Moderating Effects of Perceived Social Benefits on Inhalant Initiation Among American Indian and White Youth

Abstract

This study examined whether perceived social benefits moderated the relationship between social influence variables (school attachment, peer inhalant use, perceived family caring, parental monitoring) and stage of inhalant initiation (Study 1), and lifetime inhalant use (Study 2). Participants were 7^th–12^th grade students attending schools on or near American Indian reservations with comparisons made between American Indian and White students. A total of 3498 American Indian and 1596 White students were surveyed. Differences in mean levels of social influence variables were found across ethnicity and stage of inhalant initiation and lifetime inhalant use. SEM models were evaluated to examine variable relationships for the two studies. For Study 1, social influence variables did not clearly differentiate early versus later inhalant initiators, and perceived social benefits failed to serve as a moderator. More differences were observed between users and non-users across measures of social influence (Study 2). Perceived social benefits generally did not moderate the relationships with two exceptions. Low perceived social benefits provided greater protection against the influence of peers on lifetime inhalant use among White students, while high perceived social benefits increased risk of peer influence among American Indian students.

The use of inhalants by school-age youth has been in general decline for a number of years, despite being a readily cheap and available abusable substance (Johnston, O’Malley, Miech, Bachman, & Schulenberg, 2015). However, there is variation across ethno/cultural groups. Compared to students from Monitoring the Future, American Indian students living on or near reservations report higher levels of both lifetime and last month inhalant use (Stanley, Harness, Swaim, & Beauvais, 2014). For use in the last month, relative rates range anywhere from 1.3 to 2.1 times higher for American Indian youth. This finding becomes even more troubling when age of onset is considered.

The risks associated with early initiation of substance use (12 or younger) are numerous, with negative outcomes that include development of substance use disorders (King & Chassin, 2007; Lynskey et al., 2003; Sung et al., 2004), greater affiliation with deviant peers and other conduct problems (Fergusson & Horwood, 1997), school dropout (Fergusson & Lynskey, 1998; Verweij et al., 2013), cognitive deficits (Pope et al., 2003), and mood disorders (Lynskey et al., 2004).

While most prior initiation research has focused on early-onset drinking (Hingson & White, 2014) and cannabis use (Agrawal & Lynskey, 2014), there is considerable evidence regarding the risks associated with early inhalant use, including neurological, renal, cardiovascular, and pulmonary damage (Brouette & Anton, 2001). Heavy/chronic inhalant use leads to multiple psychosocial problems that include conduct disorder and other behavioral problems including antisocial and violent characteristics, and use of other drugs (Hopfer et al., 2013; Howard et al., 2010).

A variety of risk factors are predictive of adolescent substance use (Hawkins, Catalano, & Miller, 1992; Swaim, 1991). Beyond the straightforward effects of these various factors, however, other research has considered the role of cognitive appraisal in the form of expectancies and perceived outcomes of substance use within the framework of subjective expected utility (Bauman, Fisher, Bryan, & Chenoweth, 1985, Copeland, Kulesza, Patterson, & Terlecki, 2009; Patrick & Maggs, 2011). The degree to which substances are presumed to result in positive or negative outcomes may serve as a motivating and potentially moderating factor either to use or avoid use.

A recent study (Swaim, 2015) examined the potential moderating effects of perceived emotional benefits on early versus later inhalant initiation among American Indian and White youth attending schools on or near reservations. It was hypothesized that emotional factors (low self-esteem, depression, and anger) would be associated more strongly with early versus later inhalant initiators and that perceived emotional benefits of inhalants would moderate the relationship between emotional factors and inhalant initiation. Little support was found for either of these hypotheses. An exception was the relationship between stage of initiation and low self-esteem which was related to early initiation only among American Indian students. Furthermore, perceived emotional benefits were found to moderate only this relationship.

While a number of studies have identified relationships between emotional distress and adolescent substance use, social influences are consistently more strongly related to drug and alcohol use than variables such as depression, anxiety, or low self-esteem for this age group (Becker, Curry, & Yang, 2011; Swaim, Oetting, Edwards, & Beauvais, 1989). However, the relationship between emotional distress and social influence variables is complex with emotional variables potentially serving as a mediational influence between social influences and substance use (Chassin et al., 1993), or interactively (Hussong & Hicks, 2003). Among American Indian adolescents, social influences appear to be stronger predictors of substance use than emotional distress factors (Oetting, Swaim, Edwards, & Beauvais, 1989; Swaim, Oetting, Thurman, Beauvais, & Edwards, 1993).

The current study considers the potential moderating influence of perceived social benefits on the relationship between four social influence factors and inhalant use. Each of the four social influence factors (school attachment, peer inhalant use, perceived family caring, and parental monitoring) is related to levels of substance use among adolescents in general (Henry, 2008; Ewing et al., 2015) as well as among American Indian youth (Boyd-Ball, Veronneau, Dishion, & Kavanagh, 2014; Moon, Blakey, Boyas, Horton, & Kim, 2014; Swaim, Oetting, Thurman, Beauvais, & Edwards, 1993).

Two studies were conducted. The first (Study 1) considered the moderating effects of perceived social benefits on the relationship between early (12 or younger) versus later (13 or older) onset of inhalant use. It was hypothesized that in comparison to an earlier study (Swaim, 2015) which found little support for emotional factors to predict such a relationship, stronger support would be found for the influence of social factors, with perceived social benefits serving as a moderator. A follow-up study (Study 2) considered whether even stronger support would be found in differences between users and non-users of inhalants. Two groups of students were compared, American Indian and White students. As reported above, American Indian students report higher levels of inhalant use compared to students in general. With Native adolescent rates that can double that of national youth (Stanley et al., 2014), identifying differing patterns of risk for this cultural group is needed. The current sample provides a unique opportunity to compare American Indian students to White students (the largest non-Indian group in reservation samples). While comparisons to national samples are useful, evaluating differences across ethnicity among students who live and attend schools in the same locations allows for at least partial control for socioeconomic conditions that vary substantially between reservations and other population areas. To date, this author is aware of no other study that has considered these relationships across ethnically diverse groups of adolescents.

Method

Participants

Seventh through 12^th grade students from 32 schools on or near American Indian reservations comprised the sample. This study is part of an ongoing study of the epidemiology and etiology of substance use among reservation-based American Indian youth. Schools on or near reservations that included at least 20% American Indian youth were sampled across six geographic regions (Northwest, Northern Plains, Southeast, Southern Great Plains, and Southwest). The sample is based on four years of data (school years 2009–2012). Schools were paid $500 for participating and were given a comprehensive report of their school’s data. Names of specific tribes are kept private to preserve confidentiality.

Students self-identified their ethnoracial category and the current study was limited to those who self-identified either as American Indian or White. While self-identification can lead to some misrepresentation of one’s ethnoracial group, identification based on tribal enrollment is infeasible for a nationally representative study of this size. As described by Gone and Trimble (2012), we adopted the method of self-identification. These authors also used an additional criterion of “affiliation with enduring tribal communities” which accurately describes our sample of reservation-based youth. The total sample size of all 7^th through 12^th grade students was 5094, 68.7% American Indian (n=3498), and 31.3% White (n=1596). Study 1 included only those students who reported having ever tried inhalants. For the outcome variable of Study 1, a dummy variable was created with 0 representing later inhalant initiation (13 and older) and 1 representing early initiation (12 and younger). A total of 676 American Indian students reported having ever used inhalants (19.3% of all American Indian students). Of this total, 349 (51.6%) reported early initiation, and 327 (48.4%) reported later initiation. A total of 155 White students reported having ever used inhalants (9.7% of all White students). Of this total, 66 (42.5%) reported early initiation, and 89 (57.4%) reported later initiation. For Study 2, a dummy variable was created with 0 representing no lifetime inhalant use and 1 representing having ever tried inhalants. Among American Indian students, there were 676 ever tried (19.3%), and 2822 never tried (80.7%). Among White students, there were 155 ever tried (9.7%) and 1441 never tried (90.3%).

Procedure

All data collection procedures were approved by the university Institutional Review Board. In addition, resolutions of support were obtained from participating tribal authorities or local school boards according to local requirements. Teachers or school staff members were trained in human subjects procedures through online or telephonic IRB training prior to all data collection. The school staff member was provided written instructions for all survey administration procedures. Prior to survey administration, parents/guardians were notified of the study through a parent notification letter and a broad media release, and were given the opportunity to remove their child from the project. Less than one percent of parents declined to have their child participate.

Students were informed that their participation was voluntary and that they could leave any survey item blank, or discontinue taking the survey at any time. No identifying information was collected to maintain the anonymity of all participants. Surveys were completed during one class period and those students for whom parental consent was not obtained were moved to another school area. Once surveys were completed, students placed their surveys in a large envelope in random order which was then sealed and turned in to the school staff member. Staff administering the survey were instructed to remain in an area of the classroom during administration that precluded their observation of student responses.

Measures

Survey data were taken from a revised version of The American Drug and Alcohol Survey™, designed specifically for the ongoing study of substance use epidemiology among American Indian youth. Versions of this survey have been given to American Indian youth for several decades, and it has been validated for use with both majority and minority youth (Oetting & Beauvais, 1990). Two inhalant use items, one assessing lifetime use and the other assessing age of first use were asked. The lifetime measure asked, “Have you ever sniffed” or “huffed” glue, gas, sprays, or anything like that to get high?” The age of first use item asked, “How old were you the first time you “sniffed” or “huffed”? Response alternatives for this item included “never used,” 7 or younger, individual alternatives from age 8 to 18, and 19 and older.

The four measures of social influence (school attachment, peer inhalant use, perceived family caring, and parental monitoring) were assessed with Likert items that included four response alternatives. School attachment was measured with four items (e.g., “I like school,” “School is fun”; “a lot, some, not much, not at all”), with a Cronbach alpha of .86. Peer inhalant use was measured with three items (e.g., “How many of your friends sniff glue or gas, etc.?, “all of them, most of them, a few, none”), with a Cronbach alpha of .75. Perceived family caring was measured with three items (e.g., “How much does your family care about you?” “a lot, some, not much, not at all”) with a Cronbach alpha of .89. Parental monitoring was measured with four items (e.g., “My parents allow me to go out as often as I want,” “very true, mostly true, somewhat true, not at all true”), with a Cronbach alpha of .84.

The moderating variable, perceived social benefits of inhalants, was measured with four items (e.g., “Sniffing inhalants with friends is part of being in a group,” “strongly agree, agree, neutral, disagree, strongly disagree”) with a Cronbach alpha of .97. This scale score was dichotomized into high and low perceived benefits.

Missing Data

Missing data were handled using a multiple imputation approach (Shafer & Graham, 2002). Five imputed data sets were created using IBS SPSS Statistics 22, and parameter estimates were combined using procedures described by Rubin (1987). For Chi-square and other measures of fit, there is no known theoretical approach for combining statistics. Accordingly, means of these measures across the five imputations are reported.

Analytic Approach

Mean levels of the four measures of social influence were compared across early and later inhalant initiation (Study 1) and across ever tried and never tried inhalants (Study 2) for both American Indian and White adolescents. Differences across groups (ethnicity by stage of inhalant initiation) were assessed with univariate analyses of variance. Relationships between social influence variables for both Study 1 and Study 2 were conducted with an SEM approach using EQS 6.1 (Bentler & Wu, 2010). Measurement models were constructed first to evaluate the measurement quality of the four social influence variables with variances fixed to 1 and all factors allowed to freely correlate. This was followed by two sets of structural models that comprised Study 1 and Study 2. Study 1 evaluated early versus later inhalant initiation with stage of initiation regressed onto the four social influence variables. Multiple group models were conducted across those students reporting low and high levels of perceived social benefits of inhalants and across ethnicity to determine potential moderating effects for the association between the four social influence and stage of inhalant initiation. Study 2 followed the same procedures with the only difference being the outcome variable (never versus ever tried inhalants).

Results

Mean differences in social influence variables

Mean levels and standard deviations of the social influence variables by ethnicity, stage of inhalant initiation, and ever versus never used inhalants are presented in Table 1.

Table 1.

Mean (SD) Levels of Social Influence Factors

		American Indian		White
Variable	Inhalant Initiation	Mean	SD	Mean	SD
Study 1	Inhalant Initiation
School Attachment	13 or older	10.77	3.02	10.53	3.24
	12 or younger	10.33	3.23	9.00	3.58
Peer Inhalant Use	13 or older	5.57	2.39	5.72	2.38
	12 or younger	5.94	2.79	6.14	3.11
Perceived Family Caring	13 or older	11.06	1.81	10.40	2.13
	12 or younger	10.91	1.93	9.76	3.18
Parental Monitoring	13 or older	11.57	3.34	11.51	3.68
	12 or younger	11.81	3.28	11.00	4.16
Study 2	Ever Inhalants
School Attachment	Never	11.14	3.18	11.48	2.86
	Ever	10.54	3.13	9.83	3.46
Peer Inhalant Use	Never	4.30	1.89	4.58	1.83
	Ever	5.75	2.60	5.94	2.77
Perceived Family Caring	Never	11.22	1.78	11.37	1.64
	Ever	10.98	1.87	10.10	2.67
Parental Monitoring	Never	12.29	3.11	12.52	2.83
	Ever	11.70	3.31	11.23	3.91

Study 1

Among students who had tried inhalants, for school attachment, significant main effects for both ethnicity (F [1,827] = 7.486, p = .006), and stage of inhalant initiation (F [1,827] = 11.842, p = .001) were observed with no significant ethnicity by stage of inhalant interaction. American Indian students reported higher levels of school attachment and students initiating inhalants at ages 13 and older reported higher levels of school attachment. No significant differences by ethnicity or stage of inhalant initiation were found for peer inhalant use or parental monitoring. For perceived family caring, significant main effects were found for both ethnicity (F [1,827] = 24.340, p = < .001) and stage of inhalant initiation (F [1,827) = 4.815, p = .028, with no significant interaction. American Indian students reported higher levels of family caring and the difference in perceived family caring between early versus later initiators was greater among White students.

Study 2

With comparisons for never versus ever tried inhalants, a significant main effect for school attachment was found for ever tried (F[1,5093]) = 59.53, p < .001. Students who never tried inhalants reported higher levels of school attachment. A significant ever tried by ethnicity interaction (F[1,5093]) = 12.94, p < .001 indicated that among White students there was a greater difference in school attachment between never and ever users of inhalants. For perceived peer inhalant use, significant main effects were found for both ever tried (F[1,5093]) = 219.80, p < .001, and ethnicity (F[1,5093]) = 6.16, p = .013, with no significant interaction. For both ethnicities, those students who had tried inhalants reported more friends also using inhalants. For family caring, significant main effects were found for ever tried (F[1,5093]) = 79.14, p < .001, ethnicity (F[1,5093]) = 18.82, p < .001, and the interaction (F[1,5093]) = 37.32, p < .001. Students who had never tried inhalants reported higher levels of family caring. The significant interaction was accounted for by a greater difference in perceived family caring among White students who had never versus ever tried inhalants, with those having never tried reporting higher levels of family caring. For parental monitoring, a significant main effect was observed for ever tried (F[1,5093]) = 41.49, p < .001, along with a significant ever tried by ethnicity interaction (F[1,5093]) = 5.75, p = .016. Higher levels of parental monitoring were found among those students who had never tried inhalants. The disparity in levels of perceived parental parenting between never and ever users of inhalants was greater among White students, with never users reporting higher levels of parental monitoring.

SEM Models

Measurement models of the four latent variables for social influence were conducted across ethnicity and level of perceived social benefits of inhalants. For American Indian students reporting low perceived benefits, fit statistics were S-B Chi-square (71) = 302.19, p < .001, Standardized RMR = .062, Robust CFI = .90; for American Indian students reporting high perceived benefits, S-B Chi-square (71) = 121.91, p < .001, Standardized RMR = .061, Robust CFI = .93. Among White students reporting low perceived benefits, S-B Chi-square (71) = 111.96, p < .001, Standardized RMR = .078, Robust CFI = .91, and for White students who reported high perceived benefits, S-B Chi-square (71) = 157.19, p < .001, Standardized RMR = .096, Robust CFI = .87. All factor loadings were significant (p < .001) using robust standard errors with the exception of “perceived number of friends who use inhalants,” which was significant at the < .05 level. Standardized factor loadings and residuals are presented in Table 2, along with structural coefficients for Study 1 and Study 2.

Table 2.

Standardized Factor Loadings, Residuals, and Structural Paths by Group

	American Indian		White
	Lo Soc Ben	Hi Soc Ben	Lo Soc Ben	Hi Soc Ben
Measurement Model
School Attachment
I like school	.75(.66)	.71(.71)	.85(.53)	.87(.49)
My teachers like me	.76 (.65)	.74(.67)	.65(.76)	.95(.32)
I like my teachers	.83(.56)	.92(.39)	.80(.60)	.79(.61)
School is fun	.73 (.68)	.65(.76)	.78(.63)	.87(.50)
Peer Inhalant Use
# friends who use	.41(.91)	.53(.85)	.37(.93)	.51(.86)
# times average student uses	.69(.72)	.95(.31)	.63(.78)	.93(.36)
# students use at least 1/mo	.90(.43)	.76(.66)	.89(.46)	.89(.45)
Family Caring
How much family care/you	.93(.37)	.91(.43)	.92(.38)	.96(.29)
How much you care/family	.81(.59)	.87(.49)	.79(.61)	.87(.49)
How much family care/do	.72(.69)	.67(.74)	.83(.57)	.74(.67)
Parental Monitoring
Parents allow go out often	.71(.71)	.77(.64)	.77(.64)	.93(.37)
Parents let go w/o asking	.83(.56)	.80(.61)	.80(.60)	.89(.45)
Parents less strict	.71(.70)	.72(.79)	.82(.57)	.92(.38)
Parents let stay out	.83(.56)	.79(.62)	.74(.68)	.96(.29)
Structural Model
(Study 1 - Early vs. later)
School Attachment to Inh-Ini	−.06(.026))*	−.01(.044)NS	−.09(.067)NS	−.01(.080)NS
Peer Inhalant Use to Inh-Ini	−.01(.027)NS	.06(.040)NS	−.01(.054)NS	−.05(.070)NS
Family Caring to Inh-Ini	−.02(.026)NS	−.02(.045)NS	.03(.051)NS	−.07(.076)NS
Parental Monitoring to Inh-Ini	−.05(.024)*	−.04(.044)NS	.01(.045)NS	−.04(.068)NS
Structural Model
(Study 2 - Ever vs. never)
School Attachment to Inh-Ini	−.03(.008)***	−.01(022)NS	−.02(.009)*	−.05(.035)NS
Peer Inhalant Use to Inh-Ini	.07(.009)***	.17(.023)***	.02(.009)*	.11(.033)***
Family Caring to Inh-Ini	−.01(.010)NS	.01(.02)NS	−.03(.011)**	−.07(.033)*
Parental Monitoring to Inh-Ini	−.02(.008)**	−.04(.023)NS	−.02(.009)*	−.08(.034)*

Note. Lo Soc Ben = low social benefits of inhalants; Hi Soc Ben = high social benefits of inhalants; Inh-Ini = stage of inhalant initiation.

^*(p<.05),

^**(p<.01),

^***(p<.001).

NS=non-significant.

Tests for measurement invariance across ethnicity and level of perceived social benefits of inhalants were conducted next. With all factor loadings constrained to be equal, Chi-square (284) = 745.63. Equality constraints were then freed resulting in Chi-square (326) = 870.41. The Chi-square difference test, Chi-square (42) = 124.78, was significant (p < .001). Factor loadings found to be different across groups based on the LM test were freed one at a time with subsequent Chi-square difference tests conducted. A total of 13 of the 42 equality constraints were freed. Examination of these constraints revealed that nine involved Group 4 (high perceived social benefits, White students), the smallest group. The final difference test was nonsignificant, Chi-square (29) = 40.97, p = .07 and all remaining constraints were non-significant.

For Study 1, the next step tested for structural equivalence across ethnicity and perceived social benefits of inhalants. With the measurement model constrained as described above, all structural coefficients were freely estimated resulting in fit of Chi-square (344) = 819.12. This was followed by constraining all structural coefficients to equality, resulting in Chi-square (356) = 834.96. The Chi-square difference test, Chi-square (12) = 15.84, was non-significant (p = .20).

For Study 2, the same procedure as described above was used to test for structural equivalence. With all structural coefficients freely estimated, Chi-square (344) = 3665.90. A subsequent test with all structural coefficients constrained to be equal resulted in Chi-square (356) = 3717.11 and a Chi-square difference test (12) of 51.21, p < .001. Examination of the LM test indicated that the constraint from ever used inhalants to peer inhalant use should be freed for White students reporting low social benefits of inhalants. With this constraint released, Chi-square (355) = 3688.46 and a Chi-square difference test (11) of 22.56, p = .02. This was followed by releasing an additional constraint from ever used inhalants to peer inhalant use for American Indian students reporting high social benefits of inhalants. This resulted in a Chi-square (354) of 3682.20 and a Chi-square difference test (10) of 16.30, p = .09.

Discussion

In a previous study (Swaim, 2015), little support was found for the moderating effects of cognitive appraisal of emotional benefits on inhalant initiation among both American Indian and White students. This follow-up study evaluated whether more evidence would be found for moderating effects of cognitive appraisal based on social benefits. Four measures of social influence (school attachment, peer inhalant use, family caring, and parental monitoring) were considered as factors related to stage of inhalant initiation (Study 1, early vs. later) and ever tried inhalants (Study 2, ever vs. never), along with the potential moderating effects of perceived social benefits.

Mean differences in the social influence variables revealed both ethnic differences as well as interactive effects across ethnicity and inhalant use variables. The first factor, school attachment, has demonstrated its ability to predict levels of use in American Indian students (Dickens et al., 2012; Galliher, Evans, & Weiser, 2007) and inhalant use specifically (Bates et al., 1997). It is of interest that American Indian students reported higher levels of school attachment compared to their White counterparts which suggests that this may be an important target for prevention efforts, particularly for reservation youth. Their higher rates of attachment may be due to attending schools with higher rates of American Indian students, a factor that has been shown to increase school attachment among African American students (Walsemann, Bell, & Maitra, 2011) and may also reduce rates for substance use (Swaim & Stanley, 2011). However, when users versus nonusers were compared (Study 2), there was a greater disparity in levels of school attachment for White students, with nonusers reporting higher levels of attachment.

No mean differences were found across ethnicity or stage of inhalant initiation (Study 1) for perceived number of friends who use inhalants. This is likely due to the nature of some inhalant use being a solitary behavior (Matthew, Kapp, & Jones, 1989; McGarvey et al., 1999) compared to alcohol and marijuana use which tend to be more social phenomena. There is also some evidence that the solitary versus social nature of inhalant use may depend upon the types of inhalants used (Matsumoto et al., 2001). But differences were found in perceived number of peers who used inhalants between users and nonusers (Study 2). This suggests that between early and later initiators, there are less differences in number of inhalant-using peers.

Similar to peer use, no mean differences were found for parental monitoring, both across ethnicity and stage of initiation (Study 1). But, nonusers did report higher levels of monitoring than nonusers (Study 2). Prior research has found that there are diminishing effects for both parental and peer influences by the beginning of the middle school years for inhalant use (Ober et al., 2013), but there may be an additive effect when higher levels of parental monitoring are combined with strong family ties (Ramirez et al., 2004).

For perceived family caring, similar to results for school attachment, American Indian students reported higher levels of family caring. However, among White youth, there was a greater disparity in family caring between early and later initiators (Study 1), with later initiators reporting higher levels of family caring. Thus, while family influences may diminish as youth age, strong family bonds may be able to forestall early inhalant initiation. Additionally, nonusers reported higher levels of family caring than users (Study 2). One general finding was that larger differences were found across stage of initiation and user/non-user status for White students which suggests that the social influence variables examined here may be more protective for White compared to American Indian students.

The first SEM model which considered effects of social influence on early versus later initiators of inhalants (Study 1) produced minimal significant effects. With the exception of school attachment and parental monitoring for American Indian students reporting high social benefits of inhalants, social influence variables failed to differentiate the two groups. This was similar to results obtained in the earlier study of emotional distress variables on stage of inhalant initiation (Swaim, 2015). It would appear that from these two sets of results, variables that are typically good predictors of levels of substance use do not discriminate strongly between early and later initiators of inhalants. This seems to be in conflict with prior studies that find higher risk for physiologic, behavioral, and psychological problems among early initiators (Brouette & Anton, 2001; Hopfer et al., 2013; Howard et al., 2010). Two factors may help explain this apparent contradiction. First is stage of development. The articles just cited focus largely on populations older than those studied here. In Howard et. al. (2000), retrospective studies of adults comprised the sample. In Hopfer et al. (2013), older adolescents and young adults were studied. In both cases, the developmental stage of the participants allowed more time for inhalant-related problems to develop, in comparison to the younger sample in this study for whom such problems may not have yet emerged. The second factor is the academic standing of the participants in the present study, all of whom were students in school. It is well established that risk for both substance abuse and related problems increases dramatically with school dropout (Maynard, Salas-Wright, & Vaughn, 2015; Swaim, Beauvais, Chavez, & Oetting, 1997). Stronger relationships to social influence variables may well be observed among those youth who initiate inhalant use and leave school early. Additionally, this study considered only initiation of inhalants, not levels of sustained use for which negative outcomes are more likely (Brouette & Anton, 2001).

Finally, for Study 1, based on the non-significant Chi-square difference test, no moderating effects of social influence variables on stage of initiation were found for cognitive appraisal of social benefits. For Study 2, two structural equality restraints were removed. Examination of these constraints indicates that there were differences across levels of perceived social benefits in the relationship between ever used inhalants and perceived number of inhalant using peers. Low perceived social benefits provided greater protection against the influence of peers among White students, while high perceived social benefits increased risk of peer influence among American Indian students. These results suggest that the influence of peers on use of inhalants is not uniform across all students. Rather, the cognitive perception of social benefits may either increase or decrease risk related to this social influence factor. While changing peer relationships can be difficult, a more modifiable target for prevention may be to address perceived social benefits. According to Study 1, this factor is not likely to be as effective among those students who have already initiated inhalant use, but may be more useful in delaying or preventing initiation.

Implications

Results from this and a prior study (Swaim, 2015) suggest that it may be difficult to distinguish early versus later inhalant initiators among middle and high school students based on emotional and social influence risk factors. Furthermore, this difficult distinction is true both for American Indian and White youth. Given the association of early inhalant initiation with many negative outcomes (Brouette & Anton, 2001; Hopfer et al., 2013; Howard et al., 2010), it would be reasonable to suspect that key risk factors would help differentiate these two groups. But the current results do not support this model. While continuing inhalant users are at risk to experience a number of negative outcomes, social influence variables are not likely to identify differences between early and later initiators. As might be expected, measures of social influence did a better job of distinguishing between inhalant users and non-users. Perhaps of greatest interest to prevention scientists and clinicians is the finding that low perceived social benefits protected against peer influences on ever trying inhalants among White students, while high perceived social benefits increased the risk for peer influence for lifetime use among American Indian students. These two findings indicate that future study should explore the role and changing trajectories of cognitive appraisal of substance use on peer influences.

While the failure to find more significant results for the social influence variables was somewhat surprising, it should be noted that other measures of social influence have similarly failed to relate to use. A number of past studies have attempted to explain varying patterns of substance use among American Indian youth by incorporating cultural identification into explanatory models, with minimal to no significant relationships observed (Baldwin, Brown, Wayment, Nez, & Brelsford, 2011; Beauvais, 1998; Trimble, 1996). These and our findings call into question the centrality of social influence models when investigating differences between early and later initiators of inhalants. But this does not invalidate the importance of these variables. It is of note that American Indian students showed higher levels of some of these measures (school attachment, family caring) compared to White students. These variables may help explain the ability of some American Indian youth to negotiate high levels of risk and exhibit resilience and positive behavioral outcomes.

It was suggested earlier that the lack of differences found between early and later initiators may be due to developmental stage (middle and high schools students versus young adults) and academic standing (students versus dropouts). Another consideration that is likely related to academic standing is substance use trajectory. While we are currently under DSM-5, Oetting and Beauvais, who wrote the diagnostic criteria for inhalant disorders for DSM-III-R (APA, 1987), identified three types of inhalant abusers (Beauvais & Oetting, 1987; Oetting, Edwards, & Beauvais, 1988): 1) young inhalant users, 2) polydrug users, and 3) inhalant-dependent adults. Each of these types of users was rated for prognosis, level of dysfunction, and consequences of use, with inhalant-dependent adults having the poorest outcomes, young inhalant users having the best outcomes, and polydrug users falling between these two groups. It is likely that the majority of inhalant-using students surveyed in this study, both early and later initiators, would fall into the young inhalant using group, and thus would be less susceptible to more negative outcomes in the future. But the findings reported here emphasize the importance of conducting more in depth studies of the trajectories of inhalant-using youth, to help identify those youth who may move into one of the other groups described by Oetting and Beauvais.

Limitations

This study was limited by several factors that should be considered when interpreting the results. First, the sample was of American Indian and White students attending schools on or near reservations. Because it was part of a larger study of substance use epidemiology among reservation youth, no data were obtained from urban American Indian students. Patterns of risk and comparisons to non-Indian urban students may differ for urban participants, but this was beyond the scope of the current study.

Second, while retrospective accounts of age of first inhalant use were obtained, providing a proxy for temporal relationships, the data were essentially cross-sectional, precluding causative conclusions. Additionally, other factors could have intervened between onset and report of current social influence variables. As suggested above, prospective studies are needed that will not only allow for tracking and identification of unique inhalant use trajectories, but will also permit causal interpretations related to key elements of risk or protection.