Sampling and recruitment
In both studies, sampling was driven by quantitative and qualitative study aims. We restricted sampling to PWID because they represented a group of drug users at elevated risk for blood-borne infections and tend to be more heavily affected by changes in the drug scene (e.g., new drug formulations that could change as a result from the drug policy reforms). Rather than continuing to follow El Cuete IV participants from the previous phases of our binational research agenda, we decided to recruit different individuals due to attrition, mortality, and other temporal trends that affected our previous cohorts.
We sought to obtain representative samples of PWID in each site. Although we initially considered recruiting new cohorts of PWID using respondent driven sampling (RDS), due to issues of cost and the limited effectiveness of RDS at recruiting female PWID [38, 54], we instead used targeted sampling consisting of street-based outreach in diverse geographic areas . For example, El Cuete IV outreach teams established temporary mobile recruitment sites (e.g., vans and tents) in ten distinct colonias (neighborhoods) characterized by different physical risk environments and where PWID were known to spend time. Once situated in these neighborhoods, outreach workers attempted to engage individuals in conversation, sometimes by offering HIV prevention materials or information (e.g., condoms, educational pamphlets). Due to the drug related violence throughout Mexico, we developed an approved safety protocol in which outreach workers conducted all field-based recruitment activities in pairs or teams rather than alone and always carried identification cards to provide to potential participants and authorities. While all outreach workers in Tijuana spoke Spanish fluently, we also included at least one English speaker in each outreach team given the high numbers of U.S. deportees in our target population.
STAHR II also used targeted sampling methods . Recruitment involved direct street- and venue-based outreach (e.g., outreach workers passed out recruitment cards at parks, beaches, local syringe exchange vans and other areas where PWID congregate); targeted advertising through local newspapers, websites (e.g., Facebook, Craigslist), and posting flyers in neighborhoods and venues with a high concentration of PWID; and social networking strategies (e.g., peer referrals) within the target population. In addition, the study developed a website (http://www.ucsd-stahr.com) for potential participants to access information about the study. All written materials were available in English and Spanish. The outreach coordinator also maintained close contact with health centers and community agencies serving PWID (e.g., local syringe exchange programs). To better reach areas outside of the downtown area where STAHR II maintained a primary study office, a mobile outreach van and temporary office spaces were used to recruit PWID in northern, eastern, and southern San Diego County communities. These mobile sites were equipped to conduct all of the screening and data collection procedures described below. The STAHR II outreach staff also worked in pairs or teams, received study related trainings, and followed specific safety precautions described in an approved injury and illness prevention plan.
In both studies, eligibility criteria included the following characteristics: 1) being at least 18 years of age, 2) having evidence of injecting illicit drugs within the past month (i.e., confirmed by observation of track marks or other physical evidence of injecting), 3) being able to converse in English or Spanish, 4) currently residing in the study city with no plans to move away within 24 months from enrollment date, and 5) not currently participating in any intervention studies (although none to our knowledge were being conducted). Individuals with severe cognitive deficiencies or who were unwilling to provide informed consent were excluded, and PWID who met eligibility criteria but were too intoxicated to provide informed consent were rescheduled for rescreening at a later date.
Eligibility screening for both studies was conducted in private rooms in primary study offices (both located in commercial buildings) or at alternate sites (e.g., an offsite clinic, mobile van, or tent set up in outside venues). Screening began with the provision of general information about the study aims and procedures. Potential participants were asked for verbal consent before beginning the screening interview. Screening instruments were interviewer-administered, lasted approximately five minutes, and included several extraneous questions designed to prevent potential subjects from guessing eligibility criteria. Individuals were reimbursed $5 USD for their time to complete the screening, regardless of eligibility. Ineligible individuals were also offered free condoms, information and referrals for HIV testing, and reimbursement for public transportation as appropriate.
Following screening procedures, eligible PWID were asked to provide written informed consent. Interviewers handed potential subjects a copy of the consent form, read through the consent document to highlight key content, and discussed the study risks and benefits with the subject to determine understanding of the procedures and answer any questions that subjects had. In both sites, these materials were available in English and Spanish. Eligible individuals who decided to participate were asked to sign the consent form and were then invited to immediately begin baseline data collection unless they preferred to be rescheduled for a later time, which could require another eligibility screening process to ensure that time-dependent criteria were still met.
Subsamples of participants in each study were invited to undergo in-depth interviews to obtain qualitative data. Participants were purposively sampled on certain characteristics (captured in quantitative described below) to explore knowledge and behaviors in the context of Mexico’s legal reforms. Approximate sample sizes were determined based on the principle of conceptual saturation , which occurs when conducting additional interviews fails to provide major new findings . Due to the complexity of themes that we intended to explore, and the possibility of interviewing new participants and considering additional emergent themes during subsequent waves of qualitative data collection, we decided to initially conduct qualitative interviews with approximately 20 PWID per wave. During qualitative data collection, investigators held regular meetings in which interviewers presented summaries of new and recurring findings so that the research team could together evaluate saturation. Our mixed methods designs were sufficiently flexible to allow conducting qualitative interviews with additional participants when new themes were identified and required more in-depth exploration. Since the qualitative samples were fully embedded within the longitudinal cohorts, contacting additional participants was facilitated by a robust tracking system designed for retention of participants over multiple visits, as described below.
Retention and follow-up considerations
We developed active and passive follow-up strategies based on our binational team’s successful experience following PWID and other marginalized populations. During enrollment for both studies, staff members collected locator data on separate forms to facilitate participant follow-up. For example, locator forms collected information on where participants lived and spent time, their direct contact information (if available), individual physical characteristics, and family/social contacts of participants. Mailing addresses were unavailable for many of our participants, so one tracking strategy utilized in El Cuete IV involved asking participants to locate on maps where they lived (e.g., vacant lots, parks, canyons, shelters) and spent time during different hours of the day (e.g., panhandling spots, bars, clubs, food establishments). Participants also received small, wallet-sized appointment cards at enrollment that contained reminders of study visits, incentive schedules, locations and contact information for the study offices.
From our past cohort studies, we anticipated that repeated contact with participants would improve retention, so El Cuete IV and STAHR II compensated participants with $5 USD for brief “check-ins” at intervals halfway between semi-annual assessment visits (e.g., at months 3, 9, 15, etc.). Outreach workers conducted check-ins in person or by telephone, which involved asking participants to update their locator information and reminding them about future study appointments. All locator data were maintained in a password protected locator database stored separately from interview data to ensure confidentiality. Each month, data managers provided staff with calendars and lists of participants who were due for primary and locator visits. Both studies developed logos that were used throughout all methods of recruitment and tracking to enhance participants’ recognition and recollection of the study.
Staff also engaged in office- and street-based tracking of participants. Office-based tracking involved phone, text, email, postcard, and birthday card reminders, as well as phone calls to family/social contacts listed on locator forms. After three unsuccessful contact attempts, outreach workers proceeded to conduct home visits and other forms of street-based tracking (e.g., visiting locations where participants indicated that they spent time). Outreach staff also posted general notices describing the study in English and Spanish at shelters, bus terminals, airports, syringe exchange sites, health clinics, and drug treatment programs on both sides of the border. Due to the high numbers of migrants deported from the United States to Tijuana each year, STAHR II utilized administrative data to help locate participants who were incarcerated in or deported to Mexico. At the same time, El Cuete IV periodically searched for participants in neighborhoods where they were recruited, homeless day centers (i.e., desayunadores), and prisons and drug treatment centers.
Other retention strategies were specific to the contexts of each site. Based on our research experience with mobile PWID in San Diego, many of whom were unstably housed, STAHR II also utilized escalating monetary reimbursements and small gifts (e.g., toiletries, condoms, ID holders) to promote retention. For example, STAHR II reimbursed participants $25 for completing baseline surveys and testing, $25 at their 6-month visit, $30 at their 12- and 18-month visits, and $50 for their final, 24-month visit. In between these data collection visits, participants could also receive $10 for returning to receive their biological test results three weeks after each assessment visit, and additional $5 payments for completing the locator visits described above. Participants were informed that the total reimbursement amount they could receive over the two-year course of the study was $235. STAHR II also maintained the study website (http://www.ucsd-stahr.com) and accounts on popular social networking sites (e.g., Facebook) to keep in touch with participants. Finally, STAHR II gave participants wallet-sized calendars with toll-free phone number accessible from the United States and Mexico. Despite the unique contexts of the lives of PWID in both settings, both studies (and the experienced staff members dedicated to each project) were able to inform the other with respect to retention and tracking strategies.
Data collection strategies were intended to be similar across both studies, but we also recognized how studying this unique historical, binational situation would benefit from study designs that were flexible and responsive to local events and changes in risk environments over time. Both studies collected quantitative (survey) data and biological samples from the overall cohorts at baseline and semi-annual follow-up visits. Embedded subsamples of PWID were also selected from the two cohorts based on each study’s specific aims to complete qualitative interviews at multiple time points . Careful monitoring of all forms of data collection was inherent in our study designs and allowed for quantitative and qualitative instruments to be refined in-between study visits to capture new, emergent phenomena.
Quantitative data collection
For both studies, quantitative instruments were administered in English or Spanish by trained bilingual interviewers in confidential settings. Assessment instruments for both studies were developed jointly by the same team of investigators with the intention of creating identical measures whenever possible and to provide information from one study to complement data from the other study. For example, the El Cuete IV assessed experiences engaging in drug use with U.S. PWID to provide context for questions in the STAHR II assessment about injecting behaviors while in Mexico. This collaboration in measurement was intended to facilitate joint analyses including data from both cohorts. Bilingual, bicultural study staff members who were familiar with the unique language of the border region translated instruments from English into Spanish as necessary. The bilingual project director then back-translated these instruments into English to assess accuracy. Bilingual interviewers administered surveys using computer-assisted participant interview (CAPI) technology, which we have used for previous studies in Mexico and the United States.
Quantitative instruments at baseline assessed lifetime and recent experiences and behaviors, while follow-up surveys emphasized the time elapsed since the prior interview (six months). Socio-demographics measures included race/ethnicity, place of birth, education, language proficiency, citizenship and immigration status, passport ownership, marital status, living situation, binational travel, migration and deportation experiences. Measures of knowledge and attitudes focused on Mexico’s recent drug law reform, the health risks of using/injecting drugs following reforms, and, for STAHR II participants, travelling to Mexico to use drugs. For example, due to recent drug-related deaths in Mexico and shifts in methamphetamine production from California to Baja California [32, 33], knowledge and perceptions among PWID in San Diego regarding drug manufacturing could influence their perceptions about the risks and benefits of using drugs in Mexico.
Drug use behaviors included lifetime and recent use of specific drugs and routes of drug administration (e.g., sniffing, smoking, swallowing, injecting), including routes that could increase TB exposure (e.g., sharing pipes and “shot-gunning,” the exhalation of smoke directly into another person’s mouth) [84, 85]. We also assessed syringe and drug acquisition and periodically added new items in response to reports about emerging trends were worth monitoring in high risk communities (e.g., synthetic drugs such as “bath salts” [i.e., synthetic cathinones]). Drug treatment measures included lifetime and recent experiences with voluntary and court-mandated drug treatment involving diverse modalities (e.g. methadone, outpatient vs. residential drug treatment, self-help groups), barriers to accessing treatment, and perceptions regarding service quality/efficacy. Sexual risk behaviors for HIV transmission included number and types of partners, exchanging sex for money or other material goods, condom use, and drug use with sex partners. Other health measures included access to healthcare, history of prior diagnosis and treatment for active TB or LTBI and HCV, treatment type and completion, and symptoms . Location data for key outcomes of interest (e.g., interactions with police) were obtained by showing participants electronic maps on laptop computers and asking them to indicate exactly where such events occurred.
Serologic counseling and testing
Both studies conducted serologic testing for HIV infection at each visit. Participants received pre- and post-test counseling according to the Mexican Ministry of Health (El Cuete IV only) and U.S. Centers for Disease Control (CDC) guidelines (both studies). Serologic testing involved blood specimens collected via fingerstick and venipuncture according to standard clinical practice by a trained phlebotomists who were experienced in obtaining blood from PWID with scarred veins. El Cuete IV used Advance Quality rapid HIV tests (InTec Products, Inc). Reactive rapid tests were repeated. Participants receiving a second reactive rapid test were considered positive and referred to nearby municipal health clinics for free care under Mexico’s universal health system (e.g., Centro de Salud No. 1 or CAPASITS).
In STAHR II, HIV testing was performed using Uni-Gold TM Recombigen® HIV rapid test on whole blood collected via finger stick. For reactive tests, confirmatory testing was performed using the OraQuick ADVANCE® Rapid HIV-1/2 Antibody Test (OraSure Technologies, Bethlehem, Pennsylvania, USA). For reactive, inconclusive, or discordant test results, additional blood samples were collected and sent for confirmatory testing at the San Diego County Public Health Laboratory. In addition to serologic testing, we collected whole blood specimens for HIV nucleic acid analysis using DNAgard® Blood Tubes (Biomatrica, San Diego, California, USA), which contain a stabilizing agent allowing blood to be stored at room temperature without sacrificing DNA integrity or recovery. Plasma and serum samples were also collected at each visit for storage for future studies.
STAHR II also tested for HCV and Mycobacterium tuberculosis (Mtb) infection. HCV testing was conducted using the OraQuick® HCV Rapid Antibody Test (OraSure Technologies, Bethlehem, Pennsylvania, USA). Given the accuracy of this test, no confirmatory HCV testing was performed. Participants testing HCV-positive were referred to their health care provider or assisted in identifying a provider who assess liver function and test for active versus resolve infection. Testing for Mtb infection was performed using the QuantiFERON® TB Gold In-Tube assay ([QFT] Cellestis, Carnegie, Victoria, Australia). QFT requires blood samples to be incubated overnight followed by enzyme-linked immunosorbant assay; thus, participants were invited to return to receive these results. Participants testing positive for Mtb infection required evaluation to rule out active TB and potentially receive treatment for LTBI. In these situations, counselors assisted them in making appointments with their own physicians or referred them to the San Diego County TB Control Program. In both studies, participants were retested for infections at each visit until receiving a positive result, after which they were no longer retested for that infection. Referrals and educational resources were also provided for substance abuse treatment, management of wounds and abscesses, liver care, domestic violence, hunger, and housing.
Qualitative data collection
In both studies, qualitative interview guides contained broad, open-ended questions addressing study aims could be amended to more specifically address emergent findings. Interview guides were intended to be unstructured enough to allow interviewers to explore these emergent issues throughout the interview. Interviewers were also extensively trained in conducting interviews in a nonjudgmental, conversational manner and probing for additional details relating to study aims. Qualitative interviews were digitally recorded and transcribed for content analysis. During and after each interview, interviewers wrote detailed summary notes on new and important findings and reflected on emergent themes across interviews to inform preliminary analyses. These notes were later presented during regular team meetings so that the research team could identify commonalities across interviews and discuss progress towards achieving conceptual saturation. Interviewers also used feedback forms that were specifically designed for each study to record additional information on interview quality, participant demeanor, nonverbal behavior or emotions, and their general thoughts following the interview. Interview transcripts, interviewer notes, and feedback forms were translated as necessary for biweekly phone and in-person team meetings and preliminary analyses.
Other data collection
The El Cuete IV study obtained permission to access administrative records from Mexican authorities at the municipal and state levels on the following outcomes of interest: 1) drug treatment programs (e.g., type of treatment, date of entry and exit, voluntary vs. court-mandated, methadone dose, etc.), 2) interactions with law enforcement (e.g., dates and types of arrests, numbers of strikes, released vs. referred to drug treatment, etc.), 3) incarceration (e.g., nature of conviction, date of entry and exit, etc.). We also searched public registries semiannually to obtain death certificates (e.g., using the Mexican Registro Civil or SEMEFO). STAHR II also obtained access to publically available incarceration and reviewed death records to determine the status of participants who were lost to follow-up.
Results & discussion
Through a detailed explanation of our methodological approach, we demonstrated the feasibility of implementing binational mixed methods studies that borrow from the principles of a “natural experiment” to evaluate the effects of a high level structural intervention (i.e., a national policy change) in two diverse but interdependent contexts. Through the unique collaboration represented by El Cuete IV and STAHR II, which involved extensive interaction between investigators and research staff across an international border, we successfully recruited two parallel cohorts of PWID in San Diego (United States) and Tijuana (Mexico). We believe that our experiences can help inform future research on the impact of drug policy reform on the health and wellbeing of drug users in other international settings.
Overall, our prospective, mixed methods study design was both exploratory (i.e., qualitative data helped develop and refine conceptual models for how drug policy reform impacts the behaviors and experiences of PWID) , and complementary (i.e., findings from different study components led to different but complementary contributions to our understanding of complex local phenomena) [92, 93]. The iterative process of using qualitative and quantitative data to inform the methods and interpretations of results from each study component [92, 93] contributed toward a more comprehensive understanding of the binational effects of Mexico’s drug policy reforms on the health and wellbeing of PWID. Our integration of multiple methods of inquiry generated important new hypotheses, refined analysis plans, and enhanced interpretations of preliminary findings. While traditional epidemiologic survey could miss the key aspects of local contexts, a small qualitative study could not necessarily capture larger trends in drug abuse behaviors across this international border. Utilizing multiple research methods allowed us to offset the weaknesses of each individual method and helped address questions that could not have been answered through single methods alone.
We integrating different research methods in several ways that have been recommended in the literature . First, as our qualitative samples were embedded within two larger cohorts, we were able to use quantitative datasets to identify potential qualitative interviewees who met our purposive sampling criteria . For example, we used quantitative data to identify STAHR II participants who lived in different geographic areas of San Diego County to compare their experiences traveling to and using drugs in Mexico. Second, quantitative data also provided strata on which qualitative themes could be compared. Third, while quantitative data allowed us to describe behaviors, geographic hotspots and associated outcomes, our qualitative data provided more rich, narrative context on the reasons, meanings, and processes underlying quantitative associations, leading to enhanced interpretations of quantitative results. Fourth, emergent qualitative findings led to the development of additional quantitative hypotheses. Finally, our utilization of HIV sequence analysis and newer phylogenetic and network based analytic tools enabled us build upon our quantitative findings regarding prevalence and incidence by providing objective evidence about cross-border population mixing and the implications for HIV epidemics on both sides of the border.
An essential feature of our research agenda was the flexibility built into our approach, which allowed preliminary findings from different study components to inform other study components (e.g., preliminary qualitative findings from STAHR II helped refine future waves of quantitative survey data collection in STAHR II and El Cuete IV) . We believe that conducting a comprehensive evaluations of high level structural interventions such as national drug policy reforms demands a research design that can be responsive to local phenomena that evolve over time. By embedding a flexible, prospective qualitative component within two larger cohort studies, we were able to conduct real-time assessments of ongoing events while also understanding evolution in the perceptions of PWID living through those events.
In both studies, flexible sampling strategies have allowed the following-up on emergent themes during subsequent waves of data collection. For example, in El Cuete IV, we initially sought to conduct qualitative interviews with participants who had received varying numbers of drug possession apprehensions. Upon learning that implementation of the reforms had been severely delayed and few PWID in Tijuana had received any strikes at baseline, we modified our qualitative sampling criteria to include any participants who reported being stopped with and without being arrested. Our ability to proactively seek additional qualitative respondents meeting the new criteria was heightened by our use of quantitative survey data . Flexibility has also afforded us the ability to follow-up on preliminary quantitative findings using subsequent rounds of “member checking” qualitative interviews that solicit participants’ feedback on findings and investigators’ interpretations .
This type of flexible and iterative sampling is a hallmark of traditional qualitative designs  that is rarely integrated into large epidemiologic cohort studies. We argue that there are multiple benefits of having a highly responsive qualitative component but also acknowledge that it requires additional inputs from investigators and staff . In our case, to meet this challenge, we have held regular meetings to discuss emergent findings, inform investigators and staff working on the other parallel study, and revise quantitative instruments accordingly. Investigators and analysts also communicate across sites regularly to share and confirm preliminary findings across studies and identify emergent topics that warrant additional investigation. Although time-consuming, we agree with others that this iterative, flexible, and collaborative approach yields more inclusive and nuanced findings with important program and policy implications [98, 99].
A final strength of our research agenda is our ability to train a diverse new generation of harm reduction researchers and practitioners by leveraging support from the National Institutes of Health and the Fogarty International Center. Examples of funding that we have obtained include career development grants (e.g., Mentored Research Scientist Development Awards [K01 grants]), institutional research training grants (e.g., T32 for pre- and postdoctoral trainees), and diversity-promoting fellowships and supplements. To date, the El Cuete series has trained over 50 graduate students, medical students, fellows and junior faculty from institutions on both side of the border including the two largest public universities in this region: the University of California, San Diego (UCSD) and the Universidad Autónoma de Baja California (in Tijuana). The STAHR series has utilized similar funding opportunities to train 16 students from the high school through postdoctoral levels, as well as medical residents and junior faculty from UCSD and San Diego State University. Eleven of these trainees belong to racial/ethnic minority groups. While the primary objective of our training agenda has been to provide interdisciplinary research opportunities for students and new investigators, particularly those from underrepresented minority groups in the sciences, including diverse perspectives and expertise within our team has also enhanced the quality and meaningfulness of our research.
In designing our binational protocol, we considered several potential limitations. First, our studies depend largely on highly sensitive, self-reported behaviors. However, there is an extensive literature finding drug users’ self-reported behaviors to be valid [100–102]. As an additional strategy for guarding against underreporting of risky behaviors, we consistently assure participants that all data is confidential and that we do not report any behaviors or identities to authorities. To provide U.S. participants with greater assurance that their data will be kept confidential, STAHR II obtained a Certificate of Confidentiality. To the extent possible, we attempt to confirm some behaviors and experiences using objective administrative data (e.g., from drug treatment programs, police records, and jail/prisons). We also provide staff members with extensive training to increase their familiarity with the drug scene (e.g., local drug abuse trends, slang currently used on the street), which we believe enhances the rapport that they are able to develop with participants. We believe that this training and the resulting ability of staff members to interact comfortably with participants in a nonjudgmental manner serves to improve the reliability of self-reported behaviors.
Other important limitations of our research relate to sampling. Due to our nonrandom sampling strategies, our results cannot be interpreted as representative of the entire populations of PWID in Tijuana or San Diego. Our data also have limited generalizability with respect to other populations of PWID internationally, especially in regions with different policies and where heroin and methamphetamine are not the major drugs of abuse. Sample size constraints may limit our ability to identify all hypothesized effects. However, to minimize attrition, our outreach teams were extensively trained in office- and street-based tracking and the use of multiple active and passive follow-up techniques. As part of our binational collaboration, we also share information about successful strategies between studies. We believe that our collaborative, mixed methods approach, combined with the innovation and fortuitous timing of our binational research, will offset many of these limitations.