Study area
The study was carried out in Bensa district, Sidama region, South Ethiopia. The district is located 400 km from Addis Ababa, the capital of Ethiopia. It is also 131 km from Hawassa, the capital of Sidama Region. According to the central statistical agency report of Ethiopia, the total population of the district was estimated to be 310, 952 (8.1 % urban and 91.9% rural). Of these, 13.94 % were children in the age group of 6-59 months. The district has consists of 03 urban and 34 rural Kebeles (smallest administrative unit of Ethiopia). The health service coverage of the district was 94%. There are one-government primary hospital, 11 health centers and 37 health post. It has also consists of 4 private clinics and 12 pharmacies in the districts. According to the health department report, the distribution of stunting affects almost all Kebeles of the district. Agriculture is the main source of income in the district; inhabitants of the district mainly produce enset, cash crops (coffee) and livestock. The total area of the districts is 732.74 . Astronomically, the district is situated between 6.23-6.88 degrees to the north of equator and 38.74-39.09 degrees to the east. The altitude of the district is 1001-2650 meters above sea level. The climatic conditions of district are 47% highland, 35% midland and 18% lowland. Annual temperature is estimated to be between 12.6 -22.5 degree centigrade. Annual temperature is estimated to be between 12.6 -22.5 degree centigrade. Annual range of rainfall is 900-1450 mm with average of 1125 mm [18].
Study design and population
A facility-based unmatched case control study was carried out in Bensa district from January 10 to March 10, 2018. The source population of this study were all children in age group of 6-59 months and mothers/caregivers who utilized EPI and under-five OPD service in all health facilities. The study population were all selected children in age group of 6–59 months and mothers/caregivers in selected health facilities who have lived with the child at least for 6 months. Those children whose family lived less than 6 months and children who were very sick requiring emergency treatment were excluded from this study.
Sample size determination and sampling technique
The sample size was calculated by using a double population proportion formula in consideration of the following assumptions. The proportion of control in exposure 14.6% and cases in exposure 32.4%, level of confidence 95%, power of the study 80%, ratio of controls over cases 2:1(r=2). Thus, the final sample size after adding a 10% non-response rate is 237 (79 cases and 158 controls). In Bensa district, there is 1 primary hospital and 11 health centers. One primary Hospital and three Health centers were selected out of eleven health centers by simple random sampling technique (lottery method). The calculated sample size (237) was proportionally allocated to the selected health facilities. A consecutive sampling technique was used to select the study participants until the calculated sample size attained. All children aged 6 to 59 months visiting hospital and health centers during the data collection period were measured for their height. Then, children were categorized as stunted or non-stunted based on calculated z-score value. First, stunted children were identified and then selected as cases. The controls were children aged 6 to 59 months without stunting from the same facility cases where selected.
Study variables and data collection technique
The outcome variable was stunting. The independent variables were: socio-demographic variables such as marital status, residence, ethnicity, religion, number of under-five children, family size, parent’s education status, occupation, and economic status: Child characteristics like age, sex, birth order, birth interval, place of delivery, types of birth, and morbidly status (fever, diarrhea and ARI): Child caring practices; such as feeding and immunization: maternal characteristics such as age, mothers’ age during first child, number of children ever has born, ANC visits, use of extra food during pregnancy or lactation and family planning: environmental Health condition like water, hygiene and sanitation.
The data collection was administered by 9 Bsc nurses. One health officer and the principal investigator intensively supervised the data collection process. The anthropometric data of children were collected by using the measurement of age and height/length. Height was measured using a measuring board by appropriately trained nurses and the child’s head, shoulders, buttocks, knees and heels touched the board. Data collection was conducted in a stepwise manner in each health facility in their respective schedule. Measurement of height was done without shoes; to the nearest 0.1 cm. The raw anthropometric data of the studied children were converted to nutritional indicators using WHO Anthro Software (HFA) by taking sex into consideration. Accordingly, a low height for age, less than -2 SD of the reference population indicates stunting, while less than −3SD indicates severe stunting [3].
Operational definitions
Stunting (chronic malnutrition): means HFA is below -2 SD of the reference population while below -3 SD indicates severe stunting.
Acute Respiratory illness: child with cough and fast breathing or difficulty in breathing.
Duration of breastfeeding: the number of months of breastfeeding among children.
Pre-lacteal feeding: a child had given anything to drink other than breast milk in the first three days after delivery.
Complementary foods: are foods which are required by the child, after six months of age, in addition to sustained breastfeeding.
Diarrhea: a child with loose stools for three or more times in a day.
Family size: refers total number of people living in a house during the study period.
Fever: a child with elevated body temperature than usual.
Data quality control
Data were collected using a structured, face-to-face interviewer-administered questionnaire and standard physical measurements. Firstly, the questionnaire was prepared in English. Secondly, it was translated into Sidama language. Finally, it was retranslated back to English to keep its consistency. The comparison was done to assess the inconsistency and non-accuracy between the two versions of the questionnaire. It was pre-tested on 5% of samples in health facility other than actual study area. Then, any inconsistency and non-accuracy was corrected accordingly. Training was given for data collectors and supervisors by the principal investigator for two days. The training was focused on the objective, methods and data collection process. Regular checkup for completeness and consistency of the data were made on a daily basis.
Data processing and analysis
The data were entered into EPINFO version 7 and WHO Anthro software and analyzed using SPSS version 20. Wealth index was constructed by using Principal Components Analysis in SPSS. All required variables recoding and computations were done prior to the main analysis. Descriptive analyses were conducted to obtain descriptive measures for the socio-demographic characteristics and other variables. Chi-square(X2) test was used to determine the overall association between explanatory and outcome variables. Cross tabulation was used to test the assumption of X2.
Binary logistic regression was used to identify predictors of stunting. The bi-variable logistic regression analysis started with unadjusted analysis in which each potential predictor was assessed separately for its association with stunting. Variables with p-values < 0.25 on the unadjusted analysis were entered into a multivariable logistic regression model to find out independent predictors of stunting adjusting for other factors in the model. The variables were entered into the multivariable model using the backward stepwise regression approach. The main assumptions of the logistic regression model (absence of outlier, multicollinearity and interaction among independent variables) were checked to be satisfied. Accordingly, none of the interaction terms was statistically significant indicating absence of a significant effect modification. Multicollinearity between the independent variables was also assessed using multiple linear regression. No evidence of multicollinearity was found as the variance inflation factor (VIF) for all variables was less than 5 and the tolerance statistic was greater than 0.1. The fitness of the logistic regression model was also evaluated in the model using the Hosmer-Lemeshow statistic and greater than 0.05. The presence and strength of association between stunting and the predictors were assessed using adjusted odds ratios (AORs) with a 95% CIs. A statistically significant association was declared when the 95% CI of the AOR did not contain1.
Ethical considerations
Ethical clearance was obtained from the Institutional Review Board (IRB) at the College of Medicine and Health Sciences of Hawassa University before commencing data collection (Ref. No: IRB/145/11). An official letter of permission was obtained from the Department of Public Health to the respective district health office. Informed written permission was also obtained from district health office. Informed written consent and child assent was also obtained from each study participant after explaining the objectives, risks/benefits, rights, confidentiality, nature of the study and the scope of their involvement in the study.