Culturally Adapted Brief Intervention for Heavy Drinking Hispanic Men

Overview

In this comparative-effectiveness study, investigators will recruit 400 English-speaking, Spanish-speaking, or bilingual heavy-drinking Mexican-origin men admitted to a community hospital for medical treatment of an alcohol-related injury or heavy drinking. Participants will be randomized to receive a culturally adapted brief motivational intervention (CA-BMI) or a non-adapted brief motivational intervention (NA-BMI). The primary outcomes of interest include alcohol use, alcohol problems, and treatment utilization. Secondary outcomes include therapeutic alliance ratings and social support. Telephone follow-up assessments will be completed at 3, 6, and 12 months post-treatment.

Non-Adapted Brief Motivational Intervention: The core components of NA-BMI are consistent with the person-centered approach of MI and include 1) providing personalized feedback based on screening and baseline assessment results; 2) exploring decisional balance (pros and cons) of alcohol use from the patient's perspective; 3) building motivation for change through the assessment and discussion of the patients' selfreport of levels of importance, confidence, and readiness to change drinking; 4) enhancing commitment to change by exploring the patient's options for change and developing a change plan if indicated or desired; and 5) providing referrals for formal treatment of alcohol problems and other community resources. The NA-BMI will not specifically target cultural risk or protective factors beyond any normal tailoring that may occur in a standard BMI as described in the current literature. In NA-BMI, personalized feedback will be based on drinking norms and frequency of alcohol problems from the U.S. general population. Culturally Adapted Brief Motivational Intervention: CA-BMI also adheres to the core principles of MI and practice of BMI. In CA-BMI, the core components of NA-BMI are adapted to be culturally responsive to the unique risk (acculturative stress) and protective (familism) factors associated with heavy drinking, alcohol problems, help seeking, and treatment utilization among Latinos. It is important to note that CA-BMI goes well beyond any tailoring that may occur in NA-BMI by targeting factors that are important predictors of drinking among Latinos. Specifically, there are two primary adaptations to the CA-BMI: 1. CA-BMI will incorporate the assessment and personalized feedback on the impact of acculturative stress on drinking so as to decrease temptation to drink and increase confidence to avoid drinking. Specifically, participants will receive feedback about the types and intensity of acculturative stress they may experience (e.g., issues related to immigration, cultural congruity, language barriers, and employment discrimination), and clinicians will elicit the relationship of acculturative stress to temptation and confidence to avoid drinking. 2. CA-BMI will also integrate family and community as reasons for change and as agents of behavior change when considering the impact of drinking, plans for changing drinking behavior, and engagement in help-seeking behaviors. Following methods developed by Lee et al. (2011) and Añez et al. (2008), consultants on the grant, investigators will incorporate a discussion of how social context and family dynamics are related to drinking. These modifications result in a culturally adapted intervention that is substantially distinct in its content and focus (e.g., deep structural changes) from a non-adapted intervention, while maintaining consistency with motivational interviewing and its application in brief alcohol intervention. In accord with the two central adaptations, investigators anticipate that the potential mediators or mechanisms of behavior change specific to CA-BMI are 1) temptation to drink and confidence to avoid drinking and 2) increased support from family and friends in general as well as specific support to change drinking behavior and seek treatment. Finally, investigators will also evaluate a definition of treatment utilization that is more comprehensive than that in the investigators prior study, which assessed the use of formal inpatient and outpatient substance abuse treatment and attendance to self-help groups such as Alcoholics Anonymous (Field, et al., 2010). In the current study, investigators will assess engagement in formal treatment networks as well as informal help-seeking common among Latinos (e.g., seeking help from family, religious leaders, or respected elders in the community). Statistical Analyses Preliminary Analyses: Standard examinations for outliers, data distribution, and internal consistency of measures will be conducted. For mixed models, investigators will assess the homogeneity of error and normality of residuals at all levels of the model, test for multivariate normality of random effects, examine linearity, and identify outliers. For structural equation models (SEM), investigators will follow the best practice guidelines outlined by Boomsma (2000) for analyzing and reporting SEM models. Investigators will also compare groups on all demographic and pretest variables to assess whether randomization produced equivalent groups; in the event of nonequivalent variables, these variables will be included as covariates in models. Data Analysis for Specific Aim 1: Analyses investigating group differences in alcohol problems and treatment utilization will use random coefficient models (Raudenbush \& Bryk, 2002; Singer \& Willett, 2003). Investigators will construct longitudinal models using the following sequence of analytic steps recommended by Singer and Willett (2003): 1) examine empirical growth plots; 2) fit an unconditional means model; 3) fit an unconditional linear growth model; 4) fit an unconditional non-linear model (e.g., piecewise model); 5) determine the best model of longitudinal change by comparing models in the previous two steps using the Akaike information criterion (AIC); (f) select the most appropriate error covariance structure using AIC; and 6) add level-2 predictors (e.g., intervention conditions). Models for binary outcomes (e.g., treatment utilization) will use generalized linear mixed-effects models assuming a binary distribution with a logistic link function. Data Analysis for Specific Aim 2: Potential moderators will be examined by constructing interaction terms between treatment and a priori moderator variables (e.g., acculturative stress) to examine the possibility that the relationship between a putative moderator and outcome differ across treatments (Aiken \& West, 1991). In the event of a significant interaction that indicates moderation, investigators will probe the relationship methods appropriate for multilevel models (Bauer \& Curran, 2005). Mediation analysis will be conducted using a growth-curve framework implemented in an SEM. Models will be constructed by first fitting growth models for mediators and outcomes and then fitting mediational growth models. Investigators will follow the same sequence described above for establishing the best model of longitudinal change for Aim 1. Latent growth models will be comprised of at least two latent factors; one factor will represent the initial status, and one or more factors will represent the growth rate of a variable, where more than one factor will be required in the event of non-linear change (e.g., a quadratic term). Mediation will be examined following recommendations by MacKinnon (2008) for assessing mediation in the growth models context. The growth factor of the mediator will be regressed on the initial status of the mediator, the outcome, and the intervention group. A significant effect for the intervention group establishes a relation between the intervention group and the mediator, controlling for baseline levels of the mediator and outcome. Next, the growth factor will be regressed on the initial status of the mediator, the outcome, the slope of the mediator, and the intervention group. A significant effect of the mediator growth factor establishes a relation between change in the mediator and change in the outcome, controlling for baseline levels of the mediators and outcome. Data Analysis for Aim 3: Responses to patient and interventionist satisfaction and assessment of working alliance will be compiled in aggregate form. The frequency of responses to individual items will be reported for patients and interventionists. Likewise, scale scores for patients and providers will be reported using means and standard deviations. Comparison of responses of patients and interventionist will be made using chisquare in the case of frequency data and t-tests in the case of scale scores. Organizational readiness will be assessed using a pretest-posttest design. The analysis of pretest-posttest comparison will employ Analysis of Covariance or ANCOVA. In this nonrandomized design, the main purpose of ANCOVA is to adjust the posttest means for differences among groups on the pretest, because such differences are likely to occur. The purpose of using the pretest scores as a covariate in ANCOVA with a pretest-posttest design is to (a) reduce the error variance and (b) eliminate systematic bias.