Study design
Study 1 will employ a cross-sectional observational design to investigate the patterns of alcohol consumption and domestic violence using a household survey of women. Study 2 will be a randomized clinical trial (RCT) nested within Study 1 (i.e., the women identified as having harmful alcohol consumption in Study 1 will be invited to take part in the RCT.
Setting
The current study will be conducted in Juiz de Fora, Minas Gerais, southeastern Brazil; this city has approximately half a million inhabitants. The study will randomly select women living in two neighborhoods with different average monthly incomes per capita according to the Instituto Brasileiro de Geografia e Estatística (Brazilian Institute of Geography and Statistics; IBGE) census conducted in 2000. This categorization will be used to investigate two populations with different income profiles.
Sampling
To calculate the sample size needed for this cross-sectional study, the data from a previous study (which estimated the prevalence of harmful alcohol consumption among women in Juiz de Fora as approximately 9.1 %) will be considered [48]. In addition, this study considered a maximum measurement error of 3 %, 95 % confidence intervals, a finite population correction based on the number of women residing in the randomly selected neighborhoods (consistent with the 2000 IBGE census), an effect of the sampling design equal to 2, and a 25 % refusal to participate rate. Thus, the initial sample size was estimated as 824 respondents (412 in each neighborhood).
To calculate the clinical trial sample size, the following considerations will be made: (i) a significance level of 5 %; (ii) a statistical power of 80 % (to compare proportions of the two independent groups); (iii) a non-interview rate of 15 % due to refusal to participate or failure to locate the informant; (iv) the effect of the BI after 3 months (i.e., an expected reduction of at least 20 % in the number of women with harmful alcohol consumption; and (v) an effect of the sampling design equal to 2. Thus, the calculated sample size for each group (i.e., the treated group [TG] and control group [CG]) will be 73 women.
To make the second study possible, considering the prevalence of women with risky alcohol consumption in the two neighborhoods surveyed around the estimated above (ie, 9.1 %), there will be necessary an increase in the sample size for Study 1, which now has approximately 1600, with 800 in each of the neighborhoods.
To recruit the women who belong to the target population, a two-stage cluster sampling method will be adopted. The streets belonging to each neighborhood will be considered as primary sampling units (PSUs), and households will be considered as secondary sampling units (SSUs). To select PSUs, a simple random sampling method without replacement will be adopted, whereas the SSUs will be selected using systematic sampling. In each selected household, the field researchers will invites all women who met the inclusion criteria to participate in the study.
Inclusion criteria
Female residents in the two neighborhoods who are aged 18–60 years, literate, and without obvious cognitive impairments will comprise the target sample. Domestic violence behaviors will be investigated among women who have children of aged up to 18, husbands or partners, or both residing in the same household.
Measurements
Participant sociodemographic characteristics
Information concerning ethnicity, age, education level, religion, occupation, marital status, and number of children will be obtained. Among women who are married or cohabitating, information regarding the duration of their relationship will be obtained along with the sociodemographic characteristics of their partners (e.g., age, education level, and occupation).
Alcohol consumption
Alcohol consumption will be assessed using the Alcohol Use Disorders Identification Test (AUDIT), a screening tool developed by the WHO [49] that consists of 10 questions: 3 concerning consumption amount and frequency, 3 regarding dependence symptoms, and 4 concerning personal and social problems related to alcohol abuse. The final score reflects the following alcohol consumption levels or patterns: abstainers/low risk (0–7 points), hazardous use (8–15 points), harmful use (16–19 points), and probable dependence (20–40 points). The AUDIT has been previously validated in several countries, including Brazil, and shows good sensitivity (mean = 0.90) and specificity (mean = 0.80) regarding the detection of harmful alcohol consumption [49, 50].
Information concerning the respondents’ use of any healthcare service over the previous 3 months, the approach of health professionals regarding respondents’ alcohol consumption, respondent participation in any treatment for alcohol consumption, and the respondents’ perceptions of their partners’ alcohol consumption will also be obtained.
Consumption of other substances
Information regarding the consumption of tobacco, marijuana, amphetamines, benzodiazepines, antidepressants, and cocaine over the three months preceding the interview will be obtained.
Health
A structured 15-question instrument concerning the respondents’ general health and the presence of specific diseases such as hypertension, diabetes, and heart problems will be administered.
Depression
The population-based screening scale for depression from the Center for Epidemiologic Studies (i.e., the CES-D) will be used. This screening instrument seeks to identify depressive symptoms within adult population-based studies [51]. Consisting of 20 items, its total score ranges from 0 to 60 points, with the higher scores indicating a greater amount of depressive symptoms. The cutoff of 16 points is often used to classify individuals as having possible depression. The CES-D has been previously validated for use in Brazil and shows satisfactory levels of reliability [52, 53].
Domestic violence
The Revised Conflict Tactics Scales (CTS2) will be employed to collect information on violence between intimate partners, whereas the Parent–child Conflict Tactics Scales (CTSPC) will be used to assess domestic violence by women against their children.
The CTS2 was designed to assess violence within couples, and it provides data on the respondent and her partner. This survey consists of 78 items, and each item is displayed in a set of two questions. The first question of each set refers to a possible behavior of the respondent, whereas the second refers to the same action but experienced by the partner. This instrument consists of 5 subscales that address the occurrence of negotiation, psychological aggression, physical violence, consequences of violence affecting the health of the respondent and her partner (i.e., injuries), and sexual coercion within the relationship. The CTS2 shows an internal consistency between 0.65–0.86 and an intra-observer reliability (kappa) above 0.75 when evaluated in Brazilian samples to assess its conceptual equivalence, semantic equivalence, and psychometric properties [54–56].
The CTSPC was designed based on the refinement of the previous instrument, and it addresses the parent–child relationship. This instrument is composed of questions regarding the occurrence of certain behaviors directed at children. Its 22 items are divided into three levels: nonviolent discipline (e.g., explaining errors and applying punishment), psychological aggression (e.g., swearing, screaming, threatening to kick the child out, or hitting), and physical assault (e.g., face slapping, throwing the child on the floor, or threatening with a knife); the latter was subdivided into corporal punishment, physical abuse, and severe physical abuse. The CTSPC had previously been cross-culturally adapted for use in Brazil; furthermore, prior work found that its internal consistency ranged from 0.49–0.68, and estimates of its intra-observer reliability (kappa) were above 0.75 [57, 58].
According to Straus [59] (the author of the CTS2 and CTSPC), these instruments investigate the occurrence of behaviors that, unlike emotions, attitudes, and beliefs, are less susceptible to distortions with regard to the interpretation of facts. Importantly, the methodological choice for these assessments reflects an understanding of domestic violence as a complex phenomenon that tends not to remain restricted to only one member of the family [58]. Furthermore, research indicates that violent acts are typically reciprocal, i.e., the respondent might be both the aggressor and the victim [60].
The information regarding alcohol consumption provided by the respondent and her partner during episodes of violence will also be incorporated by adapting the violence scales. After each CTS2 and CTSPC item, the interviewer will ask whether one of the involved individuals was under the influence of alcohol when the event occurred.
The CTS2 and CTSPC will be self-applied, whereas all of the other instruments will be administered to participants during interviews. Except for the CES-D, which refers to the events of the previous week, all of the instruments will be adapted to consider a time period of three months.
Procedure
Study 1: Household survey
A researcher will approach the participants at their homes; all of the instruments will be administered there. The interviews will take approximately 30 min to complete and will be held in a place as private as possible to allow the interviewees to freely answer questions and increase the credibility of their answers.
The researchers will be properly trained women who will initially approach the participants to invite them to volunteer for the survey, delivering a document that describes the study and its objectives. The training of the team will consist of attending lectures on domestic violence, alcohol consumption, BIs, the project details, a clarification of the study’s procedures, and an explanation of the questionnaires. Furthermore, the importance of research ethics will be emphasized. The researchers will gain familiarity with the instruments by administering them to a colleague. Training will be provided on the application of a BI via videos and role-playing techniques.
To locate the women living in the randomly selected households, at least three visits will be made to each household. When a resident is not found at her supposed address, the researchers will seek information from the neighbors to identify the existence of any women who meet the eligibility criteria for the study and the best days and times to approach them.
Study 2: Random clinical trial
The data collection for Study 2 will be conducted concomitantly using the women detected in Study 1. Women with AUDIT scores equal to or above 8 will be considered as hazardous alcohol users and randomly allocated into one of two groups: the TG or the CG. The TG will receive a single BI after the instrument is administered, whereas the CG will not receive BI until after the end of the study.
The participants will be allocated using a list of random numbers. Three months after the BI, all instruments except for the sociodemographic questionnaire will be re-administered to both groups. Because of ethical reasons, a BI session will also be offered to the CG after the second evaluation, at the end of the study. Independent interviewers will perform the baseline and the follow-up interview.
The intervention
The BI will be performed at the participants’ households in accordance with the model proposed by Miller and Sanchez [61], the principles of which are summarized by the acronym “FRAMES”. This acronym stands for “Feedback” (provide information regarding the score obtained on the screening instrument), “Responsibility” (the patient must take responsibility for changing their behavior), “Advice” (guidance on issues related to the use of substances and coping strategies), “Menu of options” (a list of options or alternative behaviors to substance use), “Empathy” (putting one’s self in another’s shoes based on one’s own assumptions to try to understand their behavior), and “Self-efficacy” (patients’ belief in their ability to change).
The timing of the BI will be controlled, ranging from 15 to 40 min. After administering the procedure, the researcher will record the approach used and the participant’s behavior during the intervention.
Ethics
This study has been approved by Ethics Committee of the Federal University of São Paulo (Universidade Federal de São Paulo; UNIFESP- registration number 0699/10). The study has also been registered in the Brazilian Clinical Trials Registry (ReBEC; registration number: RBR-7rjt4t).
All participants will receive information about the study, and their participation will be voluntary. Each participant will be required to sign an informed consent document. At the end of the interviews, all participants will receive an informational leaflet regarding the use of alcoholic beverages, as well as a guide concerning health services and psychosocial assistance in cases of violence and alcohol dependence. After the follow-up assessment is performed three months after the screening, the BI will be applied to participants in the CG. Women who score in the range of “probable alcohol dependence” (i.e., 20 points or more on the AUDIT) will be instructed to seek specialized treatment, regardless of the group to which they are allocated.
Statistical analyses
Given the characteristics of the adopted sampling design, the data will be processed and analyzed using the statistical software STATA version 11.
Initially, exploratory bivariate analyses will be conducted to assess the associations between the variables of interest and the possible risk factors. Chi-square tests will be used to test the association between categorical variables, whereas Student’s t-test will be used to compare the means of continuous variables.
A logistic regression model will be used to examine the patterns of alcohol consumption with the response variable and the sociodemographic and health characteristics as explanatory variables. A separate logistic regression model will be used with the type of violence as dependent variable and the sociodemographic and health characteristics, and alcohol consumption patterns as explanatory variables.
The TG and CG will be compared using t-tests and generalized linear models with repeated-measures.