Study site and recruitment
At present, approval has been obtained from the Institutional Review Boards from both The University of Texas at El Paso and University Medical Center (UMC) of El Paso. All data collection will be conducted at UMC of El Paso in El Paso, Texas. The ethical guidelines for research with human subjects followed by the present study protocol are compliant with the Helsinki Declaration. UMC is the largest public hospital (not-for-profit) located on the U.S./México border. It is a 396-bed, licensed, acute care facility and the only teaching hospital in the West Texas/Southern New Mexico region. UMC is also the only Level I Trauma Center within a 280 mile radius of El Paso. Eighty five percent of all trauma cases in the region annually are transported to UMC. In 2014, approximately 2,750 trauma patients were admitted to the UMC Trauma Center and approximately 75 % were Hispanic. The Emergency Department at UMC is also one of the busiest in the region, treating more than 55,000 patients annually, of whom approximately 80 % are Hispanic.
Design and procedure
In this comparative effectiveness study, we will recruit 600 English- or Spanish-speaking heavy-drinking men of Mexican origin who are admitted to the hospital for medical treatment of an alcohol-related injury or heavy drinking. Participants will be randomized to receive a CA-BMI or NA-BMI. The primary outcomes of interest are alcohol use, alcohol problems, help-seeking, and treatment utilization. Follow-up assessments will be completed at 3, 6, and 12 months post-treatment. See Fig. 1.
The UMC Clinical Coordinator and Screening Coordinator, in collaboration with other study staff, will ensure that comprehensive and standardized screening procedures are used to identify potentially eligible patients. At UMC, patients screen positive for an alcohol-related injury or heavy drinking based on 1) positive blood alcohol concentration (BAC), 2) initiation of a medical protocol for alcohol detoxification, or 3) a score ≥ 4 on the first three items of the Alcohol Use Disorders Identification Test (AUDIT-C), which is a gender-specific cut-off for heavy drinking among men. Based on current medical standards, all injured patients who have blood drawn have their BAC tested as part of routine laboratory tests. Once medically stable, all injured patients are screened by the UMC Screening Coordinator using the AUDIT-C, a standardized, internationally validated self-report measure of alcohol use, and heavy drinking [52, 53].
Eligibility criteria for this study are the following: 1) aged 18 or older, 2) treated for injury associated with a motor vehicle collision (involving driver, passenger, or pedestrian), a violence-related injury (i.e., gunshot, stab wounds, other assault related injuries), or a fall based on e-codes from the 10th edition of the International Classification of Diseases, and 3) screen positive for an alcohol-related injury or heavy drinking as described above. Exclusion criteria are: 1) traumatic brain injury as indicated by a Glasgow Coma Scale score < 15 , or 2) cognitive impairment as indicated by a score of ≤ 24 on the Mini-Mental Status Exam (MMSE) [55, 56]. Patients who do not successfully complete the MMSE will be monitored regularly by the Screening Coordinator to reassess participation. Intoxicated patients at admission will be approached during their hospital stay once medically stable.
Ethnicity will be confirmed by study staff using procedures similar to general population surveys (i.e., 2010 census). Response options include: 1) no, not of Hispanic, Latino, or Spanish origin, 2) Yes, Mexican, Mexican American, Chicano, 3) Yes, Puerto Rican, 4) Yes, Cuban, or 5) Yes, another Hispanic, Latino or Spanish origin . Patients who identify themselves of Hispanic, Latino, or Spanish origins, and only describe themselves as Mexican, Mexican American, or Chicano will be eligible to participate. In our prior study, 82 % of Hispanics identified themselves as Mexican, Mexican American, or Chicano only .
After obtaining informed consent, patients will complete the baseline assessment (described below). Subsequently, participants will be randomized to the CA-BMI or NA-BMI. Study staff will be blind to treatment assignment during the baseline assessment, and staff responsible for the assessment will not provide the intervention (vice versa). We will use an adaptive randomization procedure that minimizes imbalances in covariates  using a SAS macro . This procedure ensures adequate distribution of key characteristics (i.e., acculturation, nativity, language of assessment) which have been found to influence outcomes .
Brief motivational interventions
Both interventions (NA-BMI and CA-BMI) will adhere to MI principles [16, 17, 60], and practice of BMI. To help maintain the integrity of the two interventions, clinicians will be trained and supervised in the provision of only one of the interventions (NA-BMI or CA-BMI). As part of standard BMI protocols, all participants will receive personalized feedback based on results from their screening, as well as a personal drinking profile based on the results of the baseline assessment [61, 62].
Non-adapted brief motivational intervention (NA-BMI)
Consistent with MI, the core NA-BMI components include: 1) personalized feedback based on screening and baseline assessment, 2) exploring decisional balance of alcohol use from patient’s perspective, 3) building motivation for change through the assessment, and discussion of patients’ self-report of levels of importance, confidence, and readiness to change, 4) enhancing commitment to change by exploring patient’s options for change, and developing a change plan if indicated, and 5) providing alcohol treatment referrals. The NA-BMI will not target cultural risk or protective factors beyond any normal tailoring that may occur in standard BMI. Personalized feedback will be based on U.S. general population drinking norms, and frequency of alcohol problems.
Culturally adapted brief motivational intervention (CA-BMI)
CA-BMI also adheres to the core principles of both MI and BMI. In CA-BMI, core components of BMI are adapted to be responsive to the unique risk (i.e., acculturative stress) and protective (i.e., familism) factors associated with heavy drinking, alcohol problems, help-seeking, and treatment utilization in Hispanics. CA-BMI goes beyond any tailoring that may occur in NA-BMI by targeting important predictive factors of drinking in Hispanics. Two primary adaptations to CA-BMI will be made.
First, CA-BMI will incorporate the assessment, and personalized feedback on the impact of acculturative stress on drinking in order to decrease temptation to drink, and increase confidence to avoid drinking. Participants will receive feedback about the types, and intensity of acculturative stress they may experience (e.g., immigration-related issues, cultural congruity), and clinicians will evoke the relation of acculturative stress to temptation, and confidence to avoid drinking. Second, CA-BMI will integrate family, and community as reasons for change, and as agents of behavior change when considering the impact of drinking, plans for changing drinking, and engagement in help-seeking behaviors. Similar to methods developed by Lee et al. (2011)  and Añez et al. (2008) , the CA-BMI will incorporate a discussion of how social context and family dynamics may affect drinking. Feedback, and a discussion of social pressures (e.g., cultural gatherings) to engage in heavy drinking as well as changes in drinking patterns (e.g., concerns about drinking from family) will also be included. When developing a change plan, there will be an explicit discussion, and elaboration of changes in family or community networks that may facilitate reductions in drinking. Participants will be encouraged to identify, and actively engage family and community members as helpers in their efforts to change their drinking. For example, they may identify a respected member of the family or community with whom they can talk about their commitment to change. Participants’ cultural norms, expectations, and personal values will be discussed to illuminate discrepancies with current drinking behavior, which may increase motivation to change (see Añez et al., 2008 ).
These two central modifications result in a culturally adapted intervention that is substantially distinct in content, and focus from a non-adapted intervention, while still maintaining consistency with MI, and its application within brief alcohol interventions. Based on adaptations, we anticipate the potential mediators or mechanisms of behavior change specific to CA-BMI to be: 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, and seek treatment. We will evaluate a definition of treatment utilization that is more comprehensive than that in our prior study, which assessed the use of formal inpatient/outpatient substance abuse treatment, and attendance to self-help groups such as Alcoholics Anonymous . We will assess engagement in formal treatment networks as well as informal help-seeking common among Hispanics (e.g., family, religious leaders).
Training and treatment fidelity
We have determined that Hispanics respond more positively to BMI, and that patient—provider ethnic matching significantly influences drinking outcome above and beyond the effect of the intervention. Both NA-BMI and CA-BMI will be provided by bilingual, bicultural, and same gender interventionists. Therefore, gender, ethnicity, and ability to speak both Spanish, and English will be held constant across interventions. By holding these three factors constant we can determine the differential influence of the two interventions on alcohol-related outcomes. To avoid contamination, providers of CA-BMI and NA BMI will be trained, and supervised separately throughout the study period. Study staff who are responsible for the intervention will not be responsible for baseline or follow-up assessments.
Interventionist training and supervision
Dr. Field, the Principal Investigator (PI), will be primarily responsible for oversight of the training, and supervision of interventionists. Dr. Field has received advanced training in MI (including the use of treatment fidelity measures for supervision), and is a member of the Motivational Interviewing Network of Trainers (MINT). Dr. Castro, a Co-investigator (Co-I), will facilitate training, and supervision. Dr. Castro is a bilingual, bicultural clinical psychologist with expertise in culturally sensitive service provision, and extensive experience in providing Spanish-language services to Hispanic populations. Dr. Castro has experience with cultural adaptations of smoking cessation interventions based on MI . Training manuals for CA-BMI will be adapted from existing brief intervention manuals from previous trials (R01-AA13824; R01-AA015439) conducted by the PI. As part of the training process, the PI’s research team has developed a set of training videos that will serve for the purposes of both NA-BMI and CA-BMI training. These short videos portray typical BMI sessions through role plays including study investigators providing BMI. Prior to training, interventionists will read Motivational Interviewing , read the intervention manual (CA-BMI or NA-BMI), and watch the aforementioned videos. A two-day interactive training on MI principles, and the practice of brief interventions will be provided by a bilingual trainer who is a member of the MINT. In addition to the study PI and Co-I, a Training Coordinator, Patricia Juárez will be primarily responsible for the training and supervision of interventionists. Ms. Juárez has received advanced MI training, and is also a member of the MINT. Supervision will be provided by Ms. Juárez throughout the study, and will include the following: 1) weekly supervision, 2) on-site observation, feedback, and/or coaching sessions (at least one per month), and 3) a one-day booster training provided every six months. This provides trainees the opportunity to practice, receive feedback, and refine their skills.
Procedures to ensure treatment fidelity are based on recommendations from the NIH Behavior Change Consortium , which established best practices in treatment fidelity of behavioral interventions. Fidelity will consist of two components: 1) adherence, or whether the interventionist carried out specified procedures of CA-BMI or NA-BMI, and 2) competence, or the interventionist’s level of skill in implementing the intervention using MI. Adherence to CA-BMI or NA-BMI protocols, and competence in MI will be assessed throughout the study by reviewing and coding recorded interventions.
Adherence to CA-BMI and NA-BMI
Intervention protocols will be codified using behavioral checklists, which have been found to be useful in monitoring intervention delivery . These will be based on checklists from our prior studies (R01- AA13824; R01-AA015439 and R01-DA026088). The CA-BMI checklist will be adapted from a current checklist used to assess BMI adherence . Interventions will be coded using a checklist based on prescribed, and proscribed components of the respective BMI. Adherence checklists will be used in supervision, and to establish adherence and discriminability between CA-BMI, and NA-BMI.
Competence in MI
Interventionists’ competence will be assessed in both BMIs using the Motivational Interviewing Treatment Integrity (MITI) v3.1 . The MITI is a behavioral coding system that measures competence in MI by assessing the degree to which therapist behaviors are adherent or non-adherent to MI. Supervisors will complete the MITI on audio recordings of BMIs throughout the recruitment period. The MITI will be used to derive summary scores with benchmarks for basic, and expert proficiency in MI. We will use expert-level performance as the threshold for establishing, and maintaining competency. MITI summary scores will be used to provide feedback during supervision, and to demonstrate competence in MI.
Assessment of patient-oriented outcomes
This study will employ bilingual staff to conduct the assessments, which facilitates the option of reading questions aloud to participants, thereby assisting patients to complete the in-person baseline assessment during their hospital stay . All measures used were developed or tested in Spanish, and are available in both Spanish, and English. Staff responsible for baseline and follow up assessments are blind to participants’ intervention condition, and will be trained and supervised throughout the recruitment period.
The Timeline Follow Back (TLFB) will be used to obtain data on the prior month’s days of alcohol use, calculated average weekly drinks, days of binge use, average number of drinks per drinking day, maximum number of drinks on any day, and drinking above at-risk levels [69, 70]. The Short Inventory of Problems (SIP + 6) will be used to assess the number of alcohol-related consequences as an outcome variable. The SIP + 6 is a brief version of the Drinking Inventory of Consequences (DrInC) . The SIP + 6 contains five 3-item scales related to consequences in the physical, social responsibility, intrapersonal, impulse control, and interpersonal domains. This scale has been validated with both English and Spanish speaking Hispanics .
The Treatment Services Review (TSR) will be used to collect information regarding the receipt of services for seven potential problem areas—medical status, employment and support, drug use, alcohol use, legal status, family/social status, and psychiatric status [73, 74]. To complement this measure, the Mexican American Prevalence and Services Survey (MAPS) will be used to assess help seeking  through other methods, such as chiropractor, homeopath, minister, folk healer (i.e., curandero), spiritualist, santero, and sobador.
Social support will be measured using four subscales including social cohesion, support from friends and family, and family cultural conflict . The Important People and Activities (IPA) interview will complement the assessment of general social support by assessing social support specific to alcohol use [76, 77]. The IPA assesses support from family and friends through both structural, and functional processes. The IPA also assesses how people within the patient’s social network would react to the patient’s drinking on a five point Likert type scale ranging from 1 (left or made you leave) to 5 (encouraged your drinking).
Temptation and confidence
Situational temptation to drink will be measured using the 20 item Temptation to Drink Scale. This measure assesses how tempted an individual is to engage in a variety of health behaviors such as substance use. Confidence to avoid drinking will be measured using the 20 item Abstinence Self-efficacy Scale. This scale assesses an individual's confidence to abstain from drinking alcohol in various situations. Both scales assess temptation to drink or confidence to avoid drinking within these contexts: negative affect, social/positive, physical, and other concerns, and cravings/urges which may precipitate drinking [78, 79].
Readiness to change
The Readiness to Change Questionnaire (Treatment Version; RCQ [TV]) assesses a patient’s level of readiness to change their alcohol use (e.g., reduce, quit) with respect to the precontemplation, contemplation, and action stages of change . The RCQ (TV) includes 12 Likert-type items (3 per stage) ranging from −2 (Strongly disagree) to +2 (Strongly agree). A score is computed per subscale/stage, and the participant is assigned to the one with the highest score (in case where two or more scores are the same, participants are assigned to the stage that is furthest along the staging continuum [i.e., closer to action]).
Acculturation will be measured with the Multidimensional Acculturation Scale II (MAS II) . The MAS II includes 22 items (from 0 [does not apply to me] to 5 [very much/very well]) assessing 4 factors: English proficiency, Spanish proficiency, identification with Mexican cultural identity, and identification with American cultural identity. MAS II factors have been found reliable (α = .78-.93).
Acculturative stress will be measured with the Multidimensional Acculturative Stress Inventory (MASI), which lists situations that Hispanics may have experienced within past 3 months . The MASI includes 36 items ranging from 1 (never) to 5 (daily or almost daily). Identified factors are: Spanish competency pressures, English competency pressures, pressure to acculturate, and pressure against acculturation. Subscales (αs = .77 to .93), and the overall scale have demonstrated adequate reliability levels (α = .90) .
The Helping Alliance Questionnaire (HAQ-II) is part of the Penn Helping Alliance Scales [83, 84]. This self-report scale assesses experiences in therapy from both the patient and therapist's perspective. Thus, there is a patient version, and a therapist version. The HAQ-II includes 11 Likert-type items from 1 (completely disagree) to 4 (completely agree).
Approach to analyses
Analyses investigating group differences in alcohol problems, and treatment utilization will use random coefficient models [85, 86]. Longitudinal models will use the following sequence of steps recommended by Singer and Willett : 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, 5) determine the best model of longitudinal change by comparing models in the previous two steps using the Akaike information criterion (AIC), 6) select the most appropriate error covariance structure using AIC, and 7) add level-2 predictors (e.g., intervention group).
Potential moderators will be examined by constructing interaction terms between treatment, and a priori moderator variables (e.g., acculturative stress and familism) to examine the possibility that the relation between a putative moderator, and outcome differ across treatments . Mediation analysis (per recommendations by MacKinnon ) 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. 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. The growth factor of the mediator will be regressed on the initial status of the mediator, the outcome, and the intervention group. A significant intervention effect 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 mediators and outcome.
Power for the hypothesized intervention effects was estimated using Monte Carlo studies. For longitudinal mixed models, a simulation was conducted (per example by Gelman and Hill)  using the R software, and power for mediation models was estimated using simulations in the Mplus software . Culturally adapted interventions for other mental health problems have produced effect sizes of .45 . The proposed effect size is informed by our previous study, which produced effect sizes between .25 and .32 with an average effect size of .28 [20, 92]. For this study, we estimated effect sizes equivalent to d = .28 for continuous outcomes and h = .28 for binary outcomes (nearly a 50 % increase in treatment utilization), using effect sizes for mixed models that are equivalent to these classical effect sizes .
Power estimates were derived as the proportion of significant effects (two-tailed α = .05) across like simulations . We simulated 10,000 data sets of 400 observations under the assumption that missing data will be ameliorated with the strategies described in the missing data section, and all participants will contribute data to the analysis. Intraclass correlation was simulated at .65 based on the intraclass correlations observed in our prior study. Power was .81 for continuous outcomes, and .83 for binary outcomes. A power analysis for a moderation effect by simulating a three-way interaction between treatment, time, and a moderator (e.g., familism) was conducted. Per Kraemer and colleagues’ recommendations, the interaction was examined using plots, confidence intervals, and p values . We investigated the power for a hypothesized contrast at high, and low levels of a moderator (e.g., low v. high acculturative stress in participants in CA-BMI) through a simple slope contrast at 1 SD ± the moderator mean for participants . Power for an effect size equivalent to d = .29 was .78 for a continuous outcome, and power for an effect size equivalent to h = .27 was .82 for a binary outcome. Power analysis for the proposed mediation model was conducted using a Monte Carlo simulation . Using standardized regression coefficients, we conducted a simulation for 10,000 data sets in which the path from the intervention to the mediator was .12 (equivalent to d = .22), and the path from the mediator to the outcome was .30 for standardized variables, reflecting the stronger relation that we anticipate between the mediators and outcomes. The Monte Carlo simulation indicated that power for the indirect effect was .77. Thus, accounting for a potential 20 % participant attrition, the proposed sample size should provide sufficient power for all planned analyses.