Preventing substance use among Latino youth: Initial results from a multistate family-based program focused on youth academic success

ABSTRACT

Background: Early initiation of alcohol or other substance use places adolescents at high risk for subsequent substance use disorders. Research on preventing substance use among Latino youth significantly lags behind the growth of this population.Objectives: To assess the effects of a family-based intervention on past 30-day substance use in a population of Latino early adolescents (ClinicalTrials.gov Identifier: NCT03642106).Methods: A first study followed a sample of 265 Latino adolescents (51% female) over 4 years (7^th thru 10^th grades) using an interrupted time series design to compare pre- to post-intervention trajectories. The second study compared post-trajectory slopes from the intervention group to a subsample of participants from the National Longitudinal Study of Youth (NLSY) who identified as Latino and were matched on age and gender. Both studies used a zero-inflated Poisson modeling approach.Results: A piecewise random intercept growth model showed non-significant differences between pre- to post-intervention trajectories for both the probability and frequency of alcohol (p = .30, .47) and tobacco use (p = .10, .37), and a significant increase in the probability of illegal drug use (p < .01) but not frequency (p = .65). The NLSY group significantly increased their probability of use across substances (all p < .01), and increased their frequency of use for alcohol (p < .05) and tobacco (p < .01).Conclusion: Longitudinal assessments comparing Latino youth to a non-equivalent control group indicate that strengthening family involvement in youths’ schooling and promoting youth personal agency can prevent and/or reduce substance use during a developmental period in which use traditionally increases.

KEYWORDS:

• Latino adolescent
• Immigrant
• alcohol tobacco and other drug use
• Prevention
• Parental School Involvement

Introduction

Earlier age of initial substance use (SU) signals future deficiencies in social functioning, physical and mental health, and SU disorders in adulthood (1,2). For Latino youth, this is particularly worrisome for two reasons: (a) Latino population growth is among the fastest of any ethnic group in the U.S (3,4)., and particularly so for individuals under the age of 19 (5,6); and (b) early adolescent Latinos have an elevated rate of SU relative to non-Hispanic White youth (7). Latino immigrants are increasingly migrating to new areas (e.g., Midwestern and Upper South Central states) where there are scant bi-lingual and culturally appropriate resources (8). Developing and testing novel approaches to prevent or delay initiation into SU that are effective within this context are imperative. This paper reports findings from a pilot study examining whether a family-based intervention designed to increase academic performance and developmental assets among Latino immigrant 1^st (foreign-born) and 2^nd (U.S.-born to at least one foreign-born parent) generation youth can also prevent or delay initiation into alcohol, tobacco, and other drug (ATOD) use from the 7^th to the 10^th grades.

Background

Interventions that address antecedents common to multiple problems are gaining momentum over outcome-focused approaches that propose a unique intervention for distinct outcomes (4,9). One such common antecedent, positive parenting, has been inversely associated with multiple negative outcomes among Latino youth such as mental or behavioral health problems (e.g., ATOD use) (10,11); aggression and externalizing behaviors (12); hyperactivity, impulsivity, and anxiety (13); and the negative effects of stress on Latino adolescents (14,15). Parental school involvement, a component of positive parenting, plays an essential role in academic achievement and dropout rates among Latino youth (16–19). Low academic achievement is also strongly associated with adolescent health risk behaviors. Bradley and Greene (20) reviewed 25 years of publications on the relationship between academic achievement and the six leading causes of disability, social problems, and death among adolescents and young adults in the U.S. and found an inverse relationship in 97% of the studies. Furthermore, studies find that interventions that target general parenting skills and child school performance reduce health risk behaviors such as SU among Latino youth (11,21), suggesting that parental school-involvement may be a mechanism by which adolescent health risk behaviors can be reduced.

Adolescents who have more positive developmental assets (e.g., personal agency, positive peer influences) engage in less risky behaviors such as SU (22). Research has revealed that Latino youth had lower levels of self-esteem and fewer positive peer influences compared to other youth (23); this helps explain the higher rates of SU among Latino youth relative to other groups. Including mentoring or success coaching in prevention programs for youth increases developmental assets, which are associated with delayed or reduced SU (24).

Theories also support a focus on common antecedents rather than developing interventions focused on specific outcomes. Problem Behavior Theory (25) suggests that adolescents who have difficulties in one area of life are also likely to have difficulties in other areas. “Developmental cascades” (26) refers to the idea that parent-child interactions in one domain can influence other domains. Thus, competence in one domain of life becomes the scaffold for competence in other domains (27). Hirschi’s Social Control Theory (SCT) also supports the idea of cross domain effects. As youth develop attachment to social institutions their commitment to positive social norms also increases. Therefore, to the degree that parental involvement positively impacts youth attachment to home and to school, SCT suggests a corresponding generalized decrease in delinquent activities, such as ATOD use (28).

A focus on common antecedents of school related outcomes may also have pragmatic advantages. A key issue that weakens implementation and scale-up attempts of programs is the lack of “buy in” from schools and parents (29–31). Schools must balance the diverse needs of competing social programs (e.g., programs focused on bullying, SU, teen pregnancy, suicide) with increasing pressure to meet benchmarks for academic performance. Consequently, programs often face resistance resulting in fragmented and inadequate implementation. A focus on school-based outcomes may increase school-researcher collaborations while also impacting other areas such as ATOD use.

Difficultly recruiting and retaining families in programs creates issues with dosage and penetration (32). This can occur when the perceived goals of a program are not aligned with families’ goals (33). For example, programs focused on strengthening general parenting skills often draw fewer than 30% of minority parents when offered in schools (34). Some Latino immigrant parents may be reticent to attend ATOD prevention programs (35), but highly value education as a way out of poverty; many cite a better education for their children as their primary motivation to migrate (36,37). Because of this intrinsic motivation to be involved in their child’s education, many Latino parents may be more inclined to participate in programs that increase their children’s educational attainment.

Intervention description and conceptual model

The intervention was delivered simultaneously in two states under two different names: ¡Unidos Se Puede! (United We Can (38); in the Upper South Central state and Juntos Para Una Mejor Educación (Together for Better Education) in the Midwestern state (heretofore referred to as the intervention). The intervention’s conceptual model (see Figure 1) posits that stress exposure from migration-related experiences (e.g., fear of deportation, language barriers), discrimination, parental adverse childhood experiences, and impoverished living conditions combine to form chronic stress that weakens youths’ commitment and attachment to schools, teachers, and families (parents). The intervention seeks to strengthen youth’s commitments and attachments to these institutions through three core components: (a) a 6‐week family workshop series and seven monthly booster sessions that promote positive parenting and school involvement, (b) individualized success coaching for youth that is informed by a personalized strengths and needs assessment to promote youth self‐esteem and personal agency, and (c) youth groups (e.g., 4-H clubs) that nurture positive peer affiliations.

Figure 1. Conceptual model of the ¡Unidos Se Puede! Intervention.

We posit that weakened attachment to these pro-social institutions is associated with increased likelihood of SU. Stress may also interrupt mesosystemic protective factors such as parental involvement in school, which in turn can reduce youths’ sense of personal agency (e.g., emotion regulation, self-efficacy, and future aspirations) leading to increased risk for SU (39). According to our model, the path to SU can be interrupted by the three components of the intervention. First, it will assist parents in strengthening positive parenting skills, connect families to resources in the community and schools, and increase parents’ involvement and advocacy for their children in schools. Second, near-peer success coaches help youth strengthen their sense of personal agency, and serve as allies for parents and adult mentors for youth. Third, youth’s exposure to positive peers promotes prosocial bonding. Overall, the intervention will counter the effects of stress exposure on youth, which in turn reduces SU. See Cox (38) for a comprehensive overview of the theoretical underpinnings and description of the intervention.

Methods

Research design

We assessed the effects of the 20-month long intervention on past 30-day SU among Latino youth from 7^th to 10^th grade in one urban and two rural school districts from 2015 to 2018. We randomly selected youth who met the study inclusion criteria (i.e., self-identifed as Latino and were at-risk for not completing high school) to participate in the intervention from a list provided by the schools (see Cox (38) for a definition of how at-risk for not completing high school was defined).

The first baseline data point (time-1) was collected upon enrollment. Data for a second baseline data point (time-2) was collected approximately 1 month later, but prior to the intervention (~October of the 7^th grade). Subsequent data points were collected in January (time-3) and May (time-4) during year one of the study, and in September (time-5) and May (time-6) during year two (8^th grade), which also marked the end of the intervention. Follow-up data points were collected in September (time-7) and May (time-8) of study year three (9^th grade) and in September (time-9) and May (time-10) of study year four (10^th grade). Youth did not receive compensation for completing the surveys. The respective university and school district IRBs approved all research protocols and documents used in the study.

Participants

Youth in this study are drawn from two samples: a subsample of the National Longitudinal Study of Youth cohort 97 (NLSY) and youth from the intervention group. The NLSY sample consisted of 1476 Latino youth matched on age (12 years at baseline) and gender (49% female); for a description of participants in the intervention group see Table 1.

Table 1. Demographics and attrition

	N = 265	%
Male	130	49.00%
Female	135	51.00%
Free reduced Lunch	239	90.00%
Mexican Origin	236	89.00%
Central American	16	6.00%
South America	13	5.00%
Survey Language – English	233	88.00%
Survey Language – Spanish	32	12.00%
Live in US whole life^a	151	57.00%
Completed Intervention	212	80.00%
Completed 4 year follow up	95	36.00%

^aThe only variable different between sites and by attrition.

Measures

Past 30-day ATOD use

SU among youth in the intervention group was measured by the following three items: “During the past 30 days, how many times did you (a) drink alcohol more than a few sips, (b) smoke cigarettes, and (c) take an illegal drug (for example: marijuana [cannabis], sniffed glue, meth, pills not prescribed to you by a doctor, etc.)?” Response options were on a nine-point scale ranging from 0 (never) to 8 (eight or more times) for each category of substance. Past 30-day use was collected at each wave described above.

In the NLSY, where data is collected annually, the same three items and response options were included, with the exception that the youth were asked about cannabis use rather than broader illegal drug use. The NLSY measure on other illegal drugs did not follow the same pattern of past 30-day use and thus it could not be combined. However, the Monitoring the Future study notes that cannabis is the major driver for illegal drug use (7). The ZIP model parses each of the three substances into a use versus no use variable that models the probability of being in the non-user group in the binary portion of the Poisson distribution (reported as odds ratios), and a frequency of use variable that models the probability of moving from one level of frequency of use to another in the tail of the Poisson distribution (expressed as incident rate ratios). Because there is a direct mathematical connection between odds ratios (OR), incident rate ratios (IRR) and probability, and because ORs and IRRs are commonly used as effect sizes, we present ORs and IRRs in the tables, but discuss them as probabilities in the text.

Analytic plan

We designed two tests to assess the effects of the intervention on past 30-day ATOD use. First, an interrupted time series (ITS) compared pre-intervention trajectories to post-intervention trajectories. ITS models a continuous sequence of observations on a sample taken repeatedly over time. In our study, two assessments of past 30-day ATOD use were taken to establish an underlying baseline trend, which is ‘interrupted’ by the intervention at a known point in time. The hypothetical continuation of the baseline trend was the counterfactual against which the impacts of the intervention can be compared. We utilized a piecewise random intercept growth model to estimate the pre-intervention trajectory separate from the post-intervention trajectory. The pre-intervention trajectory covers approximately 30 days between the first and second baseline time points and was collected at the beginning of the 7^th grade school year. The post-intervention trajectory covered the second half of the 7^th grade through the end of the 10^th grade.

The second test of the intervention compared the post-trajectory slopes in ATOD use from the intervention group to a subsample of participants who identified as Latino and were matched on age and gender from the NLSY, using it as a nonequivalent comparison group. We recognized the age of the sample, but believed that as a nationally representative sample it offered insight regarding the expected trajectories of Latino youth. Maturation would suggest (7) that if we observed an increased ATOD use in the NLSY sample and and did not observe a parallel increase in our study sample, this would be evidence for a program effect (40). The NLSY is a national representative sample of youth that uses a panel design to follow youth across time but without an intervention. To compare the slopes for the intervention group to the slopes in the NLSY, we used the same random intercept growth model as we used in the ITS, but without a piecewise component.

The SU variables in the intervention group and in the NLSY were highly skewed with an over-abundance of zeros (no use). Therefore, we estimated the models as zero-inflated Poisson (ZIP) models (41). This approach allowed us to simultaneously estimate two regressions (1): a logistic regression predicted the probability of being in the true nonuse category (i.e., a latent class of individuals who would never use that drug that year); and (2) a Poisson regression predicted the frequency or extent of use among the latent class of those who would use that drug, including users estimated to use it zero times according to the ZIP distribution. Both analyses controlled for gender and were conducted using Mplus 8.4 (42). Missing data were handled using Full Information Maximum Likelihood. The study is registered with ClinicalTrials.gov (43) (Identifier: NCT03642106).

Results

Intervention group only test

The first test of the intervention deals with the interrupted time series design (ITS). See Table 2 for details of both tests with incident rate ratios and odds ratios as effects sizes (full results in supplemental Table 1). As predicted by the ITS design, the baseline slopes for responses regarding use of the three substances (i.e., alcohol, cigarettes, illegal drugs) show no significant differences between time-1 and time-2 with the exception of a slight increased probability for youth to be in the user group for alcohol (p= .007). Among the intervention group, analyses assessing for the frequency of SU in the past 30 days showed no significant differences between baseline-trajectories and post-trajectories for any of the three substances from the 7^th to 10^th grade. Similarly, the probability of being in the SU group over the past 30 days showed no significant increases for any of the three substances between the pre- and post-intervention slopes. There was a small but significant increase in the probability of youth in the intervention group being in the illegal drug use group from the 7^th to 10^th grade.

Table 2. Results of ITS model and comparison NLSY sample

Intervention
	Tobacco			Alcohol			Illegal Drugs
Frequency of Use	IRR	SE	p	IRR	SE	p	IRR	SE	p
Pre-Slope	1.08	0.06	0.19	0.94	0.05	0.16	1.04	0.06	0.52
Post-Slope	1.01	0.01	0.37	1.01	0.01	0.47	1.00	0.01	0.65
Sex	0.52	0.16	0.00	0.73	0.20	0.13	0.59	0.23	0.02
Use vs No Use	OR	SE	p	OR	SE	p	OR	SE	p
Pre-Slope	0.96	0.05	0.37	0.86	0.06	0.01	0.91	0.07	0.20
Post-Slope	0.99	0.01	0.10	0.99	0.01	0.30	0.97	0.01	0.00
Sex	1.07	0.26	0.79	1.53	0.28	0.13	1.11	0.29	0.72
NLSY
	Tobacco			Alcohol			Cannabis
Frequency of Use	IRR	SE	p	IRR	SE	p	IRR	SE	p
Pre-Slope
Post-Slope	1.36	0.09	0.00	1.25	0.11	0.04	1.29	0.17	0.13
Sex	1.28	0.19	0.20	0.95	0.17	0.77	1.06	0.25	0.82
Use vs No Use	OR	SE	p	OR	SE	p	OR	SE	p
Pre-Slope
Post-Slope	0.61	0.06	0.00	0.55	0.06	0.00	0.52	0.08	0.00
Sex	0.90	0.22	0.64	1.01	0.17	0.97	1.06	0.26	0.82

Intervention collected from 7^th to 10^th grade, with NLSY collected across similar ages. IRR stands for Incidence Rate Ratio. OR stands for Odds Ratio.

Intervention versus NLSY test

Next we compared the three substance-use slopes from the 7^th to 10^th grade in the intervention group to the same slopes from the NLSY group matched on age and gender. For the NLSY group, the frequency of use portion of the variable showed a significant slope for smoking and drinking, but not for cannabis use from the 7^th to 10^th grade. The use vs. no use portion of the variable showed significant increases in slopes from the 7^th to 10^th grade across all three substances.

Gender

In the intervention group, among youth reporting use during the past 30 days, females were significantly less likely to increase their use of illegal drugs and smoking, but not alcohol across all time points. There were no gender differences in the likelihood to be in the non-user group for youth in the intervention. Similarly, there were no gender differences in the NLSY group for either the use vs. no use or frequency of use portion of the variable.

Discussion

Although the primary goal of the intervention focused on improving academic performance, two different analyses suggest that the program may also prevent ATOD use among Latino immigrant youth. The ITS showed no significant change between baseline and post-slopes for the intervention group with the exception of a small, albeit significant, increase in the probability of having used illegal substances in the past 30 days from the 7^th to 10^th grade. This suggests the intervention was successful in preventing both the percentage of youth who used substances and an increase in frequency of SU among youth who used substances in the past 30 days. The increase in use for other illegal substances was unexpected. Additional investigation is needed to examine potential mediators and moderators that may help explain this result. However, the decrease in abstainers (i.e., increase in users) of illegal SU among the intervention group was small (OR = 0.97, 95% CI = 0.95, 0.99) compared to the decrease in abstainers (i.e., increase in users) of cannabis use for the NLSY sample (OR = 0.52, 95% CI = 0.44, 0.61), thus pointing to the potential for the intervention to diminish the effect even if it did not remove it. Figure 2 shows that over the course of the study there is a very small increase in the probability of the intervention group to use illegal drugs, but a considerable increase in cannabis use among the NLSY group over the same number of years. The other between groups comparisons are not shown because they mirror the comparison of illegal drugs.

Figure 2. Trajectories of illegal drug use (OR = 0.97, p= .00) for the intervention group and of Cannabis (OR = 0.52, p= .00) for the NLSY group.

Studies such as Monitoring the Future (9) consistently show increases in usage across all substances as youth age. The linear relation between SU and age is also illustrated in the analysis of data from the NLSY group, which showed significant increases in the probability of youth having used all three of the substances during the past 30 days and in the frequency of use for smoking and drinking among youth who had used during the past 30 days. Only the frequency of cannabis use showed a non-significant increase over time. The results of this study suggest that youth who were exposed to the intervention did not increase their probability of being in the user group or their frequency of SU as they progressed from the 7^th to 10^th grade. We present these null findings for change over time as initial evidence for the effectiveness of the intervention to potentially delay the onset of SU among Latino youth, and to decrease the frequency of SU among youth who have already begun to use substances.

Interestingly, this intervention does not, by design, address adolescent SU. Rather, the prevention of initiation into SU or the prevention of increased frequency of SU is seemingly associated with some other component or combination of components in the intervention. Social control theory suggests that youth SU is the result of detachment from prosocial institutions such as family and school. A plethora of research has demonstrated that “positive parenting” (10,11), peers (44,45) and school connectedness (46,47) are associated with decreases in youth SU. From an ecodevelopmental perspective, others have found that the intersection of microlevel influences in the mesosystem are associated with adolescent SU among Latino youth (48). For example, Cox and colleagues (10,39) found that increases in parental school involvement are associated with decreases in adolescent SU. Moreover, when a preponderance of parents are involved in their youth’s schooling within a given school, parental involvement can exert a school-level effect that reduces risk for all youth in that school (39). This suggests that prevention approaches that focus on prosocial bonding to parents and to school and the interaction between school and parents may hold advantages over traditional approaches that employ a more direct focus on the prevention of youth SU. Latino parents and youth are more willing to participate in interventions that address academic outcomes, and school administrators may be more supportive of such efforts, both of which may affect dosage and other key factors surrounding implementation (39,46,48). Furthermore, when parents are involved in their youth’s schooling they may transmit important values and expectations to their youth that buffer against the influence of negative peer groups and antisocial behaviors such as SU (44,49,50).

Limitations

Several limitations are noted. First, Latinos are a heterogeneous population that tend to cluster in different regions of the country and can vary by nationality, exposure to discrimination, and level of acculturation. Caution should be used when extrapolating these findings to other Latino communities and nationalities. Second, the NLSY is not a current representation of Latino SU, which may affect the interpretation of our results. Still, other more current nationally representative data also show that SU among Latino youth increases as youth age (51) bolstering the adequacy of the NLSY as a comparison group.

Third, drawing the sample from an academically underperforming population may have introduced a bias into the study. On the other hand, sampling from youth who need academic assistance may add ecological validity to the findings of this study in that it more closely mimics what happens in actual practice (i.e., high-performing students are less likely to engage in remedial programs).

Fourth, an additional limitation of the NLSY is that it is a national sample and may not reflect SU in the communities from which the intervention sample was drawn. In the intervention sample, most of the youth were first or second-generation immigrants, and acculturation may play a role in the differences found among the two groups. Latino youth who are less acculturated to US culture (i.e., intervention sample) may be less likely to use substances than more acculturated youth (i.e., NLSY sample). Thus, alignment with traditional Latino cultural values, expectations, etc. may play a protective role for less acculturated youth.

Fifth, the comparison between illegal drug use (including cannabis) in the intervention group and only cannabis in the NLSY is not a just comparison. However, the additional substances captured in the intervention group measure should bias against the intervention. This may help explain the small difference in the pre- versus post-slopes in the intervention group. Finally, although interrupted time series designs are one of the more robust quasi-experimental designs, they have inherent limitations that should be considered. For instance, ITS designs are susceptible to what Bronfenbrenner termed the Chronosystem influences (52) or effects due to historical time, such as changes in national policy around immigration. The threat of bias is ameliorated due to the longitudinal design spanning four years, our non-probability sampling methods, and the addition of a comparison group.

Strengths

The present study has several strengths that enhance the field of prevention science. First, this pilot study uses two sources of evidence to demonstrate the effects of the intervention: an interrupted time series design that followed Latino youth across four years; and a comparison to a matched sample of Latino youth from the NLSY. Second, most studies are conducted with Latino populations in established enclaves where Latinos have resided for decades. As Latinos move to other areas of the country, it is important to conduct studies in these different environmental contexts. Third, the prevention strategy used in this intervention is somewhat novel in that it was developed for Latino immigrant families residing in new settlement areas, does not rely on Master’s level clinicians or teachers, and can be delivered by community health workers or “promotors” (42), with the objective to improve academic outcomes rather than a focus on drug use. Thus the change in the typical trajectory of SU was an unintended, but welcomed result.

Conclusions

Reducing the instance of adolescent SU continues to be a major public health concern in the U.S. and is particularly salient among the fast-growing Latino population. This study tested the ability of a family-based intervention focused on youth academic success to reduce past 30-day SU among a sample of Latino youth. The approach embodied in the intervention described in this study incorporated specific elements that may have had a spillover effect that led to the prevention or reduction of SU. Such elements included: (a) tapping into the high value that Latino parents place on education; (b) activities to strengthen parent involvement in youths’ schooling, family communication and cohesion, youth confidence and competence, and a sense of community and support; (c) assistance to families in navigating information and resources; and (d) aligning with priorities held by school administrators. As the Latino population continues to increase in the U.S., it is important to develop and implement effective approaches and strategies that promote health and well-being and lessen health disparities, which includes the prevention of SU among Latino youth.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by funding from the United States Department of Agriculture, National Institute of Food and Agriculture, Children Youth and Families at Risk Grants [2013-41520-21026 to Dr. Ronald Cox, and 2014-41520-22189 to Dr. Kimberly Greder], and the National Institute of General Medicine [1 P20 GM 109097-01 A1 to Dr. Ronald Cox], the National Institute on Drug Abuse [R15DA049232 to Dr. Ronald Cox] and the George Kaiser Family Foundation.