Background
The annual number of persons experiencing homelessness in the United States is estimated to be approximately 3.5 million, and this number is increasing [
1,
2]. The rate of homelessness is increasing due to both individual challenges (e.g., poverty, mental health and substance use disorders) [
3,
4], and structural factors (e.g., lack of affordable market-rate housing, low-wage jobs, diminishing housing voucher program (Section 8), and changing subsidized housing eligibility guidelines) [
5]. Homeless individuals experience higher rates of physical and mental illness leading to higher rates of hospitalization and mortality compared to those in the general population [
6].
Homeless individuals have a mortality rate three times higher than the general population [
7]. One study showed that the average homeless person has a life expectancy of only 47 years, compared to an average of the general population, 77 years [
8]. Many factors may contribute to the high mortality rates among homeless adults, including affective disorders [
9], chronic illnesses (e.g., HIV, heart disease, and cancer) [
10], and excess alcohol and drug use [
11]. This population also has poor medication adherence and low self-efficacy which, along with the many practical challenges of homelessness, limits the ability to adopt behaviors that lead to better health outcomes [
12]. Several studies show that homeless individuals in the United States have elevated rates of chronic illnesses such as human immunodeficiency virus (HIV) infection, tuberculosis (TB), and mental health and substance use disorders [
13‐
15]. Approximately 33% of adults who are homeless suffer from some form of severe and persistent mental illness [
16] as compared to the general population estimates of 18.1% [
17]. These findings highlight the need for clinical research to identify effective interventions to improve health and prevent disease in this vulnerable and underserved community [
18]. Despite the high prevalence of medical and mental health illnesses in this population, very few intervention studies are conducted among homeless adults [
19] and little is known about effective interventions to address both acute and chronic conditions in these vulnerable adults. Homeless individuals are generally excluded from research studies because they are perceived as “hard to reach and retain” [
20‐
22]. This assumption stems from the fact that this population is often transient which may lead to high study attrition [
23]. While this is true in some cases, some studies have demonstrated that high retention can be achieved in programs tailored to address the unique challenges inherent in conducting research among persons who are homeless [
23,
24].
However, high attrition rates due to participant withdrawal, relocation, and death can threaten the integrity of study results which may contribute to researchers’ reluctance to engage with homeless populations. In order to decrease attrition, increase protocol adherence, and to protect study staff and homeless participants, it is important to address the issues that are germane to homeless populations.
Study context
The first smoking-cessation randomized controlled trial (RCT) designed for homeless smokers was titled “Power To Quit” (PTQI) [
24]. The PTQI study was a community-based RCT that enrolled 430 adult smokers who were homeless. PTQI compared Standard Care (one-time brief advice to quit smoking) to six Motivational Interviewing (MI) counseling sessions. The details of the study design and recruitment procedures have been published [
25]. All participants received 21-mg nicotine patches for 8 weeks. Assessments were conducted at post-randomization weeks 8 and 26. Overall, this study found that cotinine-verified 7-day quit rates were 9.3% for MI and 5.6% for control at 26 weeks (
p = 0.15) [
24]. These quit rates are low compared to the general population and may be associated with the observation that many participants reported concurrent heavy alcohol use (approximately 46%). The results of a systemic review and meta-regression analysis show that the prevalence of alcohol dependence in the homeless population is 8.5–58.1% (
β = 0.18, SE[
β] = 0.07,
p = 0.007) [
5]. Therefore, a more intense follow-up intervention was designed to address some of the limitations encountered in the first trial and to target concurrent alcohol abuse.
PTQII included (1) a higher dose of pharmacotherapy (nicotine patch combined with nicotine gum or lozenge) and a longer duration of prescribed use (i.e., 12 versus 8 weeks), (2) an increased number of counseling sessions (12 versus 8 sessions over 3 months), (3) an increase in the duration of individual sessions from 15 to 30 min to 45 to 60 min, (4) counseling sessions utilizing cognitive behavioral therapy (CBT) strategies to enable counselors to provide more strategies during counseling sessions, and (5) counseling content targeting alcohol abstinence in addition to smoking cessation. Details regarding the design of this study have been published [
26]. Briefly, the study utilizes a three-group design that includes (1) Usual care (UC) for smoking and alcohol cessation, (2) Intensive smoking cessation plus UC alcohol abstinence counseling (IS), and (3) Integrated Intensive Smoking and Intensive Alcohol Counseling (IntS + A). All participants were invited to receive 12 weeks of nicotine replacement therapy utilizing both nicotine patches (tailored to their baseline cigarettes smoked per day), plus their choice of nicotine gum or lozenge. Integrating alcohol treatment with the intensive smoking intervention helped to assess whether addressing alcohol use concurrently with smoking cessation will result in improved smoking abstinence and/or reduced alcohol use, an important scientific question that has never been studied in homeless populations.
Both PTQI and PTQII were approved and monitored by the University of Minnesota Institutional Review Board. The objective of this paper is to describe several practical lessons learned while conducting two large RCTs targeting smoking cessation among persons who are homeless. Our intention is to provide information that may assist other investigators who are interested in developing and testing the efficacy of interventions designed for homeless populations. Both studies were funded by the National Institutes of Health (USA).
The critical issues addressed in this paper are organized into four domains: (1) study settings, (2) participants, (3) data collection and management, and (4) staffing issues. For each domain we identify the central challenge, discuss the strategy used to address it, and highlight the potential implications for future research trials.
Intervention
In PTQI and PTQII we had limited availability of times to see participants at the shelters. The shelters allotted two 2-h blocks per day. In our initial protocol, we planned on using group counseling to supplement individual counseling to increase social support among participants and increase retention. However, we found that homeless individuals in these particular studies were not interested in meeting in groups. In addition, scheduling groups around the transient individuals in this population made finding appropriate times for all the participants in the group difficult. In addition, it was sometimes challenging finding a private space at the shelters to conduct counseling; as a result, participants had to be rescheduled within their specific time window.
Designing a feasible and practical intervention to accommodate missed visits and maintain a flexible schedule is key to maximizing adherence to the intervention protocol. It is important to keep in mind that delaying treatment initiation may contribute to attrition. For example, retention increased when NRT was administered at the baseline visit rather than waiting to the 1-week visit.
Retention
In PTQI and PTQII we collected multiple contact sources from the participants, this allowed the team to reach those with government-issued phones, no phone number or contact information; or extremely transient individuals. Community mobilizers (see below) played an important role in enhancing recruitment and minimizing attrition. In addition, the mobilizers were tasked with the responsibility of reminding participants of the location and time of the next study appointment, and giving them reminder calls. They made reminder calls to participants during the week prior to each appointment and continued to call participants until the window for completing a given appointment closed. Calls were placed from the project office and made either to each participant’s cell phone or to the shelter identified as the most recent nighttime residence in the participant’s file. In order for study staff to retain participants it was necessary to maintain an active log of phone numbers, emails, alternate addresses, and shelter contacts for each participant. This information must be retained separately from other study documents to protect privacy.
Compensation
In the PTQI study, the participants were compensated up to US$275 over 6 months [
23]. The value of the incentives given out at any particular time was capped at US$20 with the exception of months 6 and 12 which were data collection visits in which we provided US$50. Lottery-style drawings were implemented in PTQII to bolster retention and incentivize participants to return. This type of blind incentivizing rather than a fixed-ratio compensation has been shown to increase retention [
40]. Participants enrolled in both studies were compensated at every intervention and data collection visit. This incentive scheme was necessary to increase enrollment, treatment delivery, and data collection. It was also important to give compensation that was tailored to this population’s competing needs. Nonmonetary incentives were also provided; these include bus passes, tote bags, phone cards, and calendars. Calendars and pens were used with the goal that they would help with keeping study appointments.
Adverse events
In both studies, the study coordinator ensured that adverse event documentation was thoroughly reviewed by the principal investigator (a practicing board-certified family physician). In addition, participants were given the study office phone number (24-h coverage) to contact study staff and/or the investigators to report adverse events. The study followed the NIH guidelines for reporting adverse events to the Institutional Review Board. Any problems needing medical attention were referred to the licensed provider and clinics, which are federally qualified community health centers that provide medical care and social services for homeless persons. The study was discontinued if participants became pregnant or developed a contraindication to continuing in the study. Documentation of adverse events is especially important in this population due to the high prevalence of medical and psychiatric health problems. In order to accurately assess adverse events, participants were asked questions to elicit responses concerning recent hospital visits and any other pertinent events or information at every visit.
Data collection and management
In PTQI and PTQII, screening for alcohol intoxication was conducted before each appointment using with Alco-Sensor III breathalyzers (Intoximeters, Inc., St. Louis, MO, USA) testing [
41]. Any participant who showed a blood alcohol level (BAL) reading of 0.08 or above was unable to participate in activities for that visit. Some other studies in the homeless have used a cutoff mark of a BAL of 0.05 [
42]. A criterion for proceeding or rescheduling data collection visits due to intoxication or lack of sleep was required. Study staff used the Short Blessed Test [
30] to evaluate cognitive impairment prior to data collection at all visits. Breathalyzer screens were conducted before each appointment. Timing of data collection can have an important impact on participant recruitment and retention. For example, baseline data collection can occur at the time of enrollment (which improves follow-up after enrollment) or some days later (which may compromise retention). Biospecimen collection protocols should be as simple as possible and can be a sensitive requirement due to concern about using biospecimens for drug monitoring.
Discussion
This paper reports on the complexities of conducting two RCTs of smoking interventions conducted among persons who are homeless. The value and importance of developing protocols that support the inclusion of homeless populations in trials cannot be overemphasized. There are key research questions to address but the more inclusive a study is, the more planning is needed in order to support the participation. Homeless individuals constantly deal with problems that may make it particularly difficult to participate in research protocols, including the need for negotiating daily access to food, clothing, and shelter. Research protocols are often ill prepared to address the daily challenges that homeless individuals encounter which makes it difficult to execute clinical trials in this underserved community. Many clinical trials are avoided in this high-risk population due to concerns about safety, recruitment, and retention. Designing a comprehensive study protocol that addresses these concerns can increase the effectiveness of executing a controlled trial in homeless populations.
This study has some limitations. First, the two studies were conducted at a single metropolitan area in the upper Midwest of the United States, and the challenges faced may not generalize to all homeless persons. Secondly, limited data on this topic are available from other randomized clinical trials. Therefore, the research experience is descriptive and the recommendations are based on the experience of the research team at a single institution conducting research in this community over the past decade.
Acknowledgements
The authors would like to thank the directors and staff of the participating shelters and express gratitude to the members of the Community Advisory Board and the study participants. Research reported in this publication was supported by the National Cancer Institute of the National Institutes of Health under Award Number 2T32CA163184 (Michele Allen, MD, MS; PI). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.