Recruiting the Hidden Population of Problem Drug Users in East Renfrewshire
Table of Contents
1 – Introduction
2 – Methods for Recruiting Hidden Populations
2.1 Ethnographic Studies
2.2 Snowball Sampling
2.3 Respondent Driven Sampling
2.4 Privileged Access Interviewing (PAI)
3 - Recruitment Method(s) in East Renfrewshire
3.1 Selecting a recruitment method
3.2 Recruiting Peer Researchers
3.3 Training Peer Researchers
4 – The Data Collection Process
4.1 Preparatory Work
4.2 Recruitment Methods in Practice
4.3 Implementing Alternative Methods to Data Collection
4.3.1 Teen Challenge Bus
4.3.2 Twelve Step Meetings
4.3.3 Targeting “Well Known” Service Users
4.3.4 Placing Advertising in Needle Exchanges and Local Press
5 – Concluding Comments on Recruitment Methods
5.1 Peer Researcher Recruitment
5.2 Safety and Ethics
5.3 Cultural Issues
References
1 – Introduction
The East Renfrewshire Hidden Population Needs Assessment was undertaken in order to investigate the prevalence rates of drug use in the area that suggested roughly only 30% of problem drug users were accessing treatment for their drug use. The aim of the Needs Assessment was to recruit and gather data from individuals who were experiencing problems with their drug use but were not currently engaged in treatment to explore the barriers to accessing treatment services and what improvements could be made.
An overview of the design and issues considered to undertake the Needs Assessment is outlined in the Research Protocoland the background and findings are outlined in the Main Report. However the purpose of this supplementary document is to provide an overview of the recruitment techniques available for the study of hidden or hard to reach populations and the rationale as to why specific methods were chosen to undertake the Needs Assessment. This leads to an in depth overview of how the recruitment techniques were implemented and alternative methods that were required in order to recruit a sample. It is anticipated that this detailed overview of methods may provide insight to others who wish to undertake work to recruit and sample hidden populations.
To provide a narrative to the methods involved in recruiting members of the hidden population, the report is written specifically in 1st person.
2 – Methods for Recruiting Hidden Populations
‘Hard to reach’ or ‘hidden populations’ are terms used to describe populations that may be difficult to reach due to physical locations, but more often are those who do not wish to be contacted, are socially invisible or may actively try to conceal their identity. Such populations may include the homeless, faith based communities, migrants and drug users (Wiebel, 1990). Given that these groups are difficult to locate and may exist in specific social worlds, traditional sampling techniques are not useful when attempting to recruit for research purposes. Despite this, due to their under researched nature, they may be the populations who would be at most benefit of research findings.
Given this, a range of alternative sampling techniques have been devised and utilised in order to successfully access and recruit members of hidden populations. These techniques have been reviewed in previous literature (Shaghaghi, Bhopal & Shiekh, 2011), however for the purposes of the Needs Assessment, recruitments methods specific to targeting problem drug users were required. The main techniques considered for the Needs Assessment are discussed below.
2.1 Ethnographic Studies
As Schensul (1999) highlights, work exploring hidden populations is not new and a specific method for exploring drug use and drug using cultures that has been utilised historically is ethnographic research. This method typically involves establishing links with population, gaining their trust and thus, as a researcher, be able to immerse into the day to day lives of the population and gain an understanding of their lives and culture (Creswell, 2006). This is typically done through establishing various contacts within the population in question and spending extensive periods of time observing the group through participant observation and through informal interviews. This approach allows the researcher to gain an in depth insight into the culture of the population, the language used and social processes that occur within the group (Creswell, 2006).
Ethnographic studies into drug using populations have been extremely insightful and have provided information on the gender differences associated with drug use, the heterogeneity of drug using populations and the health issues and concerns that arise in drug using populations (Carlson, Singer, Stephens & Sterk, 2009). Despite this, undertaking ethnographic research is extremely demanding and involves spending extended periods of time in the field collecting data as well as time developing and maintaining contacts (Cresswell, 2006).
Similarly as Moore (1993) discusses in relation to his own ethnographic research into young, recreational drug users in Australia problems can arise using an ethnographic method. For example in establishing contact with the group he gained the help of a young female known to him through a mutual friend. This contact provided him with a means of accessing the population of interest but when interacting with individuals, he was typically known as “this guy who wants to watch us use” (Moore, 1993: 14). This status thus rendered it difficult for the researcher to gain the trust of group members with regards to confidentiality and safety and thus may have affected the depth of the information gathered.
A final issue concerning ethnographic research is questions around morality, ethics and safety. For example as Moore (1993) highlights, a question that ethnographers immersed in the social worlds of drug users face is whether to consume illicit drugs during the course of the fieldwork. While this may be considered a personal choice of the researcher is may have implications for properly conducting research and more importantly compromising researcher safety and well-being. This is also true for the reporting of significant illegal activity or harm to others as while the researcher may wish to make personal judgements, they may also be adhering to ethical guidelines set out by their research funder (United Nations, 2004).
2.2 Snowball Sampling
While ethnographic research has provided unique and useful information to develop knowledge in relation to drug users, it is usually presented in a narrative and may not appeal to social scientists wishing to conduct more systematic research investigating a particular research question or hypothesis. As such other methods to collect information from hidden populations have been used.
One such method is snowball sampling which involves acquiring a sample through referrals from individuals who know others or who share the same characteristic of interest, who will then in turn refer others, thus developing a chain of recruits (Biernacki & Waldorf, 1981). It is considered to be a particularly good method when gaining information on a sensitive issue such as drug use that requires insiders to locate individuals of interest.
It has been a very popular technique in hidden population research of drug users (Biernacki & Waldorf, 1981); however it does have certain limitations. For example as the technique relies upon key informants recruiting others who are eligible to participate in the study, who then in turn recruit others, the sample gained will not be random, thus making it difficult to make any generalisation from findings to the whole population (Shaghaghi, Bhopal & Sheikh, 2011). Similarly as referrals made will likely be friends or associates, those who are isolated from social networks are likely to be under-represented in the sample.
In an attempt to address some of these issues, Eland-Goossensen et al (1997) used a form of snowball sampling with random nominee selection, in an attempt to generate a more representative sample of out of treatment individuals dependent on opiates. This involved having key informants form a zero stage sample that were from as diverse networks as possible. These informants were then required to create a network of individuals they knew that met the study criteria and recruitment was made by randomly selecting a member of this network and then repeating the process until no more nominations could be made or a person could not be contacted. It was anticipated that having a diverse initial sample coupled with randomly selecting nominated individuals would make the sample more representative.
While the study could not conclude that this method made the sample more representative, it did suggest that it reached a wider segment of the population of interest. However it also raised other issues in relation to sampling. For example it was found that informants often had difficulty describing or were not as forthcoming about members of their network as initially anticipated. Similarly when they did refer individuals, as they were approached for interview, they often refused. To rectify this, the individual who provided the referral was paid to approach their associate about participating in the research.
These are issues that other studies have raised. For example Goode (1999) highlights in her study of drug taking mothers that snowball sampling was ineffective as the women that had been recruited reported that they could not refer anyone as they were either isolated or knew each other due to meeting in specialised drug treatment agencies. Similarly Biernacki & Waldorf, (1981) state that starting off the process of referrals can be difficult and requires finding and building rapport with key network members and may not necessarily lead to large, uniform samples being generated.
Despite these limitations, snowball sampling has been widely used and has been successful in recruiting samples of out of treatment drug users (Robson & Bruce, 1997; Shewan, Dalgarno, Marshall, Lowe, Campbell, Nicholson, Reith, McLafferty & Thomson, 1998). This is perhaps through the careful planning of selecting initial participants to begin the referral process and also drawing upon other resources such as advertising and accessing multiple sites in order to initiate a referral process.
With respect to snowball sampling only providing a convenience sample, this is only problematic when the aim if the study is to produce findings that can be generalised to the whole population. When a study is more exploratory in nature, often the aim is not to get a representative sample but to gain an insight into the lives, opinions and experiences of the individuals being studied (Elliot, Watson & Harries, 2002). However this limitation should still be noted.
2.3 Respondent Driven Sampling
In an attempt to address the issues of bias that are evident in chain referral methods such as snowball sampling, a newer technique known as RDS was developed by Heckathorn (1997). The principles of this method are similar to snowball sampling, however through more rigorous implementation and analysis, the aim is to generate a representative sample whose characteristics are independent of the “seeds” who were the initial respondents asked to begin the referral process (Johnson & Sabin, 2010).
To implement RDS, the initial “seeds” are expected to have large social networks, be respected by members of the population of interest, have an ability to initiate recruitment and come from diverse networks. This is essential for RDS methodology so that the sample characteristics comprise a complete social network and form a stable composition (Johnston & Sabin, 2010). To allow recruitment to function, networks must already exist and the seeds are given a fixed number of coupons each to recruit members of their network. Once new seeds have been recruited, they, in turn are given a fixed number of coupons to recruit others, with the expectation that several waves of recruitment will occur to penetrate as many key sub-populations as possible. Both the recruits and those who were responsible for recruiting are provided with an incentive, which is usually monetary, to encourage both recruitment and participation in the study.
In addition to these criteria, in order to analyse recruitment from RDS, information must be collected on the social network size of each participant and the connections between the recruiters and the recruits. This data is required in order to weight the data to account for the possibility that some respondents in the study will have larger network sizes than others.
By adhering to this recruitment style and collecting all the necessary supplementary data for analysis, it is expected that RDS will provide a sample from which generalisations can be made to the population as a whole (Johnston & Sabin, 2010).
While this technique has been implemented successfully to recruit large numbers of drug users within a relatively short time (McKnight, Jarlais, Bramson, Tower, Abdul-Quader, Nemeth & Heckathorn, 2007), it could be argued that method, while useful in theory may not necessarily translate well in practice. For example, for RDS to make accurate inferences about the general population, participants are required to provide detailed information about their network. This is somewhat problematic as individuals may not be forthcoming about their network or may not be able to describe it in detail which would lead to the potential for error when generating estimates for the networked population (Platt, Wall, Rhodes, Judd, Hickman, Johnston, Renton, Brobrova & Sarang, 2006). Secondly, RDS relies heavily on monetary incentives, both for participation and as an incentive for participants to recruit others. However some studies have questioned the ethical practice of using monetary incentives, with the finding that underground networks used violence and coercion in the trading of study coupons. This often led to “participants” presenting at data collection sites who were not eligible for the study (Scott, 2008).
2.4 Privileged Access Interviewing (PAI)
Another sampling method that has been popular in hidden population research for problem drug users is PAI or peer researching. This method is based on the premise that those who are involved in drug use or have associates that are members of such networks will have “privileged access” to these groups. Not only will they have “inside knowledge” to be able to locate individuals within the populations of interest, they will also have an already established rapport. Thus, instead of a researcher having to gain the trust of those they are interested in researching, it is expected that a PAI will have already established this (Griffiths, Gossop, Powis & Strang, 1993).
Griffiths et al., (1993) defined the criteria for a PAI’s as being:
- They had existing contacts within the subculture
- They had personal attributes/experiences that made them non-threatening to the participants
- Were suitably equipped to understanding training and undertake an interview
- They had a stable enough lifestyle and conducting the interview/making contact with drug users would not be damaging to them
This method has been successful in the recruitment and collection of data from hidden populations of drug users. For example Griffiths et al. (1993) through careful implementation were able to recruit a large number of out of treatment drug users in a relatively short space of time. Similarly, Keubler and Hausser (1997) also found that, while time consuming to locate and train PAI’s the method was useful and they were able to access 508 hidden drug users and successfully administered a standardised questionnaire.
Despite this success, other research has highlighted the advantages and disadvantages of using this method. For example Power (1994) discusses the issue of ensuring that the data collected from PAI’s is of good quality and is as valid and reliable as possible. This typically involves monitoring and carrying out checks on interviews carried out by PAI’s. For questionnaires, this may involve including questions that ask the same question but in a different manner to identify consistency in the participants’ account. Similarly, a selection of participants may be selected to be re-interviewed, both as a way of clarifying information collected but also to alert PAI’s that the information they gather will be checked (Power, 1994).
Elliot et al. (2002) also encountered difficulties in ensuring data quality when using PAI’s. They recruited four PAI’s to conduct qualitative interviews with drug using parents, however they found that three of the PAI’s did not record the interview despite receiving training on using recording equipment. They stated that this was because the participants they had recruited refused to be recorded. It has been suggested however that a successful way of overcoming this problem early in the data collection stage is for PAI’s to say the main reason for recording the interview is to ensure that the interview was carried out correctly (Griffiths et al., 1993).
The failure to record the interviews also led to another issue of using PAI’s, which was the researchers feeling that they had little connection to the data and had to rely on second hand accounts of the interviews. However, while this clearly is an issue, Elliot et al. (2002) point out that if it were not for the PAI’s, they may not have got any access to the participants of interest in the first place. Similarly they also note that in many circumstances, due to the shared experiences PAI’s had with the participants they used a common language that would have been somewhat alien to researchers, allowing a more free flowing discussion to ensue.