Effects of Alcohol Interventions on Other Drug Use in the Cherokee Nation

Abstract

Objectives. To evaluate effects of 2 alcohol prevention interventions—Communities Mobilizing for Change on Alcohol (CMCA), a community organizing intervention designed to reduce youth alcohol access, and CONNECT, an individual-level screening and brief intervention approach—on other drug use outcomes.

Methods. We conducted a community intervention trial with quarterly surveys over 3 years (2012–2015) of high school students living within the jurisdictional service area of the Cherokee Nation in Oklahoma. We used generalized estimating equations and linear probability models to examine intervention spillover effects on other drug use.

Results. We found significant reductions in drug use other than alcohol attributable to CMCA and CONNECT. CMCA was associated with a 35% reduction in chewing tobacco use, a 39% reduction in marijuana use, and a 48% reduction in prescription drug misuse. CONNECT was associated with a 26% reduction in marijuana use and a 31% reduction in prescription drug misuse.

Conclusions. Nonalcohol drug use was consistently reduced as a result of 2 theoretically and operationally distinct alcohol prevention strategies. Evaluations of alcohol prevention efforts should continue to include other drug use to understand the broader effects of such interventions.

Alcohol is the most used drug by adolescents, followed by marijuana and tobacco. Polysubstance use is common, with more than a third of adolescents reporting recent use of both alcohol and marijuana or use of alcohol, marijuana, and cigarettes.¹ Among users of both substances, alcohol and marijuana are more likely to be used concurrently than alone.² Concurrent use of tobacco, alcohol, or other drugs is a concern because use of one influences consumption of the other and leads to negative health outcomes and reduced rates of cessation.³ Polysubstance use among adolescents is strongly associated with substance use problems, including substance use disorder diagnoses,¹ particularly among American Indian youths (with 21% prevalence of polysubstance use).⁴

In addition to epidemiological data showing co-occurrence of alcohol and other drug use, both the intervention and the economic literature point to the possibility of spillover effects of alcohol prevention on other drugs. A previous alcohol prevention trial of White rural adolescents found that the intervention reduced alcohol, cigarette, and marijuana use among youths who had not initiated alcohol use by age 12.⁵ However, 2 recent meta-analyses found no evidence that alcohol-specific screening and brief interventions affect other drug use.⁶ The economic literature reflects some disagreement on whether alcohol and other drugs act as complements or substitutes, although our read of the epidemiological and economic literature suggests a preponderance of evidence for complementarity.^7–9

We previously presented results from a randomized trial of 2 alcohol prevention strategies among youths in the Cherokee Nation.¹⁰ Both Communities Mobilizing for Change on Alcohol (CMCA), a community organizing intervention designed to reduce youth alcohol access, and CONNECT, an individual screening and brief intervention strategy, were effective in reducing adolescent alcohol use and related consequences. Given the prevalence of early-onset polysubstance use, simultaneous use patterns, and important health sequelae of polysubstance use, we evaluated possible spillover effects of CMCA and CONNECT on other drug use.

METHODS

We conducted a community intervention trial with quarterly data over 3 years (2012–2015; 12 waves) to examine the effectiveness of 2 alcohol prevention interventions implemented alone or in combination. We purposively selected 6 of 12 potential study communities in the Cherokee Nation in Oklahoma based on similar characteristics and with higher substance use risk profiles compared with other candidate communities^10,11 and randomly assigned them to 1 of 4 study conditions: CMCA only (n = 1 community), CONNECT only (n = 1), both interventions combined (n = 2), and delayed intervention control (n = 2). Details on the theory, interventions, trial design, and data collection were published before outcome data collection.¹¹

Outcome Measurements

To assess effects of our alcohol prevention interventions on other drugs, we dichotomized each standard survey item on self-reported use or misuse in the past 30 days of cigarettes, chewing tobacco, marijuana, prescription drugs, and other illegal drugs into categories of any use versus no use in the past 30 days. Then, we summed these variables to measure the total use of nonalcohol substances reported in the past 30 days.

Statistical Analysis

We used generalized estimating equations to estimate intervention spillover effects on the number of other reported drugs across follow-up measurements. For drug-specific models, we estimated linear probability models with intervention-by-time interactions to estimate additive effects of our interventions on use of each drug over time. Average treatment effects were estimated by the average mean difference in the number of reported drugs used in the past 30 days and the mean difference in the probability of using a specific drug between each intervention and the control condition, accounting for baseline differences between intervention conditions and control.

Multiple imputations (m = 20) accounted for nonresponse over time, and combined treatment effects were estimated with PROC MIANALYZE (SAS version 9.4, SAS Institute, Cary, NC). We combined χ² statistics across imputations with miceadds in R version 3.2 (http://www.R-project.org).

All models controlled for baseline differences in the outcome, gender, race, and age. All analyses were conducted with GENMOD in SAS version 9.3, with the continuous-time first-order autoregressive correlation structure, denoted AR(1), for the residuals. We subsequently verified results from all linear probability models with corresponding logistic models.

RESULTS

Among the 1623 students surveyed, 25% reported using 1 or more nonalcohol substances during the past 30 days. Most commonly reported was cigarette use (14%), then chewing tobacco (10%), marijuana (9%), prescription drug misuse (8%), and other drug use (3%). No statistically significant differences were found between study conditions on baseline chewing tobacco, marijuana, prescription drug misuse, or other drug use. Students in the CONNECT-only condition reported lower cigarette use (8%) compared with other conditions (approximately 14%; χ² = 8.03; P = .046). Each individual drug item had less than 1% missing data at baseline; wave-specific response rates varied from 83% to 90%.

Both interventions were associated with statistically significant decreases in the number of nonalcohol drugs used in the past 30 days. Compared with the control condition, CMCA-only participants reported using 0.20 fewer substances per wave (95% confidence interval [CI] = −0.30, −0.11; t = 4.10; P < .001), CONNECT-only participants reported using 0.11 fewer substances per wave (95% CI = −0.21, 0.00; t = −2.03; P = .042), and participants from the combined condition reported using 0.10 fewer substances per wave (95% CI = −0.18, −0.02; t = −2.42; P = .016). The change in number of reported nonalcohol drugs was driven by changes in chewing tobacco use, marijuana use, and prescription drug misuse (Table 1).

TABLE 1—

Effect of Alcohol-Specific Interventions on Nonalcohol Drug Use Outcomes: Cherokee Nation, Oklahoma, 2012–2015

	CMCA (n = 208), Estimate (95% CI)	CONNECT (n = 224), Estimate (95% CI)	Combined (n = 603), Estimate (95% CI)	Intervention by Time P
Change in number of substances reported in the past 30 d	−0.20 (−0.30, −0.11)	−0.11 (−0.21, 0.00)	−0.10 (−0.18, −0.02)	.002
Change in past 30 d cigarette use, %	−1.45 (−5.64, 2.74)	−1.60 (−5.09, 1.90)	1.06 (−2.36, 4.47)	.39
Change in past 30 d chewing tobacco use, %	−4.37 (−7.97, −0.76)	−0.84 (−4.32, 2.57)	−0.35 (−3.04, 2.34)	.005
Change in past 30 d marijuana use, %	−6.75 (−10.08, −3.43)	−4.49 (−8.18, −0.80)	−4.00 (−7.10, −0.90)	.015
Change in past 30 d illegal prescription drug use, %	−5.01 (−7.57, −2.45)	−3.21 (−5.89, −0.53)	−3.74 (−6.07, −1.41)	.029
Change in past 30 d other drug use, %	−3.43 (−5.54, −1.32)	−1.58 (−3.77, 0.61)	−2.41 (−4.29. −0.52)	.06

Note. CI = confidence interval; CMCA = Communities Mobilizing for Change on Alcohol. All estimates are relative to a delayed intervention control group (n = 588). All models controlled for baseline differences in the outcome variable, gender, race, and age.

DISCUSSION

In previously reported results, we showed a reduction in 30-day alcohol use, heavy episodic alcohol use, and alcohol-related consequences resulting from both interventions among a cohort of American Indian and White high school students in the Cherokee Nation.¹⁰ Decades of research indicates that drug use among adolescents typically covaries across substances and occurs contemporaneously^1,7–9 because of numerous possible genetic, developmental, social, and environmental mechanisms.¹² Results reported here show significant reductions in nonalcohol drug use attributable to CONNECT and CMCA, despite the fact that these substances were not specifically targeted by either intervention. CMCA was associated with a 35% reduction in chewing tobacco use, a 39% reduction in marijuana use, and a 48% reduction in prescription drug misuse. CONNECT was associated with a 26% reduction in marijuana use and a 31% reduction in prescription drug misuse. No synergistic effects were observed between CMCA and CONNECT, and no differences in baseline substance use or implementation fidelity were observed that might explain this. However, findings for marijuana use and prescription drug misuse were replicated in the communities receiving both interventions.

Findings for CONNECT, the alcohol-specific screening and brief intervention strategy, run contrary to those for recent meta-analyses. However, much of the screening and brief intervention literature is restricted to interventions lasting no more than 4 weeks.⁶ Our school-based, universally implemented adaptation of screening and brief interventions involved regular screening and brief motivational interviewing over multiple years. More frequent contact over a longer time than typical may be needed to have broader effects regarding nonalcohol substances. The spillover effects of CMCA on other drug use has not been examined previously.

It is important to consider study limitations when examining these findings. Our trial was designed as an alcohol prevention trial; thus, measures of nonalcohol substance use were limited. Because of budgetary constraints, a small number of communities were randomly assigned to the treatment conditions. Nevertheless, we incorporated additional design elements, including intensive longitudinal measurements, to improve causal inference.¹¹

PUBLIC HEALTH IMPLICATIONS

Reductions in nonalcohol drug use resulted from 2 theoretically and operationally distinct alcohol prevention strategies. Adolescent alcohol prevention efforts should extend beyond alcohol-specific evaluations to understand their full effects. Further research is needed to understand the underlying mechanisms for alcohol-specific intervention spillover effects on other substances.