Effects of dietary and health factors on nutritional status of children under pastoral settings in Borana, southern Ethiopia, August–October 2015


 Background: Childhood undernourishment is a serious public health problem globally, and being responsible for higher mortalities in children and enormous health costs in sub-Saharan Africa. However, scarcity of data on the magnitude of malnutrition and underlying causes especially in the pastoral system limits the effectiveness of potential interventions. This study was intended to investigate the nutritional status and associated factors among children in Borana pastoral system, southern Ethiopia.Methods: A community based cross-sectional study, using multistage cluster sampling, was conducted from August to October 2015. Dietary diversity score (DDS), milk and meal frequencies, anthropometric measurements, and socio-economic variables were recorded for 538 children aged 6 – 59 months. Multivariable generalized linear model (GLM) with log link function was applied to ascertain determinants of malnutrition. The strength of association was assessed with prevalence ratio (PR). Results: Underweight, stunting, and wasting was prevalent in 28.3% (95% CI: 24.4, 32.1), 41.1% (95% CI: 36.7, 45.1), and 9.8% (95% CI: 7.3, 12.4) of the children, respectively. Children who consumed more diverse foods were at lower risk of being underweight (PR=0.72, 95%CL: 0.59–0.88), stunted (PR=0.80, 95%CL: 0.68–0.93) and wasted (PR=0.42, 95%CL: 0.27–0.66) compared to others. Intake of increased milk frequency was also associated with lower risk of underweight (PR=0.86, 95%CL: 0.76–0.97), stunting (PR=0.83, 95%CL: 0.75–0.91) and wasting (PR=0.73, 95%CL: 0.56–0.96). The risk of underweight (PR=1.02, 95%CL: 1.01–1.03) stunting (PR=1.01, 95%CL: 1.00–1.02), and wasting (PR=1.01, 95%CL: 1.00–1.04) was significantly increased with age whilst no difference was observed between boys and girls. Children who lived far away from health care facilities were 1.2 and 1.4 times more likely to be stunted and wasted than those residing nearby. Ownership of toilet and living close to market were associated with reduced stunting while illness was associated with increased risk of underweight. Conclusion: The high prevalence of stunting among pastoral children is a serious public health concern, and calls for urgent action. Association of nutritional status with dietary intake, health status, access to health services and toilet availability underlines the need for improved feeding practices, health care and sanitary conditions in the pastoral community.


Methodology
Study design and sampling procedure A community-based cross-sectional study was conducted in Borana, southern Ethiopia between August and October 2015. The area is characterized by an arid and semi-arid environment, and a pastoral/agro-pastoral production system. We applied a multistage cluster sampling method to recruit the study participants from the area. Six out of the 18 pastoral associations (PAs, the lowest administrative units), were randomly selected from Yabelo district. The district is located at the center of Borana zone, and was considered suitable to easily access to all randomly selected PAs and villages. Subsequently, about half of the villages (having 10 to 20 households) were randomly sampled from each selected PAs. All eligible households in a village, who had at least one child within the age range of 6 to 59 months, were sampled by cluster sampling. Finally, every mother and her one or more children from a selected village were enrolled for the study. Exclusion criteria were children with physical disability or abnormalities, mental impairment, edematous conditions, evidence of chronic disease (e.g. tuberculosis), and those with signs of fever, vomiting, diarrhea and cough.
Sample size was estimated using presumed population proportion of minimum meal frequency (42%) report of the Ethiopian demographic and health survey [11], 95% con dence level, marginal error of 5% and design effect (1.5 times) which gave a total of 561 samples. Subsequently, we were able to sample a total of 538 children with complete information while 13 children were excluded from the nal dataset: eight of them had incomplete data, three were with physical abnormalities and two were ill (showing diarrhea, fever and respiratory symptoms).

Dietary data collection
Data on dietary intake such as dietary diversity and amount of consumed food items, dairy consumption (amount and frequency) and meal frequency were collected using 24 hours dietary recalls. All food items consumed by children during the previous 24 hours were listed and qualitatively described by mothers or primary caregivers, and the food items were categorized according to the seven food groups model of the World Health Organization (WHO) guidelines [22]. Recipes and ingredients used to prepare the food items, dairy products and semisolid food preparations were also recorded. Accordingly, a child was considered to have received a minimum dietary diversity if she or he had consumed four and above of the seven food groups during the last 24 hours preceding the survey.
Similarly, minimum meal frequency was met if a child received a minimum of three meals with one to two snacks per day.
The questionnaire was used to collect data on socio-economic and demographic variables (family size, number of children, livestock ownership and species diversity, crop cultivation, income sources, possession of radio and mobile phone, distance from town and basic services), maternal characteristics (age, number of children ever born, use of extra food during pregnancy or lactation, education, autonomy in decision-making), child characteristics (age, sex, birth order, place of delivery, size at birth). Data on hygienic and sanitary conditions (access to sanitary facilities, hand washings and cleanings of utensils), water sources, water treatments, health related information (visit by health workers, vaccinations, vitamin supplement, and distance to health centers and history of illness three weeks before the visit) were also collected. Information on child feeding practices, child health care, the frequency and duration of breastfeeding and intake of animal source foods was also recorded.

Anthropometric measurements
Weight and height measurements of children wearing light cloth and no shoes were recorded according to recommended procedures of WHO guidelines. Weights were taken at standing or hanging positions using Seca scales (Seca GmbH, Hamburg Germany) with 0.1 kg digit. Heights (standing) or lengths (lying for children below 24 months) in centimeters were measured with a measuring board graduated by 0.1 cm. The age of a child (in months) was obtained either from a birth certi cate or the child's vaccination card, and mother's recall was used for those without records. The recall was assisted by referring to the Borana Gada time calendar, local events, seasons, and months to determine the birthdates. The weight and height measurements combined with age and sex of children were converted to anthropometric Z-scores of weight-for-age, height-for-age and weight-for-height according to WHO standards using Emergency Nutrition Assessment software [23]. A child with Z-scores of lower than -2 standard deviations (SD) was regarded as malnourished i.e. underweight, stunting or wasting [15]. These indices were regarded as proxy measures of nutritional status and analyzed against selected independent variables. Collection of dietary data, socioeconomic variables and anthropometric measurements were conducted by three experienced health workers with diplomas in clinical nursing.

Data analysis
Data entry, coding and checking for errors were done using Microsoft Excel spreadsheet, and imported to Stata version 14.2 (Stata Corp. College Station, USA) for all statistical analyses. Dietary intake indicators (dietary diversity score, meal frequency, milk frequency) and socio-economic variables were descriptively summarized as percentage or mean. Most of the independent variables were categorized as binary (dummy) whereas others were regarded as quantitative variables (e.g. age of children in months, dietary diversity score, milk and meal frequency, and travel time in hours to various institutions). A total of 40 independent variables related to socio-economic variables, dietary intake indicators, maternal and child characteristic, health and sanitary conditions were selected as potential determinants of nutritional status of the study children. The nutritional status (outcome variable) was categorizing as either malnourished (status =1 for underweight, stunting or wasting) if a child had Z-score value below -2 SD or regarded as normal (status=0) when Z-score value >-2SD.
Since, proportions of malnutrition of the study children are close to 10% or higher, the strength of association of factors with nutritional status was assessed with prevalence ratio (PR) rather than with odds ratio that tends to overestimate for large proportions (>10%). For this reason, we applied multivariable generalized linear model (GLM) with Poisson family, log link function, and option reporting exponential coe cient. Pastoral association (PA) was entered as a cluster variable in the clustered robust option of the model. A stepwise backward selection procedure was used to retain variables by setting p<0.15 in the model. Meal frequency and sex were retained in the nal model for comparison regardless of their p-value. Multi-collinearity among the independent variables was checked using variance in ation factor (VIF), and those with VIF <9 were kept in the model.

Socioeconomic and demographic characteristics
General characteristics of households, study children, and their dietary data are summarized in Table 1. All households were dependent on livestock production in addition to engagement in crop cultivation (88 %) and with non-pastoral income sources (25%). Study households mostly kept cattle (62%) while the remaining proportion mainly relied on camel production. Households comprised seven individuals on average, and 10% of them were polygamous families. A large proportion of the respondents had a toilet (62%) and at least one mobile phone (66%) and one-third owned one radio (35%) per family. Few households (16.2%) had access to developed water sources while the majority (84%) uses unprotected sources such as pond, deep well or springs.
Mothers or caretakers had average age of 29 years, low level of literacy (12.5%) and meager participation in deciding on household resources (18%). The average fertility rate was ve children per mother, out of which 30% of the children were under ve years. Most of the mothers (83%) had received advice on child care and feeding practices. Around 70% of them practiced hand washing after toilet use, and before preparing food or feeding their children. Some mothers (27.1%) also reported treating drinking water (at point of use) by boiling, sand ltration or using a chemical locally called "Bishangari" (aluminum sulphate and calcium hypochlorite).
The study children (270 boys and 268 girls) had an average age, height and weight of 32.7 months, 86.0 cm, and 11.2 kg, respectively. The mean Z-score of weight for age, height for age, and weight for height were -1.45, -1.66, and -0.74, respectively. Nearly 29% of the children were under two years of age, and 86% of them were breastfeeding.
About four percent of the children were living with non-biological mothers including grandmothers and relatives.
Illness (mainly diarrhea, respiratory symptoms and fever) occurrences were reported for 32% of the children during the past three weeks before the survey. The average milk and meal frequencies were 4.3 and 2.3 times per day. The average dietary diversity score (DDS) was 2.7 food groups, so that most children (82%) consumed less than four food groups of the WHO minimum dietary diversity. Majority of the children (68.8%) also did not meet the WHO minimum meal frequency of at least three meals with one to two snacks per day. In contrast, a satisfactory consumption of dairy products was observed with over 90% of the children having dairy intake more than four times a day.
The relationship of stunting, underweight and wasting was plotted against the age of the children (Fig. 1). Anthropometric indices have initially decreased with age until 40 months of age, after which WAZ and HAZ showed a slight increase.

Discussion
Here we provide empirical data on nutritional status and associated factors among children (aged 6-59 months) in the Borana community, that is undergoing a process of agro-pastoralism or settlement, and is increasingly engaged in crop cultivation [17,18]. The magnitude of observed stunting (41.1%), underweight (28.3%), and wasting (9.8%) are in the "very high, high and serious" categories of WHO prevalence thresholds, respectively [15]. Observed high prevalence of malnutrition can be attributed to the observed inadequate dietary intake, in that most of the children (82%) consumed below the WHO minimum dietary diversity of four food groups. Such low dietary diversity (below four food groups) have been also reported in pastoral [24][25][26] and agrarian communities of Ethiopia [27][28][29][30][31].
Our ndings of stunting and wasting is higher than the prevalence of stunting (19%) and wasting (below 5%) that reported by Lindtjorn et al. [32] from the same study area about two decades ago. Although comparison of two cross-sectional studies has limitations, the observed difference may indicate that the nutritional status of children in the study area is deteriorating over time, and is becoming as high as prevalence reports from mixed farming area of the country [5,7,33,34]. Observed changes in nutritional status over time (between the two studies) may be linked to the noticeable changes in Borana areas; such as increasing crop cultivation [18], climate variability and rangeland degradation [19], decreasing herd size per households [35], and weakening of pastoral lifestyle (e.g. mobility and exibility) [17], besides human population growth. These factors in one way or another can in uence household level food availability and reduce their economic access to food. In addition to low harvest rates in arid environments [18], crop cultivation has a substantial impact on livestock production due to high competition for land, and likely reduces intake of animal source foods. In line with this, two case studies comparing settled and mobile pastoral communities in Kenya and Nigeria have also documented the adverse effects of settlement on the nutritional status of children [20,21].
Occurrence of stunting suggests repeated infections and/or long-term inadequate nutrient intake that often occurs in pastoral areas during dry periods and in droughts years, when dairy production and terms of trading with food grain fall [32,36]. Thus, such very high prevalence of stunting is of great concern, as it leads to delayed motor development and impaired cognitive development that could be irreversible. Wasting, however, can be caused by acute food shortage and illness such as diarrhea, and is often associated with child mortality. Underweight is composite indicator that combines linear growth impediment and low weight for height as a result of current insu cient dietary intake and illness [1]. In general, recorded high magnitude of malnutrition in the study area calls for urgent attention, as it results in poor school performance, reduced intellectual capacity of children and leads to lower productivity. The socio-economic burden of childhood malnutrition is evidenced by causing 44% of the health costs, 28% of child mortality, 16% of all repetitions in primary school, and 67% of the adults having suffered from childhood stunting in Ethiopia [12].
Increased prevalence of underweight, stunting, and wasting with age of children is in agreement with previous ndings [5,7,34,37,38]. This could be linked to the introduction of supplemental diets of less nutrient-dense cereals. A literature review by Onyango [39] showed that stunting in African children occurs at early infancy and gets worse after two years of age, which has been hypothesized to be linked to the introduction of less nutrient-dense supplemental diets. In addition to poor nutrient contents, cereals and tubers also contain anti-nutrient factors (e.g. phytate) which interfere with the absorption of essential micronutrients from consumed food items and reduce their bioavailability. Moreover, in higher ages children start to interact with the environment, and consume contaminated food and water that increase the risk of exposure to infection and diarrhea episodes [4,13]. Hence, paying more attention to feeding practices, hygienic conditions and health care during at weaning age may contribute to improved nutrition status.
Contrasting results can be found in the literature regarding differences between boys and girls. In several studies boys were found to suffer more from undernourishment compared to girls [5,7,40,41], while others reported no difference [37,38,42] or found girls to be at a higher risk of malnourishment [43]. In pastoral communities such as Borana, care for children likely disfavors girls, and greater nutritional investment in girls than boys is unlikely. It is not clear whether mothers or biological differences compensate for such socio-cultural disparities, and favor girls to be better off. Possible reasons are differences in nutritional requirements as well as e ciency of nutrient conversions between girls and boys of the same age.
Dietary diversity and milk frequency emerged as major predictors of the nutritional status of study children, having signi cant association with lower risk of stunting, underweight, and wasting. The observed protective effects of increased dietary diversity on nutritional status of the study children con rms earlier studies' ndings [7,[44][45][46]. According to Motbainor et al. [47], low dietary diversity was signi cantly associated with higher prevalence of stunting. In another study, Steyn and colleagues [44] have demonstrated dietary diversity score was a good estimate of nutritional adequacy and nutritional status, in that children who had low DDS were at higher risk of undernutrition. These ndings imply that increasing dietary diversity results in improvement of the dietary quality of consumed food items (e.g. animal source foods) and intake of essential micronutrients that have roles in normal growth and immune system [38,42].
Signi cant association of dairy intake with improved anthropometric indices is consistent with other studies in which milk consumption was associated with reduced prevalence of malnutrition and health problems, in addition to improving cognitive functions and school performances [48]. Other studies also found milk consumption to be associated with improved nutritional status of children, and to reduce the prevalence of morbidity and mortality [24,49]. In a case-control studies among school children in Iran [50] and Vietnam [48], milk consumption was signi cantly associated with higher anthropometric measurements of children in intervention groups. Milk is regarded as an ultimate food that provides energy, protein, and several micronutrients and bioactive peptides with growth-promoting abilities, thus vitally enhancing the health and growth of children [51,52]. In our study, we did not observe signi cant association between nutritional status and number of meals consumed per day. This could be explained by the limitation of frequency based indicator i.e. the same food group might be frequently consumed.
Cereals preparations and to some extent beans have been found to be the most commonly consumed foods in Borana area [26]. Hence, meal frequency does not necessarily correlate with the extent of dietary diversity and nutritional quality.
Other notable ndings of our study were the association of nutritional status with health-related factors such as physical access to health services, availability of family toilet, and occurrence of illness during the three weeks before the survey. Households reside nearby health institutions might have better health information, and higher tendency to visit health institution and get health services compared to those living far away. It has been well documented that disease control and prevention activities through sanitation, promotion of breastfeeding, vaccination, and treatments vitally improve the health status of children, thereby contributing to normal growth and development [4,53]. Another study in Ethiopia also documented improvement of nutritional status following immunization [54].
Association of latrine ownership with reduced occurrence of stunting points to the role of improved sanitary and hygiene practices on reducing illness and malnutrition. A study in India also reported lower risk of underweight and stunting among children whose households use toilets [55]. In another study, Babatunde and Qaim [56] also observed a signi cant association of toilet use with underweight and wasting among children in Nigeria. Toilet use indicates better sanitary conditions that reduce the risk of exposure to infections and consequent effects on nutritional status of children. The Ethiopian health extension services, (encompassing various activities like toilet construction, proper hand washing, improved hygienic practices etc.) have initiated efforts to improve public health, sanitary and hygienic conditions, and child care and nutrition in rural areas. Strengthening the health extension services would enhance disease control and prevention through improved sanitary and hygienic practices, and ultimately contribute to effective micronutrient absorption and utilization.
Association of illness occurrence with higher risk of underweight can be expected as underweight re ects current poor dietary intake and illnesses; typically diarrhea and respiratory infections [1]. Bloss et al. [57] also reported children having diarrhea, upper respiratory infections or other illnesses in the past few weeks were three folds more likely to be underweight than other groups in Kenya. Another study [54] in Ethiopia also found a signi cant contribution of morbidity to the increased prevalence of wasting. Reviewed studies further demonstrated illness episodes to be the most frequently reported determinants of underweight and wasting [4], [13]. In general, infections can reduce the appetite of children (reduce intake), interfering with absorption, and result in nutrient and uid losses, so that leading to temporary weight losses.
The adverse effects of remoteness of market on nutritional status of children (increased risk of stunting) can be explained by the large share of the food supply in pastoral areas originates from markets, affecting dietary intake. A study from Mali [58] reported positive effects of close-proximity to markets on the dietary intake of children, subsequently leading to improvement in anthropometric measures. Stifel and Minten [59] also found higher consumption expenditures and more diverse diets among households nearer to markets compared to those living far-off in Ethiopia. Besides reducing transition cost, proximity to marketplaces may increase opportunities for household members to be engaged in non-pastoral income generating activities or provide better access to information that may improve their nutrition knowledge. It's worth noting that physical access to market alone may not ensure availability and access to foods, as dietary intake is mostly rooted in socioeconomic status of the households.
Surprisingly, we observed the associations of non-pastoral income sources and possession of radio with increased risk of underweight and wasting, respectively. A positive nutrition impact of off-farm income sources has been reported for farming households from Nigeria via increased household income that enables better access to food [56]. A negative outcome of off-farm activities was also observed and associated with its competition for family labor and its negative impact on farming activities in Mexico [60]. Our observation could be explained by the dearth of off-farm income opportunities in a pastoral area, which are mainly practiced by resource-poor households in order to stabilize their income and reduce distress selling of livestock. Those households mainly engaged in marginal off-farm income sources (e.g. petty trade, casual labor, and selling bush products), the earnings of which cannot have a signi cant effect on the nutritional status of children.
Lastly, this work is among the few study reports from pastoral areas and needs to call attention to policymakers in order to address the challenges of malnutrition among children given the changing environmental and socioeconomic factors that disfavor pastoral coping mechanisms. The Ethiopian government has already showed its commitment to end childhood malnutrition by 2030, so that has initiated a National Nutrition Program (2016-2020) to improve child nutrition through proven nutrition interventions. Thus, our ndings can contribute to the required data in addressing the nutrition interventions under the National Nutrition Program.

Limitations
Given the cross-sectional design of the study, it was not possible to investigate the temporal as well as causal relationship of dietary intake with nutritional status of children (although an attempt was made to collect monthly dietary diversity and dairy consumption for sub-samples). In pastoral areas, where consumption varies considerably with seasonal food availability, we selected the minor wet season, when food availability and diversity is considered to be average. Helminthes and other health problems also could have signi cant impact on the nutritional status of the children and must be taken into considerations when interpreting our ndings.

Conclusion
The magnitude of observed stunting (41.1%), underweight (28.3%), and wasting (9.8%) are in the "very high, high, and serious" categories of WHO classi cation, respectively. In particular, the high prevalence of stunting among pastoral children is of great concern, indicating long-term undernourishment, and calling for urgent intervention measures. Dietary diversity and dairy intake emerged as major predictors of the nutritional status of study children, thus underpinning the signi cance of improving mothers' knowledge on feeding practices. Association of nutritional status with access to health institutions, illness, and availability of latrine points to the simultaneous need for Tufts University. Authors are thankful to mothers and caretakers of study children, staffs of Yabelo health centers and IMMANA coordinating team whose contributions were extremely valuable.
Authors' contributions BM was responsible for the study design, data acquisition and analysis, drafting and revising the manuscript, and interpretation of results. AH and UK participated in the design of the study, drafting and revising the manuscript, and interpretation of ndings. All authors read and approved the nal manuscript.
Ethics approval and consent to participate Following submission of proposal and questionnaire formats, ethical clearance was obtained from the ethical committee of Health Bureau of South Regional Sate. A letter of permission was also received from the Borana zone health department. Verbal informed consent was obtained from mothers and caretakers of children after informing them the purpose, bene t, and con dentiality of the information. Participants were informed about their voluntary participation in study, and they were given the chance to ask any question related to the study.
Funding IMMANA postdoctoral fellowship funded by UKAID from the UK government Consent for publication Not applicable Availability of the data The datasets used and analyzed during the current study are available from the corresponding author on reasonable request.

Figure 1
Relationship of nutritional status (anthropometric indices) with age of children in Borana Ethiopia, August -October