Estimating spatial inequalities of urban child mortality

Abstract

Background: Recent studies indicate that the traditional rural-urban dichotomy pointing to cities as places of better health in the developing world can be complicated by poverty differentials. Knowledge of spatial patterns is essential to understanding the processes that link individual demographic outcomes to characteristics of a place. A significant limitation, however, is the lack of spatial data and methods that offer flexibility in data inputs.

Objective: This paper tackles some of the issues in calculating intra-urban child mortality by combining multiple data sets in Accra, Ghana and applying a new method developed by Rajaratnam et al. (2010) that efficiently uses summary birth histories for creating local-level measures of under-five child mortality (5q0). Intra-urban 5q0 rates are then compared with characteristics of the environment that may be linked to child mortality.

Methods: Rates of child mortality are calculated for 16 urban zones within Accra for birth cohorts from 1987 to 2006. Estimates are compared to calculated 5q0 rates from full birth histories. 5q0 estimates are then related to zone measures of slum characteristics, housing quality, health facilities, and vegetation using a simple trendline R² analysis.

Results: Results suggest the potential value of the Rajaratnam et al. method at the micro-spatial scale. Estimated rates indicate that there is variability in child mortality between zones, with a spread of up to 50 deaths per 1,000 births. Furthermore, there is evidence that child mortality is connected to environmental factors such as housing quality, slum-like conditions, and neighborhood levels of vegetation.

Figure 1

Accra Metropolitan Area in the south of Ghana

Figure 2

Zone counts of children ever born (CEB) and children dead (CD) for each age cohort for the pooled 2003 and 2008 data sets

Figure 3

AMA trends from 1979 to 2006 of 5q0 for the 2003 and 2008 data sets using MAC (Maternal Age Cohort), MAP (Maternal Age Period), LOESS (regression MAC 2003/2008 and MAP 2003/2008 results), and 2003 and 2008 full birth history calculations

Figure 4

LOESS trend estimates from 1980 to 2006 for each zone. Full birth-history-derived calculations of 5q0 from the 2003 and 2008 data are also shown for comparison

Figure 5

LOESS five-year 5q0 estimates from 1987 to 2006 for zones in Accra

Figure 6

Trendlines of the scaled zone slum index, housing quality, and vegetation as related to 1997-2001 zone 5q0 (left) and change in 5q0 from 1992 to 2001 (right)