Stunting and Associated Factors Among Children Aged 6-59 Months From Households in Productive Safety Net Program and Non-productive Safety Net Program in Meta District, East Hararghe Zone, Eastern Ethiopia: a Comparative Cross – Sectional Study.

Background: Undernutrition is one of the major public health problems affecting children in developing settings. Despite impressive interventions such as productive safety net program (PSNP) in Ethiopia, there is paucity of information on the prevalence of stunting and its predictors among children aged 6-59 months born to beneciaries and non-beneciaries of the program in Ethiopia. Methods: Community based comparative cross sectional study design was used among randomly selected 1555 children aged 6 to 59 months from households using and not using PSNP respectively in Meta distinct from 5 th -20 th of March, 2017. Data were collected using pre-tested structured questionnaire. Measuring board was used to measure the length/height of the children. Length/height for age z-score was generated using world Health Organization (WHO) Anthros version 3.2.2. Descriptive statistics was used to describe all relevant data. Bivariable and Multivariable logistic regression analyses were used to predictors of the outcome variable. Odds ratio along with 95% condence interval were estimated to measure the strength of the association. Level of statistical signicance was declared at p-value less than 0.05. Result: The prevalence of stunting was 47.7%, 95% CI: (44.1%, 51.5%) among children from PSNP beneciary households and 33.5%, 95% CI: (29.9%, 36.9%) from non-PSNP households respectively. Lack of maternal education [AOR = 3.39; 95% CI: (1.12, 5.11)], lack of women’s empowerment [AOR= 3.48; 95% CI: (2.36, 5.12)], non-attendance of antenatal care fourth visit [AOR= 4.2, 95% CI (2.5, 6.8)], practicing hand washing [AOR= 0.46; 95% CI: (0.28, 0.76)], childhood illness [AOR= 8.41; 95%CI: (4.58, 12.76)], non-exclusive breastfeeding [AOR= 3.6; 95% CI: (2.30, 4.80)], low dietary diversity score [AOR= 4.7; 95% CI (3.0, 7.40)] , and child age between 24 to 59 [AOR 3.2; 95% CI (1.6, 6.3)] were predictors of stunting. Conclusions: The prevalence of stunting was high among children from PSNP households in the study area. Stunting was signicantly associated with maternal socio-demographic and obstetric and child related factors. Therefore, empowering women on household’s issues and improving infant and young child feeding practices could reduce the magnitude of stunting and its impacts on child health.


Background
Undernutrition is known to be one of important causes of childhood illness, disease, and disability among children of low-income countries. Undernourished children are victims of various de ciency states such as night blindness, anemia, iodine de ciency disorder, mental retardation and risk of dying compared with well-nourished children. Globally, 22.2% of under-ve children were stunted. South Asia and sub-Saharan Africa (SSA) including Ethiopia have the greatest lion share of the burden of undernutrition [1,2 ].
Stunting was one of the major nutritional problems of public health importance in Ethiopia. where 38%, 24%, and 10% of children aged 0-59 months were stunted, underweight and wasted respectively though there is marked reginal variation within the country [3][4].
The causes of undernutrition including stunting are multifaceted in SSA including basic, underlying and proximal factors.
Different studies indicated that undernutrition is attributed to maternal illiteracy, large family size, maternal age., being male child, child birth order, amount of water (< 40 litters) for use, lesser child age, lack of extra food during pregnancy/lactation, and low dietary diversity score (DDS below four food groups) due to household food insecurity among others in developing countries like Ethiopia [3][4][5][6][7][8].
In cognizant of this, the government of has launched productive safety net program (PSNP), one of the social protection programs implemented in SSA since 2005 being implemented in different phases as a strategy to reduce the burden of undernutrition by transferring cash and food to poor food-insecure rural households through public works of able bodied labors and direct transfers for none adult able body labor in the households [9][10][11], thus improving availability and access to adequate and nutritious food.
The program has public work, direct support and Institutional support. Some of the public work activities include water point development, road maintenance, agro forestry, irrigation canal, schools and health post maintenance & construction [12]. The national nutrition program (NNP II) was planned to implement in Ethiopia from 2016-2020, including health all sectors should be increased their efforts to enhance good nutritional practices through deferent intervention in reducing and prevention of malnutrition [13].
However, there is a paucity of evidence with regard to stunting and its predictors among children aged 6-59 months who live in productive safety net program bene ciary's households and nonbene ciaries' households in Ethiopia particularly in the study setting. Therefore, this study was aimed at assessing the prevalence of stunting and its predictors among children aged 6-59 months from productive safety net program bene ciaries and non-productive bene ciaries' households.

Methods
Study setting, period, and design Community based comparative cross-sectional study design was used in Meta District of East Harerge Zone of Eastern Ethiopia from 5th to 20th of March, 2017. The district has a total population of 222,016 (28,620 urban and 193,396 rural population) [14].
Chelenko is the main town of district and about 435 km far away from Addis Ababa, the capital city of Ethiopia. The district has 39 rural and 3 urban kebeles (smallest administrative unit in Ethiopia). The rural kebeles of the district were divided in to three agro ecological conditions (12 highlands, 15 midlands and 12 low-lands kebeles). The main ethnic groups in the district are Oromo and Amhara. The rural population was dependent up on farming. The number of functional health facility available in the district are 1 hospital, 7 health centers, 39 rural health posts and 11 different private health facilities. According to the 2017 report of Meta District disaster prevention and preparedness o ce (DPPO) [15], the district has 23 foods insecure rural kebeles that has been covered with PSNP and 16 food secure households. The total number of PSNP bene ciary households was 18,948 (17.5%) of the district's population.

Study Participants and sample selection
Sample size was estimated using double population proportions formula for both PSNP bene ciaries and NPSNP households with the following assumptions: two sided 95% con dence level, 80% power, proportion of stunting among PSNP bene ciaries to be 47.0%, proportion of stunting among PSNP non -bene ciaries to be 41.8% [3], desired precision of 5%, and 5% for nonresponse yielding 1590 children aged 6-59 months from Productive Safety Net bene ciaries (788) and non-Productive Safety Net (671) bene ciaries households in the district. The sample size was then proportionally allocated for each selected kebeles in each arm. Initially, the rural kebeles were strati ed in to high land, midland and low land agro ecology selected purposively. Totally 7 kebeles were selected randomly (3 kebele from NPSNP and 4 kebele from PSNP) to take part in the study. Children from households in each kebele for both population categories from different climatic zones were selected using a strati ed random sampling method.

Data collection and Measurements
The questionnaire was initially prepared in English by reviewing relevant literature related to the study objectives and translated to "afaan Oromo" for better understanding and translated back to English by two different experts who were uent in both languages in order to ensure the consistency of the questionnaire. Fourteen data collectors and two supervisors were involved in the process of data collection after getting the required training with regard to interview and anthropometric measurements. Sociodemographic data were collected using a pretested structured questionnaire which consisted of different sections such as questions related to socio-demographic characteristics, household's assets related questions, productive safety net membership status, maternal speci c questions, child's anthropometry ,and illness two weeks before the survey Anthropometric data such as weight and height were measured following the standard procedures. Length /height was measured to the nearest 0.1 cm.
Recumbent length was measured using measuring board for children younger 24 months while height was measures for those 24 months and older in a standing position for all who can stand well by maintaining ve contact points [16]. The outcome variable in this study was stunting which stands for children having height for age Z -score (HAZ) of less than 2 standard deviation (SD) using world health organization (WHO) new child growth. Stunting was coded as "1" for children with HAZ of less than 2SD and "0" otherwise for further analysis [17]. The explanatory variables used to identify predictors of stunting included paternal education, family size, maternal decision making power and working status, antenatal care visit, place of delivery, exclusive breastfeeding status, dietary diversity, hand wash practice, child's age and illness, number of under ve children in the household, feeding order of the family members, and presence of child's own feeding plate (Table 3).

Data Quality management
Training was given for the data collectors and supervisors for two days on the objectives of the study, interview technique, and anthropometric measurements. The standardization procedure was followed to ensure reliability and validity of anthropometric measurements by computing relative technical error of measurement (TEM) using Emergency Nutrition Assessment Standardized Monitoring and Assessment of Relief and Transitions (ENA SMART) software to compare measurements done by each data collector with selected criterion anthropometrist before deploying the data collectors to the eld to minimize both random and systematic errors attributed to inaccurate anthropometric measurement. Accordingly, the relative TEM for inter observer (validity) and intra observer (reliability) for length /height measurement was 1.5% and 2.0% respectively [18]. The questionnaire was pretested one week before the actual task of data collection on 5% of the estimated sample size for the study in kebeles not included in the study. The whole process of data collection was supervised by supervisors in the eld on daily basis for completeness of each questionnaire. Data were double entered by two data clerks and consistency was crosschecked. Multivariable analysis was done to control for all possible confounders that might mask the true association between independent and outcome variable.

Statistical analysis
Data were manually checked for completeness, edited, double entered onto Epi-Data version 3.1, cleaned, and exported to SPSS version 20 statistical package computer software for further analysis. Moreover, anthropometric indices were also generated using WHO Anthros software version 3.2.2 and coded based on WHO's cut-off points to serve as outcome variables Descriptive statistics was such as numerical summary measures and frequency distribution were used to describe each independent variable relative to the outcome variable. The independent variables were coded based on previous related studies and distribution of responses in the data. Bivariable logistic regression analyses were done to see the association between dependent variable and each independent variable. All covariates with p value < 0.25 during bivariable analyses were considered for multivariable analysis to control for all possible confounders and to identify factors associated with stunting. The multicollinearity effect between independent variables was checked by looking at values of the standard error and correlation coe cient. Variables with standard error > 2 and correlation coe cient of less than 50% were dropped from nal logistic regression model. The tness of the model was tested by Hosmer-Lemeshow goodness of t test whereby p -value = 0.631 for stunting. Odd ratios along with 95% con dence interval were estimated to measure the strength of the association. Level of statistical signi cance was declared at pvalue less than 0.05.
This study also showed that male children were more stunted than female children. The prevalence of stunting among male children were 51.5% for PSNP bene ciary households and 33.8% for NPSNP households while the prevalence of stunting among female children were 44.2% for PSNP bene ciary households and 33.2% for NPSNP households. The overall prevalence was showed that 42.6% for male children and 38.8% for female children (Fig. 2).
Stunting is also more common among all agroecological settings namely highland, midland, and low land. However, stunting was more common among children from PSNP and NPNSP households living in the midland areas compared to their counterparts (highland and lowland). The prevalence of stunting in the highland, mid-land, and low -land were 32.6%, 57.1%, and 25.2%, respectively (Fig. 3).
Predictors of stunting among children aged 6-59 months.
In multivariable logistic regression analysis, lack of maternal education, women's inability to make decision on households issues, lack of fourth antenatal care visit, practicing of hand washing during the critical periods, history of frequent attack by childhood illness, lack of exclusive breastfeeding, eating less diversi ed foods, and child 's age (24-59 months) were signi cantly associated with stunting.
Children born to mothers who have not formal education were 3.39 times [(AOR = 3.39; 95% CI: (1.12, 5.11)], more likely to be stunted than children born to educated mothers. Stunting was 3.5 times [(AOR = 3.48; 95% CI: (2.36, 5.12)] more common among children whose mothers lacked decision making power at household level than those whose mothers could decide on household matters. Children who born to mothers who did not have antenatal care service (ANC IV) during their pregnancy were 4.

Discussion
The nutritional status of children is affected by different interwoven factors which may vary from location to location, and interhouseholds. This study was aimed at assessing the prevalence of stunting and its predictors among children aged 6-59 months living in households enrolled in productive safety net program and those non-enrolled in the program in Meta District of east Harerge zone of eastern Ethiopia. Accordingly, the prevalence of stunting was 47.7% among children aged 6-59 months from PSNP bene ciaries and 35.5% for children who belong to NPSNP bene ciaries' households respectively. Stunting was signi cantly associated with lack of maternal education, women's inability to make decision, non-attendance of fourth antenatal care visit, practicing of hand washing during the critical periods, child's illness, lack of excusive breastfeeding, eating less diversi ed foods, and child 's age (being 24-59 months). However, being a bene ciary of PSNP did not show an association with stunting which needs to be more investigated using large scale study.
In this study, the prevalence of stunting among children aged 6-59 months in PSNP bene ciary and NPSNP bene ciary households was 47.7% and 33.5% respectively. This nding indicates a signi cant difference among the comparison groups. Similar nding was reported from Saesie Tsaeda Emba district, Tigray region, northern Ethiopia which implies that stunting is more common among food insecure households in Ethiopia. The prevalence of stunting observed in Saesie Tsaeda Emba district of productive safety net program bene ciary and non-productive safety net program households were 52.1% and 46.1% respectively [19].
Another national based study reported that the prevalence of stunting was higher among PSNP users (47%) compared with NPSNP users (41.8%) respectively [3]. This difference might be attributed to low food accessibility during months of food gaps among children from PSNP and NPSNP households. Food insecurity may be the result of low rainfall, highly sloppy and degraded farm land, and low productivity of the farm land to support the population's food yield [20].It is important that the existing PSNP needs to be upgraded to the level which satis es the balanced food needs the of poor segment of the area's population.
The risk of stunting increases as the child's age increases. In this study, childen aged 24-59 months were 6. 9 times more likely to be stunted compared to children aged 6-23 months. This result was in line with the national gure [3] and study done in Belesa District [4], where children aged 24-35 months were more likely to be stunted when compared to earlier age group. This might be due to lack of continued child feeding practice with diversi ed diet and may be related to poor breast-feeding practice before the age of 23 months.
Stunting was also signi cantly associated with uneducated mothers/ care takers. The odds of stunting were nearly 4 times higher among chidren born to mothers who have no formal education compared to their counter parts. Similar nding was reported from South Wollo community, northern part of Ethiopia [21], Fadis District community, Hararghe Zone eastern Ethiopia [20] and study from Saesie Tsaeda-Emba District community, Tigray of north Ethiopia [19], and from Tanzania [22]. This might be attributed to the fact that child caring is mainly responsibility of the mother /care giver in Ethiopian context which is added to lack of knowledge of implementing appropriate to age child feeing practices and understanding of the adverse effects of undernutrition on children in the later life.
The odds of stunting were reduced by 62 %, among children from households who practice hand washing during the critical periods. This nding was consistent with study nding from northern part of the country [19]. This might be due to the e ciency of hand washing practice which can reduce pathogens that cause disease and thereby reduce stunting. It also implies in addition to other factors related chronic malnutrition an attempt needs to be made in scaling-up such cost effective intervention in reducing the magnitude of chronic malnutrition.
Children who had recovered from illness in the last 2 weeks of data collection periods were nearly 9 times more likely to be stunted than children who did not illness. This nding was comparable with community-based study done in, northwest Ethiopia [4]) and Latin America Amazon, Western Brazilian [23]. This might be due to increase in body's calorie requirement during illness and destruction of tissue as the result of the microorganisms among stunted children Stunting was nearly 6 times more common among chidren of mothers who have not exclusively breastfed their children in the previous six months compared with their counter parts. This nding is agreement with community based study from Somali region of eastern Ethiopia [24], Saesie Tsaeda Emba district of north Ethiopia [19] and in rural Colombia [25]. This might be explained in such a way that breast milk has substantial amount of anti-infective substances that can safeguard children from the occurrence of infection thereby break malnutrition -infection viscious cycle In this study, there was also a signi cant association between minimal dietary diversity and undernutrition. The odds of stunting was nearly 7 times higher among children who used to eat less diversi ed foods compared with their counter parts This result was comparable with ndings of studies done in Saesie Tsaeda Emba district of northern Ethiopia [19], Somali region of eastern Ethiopia [24] and Colombia [25]. This result implies that diversi ed foods are very important for child growth and development during the age of 6-59 months in availing all micronutrients and macro nutrients for the body requirements Children who born to mothers who were not able to make decision on the household issues were nearly 5 times more likely to have an increased odd of stunting compared with children born to empowered women to decide on households matters This nding was consistent with Maharashtra's study of Indians [6]. This might imply that when women are empowered to decide on household's matters the likelihood of caring for their children increases since they have access to household resources enhancing food accessibility for the household members.
Lack of proper antenatal care follow-up was also found to be associated the development of stunting. The odds of stunting were 7 times higher among children being delivered from mothers who did not attend their fourth ANC four follow-up when compared with their counterparts. Similar nding was reported from Yemen [7]. This might be due to the issues related to child nutrition often addressed during pregnancy upon consecutive ANC visit to alert pregnant women as how to feed their children after delivery in an attempt to prevent the onset of child undernutrition.
This study could have the following limitations. Firstly, some of the mother s /caregivers might have under or over reported their children's age, t which could resulted in inappropriate estimation of the respective nutritional indices. However, age was estimated carefully based on mother's /caregiver/s unforgettable events such as public and religious holidays of the year using local calendar for those who did not have scienti c background. Secondly, recall bias could be introduced upon collecting data on some past background variables such as 24 hours recall questions in which case an interactive 24 recall method was used to minimize such a bias. Thirdly, an anthropometric measurement error could potentially affect the results of this study due to both random and systematic error. Nevertheless, all important efforts such as calibration of measuring equipment, standardization procedure before deploying data enumerators to the eld, and refreshment training for data enumerators were made to minimize the errors. Figure 2 Prevalence of stunting among children disaggregated by sex across PSNP and NPSNP bene ciary households, Meta District, March 2017.

Figure 3
Prevalence of stunting among children across the three agroecological zones, Meta District, March 2017.