Spatial variations and associated factors of food insecurity and child under nutrition in east Gojjam zone, Ethiopia: a multilevel mixed effects model.
No Thumbnail Available
Date
2018-10
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Addis Ababa Universty
Abstract
Background: - Child undernutrition remains a major public health challenge. The magnitude of the
problem varies based on geographical location. In Ethiopia, spatial analysis studies were done based
on coarse spatial resolutions. To be more efficient in targeting interventions geographically, spatial
analysis using micro level spatial resolution is recommended. Accordingly, different studies were done
to identify factors associated with food insecurity and child undernutrition, but most of them ignore
either the individual or community level factors. Hence, identification of spatial variations and
individual and contextual level determinant factors of food insecurity and child undernutrition in
XVI | P a g e
relation to the agroecosystem is essential to deliver targeted, efficient and sustainable solutions to the
problems.
Objectives: - This study determined spatial variations of food insecurity and child undernutrition. The
study also identified the role of individual and community level factors of food insecurity and child
undernutrition using multilevel mixed effects regression analysis in East Gojjam Zone, Ethiopia.
Methods: - An agroecosystem linked to community based comparative cross sectional survey was
conducted among 3108 households with children aged 6-59 months. Multistage cluster random
sampling technique was used to select study participants. Data were collected on household
geographical location, socio-demographic characteristics, child and maternal anthropometry and on
potential individual and community level determinant factors. Collected data were entered using Epi
info version 3.5 and exported to World Health Organization (WHO) Anthro to determine child
nutritional status.
SaTScan software was used to determine spatial variations of food insecurity and child undernutrition
using SaTScan Bernoulli model. To identify the most likely clusters using SaTScan software, the Log
Likelihood Ratio (LLR) at 95% Confidence Interval (CI) and P value less than 0.05 as the level of
significance were considered. To identify determinant factors of food insecurity and child
undernutrition, multilevel mixed effects ordinal regression and multilevel mixed effects linear
regression analyses were used, respectively. The results of fixed effects were shown as an adjusted
odd ratio (AORs) for the multilevel mixed effects ordinal regression and regression coefficients for the
multilevel mixed effects linear regression. The results of random effects were presented as variance
and the intra-class correlation coefficient (ICC) was calculated to estimate unexplained variance
attributable to cluster level.
Results: - The overall prevalence of household food insecurity was 65.3% (95% CI: 63.5, 67.00). The
highest prevalence of food insecurity was observed from the lowlands of the Abay Valley (70.6%,
95% CI: 66.9, 74.2). Followed by the hilly and mountainous highland areas of Choke Mountain
(69.8%, 95% CI: 65.9, 73.3). Similarly, sample clusters taken from hilly and mountainous highland
areas (LLR: 11.64; P<0.01) and low lands of the Abay Valley (LLR: 8.23; P<0.05) were identified as
the most likely primary and secondary clusters, respectively.
XVII | P a g e
The prevalence of stunting 39.0% (95% CI: 37.32, 40.75), 18.7% (95% CI: 17.32, 20.0) underweight,
and 12.22% (95% CI: 11.12, 13.42) wasting were observed in the study area. The highest prevalence
of wasting (15.9%; 95% CI: 13.5, 18.8) was observed from hilly and mountainous highlands. The
highest prevalence of child stunting (42.4%; 95% CI: 38.5, 46. 6) and underweight (22.9%; 95% CI:
19.7, 26.3)) were observed from the lowlands of Abay Valley. SaTScan spatial analysis indicated that
sample clusters taken from the hilly and mountainous highlands were the most likely primary cluster
for child wasting (LLR: 13.0, p < 0.01) and underweight (LLR: 23.16, p < 0.001). Also, primary
cluster for child stunting was identified from lowlands of Abay Valley (LLR: 10.78, p<0.05).
After adjusting for both individual and community level determinant factors, 1.5% (p<0.001) of the
variance of food insecurity was attributable to the cluster level. Similarly, after adjusting for all
potential determinant factors, 2.4% (p<0.001) of the child weight for height Z score and 1.4%
(p<0.001) of the child height for age Z score variance were due to cluster level. From level one
factors, in the final model, household head being male, marital status being in union, higher parental
education, women’s participation in household decision making, having additional income sources,
better crop production in the survey year and application of chemical fertilizer have a positive
influence in mitigating household food insecurity. From community level determinant factors,
households being from hilly and mountainous highlands and lowlands of the Abay Valley were more
severely household food insecure compared to midland plain areas. Households with better farmland
size showed less severe household food insecurity in the study area.
In the study, from level one factors, the number of under five children, antenatal (ANC) follow up,
breast feeding initiation time, household dietary diversity, mother nutritional status, household food
insecurity and diarrheal morbidity were associated with weight for height Z score. From level two
factors, agroecosystem characteristics, proper household refuse disposal practice, agroecosystem
characteristics and proper latrine utilization were significantly associated with child weight for height
Z score. From level one factors, child age in months, child gender, the number of under-five children
in the household, child immunization status, breastfeeding initiation time, mother nutritional status,
child diarrheal morbidity, household level water treatment practice and household dietary diversity
showed a statistical significant association with child height for age Z score. From level two factors,
agroecosystem characteristics, proper household refuse disposal practice and proper latrine utilization
were significantly associated with child height-for-age Z score.
XVIII | P a g e
Conclusions: - The prevalence of food insecurity and child undernutrition were public health concerns
in the study area. Spatial variations of household food insecurity and child undernutrition were
observed across the agroecosystems. Households from the lowlands of Abay Valley and hilly and
mountainous highland areas were more vulnerable to food insecurity and child undernutrition
compared to midland areas. The SaTScan cluster level spatial analysis identified statistical significant
hotspot clusters for food insecurity, childhood stunting, underweight and wasting. The multilevel
mixed effects analysis indicated that the heterogeneity of food insecurity, childhood stunting and
wasting were observed after adjusting to potential individual and community level determinant factors.
Both individual and community level factors played a significant role in determining food insecurity
and child undernutrition (stunting and wasting). An agroecosystem characteristic was one of the
community level factors affecting household food insecurity and child undernutrition.
Recommendations: - The spatial variation of food insecurity and child undernutrition based on
agroecosystem characteristics should be fully understood by program implementers and policy makers
during planning, resource allocation and community mobilization in the study area. Water, sanitation
and hygiene interventions are important in the study area. Further study on spatiotemporal variations
of food insecurity and child undernutrition at different time is recommended. Also, food insecurity and
child undernutrition intervention strategies and plans designed using aggregated or macro level
evidence may not indicate the true picture of spatial distribution of the problem at lower government
administrative units. So, program level planning may take into account agroecosystem based micro
level variations to allocate resources. Policy and intervention strategies aiming at mitigating food
insecurity and child undernutrition should address the effects of lower and community level
determinant factors using the integration of individual/household level and geographical targeting.
Description
Keywords
Food insecurity, child undernutrition, spatial variations, multilevel, agroecosystem,Ethiopia.