Background
The existence of health-related inequalities across different areas as well as across different socio-economic groups is well documented [
1‐
5]. Numerous studies have demonstrated that urban areas have lower infant and child mortality rates than rural areas [
3,
4,
6‐
9]; and there is even more extensive literature showing that mortality is much lower among higher socio-economic groups than among the poor [
2,
10‐
14].
African cities are growing faster than cities in most other regions of the world [
15], and by 2020 more people are estimated to be living in African cities than in European cities. In 2010, almost 40% of Africans lived in urban areas, with about 33% of urban Africans living in cities of more than 1 million inhabitants [
15]. Living conditions in Sub-Saharan African cities remain particularly poor, and more than 60% of city dwellers live in slums [
16], meaning that they lack either durable housing, sufficient space, access to safe water or basic sanitation. In addition, economic inequality as measured by the Gini-coefficient is higher in African cities than in most other cities of the world [
17]. In Northern African cities, living conditions are better with about 13% of the urban population living in slums [
16].
In 2010, child (under-five years old) mortality in Africa was estimated to be at 111 per 1,000 live births [
18]. On average 132 children died per 1,000 live births in rural areas compared to 102 in urban areas (based on data from 28 African countries with recent (2005 to 2011) data from demographic and health surveys (DHS)) [
19]. However, while child mortality is on average lower in urban areas, a previous study and a World Health Organization (WHO) report have highlighted that child mortality of the poor is often higher in urban than in rural areas [
20,
21]; this is explained by the fact that inequalities between rich and poor tend to be larger in cities than in rural areas.
Inequalities in child mortality are generally considered to constitute inequities because they are perceived to be unfair, socially produced and potentially modifiable [
22]. A 2010 joint WHO and United Nations Human Settlements Programme (UN-HABITAT) report [
21] urged countries to disaggregate data within cities in order to unmask health-related inequities. However, available studies of urban inequalities in child mortality usually look at all urban residents within a country [
7,
9,
20,
23]. Only very few studies are available that investigate inequalities within specific African cities [
24‐
26], and systematic information comparing the magnitude of inequalities in child mortality across different cities and the development of inequalities over time is unavailable. Yet, unless inequalities are investigated and revealed, it is impossible for governments to act against them [
21].
This study aimed to systematically compare inequalities in child mortality across 10 major African cities and to investigate the development of inequalities over time (the last decade or so, depending on the availability of data). More specifically, the objectives were: (1) to calculate child mortality rates by wealth quintiles within cities; (2) to quantify the degree of inequality within cities using different established measures of inequality [
27,
28]; (3) to determine whether child mortality is more unequally distributed in some cities than in others; and (4) to assess whether certain cities have succeeded in reducing inequalities over time.
Discussion
This is the first study to systematically investigate socio-economic inequalities in child mortality within and across African cities and their development over time. We disaggregated data from two rounds of demographic and health surveys and calculated four measures of socio-economic inequality for ten cities in Africa. The results show that in most cities, child mortality is considerably higher among the poor than among the rich, with the difference between the poorest quintile and the richest quintile reaching as much as 108 deaths per 1,000 live births in Abidjan in 2011–2012. Around the year 2000, Dar es Salaam had the highest level of inequality, while Abidjan and Cairo had rather low absolute (Cairo) and relative (Abidjan) inequality. Since then, inequality appears to have reduced in about half of the included cities (Cairo, Lagos, Dar es Salaam, Nairobi and Dakar), while it appears to have increased in Abidjan. However, given the high degree of uncertainty surrounding the point estimates of child mortality (see Figure
1) and the resulting uncertainty around our measures of inequality (Table
2), results need to be interpreted with caution.
The study has a number of limitations. The most important one is the limited sample size of DHS within cities, which are not intended to be used for within-city analyses. In the earlier round of surveys, only four cities had data available from more than 1,000 children, and also in the later round, the samples of three cities were smaller than 1,000 children (see Table
1). Consequently, reliability of child mortality estimates is questionable and large confidence intervals around point estimates complicate the interpretation of results (see Table
2). Nevertheless, DHS remain the most reliable source of information currently available in most African countries, and they are used regularly by international agencies for estimating child mortality [
40]. Civil registration and vital statistics systems (CRVS) – if available at all – usually record only a small fraction of all births and deaths [
41], and child mortality data from censuses is often questionable. For example, in Abidjan, it is estimated that only about 70% of births and less than 40% of deaths (and even fewer among children) are registered by the CRVS [
42], and child mortality was grossly underreported in the last Ivorian census conducted in 1998 [
43]. DHS usually report child mortality figures for children born in the five years preceding the survey [
29]. Our study included birth histories and deaths of children born in the ten years preceding the survey in order to increase the sample size. This means that recent changes in inequality are only marginally reflected in our estimates, and they are not directly comparable with the figures reported in DHS. Furthermore, because birth histories cover the previous ten years, some of the included births and deaths may, in fact, have occurred before households moved to cities.
A second limitation of our study related to the use of DHS data is that the DHS wealth index combines information about household ownership of selected assets with the availability of basic community-level services, such as water or electricity [
31]. This can be problematic because it can lead to the misclassification of relatively rich households into the group of relatively poor households if they live in relatively poor neighborhoods [
44,
45]. The problem is thought to be particularly relevant when comparing urban areas, where more community-level services are available, with rural areas. However, in our study of major cities, where residential patterns tend to be more segregated, the misclassification of households is likely to be less of a problem, although it may still lead to an underestimation of the true extent of inequality.
A third limitation of this study concerns the comparison of inequality across cities and the assessment of trends over time. One problem is that DHS data from different cities were not available for the same years. For example, the most recent survey from Cairo was from 2008, while the most recent survey from Luanda was from 2011. Consequently, inequality is compared across cities at different points in time, and depending on the trend over time differences in inequality across cities might have reduced or increased beyond what can be seen in the data. In addition, the time period between the first and the second survey ranged from five years in Lagos to almost fourteen years in Abidjan and Dakar, and consequently, the change in inequality from the first survey to the later survey, which is shown in Figure
2 and Table
2, might appear relatively smaller in cities with a shorter time period between surveys than in those with longer periods. Another problem is that we selected only four measures of inequality to compare cities and to assess the trend in inequalities over time. A host of further measures are available [
27,
28,
33], including odds ratios, the slope index of inequality, the relative index of inequality, the generalized concentration index, and the Wagstaff index. In addition, it has been shown that measured inequality may differ depending on the chosen indicator [
12,
14,
36]. However, our selection of two measures of relative inequality (the rate ratio and the concentration index) and two measures of absolute inequality (the difference and the Erreygers index) is similar to other studies [
5] and should provide a reasonably nuanced view of the development of inequalities across cities.
Finally, a major weakness of our study is that it did not investigate the underlying reasons for the identified differences in inequalities across cities and for the differences in trends over time. Prior studies have decomposed calculated concentration indices in order to assess the contribution of different factors to inequality in infant [
46] or child [
47‐
49] mortality. However, the small size of the samples from the 10 African cities included in our study would draw into question the value of such analyses. Further research is needed to improve data availability from cities and to investigate reasons for differences in inequality and differences in trends over time.
Despite these limitations, our research has important implications for policy-makers and researchers. In 2010, a joint WHO and UN-HABITAT report [
21] urged countries to unmask health-related inequities in cities. Our study is the first to do this and to show the high level of inequalities in child mortality that exists within African cities. However, both average child mortality in the included cities and mortality of the poorest quintile are generally considerably below the corresponding national figures (see Additional file
1: Table S1 in the supplementary online material for national child mortality rates by wealth quintile). A previous study [
20] found that in nine low income countries, child mortality was significantly higher among the urban poor than among the rural poor. Our sample included no cities from these nine countries but our results seem to suggest that child mortality is lower in major African cities than in the rest of the country (see Additional file
1: Table S1), not only on average but also among the poor. This is in line with findings of another recent study [
9], which found that child mortality rates of children living in households in urban slums are higher than the rates of those living in formal settlements – but still lower than child mortality rates in rural areas.
In addition, our study reveals that the level of inequality in child mortality differs across cities and over time. Inequality in Abidjan was relatively low in 1998–1999, when compared with other African cities, but it was rather high in 2011–2012. Inequality in Cairo, Lagos, Dar es Salaam, Nairobi and Dakar seems to have reduced over (more or less) the same period of time. One explanation for these opposing trends might be that Abidjan suffered considerably during the time of political instability and civil war in Côte d’Ivoire, which lasted from 1999 to 2011 [
50]. Armed conflict has been shown to contribute significantly to increased child mortality not only during but also after the period of active warfare [
51‐
53]. Conflict is likely to disproportionately affect the poor [
54], and consequently inequities seem to be particularly severe in countries with a history of recent conflict [
55]. Another possible explanation might be that inequality of wealth decreased in some cities, and that this drove a decrease in inequality of child mortality. A decrease in inequality of mean wealth scores by income quintiles in Dar es Salaam, Lagos and Nairobi – cities where inequality seems to have decreased over time – might support this hypothesis, although there is no clear correlation, when looking at all included cities (see Additional file
1: Table S2). Child mortality is, of course, affected also by more proximate factors, including lack of shelter, sufficient nutrition or access to clean water and sanitation, as well as inadequate public health expenditures, low female education and short birth intervals [
56‐
58], and inequality in child mortality is largely related to unequal distribution of these factors [
46,
48,
49]. Also, Kinshasa, Luanda, Addis Ababa and Accra were found to have significant levels of inequality in the more recent round of surveys, and these cities may benefit from looking at cities that achieved reductions in inequality over time.
Finally, our study shows that larger sample sizes are needed in order to more reliably assess inequalities in child mortality within and across cities and in their development over time. Luanda was the only city included in this study with a DHS that contained full birth histories for more than 2,000 children (Table
1). WHO and UNHABITAT have asked countries to disaggregate available data in order to reveal health-related inequalities within cities [
21]. However, the small sample sizes of DHS mean that there is considerable uncertainty when this data is disaggregated in order to assess inequalities within cities. The need to improve data availability in developing countries has been recognized by the most important players in global health [
59]. Ultimately, civil registration and vital statistics systems will need to be strengthened. However, in the short-term coverage of health and demographic surveillance systems, which collect demographic and health data for a population living in a well-defined geographic area, could be improved in cities, whereas they are currently mostly focused on rural areas [
60,
61].
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
WQ and LS conceptualized the study, which was then designed in collaboration by all authors. WQ, JA, KD and MTB assured data collection in Côte d’Ivoire. All authors helped to interpret the data from their respective cities. WQ drafted the manuscript. All authors critically revised and approved the final manuscript.