Prevalence of Moderate and Acute Suicidal Ideation among a National Sample of Tribal College and University Students 2014-2015

Abstract

Objectives.

To examine the relationship between suicide risk and hazardous drinking, depression, and anxiety, adjusting for demographics, among tribal college students across the United States.

Methods.

We invited tribal college students enrolled in 22 tribal colleges from fall 2014 and 2015 to participate in the Creating Campus Change study, a cross-sectional online / paper survey assessing alcohol use patterns and mental health outcomes. 3,239 students participated in the survey, yielding a response rate of 31.3%. We assessed alcohol use, depression, and general anxiety, along with demographic characteristics. We used the Mini-International Neuropsychiatric Interview to assess suicide risk.

Results.

8.5% indicated moderate or high suicide risk. In the final adjusted model, moderate/high depression was significantly associated with moderate/high suicide risk (OR=6.64; 3.91 – 11.28, p<0.001), as was moderate/high general anxiety (OR=2.80; 1.58 – 4.97, p<0.001), and moderate/high hazardous drinking (OR=2.09; 1.19 – 3.66, p<0.001).

Conclusions.

Students attending tribal colleges who report moderate/high levels of depression, anxiety, or hazardous drinking have a greater risk of suicidality. Identifying factors buffering the risk of suicidality could support policy changes necessary to address this critical public health issue.

Introduction

Suicide and suicide risk represent critical public health issues in the United States, with suicide death rates rising nearly 30% in the overall population since 1999 (from 12.3 per 100,000 population to 15.6 per 100,000) (Stone et al., 2018). Among American Indian and Alaska Native (AI/AN) individuals, the age-adjusted suicide death rate was 20.7 per 100,000 population in 2015, compared to 15.0 among Whites (non-Hispanic) (Centers for Disease Control and Prevention, 2019). Certain individuals, depending on personal characteristics or region, experience higher levels of suicide risk. For example, AI/AN with mental health conditions have a two-fold higher odds of suicide completion compared to AI/AN without a mental health condition (Stone et al., 2018). In a 1999-2009 national sample, the age-adjusted death rate due to suicide for AI/AN under twenty-five years old in the Alaska region was 42.7, compared to 7.6 among Whites, per 100,000 population (RR = 5.65; 95% CI = 4.56, 7.03), and 21.3, compared to 5.3 among Whites, per 100,000 population in the Northern Plains region (RR = 4.02; 95% CI = 3.54, 4.54) (Herne, Bartholomew, & Weahkee, 2014). AI/AN mortality rates from suicide have stayed consistently higher than other races, with the national rate at 34.7 per 100,000 population for AIAN men (as compared to 23.2 per 100,000 population among White men) and 8.7 for AIAN women (as compared to 5.9 per 100,000 population for White women) from 1990-2009 (Espey et al., 2014). Clusters of suicide attempts in Pine Ridge, South Dakota (Bosman, 2015), Yu’pik communities in Alaska (Allen, Rasmus, Fok, Charles, & Henry, 2017), and Attawapiskat First Nation, Ontario, Canada (BBC, 2017), have highlighted the severe, critical nature of suicide risk in AI/AN communities.

Making sense of suicide risk remains a complex inquiry in many communities, with competing theoretical basis for heightened suicide risk including Durkheim’s sociological theory of suicide, Baumeister’s Escape Model, Schneidman’s psychache theory, Joiner’s Interpersonal Theory of Suicide, and Wenzel & Beck’s Cognitive Model of Suicide (O’Keefe, Tucker, Cole, Hollingsworth, & Wingate, 2018). Researchers have noted that these theories all include elements relevant in explaining AI/AN suicide risk, yet maintain culturally-based models offer a more comprehensive, context-driven theoretical approach to understanding suicide risk in Indian Country (O’Keefe et al., 2018). In particular, Wong’s Racial-Cultural Framework, which includes the processes of acculturation and enculturation, related stress, and the socio-cultural environment within which individuals operate, as well as cultural understandings and perspectives of suicide, aligns with existing research within AI/AN communities demonstrating a relationship between heightened suicide risk and discrimination, acculturative stress, historical trauma, poverty, and exposure to microaggressions (O’Keefe et al., 2018), yet the existing literature offers limited examples of research explicitly confirming these theories. Our approach relies on the Indigenous Stress-Coping Model (Walters, Simoni, & Evans-Campbell, 2002), which includes the AI/AN-specific stressors of historical trauma, lifetime trauma, and discrimination, and which offers important applications within the primarily reservation-based, geographically isolated communities within which TCU are located.

For nearly 200 years, since the beginning of the Reservation Era in 1851, AI/AN have experienced extreme poverty, with poor or nonexistent health care, despite guarantees through United States treaties, and ways of life destroyed or recovering as a result of the previous century of abuse and disenfranchisement (Evans-Campbell, 2008). As a result of these historical policies and experiences, AI/AN experience a myriad of health inequities, including significant factors related to suicide risk, such as post-traumatic stress disorder, anxiety, depression, and substance abuse/dependence (LeMaster, Beals, Novins, Manson, & AI-SUPERPFP Team, 2004). In the American Indian Services Utilization, Psychiatric Epidemiology, and Risk/Protective Factor Project (AI-SUPERPFP) study, participants meeting lifetime criteria for depression had 5.5 times higher odds of suicidal ideation, and those with a substance abuse or dependence disorder had a 2.3 times higher odds (LeMaster et al., 2004). Additional analyses demonstrated a difference in risk for suicidal ideation by region, with participants from a northern tribal community with depression experiencing 9.1 times higher odds of suicidal ideation, and those from a southern tribal community with depression experiencing 5.8 times higher odds (Bolton et al., 2014). Similarly, participants from a northern tribe with anxiety experienced 3.2 times higher odds of suicidal ideation and those with anxiety from the southern tribe had 4.5 times higher odds of suicidal ideation (Bolton et al., 2014).

Tribal Colleges and Universities (TCU) offer an important opportunity to promote educational progress, provide centers for future growth, and have potential to promote wellness in tribal communities throughout the nation (Duran, Magarati, Parker, Egashira, & Kipp, 2013). Developed during the 1970s self-determination movement, TCU provide higher education opportunities for geographically isolated, often times reservation-based, AI/AN students, ensuring students can maintain familial and cultural ties with their home communities (Rousey & Longie, 2001). In fall 2016, there were 16,820 certificate, diploma, and degree-seeking students enrolled in 34 reporting TCU (U.S. Department of Education, 2016). The majority of students are female (62.4 percent), AIAN (77.8 percent), and enrolled full-time (59.4 percent), with 16.5 percent of students self-identifying as White. Most TCU are tribally-chartered, offer 2-year Associates degrees, and all provide personalized attention to students to support them in overcoming the economic and social barriers they face in the educational process (American Indian Higher Education Consortium, 2018).

However, to date, there is little or no data on prior suicidal ideation, suicide attempts or current suicide risk among TCU students. While TCU students may appear to be high achievers with lower risk of suicide compared to their AI/AN peers outside TCU settings, our research suggests these students still experience two times the average suicide risk as compared to non-Indian college student populations. The relative lack of information from national datasets including factors predicting suicide risk in AI/AN populations presents challenges for tribes, practitioners, and researchers in developing, planning, and implementing programs that are informed by reliable data. Understanding the scope of the problem and factors influencing suicide risk among TCU students is thus an important first step in addressing this threat to the health and wellbeing of students. Moreover, confirming factors that are associated with lower suicide risk also remains an important priority. Given the many factors that contribute to suicide risk in Indian Country (Olsen & Wahab, 2006), a more nuanced understanding of malleable characteristics that are related to well-being within these communities would offer critical opportunities to continue to strengthen and support TCU students as well as other community members.

The Tribal Colleges and Universities Alcohol, Drugs, and Mental Health Epidemiological Study (TCU-ADME) study resulted from a 5-year research partnership between the American Indian Higher Education Consortium (AIHEC), 22 Tribal Colleges and Universities (TCU), and 3 UW research centers, focused on alcohol, drug and mental health risk and resiliency research. The primary aim is to determine the prevalence and correlates of psychiatric disorders among the TCU student population. The purpose of this study is to examine past-month prevalence of suicide risk, and the relationship between suicide risk and hazardous alcohol use, depression and anxiety, adjusting for demographic characteristics. These data comprise one of the first attempts to examine the potential relationships among the personal and environmental variables thought to contribute to risk for suicide among Tribal College students, and are derived from an administrative supplement to the parent study, which was sought in conjunction with the higher proportion of TCU students experiencing high suicide risk as compared to other studies of non-Indian college students. To begin examining the constructs within the Indigenous Stress Coping Model, we began by seeking to establish the relationships between other mood and behavior issues and suicide risk among TCU students. We hypothesized that higher self-reports of depressed mood, anxiety, and hazardous alcohol use would be associated with higher risk for suicide.

Method

The current study was part of the larger psychiatric epidemiologic survey completed through the TCU-ADME, a national study using community-based participatory research (CBPR) principles as a framework to inform study implementation (Israel, Schulz, Parker, & Becker, 1998). Based on Indigenous research principles (Parker, Pearson, Donald, & Fisher, 2019), the PI and study staff stressed the optional nature of the research, and confirmed that the research would proceed only with the consent of the TCU president and other local tribal research approvals. Of the 32 accredited TCU, 22 TCU chose to participate based on a series of presentations and discussions with TCU presidents and staff through the quarterly AIHEC Board Meetings. Colleges represented in the sample included seven from the Woodlands, seven TCU from the Montana/Pacific, four from the Dakotas, and four from the Southwest region. Those colleges opting out cited limited resources or no interest in participating in research, perhaps related to distrust of research due to historical harms perpetuated in AI/AN communities by researchers (Harding et al., 2012). Participating TCU Presidents reviewed the proposed measures for the study and provided input to address gaps in the data collection approach, stressing the need to move away from problematizing TCU students and toward understanding unique TCU and cultural contributors to student health and well-being. TCU agreed suicide risk remained a community-wide concern, and agreed that additional research to uniquely addresses the lack of information about suicide risk in the TCU population would provide insight for TCU programming and curriculum moving forward.

Procedure

Data and Sample

The sample came from 22 predominately reservation-based tribal colleges across the United States, ranging in size between 89 to 2,074 enrolled students. The TCU registrars provided student names and contact information. The research team followed approved protocols to ensure project compliance with Institutional Review Board, Tribal Research Review Board, and FERPA guidelines. A total of 10,357 students who met the recruitment criteria (the requirement that the registrar’s list would contain either student email address or both a mailing address and phone number, and enrolled either part- or full-time at the time the list was obtained from the registrar’s office) were invited to take the online survey.

All invited participants received an initial letter and e-mail inviting their participation, in addition to a unique Personal Identification Number (PIN) to enable them to login to the secure server to complete the online survey. Non-responders received reminder contacts designed to increase recruitment, including a paper survey if participants were unable to be reached via email. All participants who submitted a survey received a $40 incentive. Study participation inclusion criteria included: (1) Participants were 18 or older, (2) enrolled full or part time in the TCU at the time of obtaining student lists, and, (3) consented to participate.

The survey recruitment and data collection were conducted in two waves over 11 months between March 2015 and February 2016, yielding 3,239 completed surveys with an overall 31.3% response rate (ranging between 13% and 48% across TCU).

Measures

Dependent Variable

M.I.N.I. International Neuropsychiatric Interview (M.I.N.I.) Suicidality Module

For the present study, suicide risk was assessed with the suicidality module of the M.I.N.I. (Sheehan et al., 1997). Currently, no standard online suicide screening instrument exists as a gold standard for the AI/AN population. The M.I.N.I. is a short, diagnostic structured interview developed jointly by psychiatrists and clinicians in the United States and Europe for DSM-IV and ICD 10. If participants indicated, “Yes,” to the question, “In the past month did you suffer any accident,” they were given subsequent suicidality assessment questions such as, “In the past month did you think that you would be better off dead or wish you were dead,” and, “In the past month did you attempt suicide?” The M.I.N.I. score categories have been validated in other populations in the U.S. and Europe (Lecrubier et al., 1997; Mordal, Gundersen, & Bramness, 2010; Sheehan et al., 1997; Sheehan et al., 1998). Item scores were summed to yield a total risk score and were categorized as no (0), low (1-5), moderate (6-9), or high (10+) suicide risk, which were adjusted based on a version of the M.I.N.I. that includes sixteen items rather than the 10-item version used in other studies. For analyses, we created a dichotomous variable to indicate elevated suicide risk based on a score of 6 or higher (moderate to high risk for suicide = 1, which includes the two highest risk categories) versus a score of 5 or lower (low to no risk for suicide = 0, which includes the two categories representing “no” or “low” risk) based on results from a predictive validity study indicating likelihood ratios showed a four- to six-fold higher probability of self-harm attempts among patients with M.I.N.I. scores at the moderate and high risk levels compared with those with lower scores (Roaldset, Linaker, & Bjørkly, 2012). The research literature has limited reports on the internal consistency of the M.I.N.I., though one Norwegian study has reported a moderately high Cronbach’s alpha of 0.84 (Roaldset et al., 2012). For this sample, we confirm moderately high internal consistency for this instrument (α=0.80).

Independent Variables

Generalized Anxiety Disorder-7

To assess anxiety symptoms, participants completed the Generalized Anxiety Disorder-7 (GAD-7) (Spitzer, Kroenke, Williams, & Lowe, 2006). The GAD-7 is a brief, validated severity measure that evaluates presence of symptoms within the past two weeks with scores ranging from 0 to 21. Consistent with other research (Spitzer et al., 2006), we used a cutoff of 10 or higher to indicate moderate to severe generalized anxiety disorder for analysis (moderate to severe anxiety disorder = 1, and minimal to mild = 0). The internal consistency for the GAD-7 in the literature has been reported as high (α=0.92) (Spitzer et al., 2006), and the Cronbach’s alpha in this sample was consistent (α=0.94).

Patient Health Questionnaire-9

To assess the presence of depressive symptoms, participants completed the Patient Health Questionnaire (PHQ), a brief, validated measure of depression severity (Kroenke, Spitzer, & Williams, 2001). The PHQ is a 9-item questionnaire designed to evaluate the presence of depressive symptoms within the past two weeks, including the presence of depressed mood and decreased interest in pleasurable activities. As a severity measure, the PHQ-9 score can range from 0 to 27 and consistent with prior studies, a cutoff score of 10 or higher was used to indicate moderate to severe major depressive disorder (moderate to severe depression = 1, and minimal to mild = 0). This cutoff score has shown strong criterion validity against a standard of major depressive disorder in a number of populations (Manea, Gilbody, & McMillan, 2015). Note that for this analysis of suicide risk, question 9 in the PHQ-9 instrument, “Thoughts that you would be better off dead or of hurting yourself in some way,” was removed from the depression severity score to avoid multicollinearity. The internal consistency for the PHQ-9 in the literature has been reported as high among college students, ranging from 0.86 to 0.93 (Keum, Miller, & Inkelas, 2018), and the Cronbach’s alpha in this TCU sample was consistent with these findings (α=0.91).

Alcohol Use Disorders Identification Test

The consumption items of the Alcohol Use Disorders Identification Test (AUDIT-C) is a 3-item screening tool, which was derived from the 10-item AUDIT originally developed by the World Health Organization to assess alcohol consumption, drinking behaviors, and alcohol-related problems (Bush, Kivlahan, McDonell, Fihn, & Bradley, 1998). Responses are scored according to their frequency over the past month or on a typical day, with scores range from 0 to 12. A cutoff score of 4 or higher was used for men and a score of 3 or higher was used for women to indicate hazardous drinking or active alcohol use disorders requiring advice and brief counseling or evaluation and treatment (hazardous drinking or active alcohol use disorders = 1, below recommended drinking limits = 0). While other research has validated a cut off score on the full AUDIT of between 8-6 as having the necessary sensitivity and specificity to best assess unhealthy alcohol use (Allen, Litten, Fertig, & Babor, 1997; Babor, Higgins-Biddle, Saunders, & Monteiro, 2001; Johnson, Lee, Vinson, & Seale, 2013), we chose to examine suicide risk among those students exhibiting harmful to severe levels of alcohol use using the AUDIT-C that indicate the need for brief intervention or referral to treatment.

Demographic variables assessed included biological sex (female = 1, male = 0), dummy variables for race (AIAN only [reference], AIAN with another race, white/other), household income as an ordinal variable (<$5,000 = 0; $5,000-14,999 = 1; $15,000-24,999 = 2; $25,000-49,999 = 3; $50,000-74,999 = 4; $75,000+ = 5), and age as an ordinal variable (18-20 = 0; 21-23 = 1; 24-29 = 2; 30-39 = 3; 40-49 = 4; 50+ = 5). In additional analyses, we explored coding income and age as dummy variables, but the pattern was generally consistent with a linear relationship and thus we used the ordinal form of the variables for parsimony.

Analytic Plan

Logistic regression models estimated odds ratios (ORs) for the association between participant characteristics and elevated suicide risk. An initial set of models was performed that included each covariate in the model separately. Five additional models that included covariates were then examined. Model 1 included all demographic characteristics together. The next models included all demographic characteristics and depression (Model 2), anxiety (Model 3), and hazardous drinking (Model 4) separately. Model 5 included all covariates together in a single model.

To address possible bias due to differential non-response by participant characteristics and to account for the sampling of students from the TCUs, we calculated raking weights. These raking weights accounted for the probability sampling weights as well as additional information about the distribution of demographic characteristics at each school: age, sex, and Native self-identification. Applying these weights allowed the study sample to be more representative of the broader TCU student population. We applied these inverse sampling weights and also adjusted standard errors to account for clustering of students within the 22-school stratum in all analyses including descriptive statistics and regression models using survey commands in Stata 14.0 (StataCorp, College Station, TX).

Results

Table 1 shows weighted demographic and mental health characteristics of the participants in this study. The majority of participants were women (62.6 percent), AI/AN (73.6 percent), and younger than 30 years old (62.2 percent). Over 40 percent of participants reported an annual household income of $15,000 or less. In this sample, 63.5 percent had no to minimal depressive disorder as measured by the PHQ-9, with 21.6 percent reporting mild depressive disorder symptoms, 8.2 percent with moderate depressive disorder symptoms, and 2.9 percent reporting moderately severe depressive disorder symptoms. Nearly 66 percent had none to minimal general anxiety disorder symptoms as measured by the GAD-7. About 21.2 percent reported mild anxiety symptoms, 7.6 percent reported moderate anxiety symptoms, and 5.4 percent reported moderately severe anxiety symptoms. Over 77 percent of participants reported low risk alcohol use, with about 16 percent reporting moderate alcohol risk use who would benefit from simple advice, 3.1 percent reporting hazardous alcohol use who would benefit from simple advice and brief alcohol counseling, with 3.6 reporting severe alcohol use, requiring referral and treatment for alcohol use disorder. About 8.5% of participants reported symptoms that led to a suicide risk score of 6 or higher on the M.I.N.I., indicating a moderate to severe risk for suicide.

Table 1.

Demographic and Mental Health Characteristics of Tribal Colleges and Universities Student Participants (N=3,212)

Birth Sex		Percentage
	Male	37.4
	Women	62.6
Race
	AIAN	73.6
	AIAN Multiracial	14.9
	White/Other	11.5
Household Income
	<$5,000	21.0
	$5,000-15	20.6
	$15,000-2	15.2
	$25,000-5	23.1
	$50,000-7	11.9
	$75,000+	8.3
Age
	18-20	19.7
	21-23	18.6
	24-29	23.9
	30-39	19.9
	40-49	10.1
	50+	7.7
Suicide Risk
	No Risk - Low Risk (M.I.N.I. = 0-5)	91.5
	Moderate Risk – High Risk (M.I.N.I >=6)	8.5
Depression Severity
	None to minimal depression (PHQ-9 = 0-4)	63.5
	Mild depression (PHQ-9 = 5-9)	21.6
	Moderate depression (PHQ-9 = 10-14)	8.2
	Moderately severe depression (PHQ-9 ≥ 15)	2.9
Anxiety Severity
	None to minimal anxiety (GAD-7 = 0-4)	65.9
	Mild anxiety (GAD-7 = 5-9)	21.2
	Moderate anxiety (GAD-7 = 10-14)	7.6
	Moderately severe anxiety (GAD-7 ≥ 15)	5.4
Alcohol Use (AUDIT)
	Low risk / Alcohol education (AUDIT = 0-7)	77.3
	Moderate risk / Simple advice (AUDIT = 8-15)	16.0
	Hazardous alcohol use / Simple advice and brief counseling (AUDIT = 16-19)	3.1
	Severe Alcohol Use / Referral and treatment (AUDIT ≥ 16)	3.6

Table 2 shows bivariate associations between participant characteristics and suicide risk. Older age (OR = 0.84; 95% CI = 0.74, 0.95, p = 0.005) and greater household income (OR = 0.89; 95% CI = 0.80, 0.98, p = 0.023) were associated with lower suicide risk. Those reporting AI/AN in combination with another race had a higher suicide risk (OR = 1.81; 95% CI = 1.26, 2.62, p<0.001) than those reporting AI/AN alone (ref.). Identifying as White or some other race was not significantly associated with increased risk of suicide in this sample. There was no statistically significant difference in suicide risk between males and females. Moderate to moderately severe depression (OR = 18.27; 95% CI = 12.77, 26.13, p<0.001) was significantly associated with heightened risk of suicide, as was moderate to moderately severe anxiety (OR = 8.61; 95% CI = 6.00, 12.36, p<0.001). Hazardous to severe alcohol use (OR = 2.93; 95% CI = 1.80, 4.77, p<0.001) was also associated with elevated suicide risk.

Table 2.

Unadjusted Odds Ratios Predicting Suicide Risk from Demographic Characteristics

Factor	Unadjusted OR (95% CI)	p-value
Birth Sex	1.16 (0.80 – 1.66)	0.433
Race
American Indian	ref
American Indian Multiracial	1.81 (1.26 – 2.62)	<0.001
White/Other	1.30 (0.73 – 2.32)	0.371
Income	0.89 (0.80 – 0.98)	0.023
Age	0.84 (0.74 – 0.95)	0.005
Depression	18.27 (12.77 – 26.13)	<0.001
Anxiety	8.61 (6.00 – 12.36)	<0.001
Hazardous to severe alcohol use	2.93 (1.80 – 4.77)	<0.001

These bivariate associations generally remained after adjusting for the other covariates (Table 3). When examining demographic covariates together (Model 1), age (OR = 0.84; 95% CI = 0.74, 0.95, p=0.005) and income (OR = 0.86; 95% CI = 0.76, 0.96), p=0.009) continued to be associated with lower suicide risk and those reporting AI/AN and another race (OR = 2.03; 95% CI = 1.36, 3.01, p <0.001) continued to have higher suicide risk than individuals identifying as AIAN alone. After adjusting for the demographic covariates, associations with elevated suicide risk remained strong in Model 2 for moderate to moderately severe depression (OR = 17.65; 95% CI = 12.09, 25.75, p <0.001). In addition, moderate to moderately severe anxiety (OR = 8.47; 95% CI = 5.72,12.52, p-value <0.001) remained significantly associated with heightened risk for suicide, on average, after adjusting for demographic variables. Model 4 demonstrates that hazardous to severe alcohol use (OR = 3.64; 95% CI = 2.20, 6.01, p<0.001) remained significantly associated with high risk of suicide. When we controlled for demographic characteristics and other mental health outcomes (Model 5), the ORs for depression, anxiety, and harmful drinking were attenuated, but all remained strongly associated with suicide risk (OR>3) and statistically significant (p ≤ 0.010).

Table 3.

Adjusted Odds Ratios Predicting Suicide Risk from Demographic Characteristics

	Model 1 N = 3,101		Model 2 N = 2,922		Model 3 N = 2,856		Model 4 N = 3,099		Model 5 N = 2,760
Covariate	Adjusted OR (95% CI)	p-value	Adjusted OR (95% CI)	p-value	Adjusted OR (95% CI)	p-value	Adjusted OR (95% CI)	p-value	Adjusted OR (95% CI)	p-value
Birth Sex	1.17 (0.80 – 1.70)	0.409	1.13 (0.75 – 1.69)	0.569	1.01 (0.68 – 1.50)	0.947	1.32 (0.91 – 1.92)	0.148	1.16 (0.77 – 1.76)	0.471
Race
AIAN alone	ref		ref		ref		ref		ref
AIAN Multiracial	2.03 (1.36 – 3.01)	<0.001	1.96 (1.22 – 3.13)	0.005	1.85 (1.20 – 2.86)	0.005	2.19 (1.48 – 3.23)	<0.001	1.97 (1.21 – 3.20)	0.006
White/Other	1.77 (0.94 – 3.34)	0.076	1.56 (0.79 – 3.08)	0.200	1.56 (0.82 – 2.98)	0.177	2.08 (1.11 – 3.89)	0.023	1.72 (0.88 – 3.36)	0.112
Income	0.86 (0.76 – 0.96)	0.009	0.89 (0.78 – 1.01)	0.075	0.89 (0.79 – 1.01)	0.072	0.85 (0.76 – 0.95)	0.005	0.89 (0.78 – 1.02)	0.098
Age	0.84 (0.74 –0.95)	0.005	0.81 (0.70 – 0.93)	0.003	0.83 (0.72 – 0.94)	0.005	0.82 (0.72 – 0.93)	0.002	0.79 (0.69 – 0.92)	0.002
Depression			17.65 (12.09 – 25.75)	<0.001					11.41 (6.89 – 18.89)	<0.001
Anxiety					8.47 (5.72 – 12.52)	<0.001			2.12 (1.23 – 3.66)	0.007
Hazardous to Severe Alcohol Use							3.64 (2.20 – 6.01)	<0.001	2.02 (1.13 – 3.64)	0.019

Discussion

The current study assessed prevalence and risk factors for suicide risk among a large, national sample of TCU students. We hypothesized depression, anxiety, and harmful alcohol use would be associated with greater suicide risk based on self-reported ideation and attempts in the past month. We further evaluated the extent to which sociodemographic factors including age, race/ethnicity, gender, and income were associated with suicide risk.

Results indicate 8.5% of TCU students in this sample experienced moderate to severe suicide risk, demonstrating a significant need for prevention and treatment of suicide risk in this population. Results suggest a significantly higher level of need for services among TCU students in comparison to the general college student population, as literature from national college samples indicates 0.9% to 1.2% of college students have a plan and make a suicide attempt (Brener, Hassan, & Barrios, 1999; Taliaferro & Muehlenkamp, 2015; Wilcox et al., 2010). This is compared to a weighted prevalence of 1.8% and 0.8%, respectively, in our sample. Though these rates are not directly comparable due to differences in measures and sample demographics, it is nonetheless important to address the significant need for suicide prevention services for TCU students.

Importantly, as suggested among other student samples (Heaton, 2018; Subica & Wu, 2018), TCU students identifying as multi-racial, specifically AI/AN and some other race, exhibited higher suicide risk as compared to those identifying as AI/AN only, suggesting that identity, cultural connectedness, or other AI/AN-specific characteristics may confer protection from suicide risk relative to multiracial students, or perhaps that risk levels rise among multi-racial AI/AN students with potential negative experiences or perceptions of multi-racial individuals. The heightened risk seen among these multi-racial students aligns with findings of heightened risk of mental health problems and substance use disorders seen among other multi-racial student groups (Larimer & Dillworth, 2008). Other studies have examined suicide risk levels in relation to acculturation, finding that AI/AN tribal communities that have experienced higher levels of acculturation experiencing higher levels of suicide risk (Olson & Wahab, 2006). This finding requires further research to better understand cultural identity as a protective factor in these populations. With the changing demographics among AI/AN populations generally (52% of AI/AN identify as AI/AN and some other race) (Census 2010, 2015), closely examining the experiences and related risk among AI/AN multi-racial students and community members constitutes an important consideration.

As expected, depressed mood was significantly associated with higher suicide risk. On average, the odds of scoring within the high-risk category for suicide risk for TCU students reporting moderate to high depression were about 17.5 times the odds for students reporting no to mild depression, even after adjusting for demographic characteristics. The magnitude of the relationship between depressed mood and suicide risk demonstrates an urgent need to address depression among TCU students. Research in the general population and with students from mainstream universities indicates over half of those diagnosed with depression never receive any treatment, with lack of help seeking at 67 percent and 75 percent for Latinx and African American students, respectively (Lipson, Kern, Eisenberg, & Breland-Noble, 2018). This may be particularly true in AIAN populations, given increased barriers to treatment including financial barriers, discrimination in health care settings, lack of culturally appropriate treatment approaches, and barriers associated with the rural location of many TCU (Broffman et al., 2017; Duran et al., 2005; Rieckmann et al., 2012).

Generalized anxiety symptoms were also related to increased suicide risk among TCU students. After adjusting for demographic characteristics, the odds of scoring within the high-risk category for suicide risk for TCU students reporting moderate to severe anxiety were about 8.5 times the odds for students reporting no to mild anxiety, on average. Again, this demonstrates a need to address anxiety among TCU students in order to reduce suicide risk as well as improve quality of life overall.

Harmful drinking reported on the AUDIT was also related to increased suicide risk. On average, the odds of scoring within the high-risk category for suicide risk for TCU students reporting harmful drinking in the past year were about 3.6 times the odds for students reporting low to mild risk for hazardous drinking, after adjusting for demographic characteristics. College students in the general population have been shown to be at increased risk for harmful alcohol use (Johnston et al., 2018), and a variety of individual and environmental strategies have been evaluated and found to be efficacious for reducing harmful alcohol use on college campuses (Cronce & Larimer, 2011; Larimer & Cronce, 2002; NIAAA, 2015). However, the majority of this research has focused on majority White students at 4-year universities, and relatively few evidence-based approaches have been implemented or tested in community or technical colleges nor TCU or other minority-serving institutions. Cultural adaptation of efficacious approaches for implementation in TCU contexts, and evaluation of outcomes in this setting, are important priorities to improve health and wellbeing of TCU students and reduce suicide risk.

This study is limited to TCU students, a sub-group which may not be reflective of AI/AN community members overall. TCU students may have lower risk profiles, given their ability to matriculate high school and navigate a college environment. Moreover, TCU students may have access to more resources, including knowledge about treatment options, which may improve overall access to care. TCU students may also experience fewer stressors, compared to AI/AN community members overall. However, despite these potential advantages, in our sample, TCU students experienced high levels of poverty, suggesting that a closer examination of stressors experienced by TCU students would provide important information on how to best support them in their journey through the TCU system. The data collection efforts represent another limitation of this study. We attempted to include all students from one academic year, but experienced delays due to the multiple institutional review boards and administrative approvals within our university and the TCU systems. As a result, the data includes non-duplicated TCU attendees over two academic years, which afforded us the opportunity to follow up over a longer time period, yet does not allow for comparison with other samples, as it includes a group of students over a 2-year period. In addition, this study was cross-sectional, which does not allow us to make conclusions regarding causality. Further, we relied on self-report, which was not verified through clinical interviews. This may also have had implications for the accuracy in measurement of behavioral health outcomes.

Understanding the risk factors TCU students face represents an essential first step in understanding why prevalence rates of moderate to severe suicide risk in this sample are 8 times those found in prior national studies of the general college population. The current analyses broaden the state of scientific knowledge, as well as suggest clinical approaches to assess and address suicide risk in Indian Country by informing the development and/or adaptation of suicide prevention approaches that can be targeted to individuals at high risk, can be tailored to significant risk factors, and can be culturally and structurally adapted to the context of the TCU. For example, depression and anxiety screening, along with alcohol use assessments among TCU students may be an approach to identify higher risk individuals for improved referrals to care. Moreover, telephone follow-up to further assess suicide risk may be an important option for effective, acceptable, and efficient clinical practice within AI/AN communities, particularly given the limited access to quality behavioral health care and rurality of many TCU communities.

An absence of behavioral health services, along with culturally incompetent care, also play a role in the higher rates of suicide and suicide risk across AI/AN populations. Tribes have taken a variety of approaches to reverse these trends. Using Community-Based and tribal Participatory Research approaches to develop a culturally-adapted intervention for youth, the University of Washington and the S’Klallam tribe developed the Healing of the Canoe project (Donovan et al., 2015). Relying on the metaphor of the Canoe Journey, a cultural activity rooted in Coast Salish cultures, Healing of the Canoe addresses the disparities in the AI/AN substance abuse resources and the negative consequences experienced through tribal communities.

Exclusion from normative social structures for those experiencing mental health issues, has been shown to alienate those affected form their community, reinforcing feeling of low self-worth (Farrelly et al., 2015). Lack of adequate services and coping mechanisms perpetuate social isolation, so recognizing traits and staging interventions can become crucial. Evidence based treatments such as Acceptance and Commitment Therapy (ACT), Dialectical Behavior Therapy (DBT), and Trauma and Grief Component Therapy for Adolescents have shown effectiveness in reducing suicidal behavior in adolescents (Kaplow, Gipson, Horwitz, Burch, & King, 2014). The risk factors of general anxiety disorder, depression, and high levels of alcohol use can help tailor evidenced based interventions in tribal communities, confirming that these mental health disorders should be explored with AI/AN patients. Some prior literature has suggested that mental health outcomes vary by region and tribe. This study helps clarify that this type of generalization may be useful for broader planning purposes, but among TCU students, potentially a more diverse, intertribal sub-group, anxiety, depression, and alcohol use are all important issues to address when determining the appropriate treatment approach.

The policy implications of the research will provide ample evidence for tribal governments and communities to prioritize action on suicidality on a collective scale. Increasing access to mental and behavioral health care, and leading efforts to destigmatize treatment are preliminary steps to address the public health crisis amount AI/AN communities. Culturally adapted interventions have shown promise in their ability to mitigate risk of suicidality, and adoption of such policies have the potential for widespread impact in tribal communities.

Public Health Implications

These findings support program planning and policy changes in TCU and tribal communities to address a major public health risk for AI/AN youth and young adults. A better understanding of how students who express suicidal risk differ from those who do not provides valuable information to provide resources to high risk students, develop approaches to address suicidality across TCU, which represent a more tribally-diverse group than other population-based mental health outcomes studies, and support addressing risk factors such as general anxiety disorder, depression, and high alcohol use in addition to overall suicide risk for a comprehensive approach to eliminate this major health disparity in a high-risk population. This upstream approach to the current research in the field of suicidality will yield important predictors and further delineate key mechanisms that can then be applied in novel, culturally congruent approaches to suicidality in high-risk settings, and could be adopted and adapted across tribes and AI/AN communities to address suicidality more appropriately for this population.