Summary
Background
Latin America is one of the most unequal regions in the world, but evidence is lacking on the magnitude of health inequalities in urban areas of the region. Our objective was to examine inequalities in life expectancy in six large Latin American cities and its association with a measure of area-level socioeconomic status.
Methods
In this ecological analysis, we used data from the Salud Urbana en America Latina (SALURBAL) study on six large cities in Latin America (Buenos Aires, Argentina; Belo Horizonte, Brazil; Santiago, Chile; San José, Costa Rica; Mexico City, Mexico; and Panama City, Panama), comprising 266 subcity units, for the period 2011–15 (expect for Panama city, which was for 2012–16). We calculated average life expectancy at birth by sex and subcity unit with life tables using age-specific mortality rates estimated from a Bayesian model, and calculated the difference between the ninth and first decile of life expectancy at birth (P90–P10 gap) across subcity units in cities. We also analysed the association between life expectancy at birth and socioeconomic status at the subcity-unit level, using education as a proxy for socioeconomic status, and whether any geographical patterns existed in cities between subcity units.
Findings
We found large spatial differences in average life expectancy at birth in Latin American cities, with the largest P90–P10 gaps observed in Panama City (15·0 years for men and 14·7 years for women), Santiago (8·9 years for men and 17·7 years for women), and Mexico City (10·9 years for men and 9·4 years for women), and the narrowest in Buenos Aires (4·4 years for men and 5·8 years for women), Belo Horizonte (4·0 years for men and 6·5 years for women), and San José (3·9 years for men and 3·0 years for women). Higher area-level socioeconomic status was associated with higher life expectancy, especially in Santiago (change in life expectancy per P90–P10 change unit-level of educational attainment 8·0 years [95% CI 5·8–10·3] for men and 11·8 years [7·1–16·4] for women) and Panama City (7·3 years [2·6–12·1] for men and 9·0 years [2·4–15·5] for women). We saw an increase in life expectancy at birth from east to west in Panama City and from north to south in core Mexico City, and a core-periphery divide in Buenos Aires and Santiago. Whereas for San José the central part of the city had the lowest life expectancy and in Belo Horizonte the central part of the city had the highest life expectancy.
Interpretation
Large spatial differences in life expectancy in Latin American cities and their association with social factors highlight the importance of area-based approaches and policies that address social inequalities in improving health in cities of the region.
Funding
Wellcome Trust.
Introduction
Moreover, social inequalities linked to other factors such as race or Indigenous origin,
gender,
and education
are also prevalent. These social inequalities manifest themselves as large health inequalities, including differences in mortality and life expectancy at birth.
Spatial segregation by socioeconomic position or race and ethnicity can lead to large spatial variations in health in cities. However, spatial inequalities in health in the large cities of Latin America have rarely been described or quantified.
have been cited in many policy and media reports. During the most recent municipal election in Madrid, Spain, several candidates commented on inequalities in life expectancy at birth across neighbourhoods in the city, and the topic received extensive media coverage.
Placing health inequalities at the centre of political discourse can support advocacy and the multisectoral policies needed to address them.
Evidence before this study
Previous studies in the USA and Europe have shown inequalities in life expectancy by area in cities and that a lack of data on health inequalities, and subsequent lack of awareness of their existence, is a barrier to the design and implementation of policies to reduce them. Some of these studies have been used as powerful advocacy tools to raise awareness about the issue. Latin America is one of the most unequal regions in the world, yet evidence is lacking on the magnitude of these inequalities in urban areas of the region.
Added value of this study
To our knowledge, this study is the first to compare the life expectancy at birth in Latin American cities. A wide gap in life expectancy at birth exists in the six large Latin American cities we analysed: Buenos Aires, Argentina; Belo Horizonte, Brazil; Santiago, Chile; San José, Costa Rica; Mexico City, Mexico; and Panama City, Panama. We found cities with both wider and narrower gaps in life expectancy at birth than US metropolitan areas of a similar size. Spatial patterns in life expectancy varied by city, and we found an overall strong association with subcity unit-level socioeconomic status, proxied by educational attainment. Approaches to reduce health inequalities require data on their magnitude and distribution. This study provides some of the first estimates of the magnitude of inequalities in these six cities, inhabited by more than 50 million people overall.
Implications of all the available evidence
The presence of large special inequalities in health in large cities of Latin America highlights the fundamental role of social inequalities, residential segregation, and place-based factors in driving population health in the region. This evidence emphasises the potential crucial role of policies to reduce inequalities in urban areas, and might also be used as advocacy tools to bring social justice to people in cities in Latin America.
We aimed to examine inequalities and spatial patterns of life expectancy at birth in six large Latin American cities and the extent to which this in-city inequality is associated with the socioeconomic status of the populations residing in these areas.
Results
Table 1Population, area, and educational attainment in six cities in Latin America and their corresponding subcity units
For subcity units data are median (IQR).
Table 2Variability in life expectancy at birth and association with education in six large Latin American cities, by sex
P90–P10=life expectancy at birth between the ninth and first deciles of subcity units.
Figure 1Association of life expectancy at birth with socioeconomic status, as proxied by educational attainment, in six large Latin American cities, adjusted for the proportion of subcity unit that is built-up, by sex
Datapoint size is proportional to subcity unit population. Lines are linear regressions of life expectancy on education attainment, weighted by population and adjusted by proprotion of subcity unit that is built-up. The variables represented in the x axis and y axis are residuals of a regression, at the city level, of educational attainment (x axis) or life expectancy (y axis) on the proportion of the subcity unit that is built-up.
Figure 2Spatial distribution of life expectancy at birth in men (A) and women (B) in six Latin American cities
Maps of cities with subcity units indicated. Categories are quintiles of life expectancy at birth in each city. Red lines outline the 11 central corregimientos of Panama City, the central distrito of San José, the 16 delegaciones of Mexico City, the central comuna of Santiago, the central municipio of Belo Horizonte, and the 15 comunas of the Ciudad Autonoma de Buenos Aires (also shown in the inset).
Discussion
Our study has shown large variability in life expectancy at birth in six large Latin American cities—Buenos Aires, Belo Horizonte, Santiago, San José, Mexico City, and Panama City—large spatial inequalities in life expectancy at birth, and an association with area-level socioeconomic status. The spatial variability in life expectancy at birth differed substantially between cities, as did the extent to which subcity unit-level socioeconomic status was associated with life expectancy at birth. Inequalities were largest in Panama City, Santiago, and Mexico City, while the association with subcity unit-level socioeconomic status was strongest in Santiago and Panama City. We also found distinct spatial patterns of life expectancy at birth in every city. While the differences between the city with the highest (Panama City) and lowest (Mexico City) average life expectancy at birth were approximately 7 years for men and 11 years for women, this difference was overshadowed in both cities by a P90–P10 gap of 9·4–15·0 years in life expectancy at birth in the subcity units across both sexes.
,
,
,
however, to our knowledge, no other study has described variations in life expectancy at birth within multiple Latin American cities. A previous study in the core areas of Buenos Aires found a similar gradient in all-cause and cause-specific mortality as we found here.
Other studies in single cities in the USA and Europe have also found similar gaps. For instance, using data from the Global Burden of Disease study, the P90–P10 gap in census tracts in King County, WA, USA, was calculated as 8·3 years in men and 6·2 years in women,
11 years for both sexes combined in communities in Chicago, IL, USA,
and 11 years for both sexes combined in community statistical areas of Baltimore, MD, USA.
A study in Rotterdam, Amsterdam, found that the total inequality in life expectancy at birth between neighbourhoods was around 6 years for both men and women,
while a study of inequalities in life expectancy at birth in London, UK, found a 20-year range in areas of around of 7000 people.
These results in London have been used as advocacy tools,
while much narrower gaps in Madrid, Spain, were used as part of the political discussion leading up to the 2015 local elections.
However, our results regarding the size of inequality are difficult to compare with previous work because the geographical units we used were large and heterogeneous, which will likely result in narrower gaps. However, to our knowledge, no other study has compared gaps in life expectancy between multiple cities in any region, including Latin America. Our study serves as a benchmark for other studies looking at these gaps in other regions or contexts.
different educational systems might lead to heterogeneity in the indicator. However, in our sensitivity analysis, looking at other subcity unit-level proxies for socioeconomic status (water access and overcrowding) we found analogous results. Third, variations in the measurement of the outcome. Lack of complete coverage of deaths is an endemic issue in many Latin American countries.
,
However, we applied state-of-the-art methods to account for this phenomenon and selected cities with more than 90% coverage. Nevertheless, the possibility remains that our correction did not entirely solve this issue, but our sensitivity analyses using variations of these methods rendered similar inferences. Finally, area-level socioeconomic status might have a differential association with mortality by country. For instance, previous research has shown narrower education gradients in Costa Rica and Mexico than in the USA.
,
Additional work is needed to confirm these large differences in the degree of spatial patterning and in the associations of area-level life expectancy at birth with area-level socioeconomic status, and to investigate the reasons for these differences if they are confirmed.
Our study has several strengths. First, we have compiled and harmonised data across six different cities, ensuring the comparability of both exposures and outcomes. Second, we corrected for the lack of complete coverage using state-of-the-art demography methods at the subcity-unit level. Third, we selected cities that had a relatively high number of subcity units to enable observation of gaps in life expectancy at birth. Finally, to avoid issues with small areas we used a Bayesian Poisson model that derived improved estimates of rates for areas with small populations.
our results would represent a conservative estimate of the inequality in life expectancy at birth by education because life expectancy at birth might be overestimated in lower socioeconomic areas. Finally, our estimates of socioeconomic status rely on the latest available census before the years of the mortality data. For Santiago, Chile, the census data we used were from the census in 2002 because the 2012 census did not account for a substantial part of the population.
However, in a post-hoc analysis, we tested whether any changes were seen in the educational attainment of subcity units in Santiago by harmonising educational attainment indicators
for the 2002 and 2017 census, and comparing the proportion of people with secondary education or above in all subcity units of Santiago; we found a very high correlation between 2002 and 2017 (ρ=0·97).
Future research should expand this effort to more cities with a large number of subcity units, or collect and analyse data at smaller units of analysis (eg, census tracts) where available, allowing for a better and finer characterisation of the segregation patterns in life expectancy at birth across Latin American cities. SALURBAL
,
will be exploring variations between smaller areas (which are likely to be much larger than those reported here) when detailed geocoded mortality data become available.
Our study showed a wide gap in life expectancy in six large Latin American cities, different segregation patterns (north to south, east to west, or core periphery) in each city, and an association with subcity unit-level socioeconomic status. These results might also be used as advocacy tools for political incidence in bringing social justice to cities in Latin America.
UB, AVD-R, and DAR conceived the study. UB did the statistical analyses. UB and AVD-R drafted the first version of the manuscript. MA, WTC, NL-O, KM-F, JJM, and AV participated in or supported data collection. All authors participated in the interpretation of results and approved the final version of the manuscript.
We declare no competing interests.
