Design overview

Miracle Messages, the nonprofit that delivers the Miracle Friends and Miracle Money interventions, is headquartered in the San Francisco Bay Area but expanded to have staff members in Los Angeles for this study. Through partnerships with homeless service agencies or direct street outreach, Miracle Messages staff members engaged unhoused individuals to explain the Miracle Friends intervention, which requires having a phone, and signed up anyone who expressed interest in a phone buddy. Those who signed up also learned about a study to evaluate the Miracle Friends intervention, and those who expressed interest were referred to a study team affiliated with the University of Southern California. For those who agreed and provided written informed consent to participate in the study, a random number generator was used to determine sequentially whether a study participant would be offered the Miracle Friends intervention or put on a waitlist. Participants were told that they had a two-thirds chance to be offered Miracle Friends but were not told that approximately half of those offered Miracle Friends would be assigned to a third arm of the study that could receive basic income. For this group, receiving basic income through Miracle Money was contingent on participating in Miracle Friends. Figure 1 depicts how participants were randomized using a parallel-group trial design to one of three groups with a 1:1:1 allocation ratio: (a) those offered Miracle Friends only; (b) those who would be offered Miracle Money if they participated in Miracle Friends; and (c) those on a waitlist for Miracle Friends. Recruitment continued until at least 100 people began receiving Miracle Money.

Fig. 1

Randomized controlled trial design

Study team members interacting with participants for recruitment purposes were informed by a single study administrator via telephone just prior to enrollment about whether a participant had been randomized to receive Miracle Friends or put on a waitlist, but were blinded to whether the Miracle Friend assignment was part of the Miracle Money condition. Once enrolled, the single study administrator provided the Miracle Messages staff with the group assignment of each participant to determine whether Miracle Friends should be offered. For those assigned to Miracle Money, the possibility of receiving guaranteed income was only disclosed after it had been determined that they were participating in the Miracle Friends intervention.

Nondisclosure of the Miracle Money condition, a form of deception, was deemed necessary to avoid (a) unduly influencing people to participate in the research and (b) biasing results such that people might have engaged in the Miracle Friends program only because they were interested in guaranteed income. Because revealing that a participant may have been eligible for but was not offered basic income could potentially cause more distress than the deception, the study team will not debrief participants at the conclusion of the study about the complete study design, which is consistent with the Code of Federal Regulations on the protection of human subjects [46].

Regardless of group assignment, all participants identified by a unique study identification number to preserve confidentiality have been asked to complete a baseline survey and five subsequent quarterly surveys; participants will receive a $30 gift card incentive for each completed survey. Qualitative interviews with a subset of 20 participants receiving basic income have and will be conducted shortly after receipt of the first of 12 monthly payments, with another follow-up qualitative interview scheduled around the time of the last payment. A single qualitative interview will be conducted with 20 volunteers serving as a phone buddy for at least 6 months to understand their experience of delivering the Miracle Friends intervention. Participants will receive a $30 incentive at the end of each qualitative interview. This human subjects’ research is being performed in accordance with the Declaration of Helsinki with protocols approved by the first author’s institutional review board. Reporting of study protocols follows the SPIRIT guidelines [47], which include a schedule of enrollment, interventions, and assessments, as depicted in Fig. 2. The SPIRIT checklist for Trails can be found in the Supplementary materials.

Fig. 2

Schedule of enrollment, interventions, and assessments

InterventionsMiracle friends

Miracle Messages program staff members have recruited volunteers who want to serve as a phone buddy and unhoused individuals who expressed an interest in being matched with a phone buddy. Recruitment of volunteers has occurred primarily through people who learned about the program through media coverage, word of mouth, social media, or internet searches about helping PEH, with most volunteers signing up on the Miracle Messages website. Volunteers are required to complete an application listing any preferences for a friend (e.g., gender, language, shared interests, text or calls preferred) and attend a 30-min training call offered once a week synchronously. A recording of the training call is available as needed to those with significant scheduling conflicts. Volunteers receive a program handbook that outlines expectations for logging into an online platform to record any contact or attempted contact with an unhoused friend. After completing a waiver of liability, volunteers receive a phone number through Dialpad or similar service (which allows the volunteers to avoid divulging their personal phone numbers if they wish) and are subsequently matched with an unhoused friend, usually in a few weeks. Weekly support calls are offered to all volunteers, in addition to one-on-one support provided upon request based on completing the contact logs. Currently, there are approximately 300 Miracle Friend volunteers.

Recruitment of unhoused individuals has happened primarily through (a) site visits at local partner sites, including shelters, transitional housing facilities (e.g., tiny homes), converted hotels or motels for unhoused individuals, etc.; (b) referrals from caseworkers and social workers at these partner sites; and (c) limited direct outreach on the streets in and around partner sites or at events designed to address the needs of PEH (e.g., food pantries, homeless connect events). Miracle Friends program staff members explain the intervention and sign up anyone who expresses an interest in a phone buddy. Like with the volunteers, unhoused participants are also asked to complete an application listing any preferences for a friend (e.g., gender, language, shared interests, text or calls preferred).

Matching of unhoused individuals and volunteers has been primarily based on preferences and shared interests, as indicated on the enrollment application. Once a match is determined, volunteers receive the phone number of their assigned unhoused friend and are asked to make contact as soon as possible. The program has no time limit, and the development of friendships is expected to occur naturally over time and be unique to each matched pair. Communication is bidirectional in that both participants can call or text each other. Volunteers are encouraged to provide a friendly voice and be a compassionate listener without judgment; there is no expectation of formal counseling or case management. If a lack of fit occurs or communication is not maintained between the two individuals, the program offers to rematch the unhoused individual, volunteer, or both. If a volunteer or Miracle Messages staff members cannot contact an unhoused person after five attempts by the volunteer and one attempt by the staff, the person is removed from the list of people needing to be matched—effectively representing a discharge from the program. However, participants can reenroll at any time. Once matched, volunteers are expected to attempt weekly phone voice or text contact. Volunteers log their efforts on the program’s online platform after each attempt, helping the program monitor the progression of friendships and address any issues. Based on volunteer responses, the Miracle Friends program can provide referrals to the unhoused person if there is a more urgent need or contact a formal service provider if authorized by the unhoused individual.

Miracle money

Whereas the Miracle Money proof-of-concept provided $500 for 6 consecutive months, for this study, funds were raised so that Miracle Messages could provide up to 110 individuals with $750 per month for 12 months. This was based on pilot work that suggested that $500 per month meaningfully changed lives [41] but recognition that a living wage in these cities would be much higher [48]. As noted, Miracle Money is not advertised to those interested in the Miracle Friends intervention; unhoused individuals are only notified about the possibility of receiving basic income if they are randomized to the Miracle Money condition and participate in the Miracle Friends interventions. Although participation was defined as having had at least two contacts with a matched volunteer in a 1-month period, at the outset of the study there were efforts to ensure that the two contacts could be characterized as “meaningful” based on volunteer feedback to help ensure that the matching process would result in a prolonged friendship or relationship. Yet because many of the initial contacts could not be confirmed to be meaningful, Miracle Messages increasingly came to accept any type of contact with volunteers as meeting the criteria of Miracle Friend participation that would then allow participants to start receiving Miracle Money.

Volunteers are notified when the person with whom they have been matched becomes eligible for Miracle Money so that the volunteer can be present during the phone call in which the study participant is offered Miracle Money. Those who qualify and decide to move forward with receiving basic income payments are asked to complete an application for a cash transfer technology company known as AidKit, which has been contracted to process monthly payments sent via a debit card or direct deposit. Although we do not expect many people to turn down Miracle Money, we envision that some may decide against receiving income payments based on how it may affect other benefits that depend on income (e.g., Supplemental Security Income, Supplemental Nutrition Assistance Program), although the intent of guaranteed basic income is to supplement rather than replace existing resources. When participants are first offered Miracle Money, the Miracle Messages staff discusses benefits and encourages participants to discuss the topic with a case manager; referrals to financial coaching are also offered. Participants are told that they can use the money in any way they choose but that the program requests the money not be used for any illicit purposes.

Recruitment and enrollment

As noted, study recruitment has occurred via the Miracle Messages program, which notifies the study team about unhoused individuals interested in participating in an evaluation of the Miracle Friends intervention. This has occurred in person when the study team accompanied Miracle Messages staff members to outreach events or through a shared online referral document that provided the study team with the contact information of individuals interested in Miracle Friends and the evaluation study. Once the study team confirmed that an individual met the inclusion criteria—which included (a) being 18 years old or older; (b) speaking English or Spanish; (c) currently experiencing homelessness based on definition from the U.S. Department of Housing and Urban Development; and (d) expressed interest in the Miracle Friends phone buddy intervention—enrollment into the study occurred in person with signed informed consent or over the phone with informed consent, requiring an electronic signature completed via email using the REDCap platform [49, 50]. The informed consent process ensures that individuals understand that participation in the study may result in being assigned to a waitlist where they would be ineligible to participate in Miracle Friends for 15 months unless they withdrew from the study. Once enrolled in the study, participants complete a baseline survey and learn whether they have been randomly assigned to the waitlist or Miracle Friends. Surveyors are blinded to whether a participant is assigned to the group that will be offered Miracle Friends only or the group that will be offered Miracle Money if they participate in Miracle Friends. The Miracle Messages administrative staff is notified to which of the three groups a participant has been assigned to determine whether someone should be matched to a phone buddy and monitor which individuals become eligible for Miracle Money based on participation in Miracle Friends. Study recruitment continued until at least 100 people started receiving Miracle Money.

Quantitative data collection procedures

Upon enrollment in the study and regardless of group assignment, the participants complete a baseline survey that takes approximately 45 min and includes questions about demographic characteristics, homelessness history, physical and mental health status, health service utilization, employment, substance use, socioeconomic status, and income. Participants are also asked to provide an email address and collateral contact information for one or more people to ensure that the study team can reconnect with them when it is time to complete a shorter survey five more times, one every 3 months (i.e., quarterly) until 15 months. The quarterly survey asks about similar topics and takes approximately 25 min. Surveys can be completed in person or over the phone, with responses recorded by a surveyor directly into the REDCap data management platform [49, 50] that is also used to support data visualization to monitor data quality and study retention. Upon request, participants also receive a link to self-administer the survey in English or Spanish. Table 1 describes all study measures.

Table 1 Study measures and scales

At the end of each survey, participants are asked two open-ended questions: (1) Looking back over the last 3 months, what do you think was the most significant change to your quality of life? (2) Tell me in a sentence or two, what are your most important goals in life right now? If the survey is administered by a surveyor, responses are entered verbatim as much as possible; if self-administered, participants enter their responses directly. All surveyors complete a training on best practices for trauma-informed interviewing.

Qualitative data collection proceduresMiracle money recipients

Qualitative interviews will be conducted in English with a subset of 20 participants who receive basic income—shortly after receipt of the first of 12 monthly payments and again around the time of their last payment. During the first quarterly survey, participants who received monthly income will be purposively sampled and asked if they would be interested in participating in an additional in-depth qualitative interview to learn more about their experience with the program. Maximum variation sampling will be used to ensure differences in race, ethnicity, and gender in our qualitative subsample. Those who are interested and agree to participate will be contacted by a research team member who has been trained in conducting qualitative interviews. Interviewers will schedule a convenient time and place to meet unless participants request a phone interview. Once an addendum consent is completed, semistructured interviews will last between 30 and 60 min and include questions about how participants view their lives, general experiences with the Miracle Friends program, and the Miracle Money program, including how they use the money and any difference it has made in their lives. Interviews will be audio recorded and transcribed verbatim.

Miracle friend volunteers

A single qualitative interview will be conducted with 20 volunteers who served as a phone buddy for at least 6 months to understand their experience of Miracle Friends. Any volunteer who has been engaged in the intervention for at least 6 months will receive an email from the Miracle Messages program that informs them of the study and refers them to the study team if they are interested in speaking about their personal experience with the program. Thirty-three volunteers have indicated that they would be interested; 20 have been purposively sampled to include volunteers who had participants in the Miracle Friends program and Miracle Money program. Phone or videoconferencing interviews have been conducted using a semistructured interview guide that included questions about how and why participants became a volunteer, experiences with the program, and whether they felt that they had an impact on their unhoused friend or if their friend has affected their life. Questions about how the program could be improved were also included. Interviews have typically lasted between 25 and 35 min and were audio recorded and transcribed verbatim. Written informed consent has been waived by the institutional review board for volunteer interviews.

Quantitative data analysisSocial isolation outcomes

To evaluate the efficacy of the Miracle Friends intervention on social support, psychological distress, and loneliness, we will use responses to the Oslo Support Scale [53], Kessler Psychological Distress Scale [55], and UCLA Loneliness Scale [52], respectively, at the final survey (Quarter 5). Each measure will first be summed to develop a total score, then dichotomized based on cutoffs for each measure based in the literature—i.e., poor social support will be coded as 1 if the Oslo Support Scale sum score is less than 9 (indicative of poor social support) and 0 otherwise; psychological distress will be coded as 1 if the Kessler Psychological Distress Scale sum score is greater than 25 (indicative of high psychological distress) and 0 otherwise; and loneliness will be coded as 1 if the UCLA Loneliness Scale sum score is greater than or equal to 6 (indicative of high loneliness) and 0 otherwise. Each dichotomized score will be modeled as a function of treatment group (1 = Miracle Friends or Miracle Money group members who participated in at least one phone buddy intervention, 0 = waitlist group); covariates including demographic characteristics, physical and mental health, and substance use; and a random intercept for city in random intercept logistic regression models.

Housing outcomes

The association of Miracle Friends with housing outcomes will be evaluated using logistic regression with a random intercept, modeling the dichotomized outcome of exited homelessness (1 = individual responded “My own apartment or home” or “Someone else’s apartment or home,” 0 = otherwise) as a function of treatment (1 = Miracle Friends group members who participated in at least one phone buddy intervention, 0 = waitlist group), with the same covariates and random intercept.

Differences in treatment effect based on race

To evaluate if outcomes differ by race, all models will be run with the addition of an interaction between race and treatment.

Statistical power

To reach a target goal of 105 participants receiving basic income through Miracle Money, recruitment efforts have yielded more than 200 individuals in each treatment group. Assuming that each arm will have at least 105 participants, our study can detect the hypothesized 50% difference in proportion of people exiting homelessness at Quarter 5 between the Miracle Money and waitlist group (assuming 60% of the Miracle Money group exits homelessness and 10% of the waitlist group exits homelessness) using a p-value threshold of 0.05 with power exceeding 95%. When comparing the Miracle Friends group to the waitlist group, our study can detect the hypothesized 15% difference in proportion of people exiting homelessness at Quarter 5 between the Miracle Friends and waitlist group (assuming 25% of the Miracle Money group exits homelessness and 10% of the waitlist group exits homelessness) using a p-value threshold of 0.05 with power exceeding 60%. Differences in proportions of the social isolation outcomes will be detected at powers identical to the housing outcome, assuming identical differences in proportions and p-values. Regarding the effect modification of treatment by race in exiting homelessness, our study will be able to detect a difference in proportion of groups exiting homelessness of 35% with power exceeding 20%. Main effect calculations were done on unadjusted models using the WebPower R package, and the effect modification power calculations were performed via simulation.

Qualitative data analysis

In keeping with qualitative analytic procedures, each interview has been transcribed verbatim by the research team member who conducted the interview. Transcripts have been distributed and shared with the larger team for review. A team approach will occur in data analyses, where instruction in coding will be supplemented with test cases in which two researchers read and code a transcript and then meet to discuss discrepancies and arrive at consensus. Given the focused nature of the inquiry, the resulting codebook will be a reflection of the questions that were asked. At the same time, interviewees often have shared greater depth or alternative descriptions that will “earn their way” into the analyses and interpretation [56]. In the final stage of analysis, broader interpretation will be sought to identify recurrent themes agreed on by consensus and recorded as memos.