Study design and setting
A community-based cross-sectional survey, in-depth interviews and FGDs were conducted from February to March 2019 in Damot Woyde district, Southern Ethiopia. The district is located 313 km south of the Ethiopian capital, Addis Ababa through Butajira Halaba road. In 2019, the Ethiopian Central Statistical Agency (CSA) projected the total population of the district to be 113,823; of which 50,896 were children below two years of age.
Population and sampling
For the survey, the following assumptions were used to calculate the sample size of 634; the expected prevalence of 50% for the involvement of fathers in complementary feeding practice, 95% confidence level, 5% margin of error, design effect of 1.5, and 10% non-response rate. A total of 593 fathers completed the questionnaire in the survey, making the response rate 93.5%. Fathers who had at least one child in the age group of 6–23 months were the source population. A systematic random sampling method was applied to select fathers from selected kebeles (smallest administrative units). Fathers paired with their children were included in the study. First, 634 households were selected systematically from 6 kebeles that were randomly selected out of 25 kebeles. The total sample was allocated to these 6 kebeles proportional to population size. From each household, one father was recruited to complete the survey questionnaire. In households where there was more than one eligible father lottery method was applied to select one father.
In-depth interviews and FGDs were conducted to complement the survey data. For the in-depth interviews 5 fathers, 10 mothers, and 6 community health workers were recruited conveniently by the first author from the kebeles where the survey was conducted. The number of participants in the in-depth interviews was decided based on point of saturation where there was no new information emerging from the last interviews. Similarly, 4 FGDs were conducted with fathers and mothers that were recruited conveniently by the community health workers from the community. There were a total of 61 participants in the in-depth interviews and FGDs.
A structured questionnaire composed of socio-demographic characteristics and questions that measure the father’s knowledge, attitude, and involvement in complementary feeding practice was prepared in English and translated into the local language. Six Bachelor of Science degree holder research assistants (RAs) who speak the local language and had previous data collection experience administered the questionnaire. The RAs were also provided with additional training for two days on the different modules of the questionnaire, participants’ selection, and ethics. The questionnaire was pretested before conducting the main survey with a sample of 32 fathers in a community with similar characteristics to the actual study population but not part of the main study.
The first author, fluent in the local language, conducted all in-depth interviews using a semi-structured interview guide. Focus group discussions were conducted with fathers and mothers, and 3 RAs facilitated the FGDs. The RAs were gender-matched with the FGD participants. Both in-depth interviews and FGDs were recorded by the digital audio recorder and on average the interview took 45 min whereas the FGDs took 60 min.
Fathers were asked six questions to measure their involvement in complementary feeding practice. The questions had five Likert scale responses; always, often, sometimes, rarely, and never. A single variable was established by creating a total score from the six questions. Those who scored above the mean in the total score were labeled as having “fair involvement”, whereas those who scored the mean or below in the total score were labeled as having “Poor involvement”.
Similarly, to assess fathers’ knowledge about complementary feeding, additional six Likert scale questions with five Likert scale responses were administered. Fathers with a score above the mean were labeled as having “better knowledge”, whereas those with a score of the mean or below the mean were categorized as having “poor knowledge”.
Fathers’ perception towards involvement in child complementary feeding was measured using thirteen Likert scale questions. Fathers with a score above the mean were labeled as having “better perception”, whereas those with a score of the mean or below the mean were categorized as having “poor perception”.
The survey data were entered into Epi data software version 18.104.22.168 and exported to SPSS version 20 for analysis. Descriptive statistics were conducted for all variables. Normality was checked for quantitative continuous variables. We utilized binary logistic regression analysis since the outcome variable had binary outcomes. Exposure variables with p-value < 0.25 during bivariate analysis were considered for multivariate analysis. Multivariate logistic regression analysis was done to control for potential confounders and identify predictors of the outcome variable. Audio-recorded files of the FGDs and in-depth interviews were transcribed verbatim, translated into English, and coded. The codes were generated into common themes and sub-themes. Open code software was used to assist data organization and analysis.