Study area and period
Demba Gofa District is one of eight districts located in the Gofa Zone in Southern Nations Nationalities and the Regional State of Ethiopia. The district location lies between 8°71′81″ north and 43°89′85″ east. The district is located 305 km away from the regional capital city Hawassa and 525 km southwest of Addis Ababa (Fig. 1). The study was conducted from February to March 2021.
Study design and population
A community-based cross-sectional study was conducted among residents in Demba Gofa district, Ethiopia. The study population was mothers/caregivers paired with children 6 to 23 months old who lived in the selected kebeles for at least 6 months. Those who had mental illnesses interfering with the interview were not considered in the study. Children above 23 months and below 6 months were also excluded from this study.
Sample size and sampling technique
Sample size was determined using Epi-Info software version 7.2.2.6 with the following assumptions. The total number of children aged 6 to 23 months in the selected kebeles was 713, the proportion of regional stunting was 36% [8], the confidence interval (CI) was 95%, and the acceptable margin of error was 5%. After adding 5% for nonresponse, the final sample size becomes 362.
Primarily, the total number of children in the drought-vulnerable kebeles (small administrative unit) of Demba Gofa district was identified from each kebele’s health extension logbook. Then, the residential house of children 6 to 23 months old age was identified and coded. Then, systematic random sampling with proportional allocation was used to select infant mother/caregiver pairs from each kebele. Specifically, study participants were selected systematically according to their arrangement of the households every (k = 2) interval. Finally, if there was more than one child below the age of five within the household, one child was selected randomly (Fig. 2).
Data collection and measurements
The data were collected from both primary and secondary sources of data. Primary data were collected using a structured and pretested interviewer-administered questionnaire developed to collect information on sociodemographic, maternal and child health, environmental characteristics, household dietary diversity and anthropometric measurements. Secondary data related to child-care practices (breastfeeding, complementary feeding practices), maternal characteristics (number of children, feeding practice during pregnancy and lactation), and facilities (drinking water) were collected using a pretested semistructured questionnaire from the Demba Gofa district agricultural and natural resource department and from the district health office. One health extension worker and one nurse per kebele were allocated to collect the data.
Household dietary diversity (HDD) data were collected according to Food and Agricultural Organization (FAO) guidelines [26] by using semistructured questionnaires and microlevel data drawn from repeated 24-h diet recall surveys for daily dietary intake of household members. An individual who was responsible for preparing food or serving food for the family members was used as a source of household dietary diversity data. Household dietary diversity was assessed with a scale of seven food groups: cereals and grains, vegetables, fruits, dairy products, oil and fat, and protein-rich and discretionary calorie foods. Finally, the HDD score was found to be optimal when a child was fed foods greater than four food groups per day [27].
For animal food source (ASF) consumption, data analysis was conducted according to the method employed by Krasevec et al. [28]. The food lists on household dietary diversity questionnaire grouped into three categories of animal source foods as: (1) milk products including infant formula, yogurt cheese, butter and other locally available milk products: (2) meat group that includes any animal organ meat, poultry and fish: (3) eggs. Finally, animal source food consumption data were constructed as categorical based on the number of animal source food consumed, with a minimum zero and maximum of three [28].
Anthropometric data were collected according to the method designed by Gibson [29]. The child health card or birth certificate was used to ascertain and record the age of the index child. In situations where the mother/caregiver did not have the documents to ascertain the age of the child, they were asked to identify a child from the neighbourhood who was born almost the same time. The length was measured in the recumbent position using a sliding board by two data collectors and taken to the nearest 1 mm. Length was measured based on the standard by keeping the child sight perpendicular to the roof and the knees well stretched [29]. Measurement was taken twice, and the average result was taken to ensure accuracy.
The height-for-age index of children was calculated using growth standards published by the World Health Organization (WHO) in 2006. These growth standards were generated through data collected in the WHO Multicentre Growth Reference Study [30] and expressed in standard deviation units from the Multicentre Growth Reference Study median. The height-for-age index is an indicator of linear growth retardation and cumulative growth deficits in children. Children with height-for-age Z-scores below minus two standard deviations (− 2 SD) from the median of the WHO reference population were considered to be stunted or chronically malnourished, while children who were below minus three standard deviations (− 3 SD) from the reference median were considered severely stunted. However, the dependent variable HAZ was considered a continuous variable in the analysis.
Data quality management
The questionnaire was first translated into the local language Amharic for common understanding and retranslated back to English to check its consistency. A pretest was conducted on 5% of the households before the actual data commencement. Question sequence arrangement was performed based on the pretest results. One day of training was given for data collectors and supervisors on the objective of the study and the measurement.
Data analysis
Data were checked, coded and entered into Epi-data version 3.1 and exported to SPSS version software v.20.0 for analysis. First, frequencies and proportions were computed to present descriptive results. Anthropometric data analysis WHO Anthro v.3.2.2 software was utilized to convert raw nutritional data into Z-scores and then transferred to SPSS version 20. Normality, equal variance (homoscedasticity), linearity assumptions and multicollinearity were checked before fitting the linear regression model.
Accordingly, the assumption of linearity was checked through both scatter plots and correlation matrices and was satisfied. The assumption of normality was checked by plotting P-P plots and Kolmogorov–Smirnov and Shapiro–Wilk tests, and it was also satisfied. The assumption of homoscedasticity was satisfied by plotting a scatter plot of standardized residuals against the standardized predicted values, and it was randomly distributed. Durbin Watson statistics were used to check the assumption of independence of errors and autocorrelations. The value of the Durbin Watson statistics for these data was 1.87, which falls within the acceptable range from 1.50 to 2.50: therefore, this analysis satisfied the assumption of independence and no autocorrelations. Multicollinearity was also checked using whether the standard error was < 2, variance inflation factor (VIF) < 10, tolerance > 0.1. Hence there was no evidence of Multicollinearity.
Simple linear regression analysis was carried out to identify the variables associated with the outcome variable. Variables with p-values less than 0.2 in the simple linear regression analyses were considered candidate variables for multivariable linear regression. To declare statistical significance on multivariable linear regression, the regression coefficient at 95% CI with a p-value less than 0.05 was used. The R-square was used to report the model fitness.