Conflict in Somalia: impact on child undernutrition

Introduction In Somalia, protracted conflict and drought have caused population displacement and livelihood destruction. There is also widespread childhood undernutrition. We aimed to determine the independent effects of conflict on wasting and stunting among children aged 6–59 months nationwide in Somalia. Methods Data were from household surveys during 2007–2010, including 73 778 children in 1066 clusters, the Armed Conflict Location and Event Data project database and remote sensing. We used Bayesian hierarchical spatial-temporal regression to examine the effects of conflict on wasting and stunting. Models included individual, household and environmental covariates and recent (<3 months) or longer term (3–12 months) conflict events. Results 15 355 (21%) and 22 739 (31%) observations were from wasted and stunted children, respectively. The conflict was associated with undernutrition independently of the individual, household and environmental factors, and its inclusion improved model performance. Recent conflict was associated with wasting (OR 1.37, 95% credible interval (CrI): (1.33, 1.42) and attributable fraction (AF) 7.6%)) and stunting (OR 1.21, 95% CrI (1.15, 1.28), AF 6.9%). Longer term conflict had greater effects on wasting (OR 1.76, 95% CrI (1.71, 1.81), AF 6.0%) and stunting (OR 1.88, 95% CrI = (1.83, 1.94), AF 7.4%). After controlling for conflict, the harmful effect of internal displacement and protective effects of rainfall and vegetation cover on undernutrition were enhanced. Conclusion Conflict and internal displacement have large effects on undernutrition in ways not fully captured by simply measuring individual, household and environmental factors or drought.

the military seized power in a coup and the Barre regime billed this as a cold war socialist uprising. 1

Survey design
For nutrition survey, sample size was calculated in the number of children aged 6-59 months using Epi Info/Ena 2008 software (Center for Disease Control (CDC) in USA) with household as the sampling unit.
Therefore the final sample size was converted in terms of number of households. 5 6 This required an estimate of the average household size and the proportion of children aged 6-59 months in the population which was estimated from a previous survey or from national statistics.
In each survey, a stratified multi-stage cluster sampling design was adopted where the sampling frame of a selected district was based on the four livelihood definitions (pastoral, agro-pastoral, riverine and fishing) within which 30 communities and 30 households within each community were selected at random. 7 Respective samples sizes (number of households and number of children) were calculated using the Epi Info/Ena 2008 software (Center for Disease Control (CDC) in USA) after considering the population size, estimated prevalence and desired precision. A list of all villages and population within each of the assessed livelihoods served as a sampling frame and was used to construct cumulative population for the assessment area. Selection of respondents within the village was done randomly, preferably from a list of eligible names or a map of households. Where these were not available, the number of households in the village was estimated from the population figures (the total population divided by the mean household size). 8 Further detailed information on FSNAU surveys is available online (http://www.fsnau.org/). 8 u is a spatial structured random effect for clusteri , i v is an unstructured random effect for clusteri . To model spatial dependence between the random effects i u in neighbouring areas we assumed a conditional autoregressive (CAR) prior for these terms. 10 Each i u was modelled (conditional on its weighted neighbors) as normally distributed with mean equal to the mean of the effects of its neighbors and a variance that is inversely proportional to the number  Description of the ACLED conflict by event type categories used in this study. ACLED codes for nine types of events: battle -no change of territory, battle -non-state actor overtakes territory, battle -government regains territory, headquarters or base established, strategic development, riots or protests, violence against civilians, non-violent transfer of territory and remote violence. 12 13 These event types were further grouped into four classes for the analysis: battle, remote violence, violence against civilians and others.

Study class ACLED Class Event description Battle
This is a violent interaction between two politically organized armed groups at a particular time and location. Battle-No change of territory A battle between two armed groups where control of the contested location does not change. Battle-Non-state actor overtakes territory A battle between two armed groups where non-state actor win control of the contested location. Battle-Government regains territory A battle in which the government regains control of a location.

Remote violence
This is where one group is spatially removed but perpetrator did not require physical presence.
Remote violence Refers to an event in which the means for engaging in conflict did not require the physical presence on the perpetrator. This include bombings, mortar and missile attacks Violence against civilians This is a deliberate violent attacks perpetrated by an organized political group such as a rebel, militia or government force against unarmed non-combatants. Violence against civilians This occurs when any armed or violent group attacks civilians. This include rebels, militias and rioters among others.

Riots/Protests
This records group public demonstration, often against a government institution.

Other
This category capture non-violent events that may trigger future events.

Headquarters or base established
This is a non-violent event where non-state group establishes a base or headquarters. Strategic development This is a non-violent event but within the context of the war/dispute that records activity by rebel groups/militia/governments. For instance location and date of peace talks and arrests of high-ranking officials. Non-violent transfer of territory This event describes situations in which rebels or governments acquire control of a location without engaging in a violent act.     Adjusted odds ratio (POR) and 95% credible interval (CrI) of the effect of covariates on wasting and stunting among children aged 6 -59 months in Somalia. The estimates were derived from three models by type of event: Model that did not account for the effect of conflict as covariate; Model that accounted for the effect of any conflict and Model that accounted for the effect of recent and longer term conflict. Values in bold typeface do not contain the value 1 in their 95% CrI, and were considered statistically significant.