Study area and period
The study was conducted in Chencha Zuria district, Southern Ethiopia. Chencha Zuria district is located about 511 km Southwest of Addis Ababa, the capital city of Ethiopia. Based on the district health office report, the total population of the district was 91,677 and it has 18,710 households with 3,172 live births. It is divided administratively into 34 kebeles (the lowest administrative unit in Ethiopia). In addition, it has 5 health centers and 34 health posts to provide MNCH services to the population. And the study was conducted from August 1 to 30, 2020.
Study design
A quantitative community-based cross-sectional study supplemented by qualitative method was used.
Population
Mothers having infants aged less than six months were the source population. Those mothers who gave birth to an infant aged less than six months and lived in the study area for at least 6 months were included in the study. However, mothers who were seriously ill or unable to give the required information during the data collection period were excluded from the study.
For the qualitative part of the study, the study participants were breastfeeding mothers who have infants aged less than 12 months.
Sample size determination
The sample size was determined by using Epi info with the following assumptions: 95% confidence interval for a two-sided test; 80% of power; ratio of unexposed to exposed of 1:1; the proportion of outcome among mothers with timely initiation of breastfeeding (proportion of outcome in unexposed group) = 8.9%, and Adjusted Odds ratio of 2.4 [20]. Accordingly, the calculated sample size was 408 mothers. Adding 10% for non-response then multiplying by design-effect of 1.5the final sample size for this study was 674 study participants.
For the qualitative part, the data saturation approach was used to determine the sample size. Accordingly, eleven mothers were recruited for the interview.
Sampling procedure
From the total 34 kebeles found in ChenchaZuria district, eleven kebeles were selected by lottery method. The study subjects were proportionally allocated to each kebele based on the number of women who gave birth in the respective kebeles. Then study participants were identified by systematic sampling method using the Health Extension Workers (HEWs) birth registry book as a sampling frame. After determining the starting mother using the lottery method, every other mother (K = 2) was selected as a study participant from the registration book list.
For the qualitative part of the study, the purposive sampling technique was used to recruit the study participants. Then eleven breastfeeding mothers who have infants aged less than 12 months were recruited for In Depth Interview (IDI).
Operational definitions
Colostrum avoidance is defined as a failure to feed the first, thick, and yellowish milk to a newborn baby that is produced in the first 2–3 days after delivery. It includes squeezing and throwing out or pumping and discarding [9].
Maternal knowledge about importance of colostrum feeding
Of the five questions relating to maternal knowledge the mothers who correctly answered three or more questions were considered as having good knowledge. And the mothers who correctly answered less than three questions were considered as having poor knowledge [23].
Pre-lacteal feeding is giving a liquid or foods other than breast milk prior to the establishment of regular breastfeeding with in the first three days of life of new born infant.
Timely initiation of breastfeeding: If an infant put on mother’s breast to feed within one hour (including one hour) of birth.
Data collection instrument and procedures
The quantitative data were collected using a structured and interviewer-administered questionnaire, which was developed from previous literature by adapting and modifying contextually to fit the local situation and research objectives [15, 17, 19, 20, 28]. The questionnaire was initially prepared in English then translated to Gamo language. Then to check its consistency, language experts again back translate it to the English language. Six diploma nurses collected the data via Open Data Kit (ODK) application using smartphones.
The qualitative data were collected by using a semi-structured interview guide by the IDI technique. Trained BSc Nurse with the assistance of two note-takers collected the data. And they recorded the interviews via digital voice recorder.
Data quality control
Pre-testing of the quantitative questionnaire was carried out on 5% of the sample size before starting the actual data collection to improve the clarity and understandability of the tool. The data collectors trained for two days collected the data. The investigators supervised and coordinated the overall activity of the data collection. The investigators also reviewed and checked the collected data daily for completeness and consistency.
To assure the data quality for qualitative part data was triangulated in space by gathering information from various locations inside the study area. Additionally, in order to triangulate the data in person variety of study participants with various socio-demographic traits were selected. These ideas are now included in the recent submission of the manuscript.
Data processing and analysis
The quantitative data from the ODK briefcase were exported to SPSS version 25 statistical package for the further analysis. Frequencies, proportions, mean, standard deviation, tables, and graphs were used to describe the data. Principal Component Analysis (PCA) were applied to analyze wealth status of the study participants. Binary logistic regression analysis was used to identify associations between dependent and independent variables. In the bi-variable logistic regression analysis, variables which were statistically significant at p-value < 0.25 and biologically plausible were candidate variable for multi-variable logistic regression analysis. Confounders were controlled by running step-wise backward logistic regression analysis. Multi-collinearity among independent variables was checked by using Variance Inflation Factor (VIF) and the values of all variables found within 1.0 and 2.0. The fitness of logistic model was assessed by using the Hosmer and Lemeshow Goodness of fit test with p-value of 0.776.The degree of association of independent variables with the dependent variable was assessed using AOR with 95% confidence interval and p-value < 0.05 in the final multi-variable model.
The qualitative data was analyzed manually by using thematic analysis approach. Qualitative data transcribed word by word, translated into English language and coded. Then, coded data grouped into the three key themes guided by the literature and an interview guide.