Armed conflict and child mortality in Africa: a geospatial analysis

Abstract

Background: A substantial portion of child deaths in Africa take place in countries with recent history of armed conflict and political instability. However, the extent to which armed conflict is an important cause of child mortality, especially in Africa, remains unknown.

Methods: We matched child survival with proximity to armed conflict using information in the Uppsala Conflict Data Program Georeferenced Events Dataset on the location and intensity of armed conflict from 1995 to 2015 together with the location, timing, and survival of infants younger than 1 year (primary outcome) in 35 African countries. We measured the increase in mortality risk for infants exposed to armed conflicts within 50 km in the year of birth and, to study conflicts' extended health risks, up to 250 km away and 10 years before birth. We also examined the effects of conflicts of varying intensity and chronicity (conflicts lasting several years), and effect heterogeneity by residence and sex of the child. We then estimated the number and portion of deaths of infants younger than 1 year related to conflict.

Findings: We identified 15 441 armed conflict events that led to 968 444 combat-related deaths and matched these data with 1·99 million births and 133 361 infant deaths (infant mortality of 67 deaths per 1000 births) between 1995 and 2015. A child born within 50 km of an armed conflict had a risk of dying before reaching age 1 year of 5·2 per 1000 births higher than being born in the same region during periods without conflict (95% CI 3·7-6·7; a 7·7% increase above baseline). This increased risk of dying ranged from a 3·0% increase for armed conflicts with one to four deaths to a 26·7% increase for armed conflicts with more than 1000 deaths. We find evidence of increased mortality risk from an armed conflict up to 100 km away, and for 8 years after conflicts, with cumulative increase in infant mortality two to four times higher than the contemporaneous increase. In the entire continent, the number of infant deaths related to conflict from 1995 to 2015 was between 3·2 and 3·6 times the number of direct deaths from armed conflicts.

Interpretation: Armed conflict substantially and persistently increases infant mortality in Africa, with effect sizes on a scale with malnutrition and several times greater than existing estimates of the mortality burden of conflict. The toll of conflict on children, who are presumably not combatants, underscores the indirect toll of conflict on civilian populations, and the importance of developing interventions to address child health in areas of conflict.

Funding: The Doris Duke Charitable Foundation, and the Centre for Global Child Health at the Hospital for Sick Children.

Figure 1:

The distribution of armed conflicts in Africa, 1995-2015. The map shows the UCDP conflict location data in the continent, with changing location over time. The study countries are gray with thick borders. The map shows the regional changes in high-density conflicts, as well as the lack of any obvious decline in the frequency of conflict in more recent years.

Figure 2:

Impact of deadly conflict in first year of life on under-1 mortality. The top panel (Figure 2A) shows the effect of any conflict within 50km of the infant’s cluster (red bar, equation 1), as well as the rising risk from conflicts of increasing intensity (equation 2). The bottom panel (Figure 2B) shows the increasing risk from conflicts of increasing duration. We estimate the exposure based on whether or not the conflict lasted for less than 1 year, 1-2 years, and so on up to 5 or more years of consecutive conflict near the child’s cluster. The left Y axis indicated the increase in mortality risk in deaths per 1,000 births; the right Y axis converts that risk to the percent increase above baseline mortality risk.

Figure 3:

Impact of conflict on under-1 mortality over time. The top panel (Figure 3A) shows the coefficients from a regression with conflict exposure indicators for each year up to 10 years before birth. Each coefficient represents the independent lingering risk of mortality in the first year of life from historical conflict taking place up to 10 years before birth. Conflicts up to 8 years prior to birth show an independent relationship with increased under-1 mortality. The bottom panel (3B) shows the cumulative mortality risk over time for conflicts that occurred within 0-50km, 50+ - 100km, and 100+ - 250km from a child’s birth, estimated in one regression, and reveals an attenuated effect at farther distances. Grey bands represent 95% confidence intervals.

Figure 4:

Number of under-5 deaths due to conflict in Africa. We used 0.1 degree × 0.1 degree gridded estimates of the number of births, the infant mortality rate, and conflict timing and location to estimate the number of observed deaths in each grid cell, and then the number of deaths expected in the absence of conflict. We then aggregated the grid cell level estimates for the period 1995 to 2015. We used national infant mortality rates for countries without gridded infant mortality rate. The scale indicates the number of deaths in each grid cell (roughly 10km × 10km). The biggest clusters of child deaths are apparent in Nigeria, East-Central Africa (Uganda, DRC, Burundi, Kenya), Ethiopia, Lybia, Egypt, and West Africa (centered around Sierra Leone). Thick borders and gray background represent study countries.