Study design
Hadiya zone is located in Ethiopia’s Southern Nation Nationalities and Peoples Regional States (SNNPR). It is divided into ten districts and one city administration. It is located at a distance of 230 km to the northeast of Addis Ababa, the capital city of Ethiopia. Lemo is one of the food-insecure districts in Hadiya Zone. Geographically, it is situated between 7°22′00″–7°45′00″N latitude and 37°40′00″–38°00′00″E longitude. According to the 2007 Central Statistical Agency (CSA) report, the total population of Lemo district is 118,594. It is composed of 58,666 males and 59,928 females. Most (98.3%) of the population lives in rural settings. Consequently, most (65.5%) of the population in the district are farmers. Enset-based mixed crop-livestock production is the main agricultural production system. According to the annual report of the Lemo district health office, the estimated number of preschool children whose age was accounted for was 16,335 in 2018. Preschool children comprise 10.4% of the total population in the district. There are seven health centres and 35 health posts in the district that provide malnutrition diagnosis and treatment at the time of data collection. In addition, 65 health extension workers work in health posts.
Population
The target population consisted of children aged 24 to 59 months who lived in the Lemo district. Children of age 24-59 months who lived in randomly selected households and presented during the data collection period were the study population in this study. On the contrary, mothers, guardians, or children who were unable to give a full response or were incompetent for anthropometric measurement due to critical illness during the data collection period were excluded from the study.
Sampling and sample size determination
The sample size was determined using the single population proportion formula, taking the following assumptions: Stunting prevalence among children (p = 23.1%) from an Ethiopian meta-analysis study [21], 95% confidence level, 5% margin of error, design effect of 1.5, and 10% non-response rate. Hence, the initial sample size was computed using the standard Cochran formula, n = z2pq/d2. Then, plugging values into the formula, a total sample of 273 was computed. Using sample adjustment formulas (a design effect of 1.5 and a non-response rate of 10%), the final sample size was calculated as 450. Figure 1 displays the sampling strategy of this study. A multistage stratified cluster sampling technique was applied to select study samples in this study. In the first stage, 1 urban Kebele and 10 rural Kebele were selected using simple random sampling, following the technique of the lottery method. In the second stage, households were selected using a systematic random sampling technique. For this study, the health posts’ family folder (or “registration book”) was utilized as a sampling frame to select households that consist of children aged 24-59 months. We have utilized the updated registration book of health extension workers who gave service in the health posts, which consists of the name of the Kebele and the house number of children ages 24-59 months. The first household was identified using the lottery method (simple random sampling). Then, subsequent households were selected for every K-value (K = N/n) to reach the selected households. Where N is the total number of households in the selected Kebeles and n is the calculated sample size.
Figure 1 showed the schematic presentation of sampling strategy in this study. The sample households selected for this study were allocated proportionally to each Kebele. As a result, we used a different K-value. For households that consisted of more than one child, simple random sampling was used to select the study participants. Furthermore, data collectors visited households that were closed at the time of data collection 2-3 times before declaring them non-responsive. Sample size for the second objective, or assessment of factors associated with stunting, was determined using open-source epi-online software. However, the sample size computed for prevalence was found to be higher than the sample size computed for the second objective. Hence, the sample size for prevalence was considered for this study (See Fig. 1).
Measurement
The outcome variable of this study was stunting. It is the binary outcome variable which can be determined through anthropometric measurements. Whereas, the explanatory variables include the parental characteristics (both father and mother of children), household level characteristics, child characteristics and health behavior of parents. Figure 2 from this study indicates the complex interaction of independent variables with response variables. For example, the parents’ characteristics include: socio-demographic status, including age, marital status, educational status, religion and income. Furthermore, the child’s characteristics were age and sex of the child. The health behavior factors included the hygiene and environmental sanitation of the parents (See Fig. 2).
Height was measured with a vertical or horizontal measuring board reading a maximum of 175 cm and capable of measuring to 0.1 cm, which was used to take the height of a child. The child stands on the measuring board barefooted, with hands hanging loosely, feet parallel to the body, and heels, buttocks, shoulders, and the back of the head touching the board. The head would be kept comfortably straight, with the lower edge of the eye socket in the same horizontal plane as the outer ear canal. The head of the measuring board is then pushed gently, crushing the hair and coming into contact with the upper part of the head. The height is then computed to within 0.1 cm. Two readings were recorded and the mean calculation was used in the analysis. The length was measured by placing the kid flat on the length board. The slider is placed on the edge of bare feet when the head (crushing the hair) touches the other end of the meter. Then two surveys were carried out, and the average was calculated. The weight was measured using an easily transportable balance measuring 0.1 kg. The ladder was adjusted before weighing each child to zero. The child was lightly dressed when the weight was taken. Two readings were carried out for each child, and the average was entered on the questionnaire [21].
Operational definitions
Food insecurity
is a condition in which people experience limited or uncertain physical and economic access to safe, sufficient, and nutritious food to meet their dietary needs or food preferences [8].
Stunting is a long-term cumulative effect of poor nutrition and poor health
Short stature refers to a low height-for-age ratio that can be caused by either normal growth variation or a growth deficit. It is defined as low height-for-age at − 2 SD below the median value of the WHO international growth reference [21].
Data collection tools and procedure
A structured and pre-tested questionnaire was used to collect data. The questionnaire was initially written in English. Then, it was translated to Hadyiyissa and back to English to keep its consistency. The questionnaire was prepared through a comprehensive literature review [22,23,24]. The questionnaire consists of the following parts: Socio-demographic, reproductive, environmental, child health, and care practices A face-to-face interview was conducted in the local language (Hadyiyissa) by data collectors who speak and understand the language. Furthermore, anthropometric measurements were taken to determine the stunting status of children: The height and age of the children were measured. Children of ages 24 to 59 months were examined for signs of stunting by seven bachelor’s degree nurses who had previous experience measuring child malnutrition. The growth standard anthropometric measurement procedure was performed based on the World Health Organization’s recommendation [25].
Interrogation was used to determine children’s ages, which were then confirmed by probing mothers and guardians. Information about the age of children was collected from the health posts’ family folder and the child’s mother or caregiver. A standard calibrated machine was used to record weight in kilograms. Each child’s weight was measured barefoot and without heavy clothing during the measurement process. First, the weight scale was calibrated to zero before taking every measurement. Second, height was measured after the child was placed on the platform, barefooted, with their head upright and looking straight ahead, using a standard height measuring scale. Finally, the weight and height measurements were approximated to the nearest 0.1 value and reported.
Data quality control
A pre-test was performed on 5% of households that were not included in Kebeles outside of the selected Kebeles. Three days of training were provided to all data collectors and supervisors by the principal investigators. Furthermore, at the end of every data collection day, the completeness of the questionnaires was checked by immediate supervisors, and challenges were discussed and solved for the next day. Inconsistent data were rejected from analysis. We have checked that the reason for the non-responders’ refusal was not linked with the purpose of this study.
Data analysis
Data were entered, cleaned, and coded using Epi-Info 7 software. Then, it was exported to Statistical Package for Social Sciences (SPSS) software version 25 for analysis. The World Health Organization (WHO) anthro program version 3.2.2 software was utilized to generate the stunting index and export it to SPSS version 25. Before the actual logistic regression analysis, the necessary assumptions of the logistic regression analysis were checked: Errors are independent and exhibit linearity in the logit for continuous variables, absence of multicollinearity, and a lack of strongly influential outliers. WHO growth reference was used to report anthropometric results; individual anthropometric data was compared with reference values on a graph using sex and age-specific Z-score classification systems. The stunting status indicator of height-for-age (HAZ) was compared with the reference data from the World Health Organization standard. A cutoff of below 2 standard deviations (SD) of the WHO median value for HAZ was considered. Frequencies and cross tabulations were used to check for missed values and variables. Descriptive analysis was made using percentages, means, and standard deviations for the variables included in the study. Candidates for the multivariable logistic analysis model were variables with a P-value of 0.25. Multivariate logistic regression analysis was used to adjust for possible confounders. To assess the strength of the association, an adjusted odds ratio (AOR) with a 95% confidence interval (CI) and a P-value of 0.05 was calculated. Finally, Hosmer-Lemeshow goodness of fit tests was used to check the model’s fitness.