Caring Text Messages for Suicide Prevention in Urban American Indian Youth: Protocol for a Randomized Controlled Trial

The intervention and study design were adapted to the cultural values and context of the participating communities (eg, type of messages and topics, number of messages, and schedule) [20]. To create messaging that appealed to young urban AIs, the study teams conducted focus groups and semistructured key informant interviews with AI individuals affected by suicide, and gathered input from staff in the clinic’s behavioral health and traditional wellness departments, the project’s youth representatives, and study staff. Participants were asked to brainstorm and review text messages for their cultural appropriateness and relevance to young urban AIs in their community. The messages were then carefully reviewed and adapted by staff from the clinic’s behavioral health and traditional wellness departments, and research staff to ensure that they were consistent with the ideals of the caring texts intervention (ie, brief, supportive, and nondemanding). The messages used familiar language and were sent by research staff who had conducted the participants’ baseline assessments to establish continuity and connection.

Data and Safety Monitoring

Site and coordinating center staff monitored participant safety. To determine and address potential suicide risk during assessments, study staff administered the University of Washington Risk Assessment Protocol to participants immediately before and after baseline and follow-up assessments [21]. If this assessment indicated moderate or severe suicidal intentions, or if the assessor has other reasons to be concerned about risk, a site-specific crisis protocol is implemented to ensure the participant receives appropriate care from a trained clinician who can more thoroughly assess their level of risk and initiate appropriate next steps.

All adverse events are identified through assessments or through care text messages, then recorded in the REDCap database, which automatically sends email alerts to the team overseeing participant safety. The site leads ensure high-risk participants receive appropriate management and follow the crisis protocol, including referral and a warm hand-off to clinic staff or a crisis line. The National Institutes of Health required a Data Safety and Monitoring Board to meet twice a year and oversee recruitment, adverse events, and protocol violations. After each meeting, the board prepares a brief report with recommendations for protocol modifications. All adverse events are reported to the universities’ Institutional Review Boards and tribal review boards, in accordance with their policies.

Measures

A summary of all measures used in the study is available in Table 2. Unless otherwise noted, measures were administered as a survey in REDCap.

Primary Outcomes

Our primary outcome of interest was suicidal behavior. While suicidal ideation is more common than suicide attempts or attempts requiring hospitalization, ideation is not necessarily a concrete event and can be more difficult to measure. For this reason, ideation, attempts, and suicide-related hospitalization were all selected to measure the same underlying risk construct. The Suicidal Ideation Questionnaire-Junior, which has been validated with AIAN individuals, measured the study’s primary outcome of suicidal ideation [22]. It consists of 15 items that evaluate thoughts of death, desire to be dead, and specific thoughts of self-harm. Scores range from 0 to 90, with scores above 32 indicating active suicidal ideation [23]. The “junior” version of the questionnaire was used to accommodate participants with lower reading levels.

The SASI count (lifetime and recent) assessed the primary outcome of suicide attempts [24]. It is a concise questionnaire that determines the first, most recent, and most severe suicide attempt or nonsuicidal self-injury. It assessed the date, method used, and intent to die (ie, clear intent to die, ambivalent intent, or no intent to die), the highest level of medical treatment received (ie, none, doctor or clinic visit, emergency room, or admission to a medical unit), and lethality of each event. The SASI also posed questions about the number of attempts for each of the 17 methods, specifying intent to die, medical treatment received, and lethality for each one. Intent to die and lethality were derived from the SASI interview [24]. A lifetime version of the instrument administered at the baseline assessment captured all historical suicide attempts or nonsuicidal self-injury. A recent version of the instrument assessed only those events that occurred in the past year, both at baseline and follow-up assessments. The adapted version of the SASI used in this study simplified the language, included methods commonly used in AIAN communities, and added an assessment of drug and alcohol use at the time of the self-injury. The SASI was administered via interview.

The primary outcome of suicide-related hospitalizations and health care encounters was measured via the Youth Health Services Measure [25]. The questionnaire collected information about the use of and satisfaction with inpatient and outpatient medical care, emergency room visits, and the use of traditional healing. It could be used to identify health care encounters related to suicidal ideation or behaviors.

Secondary Outcomes

We used several measures of social connectedness. The Interpersonal Needs Questionnaire describes the extent to which individuals feel connected to others (ie, belongingness) and the extent to which they perceive themselves to be a burden on other people (ie, burdensomeness) [23,26]. Previous psychometric analyses indicated that the latent variable “thwarted belongingness” significantly predicted suicidal ideation scores, supporting the measure’s construct validity [9]. The Loneliness Scale, which is part of the National Institutes of Health Toolbox Social Relationship Scales [27], assesses feelings of loneliness and exhibits both internal reliability and concurrent validity. The Social Distress Scale, also part of the National Institutes of Health Toolbox, measures perceived rejection and hostility and has good internal reliability and concurrent validity [27]. The Social Contact Questionnaire asks participants about their use of various forms of social contact, such as text messaging, email, and postal mail [28]. Finally, the Social Connection Measure asks participants to choose which of 7 Venn diagrams, each with a progressively greater proportion of overlap, best reflects their sense of closeness to their community [29]. This measure is intended to provide participants with a means of visualizing their integration within their community.

Covariates and Other Measures of Interest

Participants were also asked to complete surveys related to several other aspects of mental health so that they could later be assessed as possible mediators or effect measure modifiers. The Acquired Capability for Suicide Scale is a 20-item measure that quantifies one’s fearlessness about and capability to die by suicide [30-32]. It has adequate to good internal consistency (α=–.67 to .83), convergent and discriminant validity, and does not correlate with measures of depression or suicidal ideation. The Exposure to Suicide Questionnaire asks about participants’ exposure to suicide and suicide attempts made by family members, friends, or coworkers to assess possible contagion effects. This information allows a posthoc examination of the impact of exposure to suicide and suicide clusters on the intervention effects.

The Breslau Post Traumatic Stress Disorder Screening is a 7-item scale that assesses internalization of feelings following an upsetting or frightening event [33]. Questions about feelings in the past month are answered yes or no and summed to create a total score ranging from 0 to 7; higher scores indicate probable posttraumatic stress disorder. The Ayers Coping Scale is an 8-item scale that has been validated among AIAN populations to assess the frequency of using healthy coping strategies when faced with a problem [34]. The Kessler Psychological Distress Scale, also validated for administration to AIANs, is a 6-item scale that assesses the frequency of feelings of anxiety, hopelessness, depression, restlessness, and unworthiness in the past 30 days [35]. The scale is sensitive (0.36, 0.08) and specific (0.96, 0.02) with respect to predicting serious mental illness [35].

The Historical Loss and Trauma Scale is a 12-item self-report measure that assesses the frequency of thoughts regarding historical trauma (eg, “how often do you think about the loss of our traditional spiritual ways” and “how often do you think about the loss of our language”), which has high internal reliability [36-38]. The Oetting and Beauvais Orthogonal Ethnic Identification Scale poses 8 questions regarding the degree to which respondents follow the tribal and white way of life and the importance of keeping their tribal identity, values, and practices [39,40].

The EuroQol 5-Dimension 5-Level (EQ-5D-5L) is a 6-item measure that assesses respondents’ capacity to perform activities related to mobility, self-care, usual activities, and health dimensions, including pain, discomfort, anxiety, and depression, as well as overall self-reported health status [41,42]. The Health Economics Analysis Questionnaire estimates use of outpatient, inpatient, and emergency room visits; out-of-pocket costs of care; health insurance; employment and schooling benefits; income; publicly funded benefits; lifestyle behaviors (eg, substance use); and costs of participating in the study.

Self-reported drug use and drug use severity are assessed using the Addiction Severity Index, Native American Version [43]. This is a comprehensive measure of lifetime and current drug and alcohol use, as well as the impact of substance abuse on medical conditions, psychiatric well-being, employment status, and encounters with the legal system. The culturally adapted version of the measure also addresses the impact of substance use on the respondent’s ability to participate in cultural or traditional activities and ceremonies.

Demographic information included gender, sexuality, relationship status, religion, number of children, living arrangement, employment, use of the native language, and education level. Demographic features that may change are reassessed at 12 months.

The COVID-19 Impact on Well-Being Survey, a brief 5-item survey, assesses the impact of COVID-19 infection, the pandemic, and related mitigation efforts on participants’ economic, social, and overall health well-being.

We also queried intervention participants at follow-up about their experience of and satisfaction with caring texts.

Treatment Retention

Retention in the SBIRT program is assessed at 12 months. It is operationalized as a 3-category indicator (full, partial, and none) defining engagement for each participant in the appropriate level of care. The appropriate level of care is determined during the initial in-person session with the SBIRT behavioral therapist before enrollment in the study and is independent of study procedures. These variables will be included in sensitivity analyses if they appear unbalanced, although randomization and the large sample size should balance the severity of suicidality and intensity of appropriate intervention services across the study conditions.

Data Analysis

Analysis plan

We expect that individuals receiving the culturally tailored text messages will have better outcomes than those in the treatment-as-usual control arm. Chi-square and t tests will be used to compare the overall percentages (binary outcomes) and mean differences (continuous outcomes) between study arms. We will conduct a sensitivity analysis using logistic or linear regression to adjust for baseline factors that are imbalanced between the intervention and control arms. Results will be presented as risk ratios and risk differences for binary outcomes, extracted from logistic regression models using marginal standardization, with 95% CIs [44,45]. Primary analyses will use an intention-to-treat approach.

Missing Data

Our primary analysis will rely on complete case analysis. We will conduct sensitivity analyses with multiple imputation methods. Multiple imputation through chained equations will be applied when data are missing at random. We will also conduct sensitivity analyses to exclude individuals who were randomized to receive the intervention but did not receive it per the schedule outlined in Table 2 due to technological errors or temporary lapses in access to cell phone service. These additional sensitivity analyses will improve confidence in our findings.

Power and Sample Size

We plan to recruit 350 people across both sites. We assume the study will have 70% retention at 12 months, yielding an analytic sample of 240 people (120 per group). With this sample, we will have 80% power to detect a 16% difference (Risk Ratio=0.47) in suicidal ideation between intervention and TAU groups, assuming that 30% of the TAU group reports suicidal ideation at follow-up. We will also be able to detect an 11% difference (Risk Ratio=0.25) in suicide attempts if 15% of the TAU group participants attempt suicide within the 12-month follow-up. There is variability in the literature regarding what constitutes a clinically meaningful effect size for reductions in suicidality and the best means of measuring such an effect (eg, degree of suicidality, reported number of attempts, etc). Regardless, we are confident in our ability to recruit a sizeable analytic sample in a difficult-to-reach population.

Ethical Considerations

All study procedures—including consent protocol, data collection tools, and intervention materials—and subsequent modifications were reviewed and approved by the Institutional Review Boards of Washington State University (#17435), University of New Mexico (#19-401), and University of Colorado Denver (#18-0186), in addition to the Great Plains IRB (#23-R-09GP). The trial is registered on the United States National Institutes of Health Clinical Trials Registry (ClinicalTrials.gov NCT03136094). To ensure informed and voluntary participation, the study staff privately obtains consent from individuals, unless they request otherwise. The coordinator verbally explains the purpose, procedures, risks, benefits, compensation, withdrawal rights, and confidentiality of the study. While the consent form includes permission to be contacted by text message, the form does not make the nature of the intervention explicit. Potential participants are told that the clinic is testing a mobile phone program to support people who are having a difficult time, and that they will receive periodic text messages over the next 12 months. The consent process omits explanations of when and how often text messages will be sent. If participants ask for more details, they are informed that the text messages will be “periodic” and that their purpose is to check in. Keeping the participant blinded during the consent process is an important methodological feature of this study to minimize biased assessment of outcomes and maximize the validity of the results. Participants are informed that their participation is voluntary, and they can withdraw at any time without consequences. Participants at the FNCH site were compensated US $75 upon completion of the baseline visit and received an additional US $75 upon completion of the follow-up visit. Participants recruited from the OHC site were compensated US $100 upon completion of the baseline visit and received an additional US $75 upon completion of the follow-up interview.

Results

Overview

Participants are still being enrolled at the site in South Dakota; the site in New Mexico has concluded enrollment but is continuing with follow-up assessments. We expect to complete data collection by the Spring of 2025.

Participant Reception

Thus far, participants have responded positively to the messages. While the messages are designed such that they do not demand anything of participants nor require a response, participants do occasionally respond. Most of these responses have been positive. Some responses are brief, with an expression of gratitude or the use of positive emojis. Other times, participants express that a particular message strongly resonated with them. Some of the messages received from participants in response to the caring texts intervention are included in Textbox 1. Names of site staff and site locations have been removed to preserve participant anonymity.

Discussion

Principal Findings

This study represents an opportunity to prevent or reduce suicidality among AI young adults in urban areas across the country [9-11,15,28]. We anticipate finding that those individuals who received the caring text messages will show reduced indications of suicidality (primary outcome) and increased feelings of social connectedness (secondary outcomes) compared to those in the TAU cohort. This is because the caring texts intervention has strong theoretical foundations [9,10] and is supported by evidence-based practice guidelines [11,15,28].

In addition to being theoretically grounded and supported by other evidence-based practices, we have endeavored to develop an intervention that could be implemented at scale with limited need for involvement on the part of behavioral health specialists on a day-to-day basis. The intervention is structured and delivered in such a way that it can be integrated into primary care, emergency departments, or other services not explicitly designed to manage mental health care. There is mounting support for integrating behavioral health services into primary care, which can reduce stigma and improve access to limited behavioral health resources [46-51]. This is particularly pertinent for young adults who may be trying to access care from an already overburdened IHS.

Dissemination Plan

We believe that disseminating findings back to the community is a crucial component of community-based participatory research. To this end, we have planned several opportunities for dissemination within the scientific community and among urban AI communities more broadly. The findings from the study will be published in an academic journal following the conclusion of data collection and analysis. In addition, we have participated and will continue to participate in health fairs hosted near both study sites. Previous attendance has included sharing information regarding this trial and other behavioral health resources. Future presentations will include sharing the results of the completed trial back to the community. In addition to these health fairs, the staff at the South Dakota site routinely participates in local radio shows to share information on various health topics and ongoing research projects. They plan to dedicate one segment to discussing the findings of the urban caring text messages study.

Conclusions

The intervention was culturally adapted by members of each participating urban native community. The inclusion of different constituents, such as patients at risk of suicide and behavioral health providers, in the cocreation of the intervention ensures the relevance and appropriateness of the intervention’s content, including the text messages. The text messages are designed to be culturally appropriate and increase social support and connectedness without imposing demands on the recipient. By studying whether caring texts help to reduce suicidal ideation and behaviors, and improve social connectedness as well as connection to Indigenous identity, this trial will provide evidence to inform whether primary care clinics serving AIAN young adult should integrate caring texts into their services.