Study setting
The study was conducted in Jimma Zone, one of 14 administrative zones of the Oromia region in the southwest of Ethiopia. Its capital, Jimma, is situated 352 km to the southwest of the national capital, Addis Ababa. Jimma Zone has 17 districts and one special zone. Based on the 2007 national census, it has a population of about 2.5 million, with a little more than half being men [22]. The rural part accounts for 89.5 % of the total population size of the zone, in which the dominant ethnic group is the Oromo. An estimated 52 % of Jimma Zone's residents have access to health posts, which is the lowest level of the primary healthcare system (Jimma Zone administration: Public Relations and Information Office, unpublished document). The healthcare delivery within the zone is carried out through 13 health centres, 26 health posts, 65 health stations, and 2 hospitals (Jimma Zone administration: Public Relations and Information Office, unpublished document). Among these, 35 are privately owned and non-governmental organization clinics.
Sampling
A multi-stage sampling design was used with districts as primary sampling units (PSU), and sub-districts (kebeles) as secondary sampling units (SSU). The study covered three districts i.e. Seka, Manna and Gomma, in which six kebeles were randomly selected: Goyoo qechema, Koffie, Gobiemuleta, Haro, Gembie and Bulbulo. In each selected kebele, a complete census of married couples was compiled to be used as sampling frame. Married couples were then randomly sampled from each locality, based on a computer generated random number list until the required size was achieved. From the three selected districts, two sub-districts were randomly taken and assigned to either a control or intervention group. The sample size was computed using Minitab version 14 statistical software in order to detect a 10 % or more decrease in an unmet need for family planning. With an alpha of 0.05 and 80 % power, the minimum required sample size was estimated at 388. By adding 10 per cent to account for non-responses, the final sample size was 427 couples. Villages included in the study were selected according to their geographical proximity in order to minimize geographical variations because each chosen village needed to be situated at a similar distance to a health facility. Considering the large distances between villages, cross-contamination was not likely.
We designed our study using a quasi-experimental design to assess the impact of an educational campaign intervention. A total of 854 married men and their wives (427 in control and 427 in intervention group) were selected using systematic random sampling. To be eligible, couples had to meet the following criteria: i) legally married men and their wives; ii) living together in the same house during the six months prior to the baseline data collection in the study area; iii) planning to stay in the area for one year starting from the time of data collection; iv) the wife's age was between 15 and 49 years; and v) the wife was not pregnant at the moment of the baseline data collection. Husbands within a polygamous marriage (who had more than one wife) were excluded from the analysis to decrease redundancy of information. Fieldwork took place from March 2010 to May 2010 (the baseline household survey) and from March 2011 to April 2011 (the follow-up survey) in the same three intervention sub-districts and three control sub-districts. The survey instruments were developed from a validated questionnaire and were considered valid and reliable as it was tested and used by other studies to obtain information on couples about knowledge, attitude and contraceptive practice [23–26]. A three-day training was given to the interviewers. After house numbers were assigned within each sub-district, random house numbers were generated using SPSS® version 16 for Windows®. Households were listed prior to the selection of eligible respondents. In total, 1,622 individuals (811 couples) were interviewed for the baseline study. The same approach was used to collect the follow-up household survey data. A total of 1,546 individuals were interviewed for the follow-up study (drop-out of 4.7 %). There was little variation in the participants due to the marginal dropout rate during follow-up.
Measurement
The researchers prepared a semi-structured questionnaire for the quantitative method. Data were collected in the local language with separate questionnaires sharing a similar core set of questions for men and women; informed consent was received before participation. The pre- and post- intervention data collection was conducted after six months of intervention with similar questionnaires. Couples who participated in the two rounds of interviews were included in the analysis. After the last instruction on family planning education and community gathering, we submitted the same questionnaire to the men and their wives. The intervention program for married groups described in this article incorporated both men and women. It was carried out by male and female community agents who used different communication materials (flyers, reading materials and leaflets).
Intervention
After the baseline survey conducted in early 2010, family planning education was given in villages located in the sub-districts assigned to the intervention arm. The intervention consisted of 1) family planning education on different methods of contraception through flyers, booklets, and face-to-face discussions, and 2) promotion of husband-wife discussions on family planning. Family planning education was given to both men and women at the household level in the intervention arm, in addition to monthly community gatherings. A health officer trained the three male and three female community agents who were hired for this study’s purpose and undertook the intervention activity. These community agents had completed high school, whereas the health officer held a degree; they all spoke the local language fluently. The community agents were involved in the community under study through family planning interventions. They were chosen to be community agents based on this extensive experience. They were assigned to provide information on different methods of contraception by several means of communication (flyers, booklets, and face-to-face discussions) in order to promote husband-wife discussion on family planning and to teach reproductive health rights to couples focusing on family planning. Some of the types of contraceptives discussed during intervention were pills, injectables, implants, condoms, and standard days method. The condom, for example, was discussed among the male contraceptive methods. The intervention was designed and executed in partnership with local leaders. During the routine visits to each household, the coordinators performed random spot checks to see how the topic was covered. In addition, the principal investigator supervised the proper implementation of activities throughout the intervention period. During this period also, the normal health care routine was carried out for both the control and intervention group, but there were no initiatives taken by other health care providers. There were some efforts delivered by mass media such as television and radio, but these were not considered as confounding factors because both the control and intervention groups were also exposed to the other citizens of the country.
Data analysis
We assessed the effect of the intervention on the use of contraception, the involvement of men in family planning, and spousal discussions on family planning issues. The dependent variables were contraceptive use among couples and the male involvement in family planning. We calculated crude odds ratios (ORs) as well as adjusted them (aORs) for survey design effects using conditional logistic regression methods. We used generalized linear models that accounted for stratification, clustering and weights (svyglm in R ‘survey’ package). Since quasi-experimental research can be subject to biases and can be confounding due to unforeseen baseline differences between the intervention and control group, we used an Inverse Probability of Treatment Weighted (IPTW) analysis to account for these differences by using propensity scores that were estimated through multivariate logistic regression. Variables that were included in this regression were the literacy of both spouses, whether or not family planning had been discussed previously, and baseline attitudes of the husband towards being involved in the couple’s family planning practices. As a result, our findings are adjusted for potential baseline differences in those variables. We used for all statistical analysis STATA ® version 10 for Windows® and R version 3.0.1.
Ethical considerations
Ethical clearance of the study was obtained from the research and ethics committee of the College of Public Health and Medical Sciences, Jimma University, Southwest Ethiopia and Ghent University’s Ethical Committee in Belgium. Written consent was obtained from each man and woman participating in the study after the data collectors explained the purpose of the study using a predefined information sheet. Written informed consent was taken from spouses on the behalf of those wives who were less than 18 years old, considering the cultural context of the study area. No compensation was rendered as a direct incentive to the participants. The ethics committees approved this consent procedure.