Determinants of stunting among children aged 6 to 59 months in pastoral community, Afar region, North East Ethiopia: Unmatched case control study

Background: Stunting is defined as a child with a height for-age Z-score less than minus two standard deviations. Globally, 162 million less than 5 years were stunted. In Ethiopia, Nationally the prevalence of stunting among under five children was 38.4% and in Afar it is above the national average (41.1%). This study was aimed to identify determinants of stunting among children aged 6 to 59 months in rural Dubti district, Afar region, North East Ethiopia, 2017. Methods: Community based unmatched case-control study design was conducted among 322 (161 cases and 161 controls) children aged 6 to 59 months from March 2-30/ 2017. Simple random method was used to select 5 kebelles from 13 kebelles. Training was given for data collectors and supervisors. Data were entered to EPI data version 3.02 and exported to SPSS version 20 for analysis. Binary logistic regression analysis was used and variables with p-value < 0.25 on univariable binary logistic regression analysis were further analyzed on multivariable binary logistic regression analysis and statistical significance was declared at 95% CI. Results: Being from a mother with no education (AOR = 4.92, 95%CI (1.94, 12.4), preceding birth interval less than 24 months (AOR = 4.94, 95% (2.17, 11.2), no ANC follow-up (AOR = 2.81, 95% (1.1.46, 5.38), no access to latrine (AOR =3.26, 95% CI (1.54-6.94), children born from short mother < 150 cm (AOR = 3.75, 95%CI (1.54, 9.18), not fed colostrum (AOR = 4.45, 95% CI (1.68, 11.8), breast fed for less than 24 months (AOR = 3.14, 95% CI (1.7, 5.79) and non-exclusive breast feeding (AOR = 6.68, 95% (3.1, 14.52) were determinants of stunting at 95% CI. Conclusion: No maternal education, preceding birth interval less than 24 months, no ANC follow-up, no access to latrine, short maternal height, not feeding colostrum, duration of breast feed less than 24 months and non- exclusive breast feeding were determinants of stunting at 95% CI. Community based unmatched case control study was conducted in Dubti district from March 2–30/2017. Dubti district is located 595 km North East of Addis Ababa. It is in zone one of Afar region. Based on 2007 Ethiopian central statistical agency population projection [9], the total population and children aged 6 to 59 months were 72,906 and 2000 respectively. According to the district health office administrative report, the livelihood of the population is pastoralist and agro pastoralist. The district has 14 kebelles (the smallest administrative unit) 1 urban and 13 rural and the total households in the district were 13,071; the health service coverage was 79%. There is 1 referral hospital, 3 functional health centers, and 11 functional health posts. Sample size was calculated using Epi info version 7. Percent of exposure among controls and cases were 5.8 and 16.3% respectively [10]. (95% CI), 80% power, case to control ratio of 1:1, odd ratio 3.77, the sample size was153 cases and 153 controls with total sample size of 306 and considering 5% possible non-response rate. The total sample size was 322 (161 cases and 161 controls). Out of the total 13 rural Kebelles five rural kebelles were selected by simple random sampling technique. A house to house census was made in 5 randomly selected kebelles (the smallest administration unit in Ethiopia) to enumerate all children of age 6 to 59 months. All children aged 6 to 59 months who lived for more than 6 months in the randomly selected kebelles were enumerated. Anthropometric measurement of the children were taken for all children of age 6 to 59 months living in selected kebelles and were measured for their z-score of height for age and categorized as stunted and not stunted to generate sampling frames for cases and controls by a census conducted prior to the actual data collection. Based on this children were categorized as cases (anthropometric reading with z-scores < −2SD) or controls (anthropometric reading with z-scores ≥ −2SD) based on the median of WHO 2006 reference population. After anthropometric measurement of all the children aged 6 to 59 months was taken, children from each selected kebelle were identified and registered sequentially and got identification number and were enrolled as cases and controls. After identification of the number of cases and controls in each randomly selected kebelle, proportional allocation of samples was made in relation to the number of sample size allocated for the study. Based on this A total of 322 (161 cases and 161 controls) were taken from the randomly selected kebelles. Finally, mother -child pairs from each selected kebelle were enrolled using simple random sampling method. Interval (K value) was determined for each kebelle by dividing the total eligible children in the kebelle to the sample proportion. The first household was selected by lottery method. In case more than one eligible child was found in a household, only one child was selected using lottery method. Dependent variable: Stunting. Independent variables: The independent variables were socioeconomic and demographic factors (age, sex, age of mother at first birth, birth order, preceding birth interval, house hold family member, parental educational status, parental occupational status, house hold income and house hold head), environmental factors (access to toilet facility, utilization of latrine, source of water, hand washing practice and waste disposal practice), disease or morbidity factors (diarrhea, fever), feeding or dietary intake factors (time of initiating breast feed, colostrum feeding, duration of breastfeeding, method of child feeding, complementary feeding, exclusive breast feeding, pre-lacteal feeding practices, minimum dietary diversity (MDDS)), nutritional factors (size of child at birth, height of mother and body max index (BMI) of mother)) and maternal and child care factors (antenatal care visits of mother, ANC nutritional counseling, postnatal care, place of delivery and child vaccination status). To arrive at the independent variables a review of different literatures on the subject area or similar studies conducted so far was made. UNICEF conceptual framework for causes of malnutrition (stunting) was also considered. Based on this immediate causes (inadequate intake and diseases), underline causes (household food insecurity, poor maternal and child care, lack of access to health service and unhygienic environment) and basic causes (political, ideological, economical…) causes of malnutrition were considered. Besides of this contextualization of the identified variables with livelihood of the people, with health service coverage, with health seeking behavior of the people in the pastoral community was also considered. Finally, based on the inputs from different literatures and the context in the study setting the independent variables listed above were used. The definition was taken from World Health Organization, WHO child growth standard 2006 field tables [11]. Stunting/cases: were defined as a children with a height for-age Z-score (HAZ) less than minus two standard deviations (<− 2 SD). Controls: were defined as study subjects who had anthropometric reading with z-scores ≥ − 2SD. Questionnaire was initially prepared in English and Amharic and translated into the local language, Afar’af. Three days training was given for data collectors and supervisors about the data collection technique of the study. Pretested structured questionnaire, standard height measuring board and weight measurement scale was used. Calibration of weight measuring instrument was done. Pre-test was done in 5 % of the total sample in non-selected kebelles of the source population. Data were collected by 4 females trained diploma health workers with strict supervision by two trained supervisors. Mothers were interviewed about their children using pre tested questionnaire. Length of children aged 6 to 23 months was measured on recumbent position to the nearest 0.1 cm using standard length measuring board without shoes. Height of children aged 24 to 59 month was measured by placing the child in standing upright position in the middle of board wearing light clothing without shoes. The child’s head, shoulders, buttocks, knees and heels was adjusted to touch the board and each measurement was taken two times to ensure reliability of the study to the nearest 0.1 cm. Mothers who didn’t know exactly the age of their child, immunization card were used or precision in age was maintained to the nearest month. Maternal weight was measured using portable weight scale to the nearest o.1 kg and mothers were allowed not to have anything that adds to the weight being recorded. The weighting scale was checked and reset at zero point for every consecutive study subject. Maternal height was measured in standing position and measurements were made by two data collectors by holding the meter from heel to the back of head and measured to the nearest 0.1 cm. Anthropometric data were calculated by using WHO Anthro2010 software and height for age Z- scores were also been generated based on the median of WHO 2006 reference population (child growth standards). Data was also entered to EpiDatav3.02 for cleaning and exported to SPSS version 20 for further analysis. Tight supervision, day to day follow up during data collection period and data cleaning before data entry were made to prevent missed data. After cleaning data for inconsistencies and missing values; descriptive statistics was done. Univariable binary logistic regression was used to assess the association of one independent variable with the dependent variable. Multivariable binary logistic regression model was used to identify potential significant determinants of stunting after control of all possible potential confounders. Variables with p-value < 0.25 in univariable analysis were a candidate for multivariable logistic regression analysis and statistical significance was declared at 95% CI. Multicollinearity was checked using Variance Inflation Factor (VIF) and there was no multicollinearity (VIF < 10). Residual plots were also performed. Hosmer-Lemeshow goodness-of-fit was used to test for the model fitness and the p-value for Hosmer-Lemeshow test was 0.46 which indicates a good model since p-value is > 0.05 which is insignificant.