Sage Journals: Discover world-class research

Abstract

Ethnographic-epidemiological (“ethno-epi”) research methodologies are increasingly being used to examine health-related issues, including the experiences of people who use drugs. However, the complementary application of random sampling from a well characterised cohort and qualitative data collection methods in a single study has not been described. We address this gap by sharing insights from the implementation of a novel random stratified sampling technique to recruit participants from two large prospective observational studies of people who use drugs into a qualitative study about impacts of the COVID-19 pandemic on their lived experience. We aim to describe how an ethno-epi approach we used can enhance the validity, reliability and generalizability of research findings in mixed methods investigations. We do so by providing a step-by-step description of the process we used to determine participant eligibility and recruitment into the qualitative study. Although the approach is not without limitations, findings underscore how ethno-epi random sampling approaches can increase the credibility and trustworthiness of qualitative findings without compromising data depth and integrity. Our study makes an important contribution to the growing number of new creative approaches being developed in the mixed methods research field and we hope that by sharing our account it will encourage and support others to consider the use of ethno-epi approaches in health-related research.

Keywords

ethno-epi mixed methods drug use research

Background

Ethnographic-epidemiological (“ethno-epi”) research approaches capitalize on the strengths of ethnographic/qualitative research methods and research strategies used in epidemiology. Almeida Filho (2001, 2020), one of the first scholars to describe and promote the integration of ethnography and epidemiology in health research, describes ethno-epi research as a “transdisciplinary approach for health research objects and methods” and an “alternative for research on social processes and practices related to health, able to competently combine qualitative and quantitative approaches” (Almeida Filho, 2020). According to Almeida Filho (2020), the term ethno-epi first appeared in the title of a 1992 research report on infant mortality among Hmong ethnic minority children in Thailand (Kunstadter et al., 1992).

In the 1990s, when mixed methods research approaches were gaining popularity (Johnson et al., 2007), scholars began using the term ethno-epi to advocate for the interdisciplinary integration and application of anthropological/ethnographic and epidemiological research (Agar, 1994; Almei da Filho, 2001, 2020; Inhorn, 1995). For example, Inhorn (1995), an early advocate of ethno-epi approaches argued that:

… epidemiology, like medical anthropology, operates largely within its own sphere—one that is somewhat peripheral to the domain of biomedicine. In this respect, epidemiology and medical anthropology share markedly similar structural positions vis a vis biomedicine—a commonality that should serve to unite rather than divide them. (p. 286)

In the last two decades, ethno-epi approaches have been used to understand a broad spectrum of health-related issues. Most scholars have described doing so in the context of blending quantitative and ethnographic/qualitative data for analysis; using an epidemiological lens to inform ethnographic research; or using ethnography to guide epidemiological research or make sense of or “unpack” findings derived from epidemiological research. Littleton et al. (2008) for example, refer to their ethno-epi approach as one that used qualitative interview data to contextualise epidemiological data of the everyday life experiences of tuberculosis among migrants to New Zealand. Lee et al. (2004) on the other hand described their study as ethno-epi, by using insights from ethnography to inform epidemiological data collection tools to investigate the sociocultural risk factors of postnatal depression. Furthermore, Thompson and Gifford’s study (2000) about the meanings of health and diabetes in an Australian Aboriginal community, was defined as ethno-epi because their ethnographic fieldwork was conducted through the lens of epidemiological understandings.

Interest in the use of ethno-epi approaches in drug-related research has been present for at least two decades. Early US researchers (Clatts, 2001; Clatts et al., 2002; Pach Iii & Gorman, 2002) described using broadly defined ethno-epi methods that involved blending data as a tool for integrating ethnographic/qualitative and secondary sources of quantitative data (e.g. police and media reports, drug treatment data, drug monitoring systems) to identify and track emergent drug use trends over time, including for people who use methamphetamine (Clatts, 2001; Pach Iii & Gorman, 2002) and crack cocaine (Clatts et al., 2002). Furthermore, Moore et al. (2009) positioned their ethno-epi technique as one that merged agent-based modelling simulations with ethnographic fieldwork to increase understandings of psychostimulant use and related harms among young Australians. In relation to their study, they noted:

… some of the limitations of the two forms of data collection are minimized (e.g., ethnography’s specificity and the limited depth of quantitative instruments), [while] some of their advantages are reinforced (e.g., ethnography’s richness and the use of large samples in epidemiology) and their interactive potential is maximized (p. 1).

More recently, Canadian scholars have described using ethno-epi methods to purposively recruit people who inject drugs from prospective observational studies to qualitatively understand experiences of HIV treatment (Fleming et al., 2021; McNeil et al., 2017; Small et al., 2016), injecting cessation among marginalised youth (Boyd et al., 2017) and hospital discharges (McNeil et al., 2014). Ethno-epi techniques used in these studies provided avenues for exploring and making sense of the associations identified in epidemiological studies (Small et al., 2016) and, at a practical level, allowing for efficient recruitment of participants with select characteristics of interest. Mayock et al. (2015), whose Irish study examined the risk environment of heroin use, described the approach as one that incorporated qualitative data collection techniques with the development of epidemiological descriptions of the types of environments in which young people first used heroin. As presented in the glossary of their manuscript, they describe an ethno-epi approach as:

… an emergent cross-disciplinary research methodology that combines the strengths of ethnographic observation and other qualitative methods for understanding social meanings and contexts as practised in anthropology with the design, sampling, data collection, and analytical strategies developed in epidemiology (Mayock et al., 2015, p. 780).

As the above examples illustrate, a growing number of scholars are using ethno-epi methods to examine health-related issues. Despite significant variation in how ethno-epi methods have been defined and used, missing from the literature are mixed methods studies that use quantitative random or probability sampling techniques to recruit participants into qualitative research. While there are many examples of qualitative research being used as a complement to randomized trials (e.g., Dixon et al., 2016; Maher et al., 2019; Page et al., 2019; Scantlebury et al., 2021), to our knowledge the complementary application of quantitative random sampling and qualitative data collection methods in a single study has not been described. We address this gap in the literature by sharing insights from the implementation of a stratified random sampling technique to recruit participants from two large prospective observational studies of people who use drugs into a qualitative study aiming to understand impacts of the pandemic on their lived experience (see Walker et al., 2023). By combining the use of qualitative research methods with observational cohort studies our work builds on the studies considered above, allowing for the efficient recruitment of participants with characteristics of interest, while minimizing recruitment biases in selecting these participants. We aim to describe how our novel ethno-epi approach can enhance the internal generalizability, rigour and trustworthiness of qualitative findings in mixed methods studies. In the following, we provide a broad overview of the study—COvid-19 Drug use Experiences (CODE)—to provide context for the mixed methods approach used. Next, we present a step-by-step description of our ethno-epi sampling process, followed by a discussion of the implications of its use, including benefits and limitations.

CoDE Study Overview

A growing number of studies have examined the unintended consequences of the COVID-19 pandemic (herein described as “the pandemic”) on the lives of people who use drugs. Although qualitative studies examining the impacts of the pandemic on people who use drugs have emerged globally (Ali et al., 2021; May et al., 2022; Otiashvili et al., 2022; Roe et al., 2021; Russell et al., 2021), fewer have focused on Australia (Brener et al., 2022; Coleman et al., 2022; Conway et al., 2023; Efunnuga et al., 2022). Furthermore, these studies are based on non-representative purposive or snowball sampling techniques. In contrast, our qualitative study randomly sampled participants with characteristics of interest determined by interrogation of the quantitative databases of two well characterized cohorts of people use drugs; the Melbourne Injecting Drug User Cohort Study (SuperMIX) (Van Den Boom et al., 2022) and the Understanding Methamphetamine Use in Victoria Study (VMAX) (Quinn et al., 2021), with the aim of recruiting a more representative sample to increase the rigour and trustworthiness of our findings.

SuperMIX (n = 1303 at the time of data collection) was established in 2008 and is focused on understanding how the health and socio-economic outcomes and drug use patterns of people who inject drugs evolve over time (Van Den Boom et al., 2022). Eligibility includes being at least 18 years of age, injecting at least monthly for six months prior to baseline interview, and residing in urban Melbourne, Victoria. VMAX (n = 853 at the time of data collection) commenced in 2016 and examines long-term patterns of methamphetamine use, and how these impact on service use and health and wellbeing outcomes over time (Quinn et al., 2021). Eligibility includes being over 18 years of age; residing in metropolitan Melbourne or one of three recruitment locations in rural Victoria; and at least monthly use of methamphetamine primarily via non-injecting routes of administration (e.g., smoking, snorting) in the last six months. The non-injecting criterion was relaxed to reach the target sample size as recruitment slowed. Both studies involve annual participation in follow-up surveys via phone call or in-person.

We added pandemic-related questions to the SuperMIX and VMAX cohort surveys at the beginning of the pandemic (March 2020); this epidemiological self-report survey data were used to select and recruit participants into the qualitative study. In-depth semi-structured interviews were conducted with 76 participants of the SuperMIX and VMAX studies. The aim of the study was to understand impacts of the pandemic on drug use; housing, employment, and income status; social relationships and supports; access to and use of health, drug treatment and harm reduction services; interactions with law enforcement; and acceptability of government public health restrictions and vaccines. Ethics study approval was received from the Alfred Health Ethics Committee in 2020 (#258/21).

Ethno-epi Sampling Approach

In the following we describe the ten steps used to determine inclusion eligibility for participation in the qualitative study, followed by the process of randomization, selection, and recruitment.

Step 1

Determining eligibility for participation in our random sampling process began by identifying SuperMIX and VMAX study participants who had completed at least one survey questionnaire since the pandemic-related questions were added in March 2020. Those who reported no drug use since their last follow-up interview prior to March 2020 were excluded. A total of 766 potentially eligible participants were identified (SuperMIX, n = 350; VMAX, n = 416).

Step 2

We created twelve ‘flags of interest’ representing epidemiological variables that corresponded to pandemic-related survey questions in the SuperMIX and VMAX studies (Table 1) that matched areas of inquiry in the qualitative study. Nine flags of interest were common across SuperMIX and VMAX studies: (1) employment and housing status; (2) drug use behaviour; (3) increased drug use; (4) purchasing drugs; (5) alcohol use and smoking; (6) overdose; (7) health; (8) COVID testing and isolation; and (9) police interactions. Three additional flags of interest were created for SuperMIX participants, to capture impacts specific to injecting drug use: (10) injecting behaviour; (11) opiate agonist maintenance treatment (OAMT); and (12) injecting health. (Table 1)

Table 1.

Examples of Pandemic-Related Questions used to Generate “Flags of Interest”.

	Flags of interest	SuperMIX/VMAX survey domains indicating changes in life circumstances due to the pandemic
1	Employment and housing	• Employment status and government benefits • Moved into homeless accommodation
2	Drug use behaviour	• Wanted to use [name of drug] but unable to e.g.,heroin, crystal methamphetamine, methadone, buprenorphine, morphine oxycodone, speed powder, pharmaceutical stimulants, cocaine ecstasy, cannabis, unisom
3	Increased drug use	• Use of [*name of drug] increased
4	Purchasing drugs	• Wanted to purchase [*name of drug] but unable to
5	Alcohol use and smoking	• Use of (⊕standard) drinks containing alcohol • Regularity of tobacco cigarette use
6	Overdose	• Accidental heroin overdose
7	Health	• Physical health • Mental health
8	COVID testing/isolation	• COVID-19 tests • COVID-19 self-isolation
9	Police interactions	• Interactions with police • Arrested for breaking pandemic rules
10	Injecting behaviour	• Regularity of injecting • Injecting cessation
11	Opioid Agonist Treatment (OAMT)	• OAMT treatment • Methadone take-away doses
12	Injecting health	• Sharing or re-use of injecting equipment • Access to syringe vending machine

Step 3

With the aim of maximising the potential for information-rich cases, eligibility criteria for participation in the qualitative study were further refined to include only those who indicated significant impacts (i.e., reported the most impacts across the twelve flags of interest), as indicated by their responses to pandemic-related survey questions within each flag. For SuperMIX participants (n = 350), eligibility included those who reported impacts across at least four of the 12 flags of interest (n = 122). For VMAX participants (n = 416), those who indicated impacts across at least two of the nine flags of interest were eligible (n = 169); a decision that was driven by the larger number of flags of interest for SuperMIX (n = 12) compared to VMAX (n = 9). (Figure 1)

Figure 1.

Flow chart of participants in randomized sampling process (n = 291).

Step 4

A total of 291 participants were eligible to participate in the qualitative study via the process described above; this became our sample frame for randomization. Six stratified lists of participants were generated to ensure participant age, sex and geographical location characteristics were homogenous in each stratum. For SuperMIX we created two lists of participants based on sex: females (n = 46), and males (n = 76). For VMAX, four participant lists were generated based on sex and recruitment location: female rural (n = 33), female metro (n = 34), male rural (n = 51), and male metro (n = 51).

Step 5

With the aim of minimising selection bias, participant identifiers in each of the six strata were assigned a random order using a random number generator in Excel. This procedure resulted in the creation of six randomized lists, which were used for contacting individuals in the sequential order in which they appeared in each list.

Step 6

Each of the six randomized lists was imported into a password protected Excel spreadsheet. Participant contact details including mobile telephone numbers, email addresses and Facebook handles were obtained from SuperMIX and VMAX contact databases, which are updated by researchers after survey interviews are completed or when researchers opportunistically meet participants in the field. Contact details were merged into the six randomized spreadsheets for the first ten participants in each of the randomized lists (n = 60), as a starting point for the recruitment process.

Step 7

Up to three attempts were made to contact the first ten participants in each of the six randomized lists. Once contact was made with a participant, brief information about the study was provided, followed by an invitation to participate in an interview via telephone call, virtual video call (facetime, zoom, WhatsApp, Facebook messenger), or in person (only if pandemic restrictions were not in place). If a participant indicated interest, a day and time was set up for the interview – up to two additional scheduled appointments were allowed if the participant did not attend the scheduled interview.

Step 8

Excel spreadsheets were used to provide an audit trail of the recruitment process, which included keeping detailed notes of all attempts to contact eligible participants. Information gathered included: the number of times individuals were contacted and when; if their phones were disconnected or no other up-to-date contact details were available; if they declined to participate; or if they did not attend their scheduled interview. Most individuals were contacted via telephone if a number was available (four out of five randomized participants with whom attempts to contact were made, had mobile telephone numbers) followed by two text messages if the number was disconnected. Seven were contacted via email or Facebook.

Step 9

Once attempts to contact the first ten participants in each of the six randomized lists were exhausted the next ten participants from the lists were contacted. The process was repeated until the target number of participants for theoretical saturation was reached (n = 76).

Step 10

After the first twenty interviews were complete, the recruitment process was reviewed by the research team, which involved researchers with extensive experience in both quantitative and qualitative research methods and research with people who use drugs. No changes were made to the process, and the remaining participants were contacted using the same sequential approach described above.

Recruitment and Participation Summary

In summary, of the 766 SuperMIX (n = 350) and VMAX (n = 416) study participants who had completed at least one follow-up quantitative survey since the pandemic-related questions were added (and had used drugs during this period), 291 were selected as eligible for randomization based on reported impacts across flags of interest. Six stratified lists of participants were created, with a simple process applied to randomize the participant list for each stratum. Of this sample, a total of 268 attempts were made to contact 176 individuals (SuperMIX n = 77; VMAX n = 99) in the order that they appeared in the randomized lists. Of the 176 individuals who were contacted, 100 did not participate in an interview for the following reasons (Table 2): phone disconnected with no other form of contact (n = 35); did not respond via telephone, text message, Facebook, or email (n = 39); did not attend scheduled interviews despite at least two attempts to reschedule (n = 5); and declined invitation to participate (n = 4). One hundred and fifteen eligible participants were not contacted at all, because they appeared lower down in the randomized lists and the target number for data saturation was achieved before they were eligible to be contacted.

Table 2.

Reasons for Non-participation for Eligible Participants.

Reason for non-participation	No. Of participants
Chose not to participate	4
Expressed interest but did not attend scheduled interview	5
Telephone disconnected	35
Did not respond to telephone calls, email, text or Facebook messages	56
Did not appear in lists until data saturation achieved, i.e., not contacted	115
Total	215

Seventy-six individuals participated in in-depth semi-structured interviews between August-October 2021 (during the final lockdown period in Victoria), thirty were conducted between December 2021 and January 2022, and 26 interviews were completed between March to April 2022. Most interviews were conducted via telephone (n = 56) or video call (n = 15). Five interviews were conducted in-person at the beginning and end of the recruitment period when no pandemic restrictions were in place. Interviews varied in duration from 15 to 80 minutes, with an average of 40 minutes.

We conducted a chi-squared test for association (Table 3) to assess if there were any statistically significant differences in socio-demographic characteristics and drug use behaviours between those who were eligible and participated in an interview (n = 76), and those who were eligible and did not (n = 215). The latter (n = 215) included randomly selected participants who were unable to be contacted, declined to participate, or did not attend scheduled interviews, and participants who did not appear in our randomized lists before data saturation was achieved. Our analysis found that the distribution of impacts across each of the flags of interest were almost the same for individuals who were interviewed and those who were not, with only a non-significant trend toward significant statistical difference (based on p-values) for one variable for SuperMIX participants and no differences evident for VMAX participants. SuperMIX participants who were interviewed were marginally more likely to be non-Indigenous (interviewed [97.06%] vs. not interviewed [80.77%], p = .076).

Table 3.

SuperMIX/VMAX Statistical Comparison of Interviewed and Non-Interviewed.

	Participant Characteristics	Interviewed	Not Interviewed	Pearson's Chisquared Test	p Value
SuperMIX participants	Median age	31 years	31 years
	Non-Indigenous	97.06% (n = 33)	80.77% (n = 63)	$χ^{2}$ (df) = 5.151 (1)	.076
	Born in Australia	91.18% (n = 31)	89.74% (n = 70)	$χ^{2}$ (df) = .055 (1)	.815
	Completed > year 10	97.06% (n = 33)	97.44% (n = 76)	$χ^{2}$ (df) = .013 (1)	.910
	Ever incarcerated	67.65% (n = 23)	66.67% (n = 52)	$χ^{2}$ (df) = .010 (1)	.919
	Unemployed	88.89% (n = 32)	94.74% (n = 36)	$χ^{2}$ (df) = .848 (1)	.356
	Unstable housing	8.82% (n = 3)	7.69% (n = 6)	$χ^{2}$ (df) = .041 (1)	.840
	Prescribed OAT	70.59% (n = 24)	62.82% (n = 49)	$χ^{2}$ (df) = .629 (1)	.428
	Ever overdosed	73.53% (n = 25)	57.69% (n = 45)	$χ^{2}$ (df) = 2.534 (1)	.111
VMAX participants	Median age	36 years	34 years
	Non-Indigenous	91.43% (n = 32	92.54% (n = 124)	$χ^{2}$ (df) = 0.048 (1)	.826
	Born in Australia	97.14% (n = 34)	88.89% (n = 120)	$χ^{2}$ (df) = 2.220 (1)	.136
	Completed > year 10	10 77.14% (n = 27)	79.26% (n = 107)	$χ^{2}$ (df) = 0.075 (1)	.784
	Ever incarcerated	37.14% (n = 13)	28.89% (n = 39)	$χ^{2}$ (df) = 1.324 (1)	.515
	Unemployed	74.29% (n = 26)	72.18% (n = 96)	$χ^{2}$ (df) = 0.062 (1)	.804
	Unstable housing	14.71% (n = 5)	17.42% (n = 23)	$χ^{2}$ (df) = 0.142 (1)	.706
	Smoking most common route of methamphetamine administration	76.47% (n = 26)	79.70% (n = 106)	$χ^{2}$ (df) = 0.170 (1)	.680
	Usually used other drugs with methamphetamine	82.35% (n = 28)	86.36% (n = 114)	$χ^{2}$ (df) = 0.352 (1)	.553

Discussion

While the use of ethno-epi approaches in drug-related research has been growing, most published studies simply involve the integration of quantitative epidemiological data with qualitative data collected via in-depth interviews. Missing in the literature is a description of the use of probability sampling techniques to aid in the selection of participants for a qualitative study. We have addressed this gap by describing the application of a novel ethno-epi random sampling approach to recruit participants from two large prospective studies to a qualitative study to explore the impact of the pandemic on people who use drugs. In doing so, we have highlighted how this mixed method approach can improve the internal generalizability, credibility and rigour of qualitative research nested in prospective observational studies.

One of the important features that distinguishes qualitative from quantitative inquiry is the sampling strategy, with qualitative research typically involving theoretical or purposive sampling, which rests on the proposition that the selection of information-rich samples will provide/yield an in-depth view of the phenomena (Coyne, 1997; Suri, 2011). By contrast, many observational epidemiological studies or randomized trials tend to rely on random (or probability) sampling techniques with the aim of obtaining a representative sample for the purposes of statistical generalization across population groups (Sandelowski, 1997). The approach we used allowed us to capitalize on the strengths and contributions of both quantitative and qualitative research techniques, which helped increase internal generalizability across the two eligible cohorts we recruited from without compromising the depth or richness of the data collected (see Walker et al., 2023). We believe this approach helped improve the rigor and trustworthiness of our study. Furthermore, our process of identifying ‘flags of interest’ within the SuperMIX and VMAX datasets before randomization (to maximize the potential for information-rich samples) also provided a more efficient recruitment method than would have been achieved by screening each individual participant for eligibility.

Generalizability has been an ongoing contentious issue among social scientists within the non-positivist paradigm (Maxwell & Chmiel, 2014). Indeed, for some, generalizability has been considered a topic that should be ignored, or at best avoided, with some rejecting generalization across samples as a legitimate goal for qualitative research (Morse et al., 2002; Sandelowski, 1993). Nevertheless, qualitative scholars have taken different positions on questions concerning generalizability and how it should be defined (Maxwell, 2021; Osbeck & Antczak, 2021; Roald et al., 2021). For example, Roald et al. (2021) make the claim that not only is generalizability a relevant and important topic for qualitative research but that it is inescapable and core to the practice of human science. Furthermore, Osbeck and Antczak (2021) argue “qualitative methods cannot avoid having implications that transcend the original context in which the inquiry was conducted” (p. 64).

It is not our contention that the ethno-epi approach we used means that our qualitative understandings extend to the entire population of people who use drugs in Victoria, Australia during the pandemic. Rather, our approach provided an opportunity to efficiently obtain a more representative sample from within the two cohorts than could have been achieved via the sampling methods typically used in qualitative research (e.g., purposive or theoretical sampling). Our approach extends the work of Maxwell (2021), who has posed a distinction between two different types of generalization: internal and external generalization. Maxwell defines external generalization as “generalization beyond the person(s), setting, case, or time specifically studied to other persons, settings, cases, or times” (p. 113). In contrast, he refers to internal generalization as “generalizing within the setting, group, or population studied to persons, events, and activities that are not directly represented in the data collected (p. 113). The ethno-epi approach described here, we argue, helped minimise selection bias, and thus enhanced the potential for internal generalizability to the two large underlying cohorts of people we recruited into our study, so that qualitative findings were more reflective of the larger cohort. This claim is further supported by our observation that, except for Indigenous status, there were no statistically significant differences between the randomly selected participants who were interviewed and those who were not.

Internal generalization, not surprisingly, has been posed as more challenging for qualitative researchers because sample sizes are often too small to allow meaningful statistical inferences (Maxwell, 2021)—traditionally recruiting only to the point of data saturation (Morse, 2002) and often involving less than 30 participants (Marshall et al., 2013). Nonetheless, growing numbers of qualitative studies are utilising large data sets because it can extend opportunities for depth and breadth in the analysis of findings (Blodgett et al., 2005; Brower et al., 2019; White et al., 2012). As Brower et al. (2019) argue:

… like binoculars that can focus near or far, large qualitative data sets enhance this function by allowing us to dial down to closely examine phenomena at the micro or individual level and then dial out to view phenomena at the macro or societal level (p. 5).

We believe our relatively large qualitative dataset (n = 76)—which was achieved by having access to two large prospective observational studies—increased the likelihood of achieving data saturation across the two stratified dimensions of sex (female vs. male) and geographical location (rural vs. metro). Moreover, as proposed by Maxwell (2021), an essential component for claiming internal generalization is “not just whether the results capture the typical views, characteristics, or actions of the persons involved but also the diversity of these within the case or population studied” (p. 114). Our simple process for selecting participants from each of the six stratified randomized lists, we argue, helped achieve this. At the same time, this technique allowed us to understand more fully the diversity of shared patterns of experiences, insights and views that differentiated participants from one another.

Transparency in sampling techniques used and decisions made along the way can increase the trustworthiness, validity and rigour of qualitative research results (Higginbottom, 2004; Tuval-Mashiach, 2017). For example, as Tuval-Mashiach (2017) has posed, “it could even be suggested that readers derive a greater benefit from seeing what has gone on ‘behind closed doors’ than from learning the study’s results” (p. 126), and that describing what was done, how and why, is critical for increasing the credibility and reliability of findings. By describing in detail, the process used to undertake our ethno-epi study, we have highlighted how quantitative sampling methods can be used to increase the dependability and robustness of qualitative data, without compromising its depth and nuance. Furthermore, we believe in doing so, our methodological descriptions provide important contributions to the literature on mixed methods research.

We acknowledge our approach is not without limitations. We understand the ethno-epi technique described here may have little relevance to the aims of some qualitative studies, particularly those where a large sample is deemed unnecessary and in-depth descriptions from a smaller number of participants may be all that is required.

In addition, we are cognisant that the technique described requires access to ongoing prospective cohort studies—a design which many regard as the gold standard for observational research (Black, 1996)—to ensure a large enough sample for randomization across strata. Although we had the benefit of being able to leverage our prospective observational cohorts with a total of more than 2000 participants, which provided sufficient numbers for random selection of characterised participants, we understand these studies are expensive to fund, and most qualitative research teams will not have access to such cohorts. We, therefore, encourage epidemiologists and others who conduct prospective observational studies keen to use this approach to explore opportunities for collaboration with qualitative researchers. We acknowledge, however, as Brower (2019) has highlighted, that requiring researchers from diverse methodological perspectives and, sometimes, different philosophical stances to work together can produce challenges if shared values and understandings of these diverse methodological paradigms are not communicated. Nevertheless, within our core research team—which comprised individuals with diverse methodological backgrounds (i.e., epidemiology, social science, biostatistics)—we found being able to draw on each other’s expertise also had the unintended consequence of enhancing the collective knowledge and skills of the team.

Although we make claims that our approach increased internal generalizability across the two large cohort studies we recruited from, we are cognizant that our sample for randomization were drawn from two non-randomized samples. Thus, despite our best efforts, it is possible that selection biases were already present in the cohort samples, therefore reducing the potential benefits of our randomized process for participant selection.

Finally, our inability to recruit all participants into the study in the order they appeared in the randomized lists reduced our capacity to achieve a pure probability sample - an issue that was precipitated by the ‘hard to reach’ nature of our population group (Bonevski et al., 2014). Furthermore, because the sampling approach involved contacting all participants in the randomized lists at least three times, much time was spent attempting to contact individuals who did not participate in the study. The social and economic circumstances of people who use drugs often present barriers to their participation in research (e.g., homelessness, mental health issues, fear of confrontation with legal authorities due the criminalised nature of drug use) (Marpsat & Razafindratsima, 2010; Shaghaghi et al., 2011). Ensuring these ‘hard to reach’ individuals have every possible opportunity to participate in research is necessary, however, for understanding their issues and needs. We argue, therefore, that although increased resources were required to accommodate the time taken to locate as many participants from the randomized lists as possible, it was a necessary component for enhancing the likelihood of achieving a more heterogonous and non-biased sample. Moreover, our sampling approach allowed for efficient recruitment of participants with the characteristics of interest to the study.

Conclusions

To our knowledge, this is the first published description of the application of an ethno-epi stratified random sampling approach to recruit participants into qualitative health research. We have highlighted how we were able to efficiently recruit a more internally representative sample into our study than would have been possible using purposive or snowball sampling approaches traditionally used in qualitative research. Through exploring and finding opportunities for embracing common ground, our collaborative cross-disciplinary approach, we believe, makes an important contribution to the growing number of new creative approaches being developed in the mixed methods research field. By sharing our account, we hope to encourage others conducting health-related studies to consider incorporating this approach as a method for gaining information-rich nuanced data, while also improving the internal generalizability, rigor and trustworthiness of the evidence gathered.

Footnotes

Acknowledgements

We would like to thank the SuperMIX and VMAX participants for the time and knowledge they contributed to the study. Thanks also to Burnet Institute and Monash University fieldworkers responsible for the recruitment and follow up of these cohorts.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the SuperMIX study is funded by the Colonial Foundation Trust and the National Health and Medical Research Council (#545891, #1126090). The VMAX study was established with a grant from the Colonial Foundation and is now funded by the National Health and Medical Research Council (#1148170). The Burnet Institute receives funding from the Victorian Government Operational Infrastructure Support Program.

Ethical Statement

ORCID iD

Walker S

References

Agar

(1994). Recasting the “ethno” in “epidemiology”. Medical Anthropology, 16(1-4), 391–403. https://doi.org/10.1080/01459740.1994.9966123

Ali

Russell

Nafeh

Rehm

LeBlanc

Elton-Marshall

(2021). Changes in substance supply and use characteristics among people who use drugs (PWUD) during the COVID-19 global pandemic: A national qualitative assessment in Canada. International Journal of Drug Policy, 93(103237). https://doi.org/10.1016/j.drugpo.2021.103237

Almeida Filho

(2001). For a general theory of health: Preliminary epistemological and anthropological notes. Cadernos de Saúde Pública, 17, 753–770. https://www.scielosp.org/article/csp/2001.v17n4/753-770/

Almeida Filho

(2020). Ethnoepidemiology and mental health: Insights from Latin America. Salud Colect, 16(e2786). https://doi.org/10.18294/SC.2020.2786

Black

(1996). Why we need observational studies to evaluate the effectiveness of health care. British Medical Journal, 312(7040), 1215–1218. https://doi.org/10.1136/bmj.312.7040.1215

Blodgett

L. J.

Boyer

Turk

(2005). No thank you, not today": Supporting ethical and professional relationships in large qualitative studies. Forum for Qualitative Social Research, 6(3), Article 3. https://doi.org/10.17169/fqs-6.3.31

Bonevski

Randell

Paul

Chapman

Twyman

Bryant

Hughes

(2014). Reaching the hard-to-reach: A systematic review of strategies for improving health and medical research with socially disadvantaged groups. BMC Medical Research Methodology, 14(1), 1–29. https://doi.org/10.1186/1471-2288-14-42

Boyd

Fast

Hobbins

McNeil

Small

(2017). Social-structural factors influencing periods of injection cessation among marginalized youth who inject drugs in Vancouver, Canada: An ethno-epidemiological study. Harm Reduction Journal, 14(1), 1–11. https://doi.org/10.1186/s12954-017-0159-9

Brener

Horwitz

Rance

Caruana

Bryant

(2022). Health worker perceptions of the impact of COVID 19 on harm reduction services for people who inject drugs. Health and Social Care in the Community, 30(6), 2320–2329. https://doi.org/10.1111/hsc.13782

10.

Brower

R. L.

Jones

T. B.

Osborne-Lampkin

L. T.

Park-Gaghan

T. J.

(2019). Big Qual: Defining and debating qualitative inquiry for large data sets. International Journal of Qualitative Methods, 18. https://doi.org/10.1177/1609406919880692

11.

Clatts

M. C.

(2001). Reconceptualizing the interaction of drug and sexual risk among MSM speed users: Notes toward an ethno-epidemiology. AIDS and Behavior, 5(2), 115–130. https://doi.org/10.1023/A:1011374710703

12.

Clatts

M. C.

Welle

D. L.

Goldsamt

L. A.

Lankenau

S. E.

(2002). An ethno-epidemiological model for the study of trends in illicit drug use: Reflections on the ‘emergence’of crack injection. International Journal of Drug Policy, 13(4), 285–295. https://doi.org/10.1016/S0955-3959(02)00123-8

13.

Coleman

Taran

Cuesta-Briand

(2022). Responding to rural adversity: aAqualitative study of alcohol and other drug service users’ experiences of service response to COVID-19 in Western Australia’s Southwest. Australas Psychiatry, 30(1), 74–78. https://doi.org/10.1177/10398562211036125

14.

Conway

Treloar

Crawford

Degenhardt

Dore

G. J.

Farrell

Marshall

A. D.

(2023). You’ll come in and dose even in a global pandemic”: A qualitative study of adaptive opioid agonist treatment provision during the COVID-19 pandemic. International Journal of Drug Policy, 114, 1–7. https://doi.org/10.1016/j.drugpo.2023.103998

15.

Coyne

I. T.

(1997). Sampling in qualitative research. Purposeful and theoretical sampling; merging or clear boundaries? Journal of Advanced Nursing, 26(3), 623–630. https://doi.org/10.1046/j.1365-2648.1997.t01-25-00999.x

16.

Dixon

Stein

Ngak

Chhim

Phou

Sokunny

Maher

(2016). Qualitative research and implementation science: Informing the acceptability and implementation of a trial of a conditional cash transfer intervention designed to reduce drug use and HIV risk. Methodological Innovations, 9, 1–10. https://doi.org/10.1177%2F2059799115622751

17.

Efunnuga

Higgs

Walker

O'Keefe

(2022). Health service utilisation and access for people who inject drugs during COVID-19. In Drug and alcohol review. https://doi.org/10.1111/dar.13456

18.

Fleming

Collins

A. B.

Bardwell

Fowler

Boyd

Milloy

McNeil

(2021). A qualitative investigation of HIV treatment dispensing models and impacts on adherence among people living with HIV who use drugs. PLoS One, 16(2), Article e0246999. https://doi.org/10.1371/journal.pone.0246999

19.

Higginbottom

G. M. A.

(2004). Sampling issues in qualitative research. Nurse Researcher, 12(1), 7. https://www.proquest.com/docview/200776754?pq-origsite=gscholar&fromopenview=true#

20.

Inhorn

M. C.

(1995). Medical anthropology and epidemiology: Divergences or convergences? Social Science & Medicine, 40(3), 285–290. https://doi.org/10.1016/0277-9536(94)E0029-R

21.

Johnson

R. B.

Onwuegbuzie

A. J.

Turner

L. A.

(2007). Toward a definition of mixed methods research. Journal of mixed methods research, 1(2), 112–133.

22.

Kunstadter

S. L.

Leepreecha

Podhisita

Laoyang

Thao

C. S.

Yang

W. S.

(1992). Causes and consequences of increase in child survival rates: Ethnoepidemiology among the Hmong of Thailand. Human Biology, 64(6), 821–841. https://www.jstor.org/stable/41464342

23.

Lee

D. T.

Yip

A. S.

Leung

T. Y.

Chung

T. K.

(2004). Ethnoepidemiology of postnatal depression: Prospective multivariate study of sociocultural risk factors in a Chinese population in Hong Kong. The British Journal of Psychiatry, 184(1), 34–40. https://doi.org/10.1192/bjp.184.1.34

24.

Littleton

Park

Thornley

Anderson

Lawrence

(2008). Migrants and tuberculosis: Analysing epidemiological data with ethnography. Australian & New Zealand Journal of Public Health, 32(2), 142–149. https://doi.org/10.1111/j.1753-6405.2008.00191.x

25.

Maher

Neale

(2019). Adding quality to quantity in randomized controlled trials of addiction prevention and treatment: A new framework to facilitate the integration of qualitative research. Addiction, 114(12), 2257–2266. https://doi.org/10.1111/add.14777

26.

Marpsat

Razafindratsima

(2010). Survey methods for hard-to-reach populations: Introduction to the special issue. Methodological Innovations Online, 5(2), 3–16. https://journals.sagepub.com/doi/pdf/10.4256/mio.2010.0014

27.

Marshall

Cardon

Poddar

Fontenot

(2013). Does sample size matter in qualitative research? A review of qualitative interviews in IS research. Journal of Computer Information Systems, 54(1), 11–22. https://doi.org/10.1080/08874417.2013.11645667

28.

Maxwell

J. A.

(2021). Why qualitative methods are necessary for generalization. Qualitative Psychology, 8(1), 111. https://psycnet.apa.org/doi/10.1037/qup0000173

29.

Maxwell

J. A.

Chmiel

(2014). Generalization in and from qualitative analysis. The SAGE Handbook of qualitative data analysis, 7(37), 540–553.

30.

May

Dawes

Fancourt

Burton

(2022). A qualitative study exploring the impact of the COVID-19 pandemic on people who inject drugs (PWID) and drug service provision in the UK: PWID and service provider perspectives. International Journal of Drug Policy, 106, 103752. https://doi.org/10.1016/j.drugpo.2022.103752

31.

Mayock

Cronly

Clatts

M. C.

(2015). The risk environment of heroin use initiation: Young women, intimate partners, and “drug relationships”. Substance Use & Misuse, 50(6), 771–782. https://doi.org/10.3109/10826084.2015.978629

32.

McNeil

Kerr

Coleman

Maher

Milloy

Small

(2017). Antiretroviral therapy interruption among HIV positive people who use drugs in a setting with a community-wide HIV treatment-as-prevention initiative. AIDS and Behavior, 21(2), 402–409. https://doi.org/10.1007/s10461-016-1470-2

33.

McNeil

Small

Wood

Kerr

(2014). Hospitals as a ‘risk environment’: An ethno-epidemiological study of voluntary and involuntary discharge from hospital against medical advice among people who inject drugs. Social Science & Medicine, 105, 59–66. https://doi.org/10.1016/j.socscimed.2014.01.010

34.

Moore

Dray

Green

Hudson

S. L.

Jenkinson

Siokou

Dietze

(2009). Extending drug ethno-epidemiology using agent-based modelling. Addiction, 104(12), 1991–1997. https://doi.org/10.1111/j.1360-0443.2009.02709.x

35.

Morse

J. M.

Barrett

Mayan

Olson

Spiers

(2002). Verification strategies for establishing reliability and validity in qualitative research. International Journal of Qualitative Methods, 1(2), 13–22. https://journals.sagepub.com/doi/pdf/10.1177/160940690200100202

36.

Osbeck

L. M.

Antczak

S. L.

(2021). Generalizability and qualitative research: A new look at an ongoing controversy. Qualitative Psychology, 8(1), 62. https://psycnet.apa.org/doi/10.1037/qup0000194

37.

Otiashvili

Mgebrishvili

Beselia

Vardanashvili

Dumchev

Kiriazova

Kirtadze

(2022). The impact of the COVID-19 pandemic on illicit drug supply, drug-related behaviour of people who use drugs and provision of drug related services in Georgia: Results of a mixed methods prospective cohort study. Harm Reduction Journal, 19(1), 1–15. https://link.springer.com/journal/12954

38.

Pach

Iii Gorman

E. M.

(2002). An ethno-epidemiological approach for the multi-site study of emerging drug abuse trends: The spread of methamphetamine in the United States of America. Bulletin on Narcotics, 54(1-2), 87–102. https://www.unodc.org/documents/archive/bulletin_2002_01_01_1.pdf#page=91

39.

Page

Carrico

Stein

Evans

Maly

Sokunny

Maher

(2019). Cluster randomized stepped-wedge trial of a multi-level HIV prevention intervention to decrease amphetamine-type stimulants and sexual risk in Cambodian female entertainment and sex workers. Drug and Alcohol Dependence, 196, 21–30. https://doi.org/10.1016/j.drugalcdep.2018.11.029

40.

Quinn

Ward

Agius

P. A.

Jenkinson

Hickman

Sutton

Dietze

(2021). A prospective cohort of people who use methamphetamine in Melbourne and non‐metropolitan Victoria, Australia: Baseline characteristics and correlates of methamphetamine dependence. Drug and Alcohol Review, 40(7), 1239–1248. https://doi.org/10.1111/dar.13194

41.

Roald

Køppe

Bechmann Jensen

Moeskjær Hansen

Levin

(2021). Why do we always generalize in qualitative research? Qualitative Psychology, 8(1), 69. https://psycnet.apa.org/doi/10.1037/qup0000138

42.

Roe

Proudfoot

Tay Wee Teck

Irvine

R. D.

Frankland

Baldacchino

A. M.

(2021). Isolation, solitude and social distancing for people who use drugs: An ethnographic perspective. Frontiers in Psychiatry, 11, 623032. https://doi.org/10.3389/fpsyt.2020.623032

43.

Russell

Ali

Nafeh

Rehm

LeBlanc

Elton-Marshall

(2021). Identifying the impacts of the COVID-19 pandemic on service access for people who use drugs (PWUD): A national qualitative study. Journal of Substance Abuse Treatment, 129, 108374. https://doi.org/10.1016/j.jsat.2021.108374

44.

Sandelowski

(1993). Rigor or rigor mortis: The problem of rigor in qualitative research revisited. Advances in Nursing Science, 16(2), 1–8.

45.

Sandelowski

(1997). To be of use”: Enhancing the utility of qualitative research. Nursing Outlook, 45(3), 125–132. https://doi.org/10.1016/S0029-6554(97)90043-9

46.

Scantlebury

McDaid

Brealey

Cook

Sharma

Ranganathan

And on behalf of the ACTIVE and PRESTO study teams . (2021). Embedding qualitative research in randomised controlled trials to improve recruitment: Findings from two recruitment optimisation studies of orthopaedic surgical trials. Trials, 22(461), 1–13. https://doi.org/10.1186/s13063-021-05420-4

47.

Shaghaghi

Bhopal

R. S.

Sheikh

(2011). Approaches to recruiting ‘hard-to-reach’ populations into research: A review of the literature. Health Promotion Perspectives, 1(2), 86. https://doi.org/10.5681%2Fhpp.2011.009

48.

Small

Milloy

McNeil

Maher

Kerr

(2016). Plasma HIV-1 RNA viral load rebound among people who inject drugs receiving antiretroviral therapy (ART) in a Canadian setting: An ethno-epidemiological study. AIDS Research and Therapy, 13(1), 1–12. https://doi.org/10.1186/s12981-016-0108-9

49.

Suri

(2011). Purposeful sampling in qualitative research synthesis. Qualitative Research Journal, 11(2), 63–75. https://doi.org/10.3316/QRJ1102063

50.

Thompson

S. J.

Gifford

S. M.

(2000). Trying to keep a balance: The meaning of health and diabetes in an urban Aboriginal community. Social Science & Medicine, 51(10), 1457–1472. https://doi.org/10.1016/S0277-9536(00)00046-0

51.

Tuval-Mashiach

(2017). Raising the curtain: The importance of transparency in qualitative research. Qualitative Psychology, 4(2), 126. https://doi.org/10.1037/qup0000062

52.

Van Den Boom

Quiroga

M. d-M.

O’Keefe

Kumar

Hill

P. L.

Scott

Dietze

(2022). Cohort profile: The Melbourne injecting drug user cohort study (SuperMIX). International Journal of Epidemiology, 51(3), e123–e130. https://doi.org/10.1093/ije/dyab231

53.

Walker

Dietze

Higgs

Ward

Treloar

Stoove

M. .

Maher

(2023). Socio-economic consequences of the COVID-19 pandemic for people who use drugs. Australian Journal of Social Issues, 00, 1–19. https://doi.org/10.1002/ajs4.289

54.

White

D. E.

Oelke

N. D.

Friesen

(2012). Management of a large qualitative data set: Establishing trustworthiness of the data. International Journal of Qualitative Methods, 11(3), 244–258.

Using Ethno-Epidemiology in a Prospective Observational Study to Increase the Rigour of Nested Qualitative Research

Abstract

Keywords

Background

CoDE Study Overview

Ethno-epi Sampling Approach

Recruitment and Participation Summary

Discussion

Conclusions

Footnotes

Acknowledgements

Declaration of Conflicting Interests

Funding

Ethical Statement

ORCID iD

References