Correlates of Solitary Alcohol and Cannabis Use among American Indian Adolescents

Abstract

Background:

Solitary substance use, or using substances while alone, is common among adolescents but understudied. This is the first study to examine solitary substance use among American Indian (AI) adolescents. The objective was to examine correlates of solitary alcohol use and solitary cannabis use that occur within the individual, family, school, peer, and cultural domains of the social ecology.

Method:

Data were from the 2009–2013 Drug Use Among Young American Indians Study, a cross-sectional cohort study. Two sets of hierarchical logistic regressions were conducted to examine solitary alcohol use (getting drunk) among lifetime alcohol users (n = 2,082; M^age = 15.12 years; SD=1.68; 54.2% female) and solitary cannabis use among lifetime cannabis users (n = 2,085; M^age = 14.99 years; SD=1.69; 50.5% female), including adjustment for level of substance involvement.

Results:

Prevalence of solitary alcohol use among lifetime drinkers was 24.9%. Among lifetime cannabis users, 53.6% reported solitary cannabis use. Regression analyses for solitary alcohol use showed statistically significant positive associations with coping motive, descriptive norms, violent behavior, depression, peer models for use, and (unexpectedly) peer sanctions against use and a negative association with family sanctions against use. Regression analyses for solitary cannabis use showed statistically significant positive associations with coping motive, violent behavior, and peer models for use and a negative association with family sanctions against use.

Conclusions:

Solitary alcohol and cannabis use are prevalent among AI adolescents and might, in particular, reflect attempts to cope with adversity. Findings could help guide the development of screening and prevention efforts.

1.0. Introduction

Alcohol and cannabis are the most commonly used substances among adolescents in the United States (Keyes et al., 2019). During this developmental period, substance use tends to be a social activity that occurs in the company of friends (Henneberger et al., 2021). Less common in adolescence, although still prevalent, is solitary substance use, or the use of alcohol and cannabis while alone. For example, a recent systematic review reported that approximately 14% of US adolescents reported solitary alcohol use and/or cannabis use (Mason et al., 2020). The adverse consequences of adolescent substance use, regardless of context, are well known (e.g., disrupted brain development; Squeglia et al., 2009). Furthermore, there is evidence that solitary use of alcohol and cannabis may be related more strongly to adverse outcomes, particularly the development of substance-related problems and disorders, than social-only use of those substances (Mason et al., 2020; Skrzynski and Creswell, 2020). Thus, there remains a need to better understand the correlates of adolescent solitary alcohol and cannabis use with potential implications for interventions designed to prevent such use.

As reviewed by Skrzynski and Creswell (2020), most studies of solitary substance use have focused on drinking among young adults and have used samples of college students (e.g., Bilevicius et al., 2018). Fewer studies have been conducted with adolescents (Mason et al., 2020), particularly those from underrepresented racial/ethnic minority groups, such as American Indians (AIs). AI adolescents display higher levels of alcohol and cannabis use compared to non-AI adolescents. For example, a prior publication from the current study indicated that the lifetime prevalence of alcohol use and cannabis use among AI 8^th graders was 37% and 56%, respectively (Stanley et al., 2014). The corresponding prevalence was 15% and 16%, respectively, among non-American Indian 8^th graders in the nationally-representative Monitoring the Future data over the same time frame from 2009–2012. To date, there have been no published studies of solitary alcohol and cannabis use among AI youth.

In extant studies of non-AI young adults and adolescents, solitary substance use has been associated positively with negative emotions (e.g., depression) and using substances (e.g., drinking) to cope with adversity and distress (Mason et al., 2020; Skrzynski and Creswell, 2020). These findings are consistent with both past (Cooper, 1994) and recent (Creswell, 2021) theory that draws on a motivational model positing that solitary substance use is rooted in stress-coping and self-medication processes (Cox and Klinger, 1988, 2011). American Indians, in particular, are subject to stressors associated with poverty, individual discrimination, institutional racism, and historical trauma (Brave Heart and DeBruyn, 1998; Goodkind et al., 2010; Sarche and Spicer, 2008; Whitbeck et al., 2002). These stressors may explain, in part, high rates of substance use among AI adolescents as a coping mechanism. A recent investigation of unique latent classes of cannabis use among this group found that using to cope was a strong predictor of class membership (Swaim and Stanley, 2021). Since solitary substance use, especially, may reflect attempts to cope with adversity, AI adolescents might experience heightened risk for using alcohol and cannabis alone.

Other robust correlates of solitary substance use reported in the literature include early substance initiation and heavier use in samples of both young adults (Gonzalez and Skewes, 2013; Kingston et al., 2017) and adolescents (Creswell et al., 2014); however, studies typically have not considered a larger array of variables drawn from across the social ecology. In addition to individual-psychological correlates, social ecological theory (Bronfenbrenner and Morris, 1998; Lounsbury and Mitchell, 2009) draws attention to multifaceted influences on adolescent substance use that occur within the family, school, peer, and cultural domains. It is unknown how these correlates relate, specifically, to the case of AI adolescents using substances while alone. In a sample of non-AI adolescents, Tucker et al. (2006, 2014) found that solitary alcohol and solitary cannabis users compared to their social-only counterparts were more likely to associate with both friends and close adults who use drugs, reflecting possible peer and parent influences. Further studies are needed, examining individual (e.g., depression, coping, norms, violence, sensation seeking) as well as family (e.g., rules or sanctions against use, communication, monitoring), school (e.g., academic performance), peer (e.g., offers and models for use, rules or sanctions against use), and cultural (e.g., ethnic identification) variables that prior studies (Connell et al., 2010; Van Ryzin et al., 2012) and reviews (Hawkins et al. 1992; Trucco, 2020) have identified as robust correlates of adolescent substance use, in general, to examine their associations with solitary alcohol and cannabis use in AI youth.

Guided by the motivational model (Cox and Klinger, 1988, 2011) and social ecological theory (Bronfenbrenner and Morris, 1998; Lounsbury and Mitchell, 2009), the current study advances prior research by considering a broad array of factors across multiple domains of influence (individual as well as family, school, peer, and culture) as potential correlates of solitary versus social-only alcohol and cannabis use in a large sample of AI adolescents. A series of hierarchical analyses culminated in a full model that examined all of the social ecological correlates simultaneously to determine their adjusted associations with solitary substance use (see Figure 1), while controlling for sex, age, early-age substance onset, substance involvement, and cohort. In general, it was expected that correlates from each social ecological domain would significantly distinguish solitary versus social only alcohol and cannabis use, with particularly strong positive associations for depression and coping motives.

Figure 1.

Conceptual model of Social-Ecological Correlates of Solitary Substance Use

2.0. Methods

2.1. Participants and Procedures

This study conducted secondary analyses of existing cross-sectional cohort data collected under the Drug Use Among Young American Indians Study (Stanley and Swaim, 2015; Swaim and Stanley, 2018, 2021), an annual surveillance study of youth living on or near American Indian reservations in the US. This study used the data collected from annual in-school, self-report surveys between the years 2009 to 2013 on 3,498 students who identify as AI in Grades 7 to 12 in 5 regions (Northwest, Northern Plains, Upper Great Lakes, Southeast & Texas, Southwest). Because our objective was to study solitary alcohol/cannabis use among substance initiators, two analysis samples were created, one representing the subset of AI adolescent lifetime alcohol users (ever drink alcohol in any context) and another representing the subset of AI lifetime cannabis users (every tried cannabis in any context). AI lifetime alcohol users (n= 2,082 within 43 schools) had an average age of 15.12 years (SD=1.68) and were 54.2% female. AI lifetime cannabis users (n=2,085 within 45 schools) had an average age of 14.99 years (SD=1.69) and were 50.5% female. The University of Tennessee Health Science Center Institutional Review Board exempted this secondary analysis study of publicly available, de-identified data.

2.2. Measures

Substance use data were collected using the American Drug and Alcohol Survey (Oetting et al., 1985). Data on social ecological correlates were collected using the Prevention Planning Survey (Beauvais & Swaim, 2013). Internal consistency (Cronbach’s Alpha [α]) is reported below for all multi-item scales based on the full sample of 2009–2013 AI adolescents (n=3,498). Use of the forward slash (“/”) in the descriptions below indicates one set of questions for alcohol/one set for cannabis with the same question stem. All analyses adjusted for biological sex (male=1, female=2) and age.

2.2.1. Solitary Alcohol/Cannabis Use

Participants indicated if they had “Ever gotten drunk when alone” and “Ever used cannabis when alone” to indicate solitary alcohol use and solitary cannabis use, respectively. Both items have dichotomous responses (yes=1, no=0).

2.2.2. Alcohol/Cannabis Involvement and Early-Age Onset

Alcohol involvement was the sum of five items tapping into “last 12 months: number of times had alcohol,” “last 12 months: number of times gotten drunk,” “last month: number of times had alcohol,” “last month: number of times gotten drunk” (response options for these four items: 1=none to 6=50+ times), and “last 2 weeks: number of times 5+ in 2 hours” (response options: 1=1 to 10=10+ times). Items were standardized and then summed to create an overall scale (α = .92).

Cannabis involvement was the sum of three items tapping into the frequency of cannabis use in the last 12 months (response options: 1=none to 6=50+ times) and in the last month (response options: 1=none to 6=several times every day) as well as youth’s perceived level of cannabis use (response options: 1=non user to 6=very heavy user). Items were standardized and then summed to create an overall scale (α = .94).

Early-age alcohol/cannabis onset was measured by one item each: “Age first got drunk” and “Age first tried marijuana.” Responses (ages) were collapsed into a dichotomous variable coded 1=before age 15, and 0=15 and older.

2.2.3. Individual Domain

Coping motive for alcohol use was measured by the item “drinking alcohol helps me feel better,” and coping motive for cannabis use was measured by the item “using cannabis helps me feel better.” Responses were made on a 5-point Likert scale (1=strongly agree to 5=strongly disagree) that was reverse-coded for both items.

Descriptive norms were the sum of two items indicating the perceived number of times an average student gets drunk/uses cannabis in a month (response options: 1=never to 5=10+ times) and the perceived number of students who get drunk/use cannabis at least once a month (responses 1=none to 5=almost all). Items were standardized and then summed to create overall scales for descriptive norms of alcohol use (α =.81) and descriptive norms of cannabis use (α = .84).

Student injunctive norms was measured by one item each for alcohol and cannabis that asked students to indicate their agreement with the statement “Most students think it is wrong for students to get drunk/use cannabis” on a 5-point Likert scale (1=strongly agree to 5=strongly disagree).

Violent behavior was the sum of four items asking, “have you ever-beaten up someone,” “have you ever-scared someone with a weapon,” “have you ever taken a gun to school,” and “have you ever-hurt someone with a weapon” (response options: 1=never to 4=6+ times; α = .67).

Depression was measured by seven items (e.g., I feel low, I am unhappy) with responses made on a 4-point scale (1=no, 2=not much, 3=some, 1= a lot). Items were reverse coded and then summed (α =.92).

Sensation seeking was measured by three items including “I like to do dangerous things,” “I take chances,” and “I would like to learn to skydive” with responses made on a 4-point scale (1=no, 2=not much, 3=some, 4=a lot). Items were reverse coded and then summed (α =.58).

2.2.4. Family Domain

Family sanctions against alcohol use was measured by four items, including “family would care if you: drank alcohol,” “family would care if you: got drunk” and “family would try to stop you from: drinking alcohol,” “family would try to stop you from: getting drunk” with responses made on a 4-point scale (1=a lot, 2=some, 3=not much, 4=not at all). Items were reverse coded and then summed (α = .91). Family sanctions against cannabis use was measured by two items, including “family would care if you: used marijuana,” and “family would try to stop you from: using marijuana” with the same response options as family sanctions against alcohol use. Items were reverse coded and then summed to create an overall scale (α = .82).

Family communication about alcohol use was measured by two items, including “family discusses: dangers of drinking alcohol,” “family discusses: dangers of getting drunk” (response options: 1=very true, 2=mostly true, 3=somewhat true, 4=not at all true). Items were reverse coded and then summed (α = .98). Family communication about cannabis use was measured by one item “family discusses dangers of using cannabis” with the same options that were reverse coded.

Parental monitoring was measured by four items, including “parents allow me to go out as often as I want,” “parents let me go any place without asking,” “parents are less strict than other parents,” and “parents let me stay out as late as I want” with responses made on a 4-point scale (1=very true, 2=mostly true, 3=somewhat true, 4=not at all true). The items were summed to create an overall scale (α = .84).

2.2.5. School Domain

The school domain was measured by two items assessing school performance, including “what kind of grades do you get” and “what kind of student are you” (response options for both items: 1=very good, 2=good, 3=not too good, 4=poor). The items were reverse coded and then summed (α = .75).

2.2.6. Peer Domain

Peer offer to use alcohol was measured by one item “how often friends asked you to get drunk” with a 4-point scale (1=a lot, 2=some, 3=not much, 4=not at all) that was reverse coded. Peer offer to use cannabis was measured by one item “how often friends asked you to use cannabis” with a 4-point scale (1=very often, 2=some, 3=not very often, 4=not at all) that was reverse coded.

Peer models for alcohol use was measured by three items, including “number of friends who get drunk once in a while” and “number of friends who get drunk every weekend” (response options for these two items: 1=none, 2=one or two, 3=some of them, 4=most of them) as well as “number of friends that get drunk” (response options: 1=none, 2=a few, 3=most of them, 4=all of them). The items were standardized and then summed (α = .82). Peer models for cannabis use was measured by one item “number of friends that use cannabis” (response options: 1=none, 2=a few, 3=most of them, 4=all of them).

Peer sanctions against alcohol use was measured by one item “how much friends try to stop you from getting drunk.” Peer sanctions against cannabis use was measured by one item “how much friends try to stop you from using cannabis.” Both items were on the same response scale (1=a lot, 2=some, 3=not much, 4=not at all), which was reverse coded.

2.2.7. Cultural Domain

The cultural domain is represented by six items that assess the internalization of American Indian Identification, based on a scale developed by Beauvais (2006) and used in prior project research (Swaim and Stanley, 2019). Items include “current activities/traditions-American Indian,” “future activities/traditions-American Indian,” “family lives by American Indian way of life,” “you live by American Indian way of life,” “family a success in American Indian way of life,” “you will be a success in American Indian way of life” and responses were made on a 4-point scale (1=a lot, 2=some, 3=not much, 4=none). The items were reverse coded and then summed (α = .91).

2.3. Analytic strategy

Two sets of hierarchical logistic regressions were conducted in Mplus 8.6 (Muthén & Muthén, 1998–2017), one set for solitary alcohol use based on the sample of lifetime drinkers and one set for solitary cannabis use based on the sample of lifetime cannabis users. Each set incorporated six steps. In the first step, solitary alcohol use/cannabis use was regressed on the demographic variables (sex, age), the respective measures of early-age substance onset, the respective measures of substance involvement, and three dummy-coded variables to capture 2009–2013 cohorts. Subsequent steps sequentially added variables from the individual domain (Step 2), the family domain (Step 3), the school domain (Step 4), the peer domain (Step 5) and the cultural domain (Step 6) in a progression that moved from more proximal to more distal influences, culminating in a final model that included all potential predictors simultaneously.

All scales were standardized (i.e., subtract the mean and divide by the standard deviation) to facilitate comparison of coefficients across scales and domains. The interpretation with respect to the standardized variables is in terms of one standard deviation unit change (increase or decrease). A school indicator (ULOCATION) that does not include grade was specified as a clustering variable to account for the nesting of students within schools using Mplus’ Cluster option and Type=Complex. Although missingness was <10% for most predictors and was a maximum of 12%, listwise deletion would have resulted in a substantial loss of cases. In order to include cases with partially missing data, full information maximum likelihood estimation (FIML) was used (Schafer and Graham, 2002).

3.0. Results

Prevalence of solitary alcohol use among lifetime drinkers was 24.9%. Among lifetime cannabis users, 53.6% reported solitary cannabis use. Table 1 provides correlations among all variables for lifetime drinkers in the upper diagonal and lifetime cannabis users in the lower diagonal. Although not shown in the table, the correlation between the two outcome variables, solitary alcohol use and solitary cannabis use, was moderate (r = .44).

Table 1.

Correlation Matrix for American Indian Lifetime Adolescent Drinkers (n=2,082; Upper-diagonal) and Lifetime Cannabis Users (n=2,085; Lower-diagonal)

	1	2	3	4	5	6	7	8	9	10	11	12	13	14	15	16	17	18	19
1.Sex		−.07	.07	−.01	.02	.11	−.01	−.23	.12	−.15	.11	.02	.19	.09	.06	.08	.06	.09	−.03
2.Age	−.02		−.20	.12	.002	.20	.05	−.02	−.10	−.03	−.05	−.002	−.09	.09	.07	.13	−.07	.03	.04
3.EAO	.002	−.41		.34	.27	.09	.11	.19	.08	.06	−.13	−.06	−.09	−.14	.29	.25	−.11	.02	.28
4.SI	−.07	.09	.21		.40	.25	.21	.32	.06	.12	−.23	−.12	−.23	−.17	.39	.42	−.14	.03	.35
5.CM	−.05	−.02	.21	.57		.20	.20	.21	.23	.17	−.18	−.17	−.17	−.17	.30	.32	−.13	.04	.31
6.DN	.13	.18	.03	.32	.26		.24	.17	.11	.10	−.03	.01	−.07	−.002	.27	.44	−.17	.11	.17
7.SIN	.02	.05	.10	.31	.29	.32		.12	.08	.10	−.15	−.16	−.11	−.15	.14	.25	−.21	−.03	.15
8.VB	−.22	−.01	.15	.30	.25	.11	.09		.14	.24	−.18	−.09	−.23	−.16	.21	.27	−.05	.07	.26
9.D	.13	−.08	.07	.07	.18	.07	.06	.13		.22	−.06	−.06	−.04	−.10	.09	.13	−.03	.07	.16
10.SS	−.11	−.05	.08	.20	.24	.12	.14	.22	.22		.01	.02	−.14	−.01	.12	.10	−.01	.12	.14
11.FS	.11	−.06	−.12	−.45	−.31	−.12	−.22	−.19	−.08	−.05		.36	.26	.12	−.14	−.17	.14	.06	−.18
12.FC	.004	−.03	−.10	−.24	−.21	−.07	−.21	−.11	−.07	−.03	.45		.15	.13	−.03	−.09	.16	.12	−.09
13.PM	.19	−.09	−.03	−.21	−.17	−.07	−.10	−.23	−.03	−.14	.27	.17		.10	−.13	−.16	.11	−.06	−.14
14.SP	.05	.07	−.12	−.17	−.17	.01	−.07	−.13	−.11	.001	.14	.15	.10		−.09	−.12	.05	.11	−.13
15.PO	.04	−.002	.19	.44	.39	.31	.22	.19	.12	.20	−.23	−.16	−.14	−.13		.50	−.18	.08	.19
16.PMU	.02	.03	.15	.43	.33	.40	.24	.24	.10	.15	−.20	−.12	−.15	−.13	.43		−.24	.09	.26
17.PS	.01	−.04	−.12	−.30	−.28	−.22	−.29	−.09	−.07	−.08	.27	.24	.11	.12	−.25	−.32		.00	−.05
18.AII	.05	.03	.02	.02	.02	.11	.01	.07	.06	.11	.01	.07	−.06	.10	.07	.10	−.01		.03
19.SU	−.08	.03	.22	.57	.44	.22	.22	.26	.09	.17	−.30	−.16	−.16	−.13	.31	.31	−.21	.03

Note. Bold=p<.05; EAO=early-age alcohol/cannabis onset; SI=substance involvement (alcohol involvement for drinkers, and cannabis involvement for cannabis users); CM=coping motive for use; DN=descriptive norms; SIN=student injunctive norms; VB=violent behavior; D=depression; SS=sensation seeking; FS=family sanctions against use; FC=family communication about use; PM=parental monitoring; SP=school performance; PO=peer offers to use; PMU=peer models for use; PS=peer sanctions against use; AII=American Indian identification; SU=solitary use (solitary alcohol use for drinkers, and solitary cannabis use for cannabis users). Although not shown in the table, the correlation between two outcome variables, solitary alcohol use and solitary cannabis use, was r = .44.

3.1. Solitary Alcohol Use

Estimates from the series of multiple logistic regression models predicting solitary alcohol use (Table 2) showed that both early-age alcohol onset and alcohol involvement were significantly related to solitary alcohol use at each step. In the final model, the odds of solitary alcohol use for early-age onset increased by 181% (OR=2.81; 95% CI [2.00,3.96]) compared with the odds for later onset. With a 1 standard deviation unit increase in alcohol involvement, the odds of solitary alcohol use increased by 7% (OR=1.07; 95% CI [1.03,1.11]).

Table 2.

Results of Hierarchical Logistic Regression Analyses of Solitary Alcohol Use

Independent variables	Step 1 Demographics & alcohol involvement			Step 2 Individual domain			Step 3 Family domain			Step 4 School Domain			Step 5 Peer domain			Step 6 Cultural domain
	ORs	OR 95% CI		ORs	OR 95% CI		ORs	OR 95% CI		ORs	OR 95% CI		ORs	OR 95% CI		ORs	OR 95% CI
		Low	Upp		Low	Upp		Low	Upp		Low	Upp		Low	Upp		Low	Upp
Sex	.81	.70	.95	.85	.71	1.03	.87	.71	1.07	.88	.72	1.07	.84	.69	1.03	.84	.69	1.02
Age	1.04	.95	1.14	1.05	.95	1.15	1.04	.94	1.15	1.04	.94	1.15	1.05	.94	1.16	1.05	.94	1.16
Early-age alcohol onset	3.14	2.27	4.36	2.87	2.04	4.05	2.83	2.01	4.00	2.82	1.99	3.98	2.82	2.00	3.96	2.81	2.00	3.96
Alcohol involvement	1.13	1.09	1.17	1.08	1.05	1.12	1.08	1.05	1.12	1.08	1.04	1.12	1.07	1.03	1.11	1.07	1.03	1.11
Individual Domain
Coping motive				1.50	1.26	1.79	1.48	1.23	1.78	1.47	1.22	1.77	1.45	1.19	1.76	1.45	1.19	1.76
Descriptive norms				1.13	1.04	1.23	1.13	1.05	1.23	1.14	1.05	1.23	1.11	1.02	1.20	1.11	1.02	1.20
Student injunctive norms				1.13	1.00	1.29	1.11	.97	1.27	1.10	.96	1.26	1.10	.96	1.27	1.11	.96	1.27
Violent behavior				1.27	1.15	1.40	1.26	1.14	1.38	1.25	1.14	1.37	1.23	1.11	1.35	1.22	1.12	1.34
Depression				1.24	1.10	1.39	1.24	1.10	1.39	1.23	1.09	1.39	1.23	1.09	1.39	1.23	1.09	1.39
Sensation-seeking				1.11	.99	1.25	1.12	.99	1.26	1.12	.99	1.27	1.12	99	1.27	1.12	98	1.27
Family Domain
Family sanctions against use							.88	.79	.98	.88	.79	.97	.88	.79	.98	.88	.79	.97
Family communication about use							1.02	.89	1.17	1.03	.90	1.18	1.01	.87	1.16	1.00	.87	1.16
Parental monitoring							1.00	.88	1.13	1.00	.88	1.14	1.00	.88	1.13	1.00	.88	1.13
School Domain
School Performance										.93	.84	1.02	.94	.86	1.03	.94	.85	1.03
Peer Domain
Peer offer to use alcohol													.99	.88	1.11	.99	.88	1.11
Peer models for use													1.09	1.03	1.14	1.08	1.03	1.14
Peer sanctions against use													1.20	1.07	1.34	1.20	1.07	1.34
Cultural Domain
American Indian Identification																1.03	.88	1.20

Note. Bold=p<.05; OR = odds ratio; CI = confidence interval. Cohort dummy variables were included in the models at each step but are not reported here.

In the individual domain, coping motive for alcohol use, descriptive norms, violent behavior, and depression were significantly associated with solitary alcohol use at each step. In the final model, one standard deviation unit increases in these variables were associated with increased odds of solitary alcohol use of 45%, 11%, 22%, and 23%, respectively (see Table 2). Even though student injunctive norms and sensation seeking were not significantly associated with solitary alcohol use, their odds ratios showed positive associations.

In the family domain, the measure of family sanctions against alcohol use was significantly associated with solitary alcohol use at all steps. With a 1 standard deviation increase in family sanctions against alcohol use, the odds of solitary alcohol use decreased by 12% in the final, fully adjusted model (OR=.88; 95% CI [.79,.97]). Both family communication about drinking alcohol and parental monitoring showed statistically non-significant associations with solitary alcohol use.

In the school domain, school performance had a non-significant negative association with solitary alcohol use. In the peer domain, both peer models for alcohol use and peer sanctions against alcohol use were significantly associated with solitary alcohol use. With a 1 standard deviation increase in peer models for alcohol use, the odds of solitary alcohol use increased by 8% in the final, fully adjusted model (OR=1.08; 95% CI [1.03,1.14]). With a 1 standard deviation increase in peer sanctions against alcohol use, the odds of solitary alcohol use increased by 20% in the final, fully adjusted model (OR=1.20; 95% CI [1.07,1.34]). Peer offer to use had a statistically non-significant negative association with solitary alcohol use. Finally, in the cultural domain, AI Identification was not statistically significantly related to the outcome at any step.

3.2. Solitary Cannabis Use

Estimates from the series of multiple logistic regression models predicting solitary cannabis use (Table 3) showed that sex, early-age cannabis onset, and cannabis involvement were significantly associated with solitary cannabis use at all steps. The odds of girls’ solitary cannabis use were 24% lower compared with the odds of boys in the final, fully adjusted model (OR=.76; 95% CI [.61,.94]). In the final model, the odds of solitary cannabis use for early-age onset increased by 117% (OR=2.17; 95% CI [1.54,3.07]) compared with the odds for later onset. With a 1 standard deviation increase in cannabis involvement, the odds of solitary cannabis use increased by 41% (OR=1.41; 95% CI [1.33,1.50]. Age did not have a statistically significant association with the outcome variable.

Table 3.

Results of Hierarchical Logistic Regression Analyses of Solitary Cannabis Use

Independent variables	Step 1 Demographics & cannabis involvement			Step 2 Individual domain			Step 3 Family domain			Step 4 School Domain			Step 5 Peer domain			Step 6 Cultural domain
	ORs	OR 95% CI		ORs	OR 95% CI		ORs	OR 95% CI		ORs	OR 95% CI		ORs	OR 95% CI		ORs	OR 95% CI
		Low	Upp		Low	Upp		Low	Upp		Low	Upp		Low	Upp		Low	Upp
Sex	.76	.63	.90	.75	.62	.90	.78	.63	.95	.78	.64	.96	.76	.61	.94	.76	.61	.94
Age	1.08	.99	1.17	1.07	.98	1.18	1.07	.98	1.17	1.07	.98	1.18	1.08	.98	1.18	1.08	.98	1.18
Early-age cannabis onset	2.54	1.77	3.63	2.21	1.53	3.20	2.24	1.56	3.21	22.3	1.55	3.20	2.17	1.54	3.07	2.17	1.54	3.07
Cannabis involvement	1.62	1.53	1.73	1.47	1.39	1.56	1.45	1.37	1.54	1.45	1.36	1.54	1.41	1.33	1.50	1.41	1.33	1.50
Individual Domain
Coping motive				1.85	1.55	2.21	1.82	1.53	2.16	1.82	1.53	2.16	1.77	1.49	2.10	1.77	1.49	2.10
Descriptive norms				1.09	1.02	1.16	1.09	1.02	1.16	1.09	1.03	1.16	1.05	.98	1.12	1.05	.98	1.12
Student injunctive norms				1.10	.99	1.21	1.09	.98	1.21	1.09	.98	1.21	1.08	.98	1.19	1.08	.98	1.19
Violent behavior				1.24	1.13	1.37	1.23	1.12	1.36	1.23	1.12	1.36	1.21	1.10	1.33	1.20	1.10	1.32
Depression				1.03	.93	1.14	1.03	.93	1.14	1.03	.93	1.14	1.02	.92	1.13	1.02	.92	1.13
Sensation-seeking				1.02	.90	1.16	1.03	.90	1.17	1.03	.90	1.17	1.02	.89	1.18	1.02	.89	1.18
Family Domain
Family sanctions against use							.86	.76	.97	.86	.76	.97	.86	.76	.97	.86	.76	.97
Family communication about use							1.06	.94	1.21	1.07	.94	1.21	1.06	.93	1.21	1.06	.92	1.22
Parental monitoring							.95	.82	1.09	.95	.82	1.09	.96	.83	1.12	.96	.83	1.12
School Domain
School Performance										.98	.90	1.07	.99	.91	1.08	.99	.91	1.08
Peer Domain
Peer offer to use cannabis													1.20	.91	1.60	1.20	.91	1.60
Peer models for use													1.26	1.10	1.43	1.26	1.10	1.43
Peer sanctions against use													1.03	.88	1.20	1.03	.88	1.20
Cultural Domain
American Indian Identification																1.01	.87	1.18

Note. Bold=p<.05; OR = odds ratio; CI = confidence interval. Cohort dummy variables were included in the models at each step but are not reported here.

In the individual domain, coping motive for cannabis use and violent behavior were significantly associated with solitary cannabis at each step. With a 1 standard deviation increase in coping motive and violent behavior, the odds of solitary cannabis use increased by 77% (OR=1.77, 95% CI [1.49,2.10] and 20% (OR=1.20, 95% CI [1.10,1.32]), respectively, in the final, fully adjusted model. Descriptive norms, student injunctive norms, depression, and sensation seeking had statistically non-significant positive associations with solitary cannabis use.

In the family domain, family sanctions against cannabis use was significantly associated with solitary cannabis use at all steps. With a 1 standard deviation increase in family sanctions against cannabis use, the odds of solitary cannabis use decreased by 14% throughout all the models (OR=.86; 95% CI [.76,.97]). Family communications about cannabis use had a statistically non-significant positive association with solitary cannabis use, whereas parental monitoring had a statistically non-significant negative association with the outcome variable.

In the school domain, school performance had a statistically non-significant negative association with solitary cannabis use. In the peer domain, the measure of peer models for cannabis use was significantly associated with solitary cannabis use. With a 1 standard deviation increase in peer models for cannabis use, the odds of solitary cannabis use increased by 26% in the final, fully adjusted model (OR=1.26; 95% CI [1.10,1.43]). Peer sanctions against cannabis use and peer offer to use cannabis had statistically non-significant positive associations with solitary cannabis use. Finally, in the cultural domain, AI Identification was not statistically significantly related to the outcome at any step.

Results were generally consistent across the steps of both sets of hierarchical models, although a few correlates became nonsignificant as variables were added. For example, sex became nonsignificant after adding individual domain variables in the solitary alcohol use model (Table 2). In the solitary cannabis use model (Table 3), descriptive norms became nonsignificant after adding peer domain variables.

4.0. Discussion

This is the first study to examine the correlates of both solitary alcohol and solitary cannabis use among American Indian (AI) adolescents. Results indicated that such use is prevalent. The rate of lifetime solitary alcohol use among AI adolescent drinkers is 24.9% in this study, which is comparable to the national rate of any lifetime alcohol use among youth (26.4%), and the rate of solitary cannabis use is extremely high at 53.6% compared with the national rate of any lifetime cannabis use among adolescents (30.2%) (Johnston et al., 2021). Consistent with theory and prior research (Cooper, 1994; Cox and Klinger, 1988, 2011; Creswell, 2021; Mason et al., 2020), our analyses provided support for prior research in that both solitary use of alcohol and solitary use of cannabis were positively related to coping; solitary alcohol use was additionally positively related to depression. These associations were evident over and above level of substance involvement, indicating that they are not accounted for by the possibility that solitary users are more heavily involved in substances than social only users. In contrast to social drinkers who use alcohol for social enhancement, this provides further evidence that solitary drinkers may use alcohol to cope with negative emotions (Creswell, 2021; Skrzynski and Creswell, 2020). Depression had a positive, but modest (r=.09) overall association with solitary cannabis use, but the unique association was not significant after adjusting for other predictors.

Analyses also extend the literature by addressing a larger array of social-ecological correlates of adolescent solitary substance use. We further found that AI adolescents with violent behavior (e.g., beaten up someone or hurt someone with a weapon) were more likely to report solitary alcohol and cannabis use, suggesting the importance of not only internalizing but also externalizing symptoms. Of course, violent behavior may either precede or follow solitary substance use. Further longitudinal research is needed to determine the direction of effect.

Social norms have been shown to be robustly associated with drinking behavior (e.g., Borsari and Carey, 2001). Likewise, in this study, descriptive norms reflecting the perceived prevalence and frequency of other students using alcohol and getting drunk were significantly and positively associated with solitary alcohol use among AI adolescents. No such associations were found for solitary cannabis use, nor for injunctive norms in relation to either outcome. Attitudes and behaviors surrounding cannabis have changed dramatically in the United States since the 2010s within a rapidly evolving legal landscape (Hudak, 2020). Alcohol use, although declining among youth (Johnston et al., 2021), has long been viewed as a relatively normative behavior for adolescents; therefore, perceived norms for alcohol use may have been better defined and more consequential than those for cannabis use during the timeframe of this study. Alcohol and cannabis initiators may already perceive that fellow students see relatively little harm resulting from use of those substances, thereby diminishing the potential influence of injunctive norms on solitary substance use among lifetime users.

In the family domain, family sanctions against alcohol use and against cannabis use were associated with a lower prevalence of solitary alcohol use and solitary cannabis use, respectively. This indicates that family sanctions against use may operate as a promotive factor to prevent solitary use. Interestingly, family communication about use and parental monitoring had statistically significant unadjusted negative associations with the outcomes, but were statistically non-significant correlates in the adjusted analyses, suggesting that it is the communication of clear rules against substance use that may have the most influence on adolescent’s solitary substance use. In the peer domain, the number of peers who model substance use was positively associated with both solitary alcohol use and solitary cannabis use. This is consistent with considerable research showing that having friends who use substances is a robust predictor of both social and solitary substance use among adolescents (e.g., Kandel, 1985; Tucker et al., 2014), although further longitudinal research is needed to determine if these associations represent a peer influence or peer selection process among AI adolescents. Unexpectedly, peer sanctions against alcohol use had a statistically significant positive association with solitary alcohol use in adjusted analyses; however, the unadjusted association was statistically significant and negative, suggesting the presence of suppression in the adjusted models of Steps 5 and 6.

Strengths of this study include the focus on an important topic, solitary substance use, in a large sample of AI adolescents, an understudied and vulnerable population, addressing an array of correlates across multiple ecological domains while accounting for level of substance involvement. Still, there are several noteworthy limitations. First, notwithstanding the social ecological framing, there was limited measurement coverage in certain domains. For example, school and culture were sparsely represented by only one construct each, and the cultural variable only measured cultural identification at the personal-level. Further research is needed to address an expanded set of measures, including those assessed at the school (e.g., school drug policies) and community (e.g., community religious/spiritual resources) levels in studies of AI adolescents. A few constructs were assessed using single self-report items with unknown reliability and validity. Second, the school-based recruitment of students necessarily excluded youth who had dropped out of school; therefore, the generalizability of findings might be limited. Third, data were collected as part of a cross-sectional cohort study, and longitudinal analyses could not be conducted to establish temporal ordering or to examine changes in solitary substance use over time. Finally, as a secondary analysis of existing, large-scale surveillance data, the analyses were necessarily limited to the data on hand. Certain omitted variables, such as SES and urban vs. rural location, should be considered in future studies, and cautious interpretation of the current results is recommended.

4.1. Conclusions

Despite these limitations, this study makes important contributions to the literature as the first investigation of solitary substance use among AI adolescents. Such use is prevalent and cause for concern, given the known adverse consequences for youth (Creswell et al, 2014; Creswell et al, 2015; Tucker et al., 2006). Statistically significant correlates across the social ecology in the individual, family, and peer domains were identified for both solitary alcohol and solitary cannabis use. In particular, high prevalence of solitary substance use among AI youth may be due to the challenges of coping with systemic poverty, institutional racism, interpersonal discrimination, and historical trauma (Brave Heart and DeBruyn, 1998; Goodkind et al., 2010; Whitbeck et al., 2002). Importantly, the differences observed between social-only and solitary drinkers were found over and above the influences of substance involvement. Findings from this study could help guide the development of screening efforts and, eventually, interventions designed to prevent solitary substance use among underserved AI adolescents.

Highlights.

• This study examined solitary alcohol and cannabis use in American Indian youth

• Prevalence of solitary use in initiators was 25% for alcohol and 54% for cannabis

• Coping motive and peer use were associated positively with both outcomes

• Violence and drinking norms were associated positively with solitary alcohol use

• Family sanctions against substance use was associated negatively with both outcomes