Early onset of cannabis use and alcohol intoxication predicts prescription drug misuse in American Indian and non-American Indian adolescents living on or near reservations

Abstract

Background:

Prescription drug misuse (PDM) is a growing issue within the American Indian (AI) population, especially in younger populations.

Objectives:

This study estimates relationships between PDM and early initiation (prior to 14 yrs) of cannabis use and alcohol intoxication for a national sample of AI and non-AI adolescents attending schools on or near reservations.

Method:

Participants were 2580 students (50.2% female; 58.1% AI), ages 15–18, attending schools located on or near an AI reservation. Four models of PDM were estimated: 1) demographic variables; 2) demographics plus cannabis use initiation status; 3) demographics plus alcohol intoxication initiation status; and 4) all variables. All analyses were conducted using multilevel modeling.

Results:

Results indicated that early onset of cannabis use and alcohol intoxication were individually significant predictors of PDM for AI and non-AI adolescents, with odds ratios (OR) of 47.00 for cannabis (p <.01) and 35.73 (p <.01) for intoxication and with no significant differences by race (AI vs. non-AI). Results also indicated a greater likelihood of PDM when a student was an early initiator of both cannabis use and intoxication than when they were one or the other. Finally, there was a significantly greater association between cannabis use and PDM (ORearlycannabis = 24.95, p <.01) than between intoxication and PDM (ORearlyintoxication = 3.98, p <.01) when both predictors were in the model.

Conclusions:

These findings suggest that for AI and non-AI youth who have some shared living experience, early initiation of cannabis use and alcohol intoxication are risk factors that are similarly related to PDM and that targeting early initiation for both groups of adolescents is critical in prevention of PDM.

Introduction

Prescription drug misuse (PDM) is a growing health concern in the United States (1,2). According to NIDA (2), PDM is “taking a medication in a manner or dose other than prescribed; taking someone else’s prescription, even if for a legitimate medical complaint such as pain; or taking a medication to feel euphoria (i.e., to get high).” PDM is seen in multiple age cohorts, with recent studies indicating continuing prevalence among younger individuals (3,4). The consequences of adolescent PDM are serious, including increased risk for substance use disorders later in life (5) and subsequent illicit opioid misuse (6).

PDM is a growing issue within the American Indian (AI) population (4,7,8), especially in younger populations. In 2008, PDM was the second most endorsed reason for treatment admissions among AI individuals in Washington State (8). Admissions for PDM rose from 1% in 2002 to 15% in 2008, and among AI admissions, 62% of individuals aged 18–29 endorsed PDM. Results from a study of a Midwestern AI reservation showed a high rate of opioid misuse, with over 50% of study participants between the ages of 18–25 reporting misuse (4).

One risk factor for PDM may be early initiation of any substance use. Research suggests that AI adolescents initiate use of alcohol and other drugs earlier than other ethnic/racial groups. For example, Dickerson et al. (9) found that American Indian/Alaskan Natives in Los Angeles County had a younger age of onset of alcohol, cannabis, methamphetamine, and other drug use compared to their African-American, Caucasian, Hispanic/Latino, and Asian/Pacific Islander counter-parts. Stanley and Swaim (10) found that adolescent AIs living on or near reservations had significantly earlier age of onset of alcohol, cannabis, and inhalants compared to White adolescents living in those same areas. Few studies have examined the effect of early alcohol and cannabis consumption on PDM specifically, and to date, no studies have examined these relationships for AIs. Previous studies for non-AI populations found that early onset significantly increased risk of substance use disorders (11; see Magid & Moreland (12) for a review), and individuals with early onset substance use reported increased use of other substances, including prescription drugs (13). McCabe et al. (14) found that, depending on the prescription drug class, approximately 75–97% of participants with onset of PDM before the age of 15 were also using alcohol or other drugs. Kropp et al. (15) found a similar result; participants with lower ages of substance use initiation were more likely to be using alcohol in combination with other drugs, including prescription drugs.

Given early initiation and evidence of PDM among AIs, the current study estimates the relationships between PDM and age-of-onset of alcohol intoxication and cannabis use for a sample of adolescents attending schools on or near AI reservations. We examine whether these estimated relationships differ by race by including in the analysis non-AI adolescents attending those same schools. The substantial number of non-AI youth, the vast majority who are white, in the sample allows for comparisons by race while controlling, at least in part, for factors related to attending the same school and, to some degree, living in a similar environment. This contrasts with studies that often compare reservation AI substance use to findings from national samples, where more factors related to risk are likely to differ, such as rurality, socioeconomic variables, and ethnic/racial makeup.

Method

All data were collected under procedures approved by the Institutional Review Board of Colorado State University. For each participating school, the appropriate tribal and school board approvals were also obtained.

Sample

Data were from 33 schools that participated in an ongoing epidemiologic study during academic years 2009–2012. Schools were randomly sampled from a sampling frame of schools with grades 7 or above, located on or within 25 miles of American Indian reservations, and having at least 20% AI students enrolled. Sampled schools were drawn to approximate the percentage of AIs residing in each of six regions (Northeast, Northwest, Northern Plains, Southeast, Southwest, and Upper Great Lakes). Depending on the year, approximately 20–40% of schools sampled agreed to participate, and on average, 79% of enrolled students (with a range of 66% to 100%) took the survey. Schools were paid 500 USD for resources used in survey administration, and they were given a detailed report of their results. Data were pooled across the four academic years. All schools are unique across time, ensuring that each participant appears only once in the data. Schools varied significantly in size and racial make-up, with the number of students in grades 7–12 varying between 13 and 852 students, and the percentage of AI students ranging from 20% to 100%. Seventy-five percent of the schools were on reservation. Free/reduced lunch rates varied between 19.4% and 100% for these schools, with an average of 79%. Though the average free/reduced lunch rate did not differ for schools on versus off reservation, there was a significant correlation (r = .60, p = .02) between rates of free/reduced lunch and percentage of AIs enrolled.

Participants

Students between the ages of 15 and 18 were selected for this study (n = 2580; female = 50.2%; mean age = 16.2 years). Approximately, 58.1% of respondents self-identified as AI, while 36.6% self-identified as White, 5.6% identified as African American, and 3.9% identified as Latino or Hispanic (students could choose more than one race/ethnicity). The regional distribution of AI geographic areas was as follows: Northwest 3.3%, Northern Plains 31.6%; Upper Great Lakes 26.8%; Southeast 18.0%; Southwest 20.3%. No Northeast schools participated in the study.

Procedure

At each site, a school official was required to receive CITI training prior to the administration of the survey. This staff member was responsible for supervising all surveying and administrative procedures associated with the survey. Parents were notified of the survey by mail and given directions on how to opt out their child. In addition, a notice of the survey was posted in local media likely to be seen by parents. Less than 1% of students did not participate due to parent opt out.

Prior to distributing the survey, a school official read instructions to the students which included disclosing that the survey was voluntary and that they could choose not to participate or to withdraw from the study at any time. The surveys contained no identifying information, and study procedures ensured anonymity for participants. Upon completion of the survey, each student placed their survey into a large envelope in random order, which was then sealed and returned to our research center.

Instrument

Data were obtained from the American Drug and Alcohol Survey (ADAS) which has been validated for use across multiple minority groups (16).

Predictive measures

Age of onset of alcohol intoxication and of cannabis use were assessed by asking participants how old they were the first time they got drunk and how old they were the first time they used “marijuana (e.g., weed, pot).” The mean age of onset for those who had been drunk was 13.9 years (n = 1416; SD = 1.7, Range = 10) while the mean age of onset for those who had used cannabis was 13.4 years (n = 1398, SD = 2.1, Range = 10). Approximately 45% (n = 1164) had not been drunk and 46% (n = 1182) had not used cannabis. Two dichotomous variables were created for each substance to measure early initiation (age 13 and younger), non-early initiation (age 14 and older), and nonuse (never initiated). Sex was measured as female = 1 and male = 0. Age was centered at age 13. Race was coded with one dichotomous variable (AI = 1; non-AI = 0).

Dependent measures

Prescription Drug Misuse (PDM) was assessed by asking participants if they had ever used prescription drugs to get high or taken extra doses to get high. There were four classes of PDM: stimulants (6.9%), tranquilizers (4.8%), OxyContin (9.7%), and narcotics other than heroin (10.4%). If a participant endorsed misuse of any of these, they were classified as endorsing PDM. Approximately half (51.5%) of those reporting any misuse used from only one of these classes; 25.3% used from two classes; and 11.6% each used from three or four classes.

Analyses

In order to account for students clustered within schools, all analyses were conducted in HLM 6.02. The multivariate hypothesis testing feature in HLM was used to examine the inclusion of random slopes, and all random slopes were subsequently fixed at zero, resulting in fixed and equal slopes across schools. The following procedure was used for model estimation. First, a model was estimated that included the demographic variables of age, sex, and race. Interaction terms between these variables were not significant and a likelihood ratio test (LRT) indicated that model fit was not improved with their inclusion (χ² = 6.24(3); p = .10). They were, therefore, trimmed from the models. Next, two models were estimated – one for age of first intoxication and one for age of cannabis initiation. Each model included the demographic variables, the appropriate initiation variables, and interactions between each initiation variable and race. A fourth model included all predictors. Based on significance of coefficients and an LRT of model improvement (χ² = 1.28(2); p > .50), a final model was estimated that excluded the two race x intoxication initiation variables. As the dependent variable was binary, models were specified using Bernoulli’s logistic regression, and estimates reflect unit-specific odds ratios (OR). Missing data accounted for 3.5% of all data.

Results

Descriptive statistics for PDM and the predictor variables can be found in Table 1. PDM was reported by 20.3% of AIs and 13.3% of non-AIs. For the full sample, early initiation of cannabis (25.7%) was higher than that for alcohol intoxication (19.3%) while the reverse was true for non-early initiation (21.3% for cannabis and 35.6% for intoxication).

Table 1.

Percentages and means^a for outcome and predictor variables by sex and by race.

	Full Sample	American Indian (AI)	non-AI
Outcome variable
Ever misuse prescription drugs	17.4%	20.3%	13.3%
Demographic variables
Age:	16.2(1.0)	16.1(1.0)	16.2(1.0)
American Indian (AI):	58.1%	–	–
Female:	50.2%	51.7%	48.2%
Initiation variables
Early intoxication	19.3%	22.9%	14.3%
Non-early intoxication	35.6%	34.9%	36.5%
Early cannabis use	25.7%	35.9%	11.6%
Non-early cannabis use	21.3%	21.0%	21.7%
Nb	2580	1498(58.1%)	1082(41.9%)

^aStandard deviation in parentheses for Age.

^bDue to missing values, some calculations are based on fewer observations.

Table 2 presents the unit-specific odds ratios (OR) and their 95% confidence intervals (CI), respectively, for the five models of PDM. For all models, the intercept represents a 13-year-old white male.

Table 2.

Odds Ratios (OR) and 95% Confidence Intervals (CI) for five models predicting Prescription Drug Misuse (PDM).

	(1) Demographic variables		(2) Initiation - Intoxication		(3) Initiation - Cannabis		(4) Initiation Intoxication + Cannabis		(5) Final Initiation Intoxication + Cannabis
	OR	CI	OR	CI	OR	CI	OR	CI	OR	CI
Intercept	0.13**	(0.09, 0.19)	0.02**	(0.01, 0.04)	0.01**	(0.01, 0.02)	0.01**	(0.00, 0.02)	0.01**	(0.01, 0.02)
American Indian (AI)	1.48**	(1.15, 1.91)	3.61**	(1.39, 9.35)	2.78**	(1.39, 5.58)	4.02**	(1.51, 10.73)	3.12**	(1.37, 7.10)
Female	1.24*	(1.03, 1.50)	1.16	(0.95, 1.41)	1.28*	(1.05, 1.55)	1.24*	(1.00, 1.52)	1.24	(1.01, 1.52)
Age	1.09*	(1.00, 1.18)	1.06	(0.95, 1.18)	1.13*	(1.01, 1.26)	1.12	(0.99, 1.26)	1.12	(0.99, 1.26)
Early intoxication	-	-	35.73**	(16.53, 77.27)	-	-	6.07**	(3.24, 11.33)	3.98**	(2.57, 6.18)
Non-early intoxication	-	-	13.95**	(5.89, 33.02)	-	-	3.46**	(1.87, 6.40)	2.37**	(1.48, 3.79)
AI*early intoxication	-	-	0.28**	(0.12, 0.64)	-	-	0.57	(0.29, 1.14)	-	-
AI*non-early intoxication	-	-	0.27**	(0.12, 0.62)	-	-	0.60	(0.36, 1.01)	-	-
Early cannabis use	-	-	-	-	47.00**	(25.10, 88.02)	21.47**	(12.44, 37.05)	24.95**	(14.73, 42.25)
Non-early cannabis use	-	-	-	-	16.41**	(8.43, 31.96)	7.76**	(4.60, 13.11)	9.22**	(5.18, 16.39)
AI*early cannabis use	-	-	-	-	0.25**	(0.11, 0.57)	0.27**	(0.11, 0.68)	0.21**	(0.09, 0.54)
AI*non-early cannabis use	-	-	-	-	0.22**	(0.08, 0.62)	0.24*	(0.08, 0.76)	0.18**	(0.06, 0.58)

^*:p <.05;

^**p <.01.

Demographic variables only

The demographic-only model shows that PDM varied significantly by age and race. As expected, the likelihood of misusing a prescription drug significantly increased as age increased (OR_age = 1.09, CI: 1.00, 1.18). As for race, the odds of AI PDM were 1.48 times (CI: 1.15, 1.91) that of a comparable, non-AI student.

Alcohol intoxication and PDM

Column 2 of Table 2 presents the results from a model that includes demographics and intoxication initiation variables. The results indicate a significant positive relationship of PDM to race but not to age or sex. In addition, tests found that neither OR_AI*early*OR_AI nor OR_AI*late*OR_AI were significantly different from 1 (χ² < .01(1); p > .50 for both tests), indicating that for students who had been intoxicated, the likelihood of PDM did not differ by race. For all students, any lifetime intoxication was positively related to PDM. Early initiation had the largest relationship to PDM (OR_early = 35.73, CI: 16.53, 77.27) followed by non-early initiation (OR_non-early = 13.95, CI: 5.89, 33.02). For those with no lifetime intoxication, AI students had a greater likelihood of PDM (OR = 3.61, CI: 1.39, 9.35) than non-AI students.

Cannabis and PDM

Column 3 of Table 2 gives the results of the model that includes demographics and cannabis initiation variables. Results mirror those for alcohol intoxication, with use of cannabis predicting a greater likelihood of PDM, and early initiators having a greater relationship to PDM compared to non-early initiators (OR_{early Mj} = 47.00; OR_{non-early Mj} = 16.41). The odds ratio for early initiation did not differ by race (OR_AI*early *OR_AI = 1; χ² = 2.32(1); p = .12). However, for non-early initiators, AI students were less likely to report PDM compared to non-AI students (OR_AI*early*OR_AI <1; χ² = 4.18(1), p = .04). For non-lifetime users of cannabis, AI students had a greater likelihood of PDM (OR_AI = 2.78, CI: 1.39, 5.58).

Alcohol intoxication, cannabis use, and PDM

Column 4 of Table 2 gives the results of the model that includes all previously discussed predictors, while column 5 gives the final model with the race x intoxication initiation variables excluded. In the final model, all initiation variables were significant predictors of PDM, indicating that intoxication or use of cannabis, whether as an early or non-early initiator, was significantly related to lifetime PDM. In addition, being an early initiator of cannabis use or intoxication had a greater relationship to PDM than being a non-early initiator. Results also indicated a significantly greater association between cannabis use and PDM than between alcohol intoxication and PDM. Specifically, the odds of PDM for an early cannabis initiator are 24.95 times the odds of an otherwise similar student who has not used cannabis (CI: 14.73, 42.25). This compares to the odds of PDM for an early initiator of alcohol intoxication of 3.98 (CI: 2.57, 6.18). Regarding race, AI students who were non-initiators of cannabis were significantly more likely than non-AI students to report PDM (OR_AI = 3.12). Conversely, AI students who were non-early initiators were less likely to report PDM than their non-AI counterparts (OR_{AI* non-early} *OR_AI <1, χ² = 5.29(1), p = .02). However, there was no difference by race for early initiators (OR_AI*early *OR_AI <1, χ² = 2.89(1), p = .09).

Discussion

Results of this study indicated that early onset of cannabis use and alcohol intoxication were significant predictors of PDM for both AI and non-AI adolescents. This is consistent with Hermos, et al. (13) who found that, for a large national sample of individuals aged 18–34 years who reported lifetime use of alcohol, early initiation of alcohol use (less than 14 years old) and early initiation of cannabis were associated with significantly greater risk of lifetime prescription drug misuse. In addition, it is consistent with other AI research that found a strong association between early initiation of cannabis and intoxication and being classified as a polysubstance user (defined as alcohol, cannabis, and at least one other of ten substances, e.g., cocaine, OxyContin (17);).

The likelihood of PDM did not differ by race for students who had either early or later onset of alcohol intoxication nor for those with early onset of cannabis use. These findings suggest that for AI and non-AI youth who have some shared living experience, early initiation of cannabis and intoxication are risk factors that are similarly related to PDM and that targeting early initiation for both groups of adolescents is critical in prevention of prescription drug misuse. However, levels of early initiation, especially for cannabis, differ greatly between these groups. For example, over three times as many AI youth were classified as early initiators of cannabis (35.9%) as compared to non-AIs (11.6%). Therefore, it is also critical to understand what factors are responsible for early initiation in these groups and whether those factors and their effects are different. Although these students attend the same schools, a greater percentage of white students, compared to AI students, attend schools in the sample that are near reservations rather than those that are on reservation, and most likely live off-reservation, where in general, substance use, norms, availability and other factors may differ as compared to on reservation.

AIs who did not report having ever been intoxicated or using cannabis were significantly more likely to report PDM compared to their non-AI counterparts who had not been intoxicated or used cannabis (Model 5 OR_AI = 3.12). The reasons for this cannot be discerned from this study but warrant additional research. For example, what other risk factors (e.g., greater availability) may be present for AI youth that are not present at a similar level for non-AI youth?

Finally, for non-early users of cannabis, AI students were less likely to report PDM compared to non-AI students. Stanley and Swaim (18) found that for each of two grade groups of AI students (grades 7–8 and grades 9–12), there was a large group of cannabis users (greater than 20%) who were unlikely to use alcohol and other substances. Swaim, Stanley, and Beauvais (19) found that reservation AI adolescents perceived weaker injunctive norms for cannabis use (i.e., less disapproval) from peers and adults in the community, as compared to white youth attending the same schools. This may indicate that cannabis use is seen as more acceptable among AIs, especially for older adolescents, and does not necessarily indicate greater likelihood for use of other substances.

Turning to the results from the model with both cannabis use and alcohol intoxication initiation, results indicated that the highest risk of PDM is for a student who was an early initiator of both cannabis use and alcohol intoxication than when they reported one or the other. In addition, results showed a significantly greater association between cannabis use and PDM than between alcohol intoxication and PDM when both were included in the same model. These results are consistent with findings from Hermos, et al. (13) who found that both early cannabis and alcohol use predicted greater likelihood of PDM, and inclusion of nonuse of cannabis and age of cannabis initiation into their model resulted in a reduction in odds ratios for initiation of alcohol at age 14 or less of over 70%. Our study is different, however, in that we do not measure alcohol use initiation but rather age of first alcohol intoxication (84.3% of those reporting they had drunk alcohol had also reported being intoxicated). Nevertheless, why age of first cannabis use would be associated with an odds ratio over 5 times that for first alcohol intoxication is not clear. Magid & Moreland (12) note that many of the predictors of later substance misuse are also predictors of early substance use initiation. Thus, there may be underlying factors that are salient predictors for both early cannabis use and PDM, such as those found for early alcohol initiation and later substance misuse (see Magid & Moreland (12) for a review). As with alcohol, these may include family practices, such as parental monitoring and family conflict (20), and behavioral disinhibition (21). Further research should explore identifying those predictors in order to determine the most effective targets for prevention. To test for synergistic relationships between early initiation of intoxication and cannabis, we included an interaction variable between these variables but found no improvement in model fit (χ² = 2.89 (1), p = .09).

Limitations of the study include that the data are self-reported, cross-sectional, and may contain bias or faulty recall. In addition, whether the respondent began misusing prescription drugs prior to, during, or after first intoxication and/or cannabis use is unknown. Future studies would benefit from developing a longitudinal approach. This would also allow researchers to track the progression of an adolescent’s substance use to ascertain the timing of substances. In addition, the two groups of students do not necessarily live in the same communities or under the same circumstances. At the same time, non-AI students are exposed to many of the same sociocultural factors present on reservations, and generally live in areas where SES is below the general population of the U.S., making these students distinct from the general population and thus, limiting the generalizability of our findings. In addition, although data were from a geographically diverse sample of reservation schools, it was not a random sample of all reservation schools due to voluntary participation. Thus, results may differ from those obtained from the entire population of reservation youth. In addition, the sample contained a greater percentage of students from the Upper Great Lakes and Northern Plains and a lower percentage of students from the Southwest compared to the distribution of youth living on reservations and tribal lands. However, the data in this study still represent one of the largest samples of reservation adolescents of which we are aware. To control for potential regional relationships, we included regional variables into the original models and found that their inclusion did not improve model fit or change any other results appreciably. Finally, these data were collected over 4 years, but temporal effects were not included in the models due to the relatively small number of schools surveyed each year.

In conclusion, this study demonstrated significant relationships of PDM with age of first alcohol intoxication and age of first cannabis use for a sample of AI and non-AI adolescents living on or near a reservation. Relationships between age of first alcohol intoxication and PDM did not differ by race (AI vs. non-AI), but relationships did differ for age of first cannabis use in that later initiation of cannabis use resulted in less likelihood of PDM for an AI student as compared to a non-AI student. The findings of this study highlight the need for early prevention of substance use for all individuals attending these schools with a special focus on delaying the onset of early alcohol intoxication and early use of cannabis. Delay of onset for both has the potential to reduce risk for PDM.