period of face-to-face follow-up will produce rich and versatile data enabling a better understanding of the natural history of injecting drug use and patterns of morbidity and mortality (overall, as well as among particular subgroups of PWID). These data will be integral to the evaluation of health and social interventions among this group.
Due to ethical considerations, we were not permitted to recruit participants younger than 18 years of age, however due to a miscommunication a small number of participants aged 16 and 17 were inadvertently recruited into the study; ethics approval has been obtained to use data from these participants. It remains unclear whether this population of adolescent PWID are being targeted effectively by research or health interventions.
Given the complexities involved with street-based recruitment across multiple field sites, involving a large research team, it was not possible to monitor how many PWID were invited but declined to participate in the study. Unwillingness to consent to the provision of Medicare information may have been associated with declining to participate in the study.
As with much PWID research, our data may be limited by selection bias, and as behavioural data were self-reported, also by recall and social acceptability bias. Future data linkage and BBV testing will enable us to assess the accuracy of some self-reported variables.