Community Based Intervention to Prevent Domestic Violence Against Women in Ethiopia; a Protocol for Quasi-Experimental Implementation Research

Sample Size Determination The sample size is determined considering both the intervention and control group proportion. The sample size for dichotomous outcome is recommended based on an assumption of superiority trial hypothesis testing. The hypothesis of the researcher is that implementation of the new community based prevention intervention program to prevent domestic violence against women will have significant effect than the standard (existing) prevention program in public health point of view. So the corresponding null hypothesis (Ho) is that: the new community based preventive intervention is not more efficacious to prevent domestic violence against women than the comparison group based on public health relevance. In addition, the sample size is determined considering an assumption of 5% margin of error, 95% significance level and 80% power.

Where; N is the sample size of each group, P is prevalence in standard intervention/control arm, d: the real difference between two treatment effect (d= Po-Px); 22%-9%=13%=0.13); δ0: a statistical acceptable margin (5%); ɑ: is the significance level which is 95% confidence level (ɑ=1-0.95=0.05=1.645), ß: is 80% power (ß=1-0.8= 0.2 = 0.845, and adding Then, considering 10% for potential non-achieved rate compensation. The calculated sample size for a group is 364. Eventually, as a result of the complex sampling design (multistage, stratified and systematic sampling), design effect is considered increase power and precision of the study. Design effect was computed using a finding from complex sampling design study conducted in Ethiopia (Fagitalekoma district) reported that the prevalence of DVAW was 78%, and a clustered randomized control trial done in Uganda indicated that 22% of women experienced physical IPV in the standard intervention group. The final sample size for the three arms was 1,758 women that mean 586 for each group. The number of participant in the qualitative study will be determined by information saturation during interviews and or focused group discussions (FGD). A minimum of two FGD (one male group and one female group) which comprise 8 to 10 discussants will be held at each selected kebeles. The local leaders or other person who are in-charge will be participated on the selection of qualitative study participants until information saturation.

Sampling Procedure

Multistage, stratified and systematic sampling will be used to select study participant to this study. First, contextual assessment will be employed among purposefully selected study participants from Awi zonal admisntration, district offices and community representatives. Stakeholders such as local politicians, health office, women and children affair office in-charges, local leaders, police, health care providers, non-governmental organization representatives and other concerned bodies will be participated to involve on advocacy meeting and contextualization. Then, multi-stage stratified probability sampling methods will be used to select study sites (kebeles) in the three districts among nine districts in Awi zone. Three districts will be selected for three experimental group. One district will be selected for comprehensive intervention. The second district will be assigned for partial intervention group, and the third district will be taken as comparison (control). Then, six kebeles (two kebeles from each district) will be selected from three districts purposefully considering appropriateness for intervention, resource, and geographical proximity to control information contamination between intervention and control group.

Then, kebeles will be stratified into urban and rural Kebele in respective district. Stratification will use to control the awareness level, access to information, and community tradition gender norm variability between urban and rural dwellers. After that, proportional to size allocation (ni) of the sample size will be done to each stratum. Eventually, systematic sampling will be employed to select actual study participants at household level based on kth interval. Confidential code will be given strictly to households that will be selected for interview for intervention monitoring and endline evaluation purpose. In the meantime, if two married women present in one household with one husband (polygamy), only one woman will be selected using lottery method for interview.

Regarding to qualitative study, purposive sampling method will be used to select study participants for in-depth interview, key informant interview (KII) and FGD. The community leaders will be consulted and actively involved on the recruitment of the most knowledgeable and respected community representatives for in-depth interview, key informant interview (KII) and FGD. The purpose of the study will be explained in detail to the community leaders who will participate on the recruitment of qualitative study participants. It will be conducted to assess individual perception towards DVAW mainly focused on sex negotiation, decision making, community gender norm, and wife beating attitudes with in the community. In addition, community sanction (response) against DVAW will be assessed qualitatively. It will mainly involve by taking from people who are in position (authority) such as local leaders, religious fathers, elders, police, women affair office, women association, health professionals, health extension workers, and women lawyers.

Data Collection

Mixed data collection method will be carried out using structured and semi structured questionnaires, that adapted from various literature. The tool will be pretested, modified to the local context and suite to the study objectives. The tool will consisted of demographic, socio-economic characteristics of couples (women self-reported profile), relationship level, community level and societal level factors of domestic violence. In addition, the tool will comprise questions about the experience of psychological, physical, sexual violence. In addition, semi structured questionnaire will be designed for qualitative study. Similarly, interview guide will be used to assess and dig-out the underlying bottlenecks in the community that hinder implementation of preventive interventions.

Data Collection Procedure

Data will be collected using sequential explanatory mixed method for baseline, intervention monitoring/tracking and endline implementation evaluation. Here below is the flow chart of the mixed data collection process adapted from other study. Two quantitative (1 and 2) and qualitative (1 and 2) process will be sequentially implemented to gather necessary data . Face-to-face interview method among married or cohabited women, key informant interview, in-depth interview, focus group discussion, meeting and training time data capturing will be carried out to have relevant information. In addition to the domestic violence assessment tool, the alternative questionnaire will be prepared and provide to data collectors to secure confidentiality issue if incase of third person might come suddenly during interview time. Very confidential place for women will be arranged by data collectors as per the training and WHO VAW research ethics recommendation.

Then qualitative data collection (KII, in-depth interview and FGDs) will be held to explore community perception, attitudes, norms and beliefs about domestic violence. However, mini (quick) qualitative data collection for contextualization has been done among officially in the districts and zone administrative level. On the other hand, it will use to find out community mobilization strategies, existing legal sanctions, available training guidelines, collaboration and integration of implementation activity. This will also help to design appropriate tool for main study, to design correct and culturally compatible intervention materials to tackle the problem. The KII will be held among women representative, district council members, women affair office, police officers, lawyers, elders, local politicians/leaders, health care providers, nongovernmental organization, and religious leaders.

Gender specific focused group discussion will be held on group of men and women to assess societal perception about domestic violence against women. Each focus group discussion will consisted of eight to ten participants. All participants' sociodemographic data will be captured using anonymously coded structured questionnaire. The FGDs will be facilitated by a group of three people (moderator, note taker and principal investigator). Voice recorder, note taking and nonverbal communication (physical gestures) during the discussion will be recorded as much as possible to capture necessary information. An estimate of 60 to 90 minutes will be taken to conduct one FGD. Moreover, implementation of domestic violence against women preventive intervention activities will be monitored/tracked using ongoing qualitative study method. Eventually, intervention implementation endline evaluation will be carried out using as same structured questionnaire, data collectors and supervisors as of the formative assessment. Data will be collected by exclusively twelve female health professionals (Nurse, Midwifery or public health).

Data Quality Assurance

Participants Allocation: Allocation of study participants in the intervention and control group will be considered geographic proximity to minimize information contamination. Representative sample will be taken using scientifically sounded procedures.

Appropriate Data Collectors recruitment: Data will be collected by twelve trained Diploma and above qualification female Nurses, Midwives and or public health officers who are not local resident in the community for trustworthiness and ethical concerns. Four supervisor who have Master of Public Health (MPH) with good previous field survey experience will be recruited.

Data Collectors Training: A five days training will be given for research assistants by the principal investigator. The training will mainly focus on the purpose of the study, data collection procedure, interview techniques, sampling methods, strict protection of confidentiality, ethical issues of VAW research and data quality assurance. Ethical sensitiveness of VAW will be explained to research assistants to maintain confidentiality and also quality of data. Brief explanation will be provided on biomedical research ethical principles and WHO (2001) gender based violence research ethical guideline recommendations called "women's safety first". Refreshment training and short orientation/debriefing will be provided regularly to research assistants during the study period.

Pretest: Pretest will be done among 5% of the sample size among married women from nearby district to check the compatibility of the data collection tool. The pretest will address sensitive words in the question, time, data collection procedure, interview techniques, sequence of questions, consent taking from study participants, and FGD facilitation skill and note taking. Necessary amendment will be taken according to feedback from study participant (verbal and non-verbal expression during pretest interview) and comment from research assistants.

Training of Implementers: intensive training and supportive supervision will be given for implementers. Two implementation facilitators and 6 community health extension workers will be recruited and hired for one year (12 months) contract employment. Implementers will be participated on the intervention tool validation to be familiar with the tool. Regular discussion and close supportive supervision will be given by principal investigator and facilitators.

Data Processing and Statistical Analysis

Baseline, monitoring and end-line gathered data will be checked and edited for incompleteness and inconsistency at field as well as at office. Explanatory and outcome variables will be clearly predefined prior to data entry. During data entry if woman will report at least one type of experience of either physical, psychological or sexual violence items coded as "Yes = 1" and "No = 0". On the other hand, women acceptance to violence or gender inequity norm will be measured using six questions/measurements, and also if women respond at least one justifies of wife beating. Similarly women autonomy will be measured in four components autonomy indicators such as women's economic decision-making autonomy, familial health care and family planning decision-making autonomy, extent of freedom of movement autonomy, and women's attitudes toward partner's violence. After completing the definition and coding of raw data. Data entry template will be designed on Epi data window version 3.5.1, and data entry will be carried out.

Then data will be exported to SPSS window version (23.0) for further analysis. Data cleaning will be performed computing frequency using box plot to explore outliers and other data anomalies. Multiple imputation will be done to handle missing data. Different statistical assumptions will be checked for data distribution and suitability for statistical analysis. Descriptive statistics will be computed to determine prevalence (proportion), mean, median, and standard deviation of some independent variables as necessary. Cross tabulation (chi-square) will be employed. Binary logistic regression will be carried out to confirm the association of explanatory variable with DVAW. Explanatory variable which will have score of p-value less than 0.25 in binary logistic regression will be included into the multiple logistic regression analysis to determine the independent determinants of domestic violence by controlling the confounding variable. In addition, Intention to Treat Analysis (ITTA) will be carried out to compute the overall changes on the level of DVAW. Moreover, Difference in Difference (DID) will be calculated to determine the net intervention effect. The DID will be computed as subtracting intervention arm outcomes from baseline data to endline minus difference of control arm outcomes from baseline to endline. Finally, statistical significance of intervention effect will be measured using adjusted prevalence ratio, adjusted odd ratio/adjusted risk ratio at 95% CI and p value less than 0.05.

Qualitative data will be transcribed through playing and replaying the voice recorder and also referring the note taken during interview or FGD. The voice recorder audio will be listen multiple times as much as possible at very quiet place. It will be transcribed independently by research assistant and principal investigator according to the verbatim of participants. Then transcribed qualitative data will be translated from local language (Amharic) into English. In the meantime, the qualitative data will be read and reread to understand what they want to say. The quotes will be narrated and cited using participant's code. All the qualitative data gathered from interview and FGD will be arranged into thematic areas. Similar ideas will be color coded, and then merged accordingly. Qualitative Data Analysis (QDA) software like Nvivo will be used to arrange or organize manually coded qualitative data.