1Despite occasional setbacks, mortality rates have steadily declined since the twentieth century in most regions worldwide. Numerous efforts have been undertaken to develop analytical frameworks that synthesize explanations for this decline and its determinants. Academics have coined the term “health transition” to describe the process of change in the epidemiological profile of a population concurrent with mortality decline [Frenk et al. 1991, 1996]. Advancement in the health transition entails: 1) shifts in age patterns, characterized by the concentration of mortality at older ages following reductions in infant mortality, and subsequently in young adult mortality; and 2) a transformation of cause-of-death patterns from communicable to non-communicable diseases. The health transition describes how demographic, economic, socio-cultural, and environmental patterns may interact to either facilitate or impede changes in disease and cause-of-death patterns [Lerner 1973; Caselli et al. 2002]. However, these changes do not uniformly benefit all countries or subpopulations, as progress is unevenly distributed. Initial changes typically occur in the most advantaged segments of the population, which benefit more rapidly from overall improvements compared to the rest of the population. Eventually, lagging subpopulations catch up, and the mortality reduction process unfolds through successive stages of divergence or convergence among subpopulations [Vallin and Meslé, 2004].
2During the last century, Latin American countries experienced a rapid decrease in mortality rates and a shift in cause-of-death structure. Infectious and parasitic diseases declined significantly, while degenerative and man-made diseases became prevalent in most countries Arriaga and Davis 1969; Preston 1976; Palloni 1981; Palloni and Pinto-Aguirre 2011]. This transition in the epidemiological profile of the region generally appeared first in urban locations [Tanner and Harpham 2014] and tended to proceed more rapidly in countries with higher levels of urbanization [Jaspers and Orellana 1994; Da Cunha and Rodríguez 2009; ECLAC 2014]. However, with the increasing prevalence of deaths due to cardiovascular diseases and external causes towards the end of the century, the previous decline in mortality slowed, and signs of stagnation in the increase of life expectancy appeared in some Latin American countries [Frenk, Frejka et al. 1991; ECLAC 2007; Sabino, Regidor et al. 2007; Di Cessare 2001].
3Given the evolution of the health transition in Latin America, we might question whether the slowdown in declining mortality in the region reflects an overall lack of progress or is primarily the result of certain subpopulations falling behind. Specifically, is it possible to trace cycles of divergence/convergence in health transition progress when considering urbanization? Which subpopulations are driving changes in mortality during the first decade of the twenty-first century? What causes of death are contributing to divergence/convergence among subpopulations beyond the country level? In this sense, we hypothesize the following: due to the tight linkage between urbanization and development in Latin America, the most urbanized areas continue to lead the largest progress in terms of health transition at the national level. This urban advantage is evident through differences in age-specific mortality trends. Consequently, the largest cities maintain their historical position as pioneers in the health transition, while less urbanized areas lag behind in attempting to catch up with this advantage. Similarly, the progress observed at the national level relies heavily on the successful catching-up process of less urbanized areas.
4To address these research questions, we gathered data from five Latin American countries: Brazil, Chile, Colombia, Mexico, and Venezuela. These nations encompass a range of circumstances regarding life expectancy at birth and the proportions of urban populations, which can encapsulate the regional trends observed (See Annexes 1 and 2). Moreover, they have accessible cause-of-death and population data categorized by age and sex. The cause-of-death data used in this study is sourced from the official vital statistics of each country and population estimates published by their respective National Statistics Institutes. Ensuring comparability across time and countries necessitates a rigorous standard of quality; hence, the data was adjusted to account for: incompleteness, identified as underreporting levels; and instances of misreporting and missing information in reported age, sex, and cause of death [Garcia 2020b].
5To characterize the spatial distribution of cause-of-death patterns, we focus on the 2000–2010 period and categorize the countries into three discernible spatial groups: main and large cities (more than 500,000 inhabitants); medium-sized and small cities (20,000 to 499,000 inhabitants); and towns and rural areas (less than 20,000 inhabitants). The delineations of these spatial groups are based on demographic similarities concerning population structures, migration trends, and living conditions across cities in the region [ECLAC 2005; Garcia 2020a]. Two statistical analyses were employed to condense the age-specific mortality patterns identified within the spatial groups: Principal Component and Hierarchical Cluster Analysis. These methods facilitate the classification of spatial groups according to their progress through the health transition. Subsequently, we investigate the contribution of cause-of-death to the disparities among spatial groups by applying the algorithm of the contour replacement decomposition method.
6The health transition in Latin America has unfolded at varying rates and with distinct characteristics in each country. Argentina and Uruguay led the way in this transition by initiating changes during the last two decades of the nineteenth century. These pioneering countries are the most urbanized in the region and host the largest proportion of their entire population in their capital cities. In Chile, Costa Rica, Cuba, Colombia, Mexico, Panama, Paraguay, and Venezuela, the decline in mortality began during the second and third decades of the twentieth century, coinciding with their accelerated urbanization processes. Subsequently, Brazil, the Dominican Republic, Ecuador, and Peru began to catch up. Countries that made fewer gains before 1980 experienced increases in life expectancy after that period [Palloni and Pinto-Aguirre 2011]. Conversely, Bolivia, Guatemala, Haiti, Honduras, Nicaragua, and El Salvador underwent a considerably delayed health transition. Even in the 1990s, when communicable diseases accounted for only 6 percent of registered deaths in Uruguay, they still represented 60 percent of total deaths in Guatemala [Chackiel 1999]. These lagging countries were also among the least urbanized in the region.
7By the 1960s, circulatory diseases began to rival infectious diseases as the leading cause of death among individuals aged 45 and over. The decline in circulatory diseases during the 1970s and 1980s, particularly in pioneer countries, significantly contributed to the decrease in adult mortality [Taucher, Albala et al. 1996]. Since the eighties, chronic non-communicable diseases have dominated age-specific mortality from ages 15 onward. In countries such as Chile and Costa Rica, neoplasms have replaced diseases of the circulatory system as the most prevalent cause of death. However, the rise of neoplasms and reduction in cardiovascular diseases occurred unevenly among countries and at a slower pace than the shift from communicable to non-communicable diseases [PAHO 1998]. Even during the 1990s, age-adjusted cardiovascular disease mortality continued to rise in Mexico, Brazil, and Venezuela [Stevens, Dias et al., 2008]. Because circulatory diseases have not dropped to the anticipated low levels, it poses the greatest challenge for most Latin American countries, especially among older age groups, indicating stagnation in the region’s cardiovascular revolution.
8In the late twentieth century, other causes of death emerged as the main drivers of change, particularly diabetes and violence. Between 1998 and 2002, diabetes-related mortality doubled in Mexico, making it the leading cause of mortality in women and the second leading cause in men [Glassman, Gaziano et al. 2010]. However, Cuba, Costa Rica, and Uruguay stand as notable exceptions, as they are the only countries where the impact of diabetes has remained steady or even declined slightly [Schargrodsky, Hernandez-Hernandez et al. 2008].
9Violence-related causes of death, which were a minor, incipient problem during the 1950s, escalated into an epidemic in several Latin American countries, disproportionately affecting young males [Ruiz and Rincon 1996; Imbusch, Misse et al. 2011]. Although countries such as El Salvador, Honduras, Colombia, Mexico, and Venezuela have seen gains in life expectancy due to reductions in other causes of death, these gains have been offset by increases in violent mortality [Ríos 2013; Garcia and Aburto 2019; Alvarez et al. 2019]. Violence has often been cited as the primary reason behind gender differentials in Latin America [Anzola-Perez and Bangdiwala 1996].
10Life expectancy in Latin America tends to be higher for women than for men, as in every world region, although the gap between male and female longevity varies among countries. For instance, in 2000, women in Uruguay outlived men by an average of eight years, whereas the female advantage in Bolivia was only three years. The sex differential in life expectancy at birth is generally more pronounced in countries with higher overall life expectancy levels [Brea 2002]. This widening gap is attributed to women’s better performance in terms of non-communicable diseases and external causes of death.
11While major cities, especially capital cities, have historically led in mortality decline and health transition progress, external causes of death threaten their superior performance. Crime is particularly prevalent in Latin America’s largest cities, where it disproportionately affects men in low-income neighborhoods [Montgomery 2008] . Furthermore, the mortality advantage at early ages provided by urban amenities may be offset by violence-related deaths among young adults (aged 15 to 34). Thus, the decline in cardiovascular diseases and rise in violence-related mortality as leading causes of death represent major 21st-century challenges to the health transition in Latin America. Understanding the evolution of the health transition at subnational levels may offer insights into overcoming these challenges and inform policy designs aimed at improving survival rates, thereby contributing to the broader discourse on the health transition worldwide.
12This research employs two levels of analysis. First, we categorized all spatial groups according to their current performance in traversing the stages of the health transition. This classification involved assessing the initial mortality levels in 2000 and the changes in age-specific mortality during the period 2000 to 2010. A continuous shift to the right in the survival curve indicates progression in the health transition. Second, we quantified (in years) the contributions of causes of death to the difference in life expectancy at birth among the spatial groups. This approach allowed us to identify which causes of death in these spatial groups were driving progress, stagnation, or deterioration towards achieving the health transition.
13Classification analysis allows identifying similarities within a set of sub-populations. In this research, two statistical analyses were employed to summarize age-specific mortality patterns in the spatial groups: Principal Component Analysis and Hierarchical Cluster Analysis (HCA). These analyses aim to reduce potentially correlated indicators into broader dimensions that are linearly uncorrelated.
14Principal component analysis (PCA) is a well-known method of multivariate analysis that uses the spectral decomposition of a correlation coefficient, also known as a covariance matrix [Yanai and Ishii 2010]. PCA is particularly valuable when variables within a dataset are highly correlated. In this study, our objective was to classify age-specific mortality across 30 spatial groups: 3 spatial groups multiplied by 2 sexes in 5 countries, (plus 5 country populations for illustration), based on similarities and differences in mortality levels and changes over a ten-year period. The indicators considered were: infant mortality rate (IMR), young adult mortality (30q15), older adult mortality (20q45), and life expectancy at age 70. We chose these indicators because they: 1) encompass mortality trends across all ages, thus providing a comprehensive view of mortality patterns throughout the life cycle; and 2) are standardized, enabling comparability across different populations.
15The active variables subjected to PCA were previously scaled through standardization to their mean at the baseline (2000), ensuring that all changes (reductions or gains) during the period 2000–2010 were equivalent to the same mean. This standardization was necessary because the variables considered are measured on different scales (rates, years of life expectancy, and conditional probability of dying). Determining the appropriate number of principal components is subjective and lacks a universally accepted method. In this study, we used a threshold of 70 percent of cumulative explained variance to indicate an acceptable number of principal components. The analyses were conducted using the statistical program SPAD.
16Hierarchical Cluster Analysis (HCA) is a statistical method used to identify relatively homogeneous groups of cases based on dissimilarities or distances between objects [Serra and Tagliaferri 2019]. HCA considers distances between cases in order to position groups within the non-correlated dimensional space previously established by PCA. The result is a hierarchical classification tree that illustrates the linkage points resulting from an algorithm. Various methods have been developed to determine the number of clusters, but assessing the sensitivity of clustering algorithms to measurement errors suggests that no method is universally superior to others [Bergman and Magnusson 2001]. This study performed agglomerative hierarchical cluster analysis using Ward's minimum variance method, which assumes that clusters are represented by their centroid, and the distance between clusters is measured in terms of increases in the sum of square errors [Roy, Kar et al. 2015].
17After classifying spatial groups according to their age-specific mortality, we proceeded to analyze the underlying causes of death and decomposed their contribution to the observed differences in life expectancy at birth during the period.
18We categorized deaths into six groups based on the 10th revision of the International Classification of Diseases (ICD-10), which we itemize in Table 1.
Table 1 Cause-of-death group
|
Cause-of-death group
|
ICD 10
|
|
Neoplasms
|
C00–D48
|
|
Circulatory diseases (heart diseases, stroke, other)
|
I00–I99
|
|
Respiratory, Digestive and Infectious Diseases and Maternal Conditions (RDI&M)
|
J00–J98, K00–K92, A00–B99, O00–O99
|
|
Diabetes
|
E10–E14
|
|
External causes (traffic accidents, other accidents, suicide, homicide, and violent causes with undetermined intention)
|
V01–X84
|
|
Others
|
|
19We employed the contour replacement decomposition method algorithm to estimate the age-specific contributions of causes of death to the differences in life expectancy at birth among spatial groups. This method incorporates three decompositions: the initial component between two populations at the initial time; and two decompositions of their changes (trend component) between the initial and final times (final difference) of the life expectancies being compared (Jdanov and Shkolnikov 2014; Jdanov, Shkolnikov et al. 2017). This approach allowed us to analyze whether current differences in life expectancy at birth among spatial groups stemmed from a catching-up process of lagging subpopulations or from persistent (dis)advantages, specifically by quantifying the effects of past mortality conditions and temporal changes within each age component.
20Alongside varying levels of urbanization, the countries analyzed in this study exhibit differing mortality levels. Among the five countries, Chile consistently maintained the lowest mortality rates throughout the entire period for both sexes (figure 1 in annexes). Chile has been a pioneer in mortality decline in the region since the 1960s, with no signs of stagnation. In contrast, Mexico and Venezuela experienced stagnation in the survival of their populations. Brazil and Colombia, situated between the aforementioned examples. Demonstrated divergent trends between sexes, with stagnating male mortality rates and declining female mortality.
21Figure 1 ranks the spatial group according to their mortality levels, as measured by the standardized rates at the beginning (2000) and the end (2010) of the period. The Chilean spatial groups exhibited the most pronounced decline in mortality. Notably, the standardized mortality rate of Chilean females residing in the main-and-large-cities was the lowest among all spatial groups throughout the entire period. Additionally, Chile and females in Venezuela demonstrated the smallest differentials among their spatial groups compared to other countries in the sample. This pattern of lowest mortality in the main-and-large-cities group held true for all countries, except for males in Brazil in 2000 and in Venezuela from 2000 to 2010. Notably, even though the main-and-large-cities group had the lowest mortality rate, not all main and large cities were regional pioneers. Nevertheless, they generally exhibited better performance compared to other spatial groups within the same country.
22Although nearly all spatial groups had reduced their mortality levels by the year 2010, mortality dispersion among subpopulations remained relatively widespread. In 2000, the spatial group with the highest mortality level (Colombian males living in towns and rural areas: CoMaTo) had a standardized mortality rate 2.8 times higher than the lowest-mortality spatial group (Chilean females living in main and large cities: ChFeMn). By the year 2010, the highest-mortality spatial group (Mexican males living in towns and rural areas: MxMaTo) had a standardized mortality rate 2.9 times higher than the lowest-mortality spatial group (still ChFeMn). Changes in the hierarchy of spatial group performance were primarily due to slower reductions in mortality in certain lagging spatial groups. The only exceptions were observed in the spatial groups of males and females living in Mexican medium-sized and small cities, as well as those in towns and rural areas, where all four groups experienced higher overall levels of mortality in 2010 than in 2000.
Figure 1: Spatial group ranking according to level of mortality measured by the standardized mortality rates in 2000 and 2010
23In order to classify the spatial groups according to their age-specific mortality, the PCA results (Figure 2) indicate that mortality patterns could be summarized into two broad dimensions. The first dimension, referred to here as the rectangularization of the survival curve on axis 1, signifies the transition from a wide dispersion of deaths to a concentration of deaths around an older modal age. This rectangularization is a consequence of improvements in life expectancy at all ages. The second dimension, on axis 2, is named the shrinking of the mortality hump and it reflects the spatial groups showing either reduction or enlargement of the mortality hump. The mortality hump refers to the excess mortality in early adulthood, which usually coincides with an increase in the proportion of external causes of death, and it is typically more pronounced (or more evident) among males. Both dimensions, the rectangularization of the survival curve and the shrinking of the mortality hump, capture changes in mortality patterns.
Figure 2: Two-dimensional principal component subspace and grouping for mortality patterns in subpopulations
- 1 A third axis, with an eigenvalue of 0.9, accounts for 11.6 percent of the variance that was disrega (...)
24These first two axes account for 72.3 percent of the variance: axis 1 has an eigenvalue of 3.647, equivalent to 45.6 percent, while the eigenvalue of axis 2 is equal to 2.134 and equivalent to 26.7 percent.1 Although the two axes may indicate an increasing or enlarging dispersion of age of death, they also underscore the impact on mortality by age due to the region’s (dis)advantages in young and old adult mortality. From the PCA results, five clusters summarize the age-specific mortality in the spatial groups (see Table 2).
Table 2: Cluster composition
Source: Own estimates based on data from National Statistical Offices in each country.
25Table 2 organizes the clusters from the best (blue color) to the worst performance (orange color) during the period. Annex 5 illustrates the differences between the clusters. The clusters are identified based on their progress in achieving the health transition. Progress is measured as the transition from a pre-transition phase of the epidemiological transition, dominated by infant and maternal mortality, to a transitional phase dominated by adult mortality, to, finally, a post-transition phase characterized by older adult survival. Thus, the cluster named Successful transition holds the highest life expectancy, contrasting with the cluster holding the lowest life expectancy, named Failed transition. Since the clusters represent not only levels but also trends in changes during the period, the life expectancy of the subpopulations named Delayed transition was not much higher than those in the Failed transition group – although the former showed constant improvements instead of stagnation, as observed in the spatial groups in the latter cluster. Similarly, the impact of the mortality hump differentiates the Delayed (shrinking) transition and Truncated (increasing) transition clusters from the others, in which young adult mortality rates remained stable during the period.
26In detail, the spatial groups in the Failed transition cluster appear to have reached a ceiling in life expectancy due to their failure to decrease the incidence of young adult mortality while also showing no improvements at any other ages. Conversely, these spatial groups exhibited an increasing mortality risk for all adults. On the other hand, the Delayed transition cluster had a much higher risk of mortality at older adult ages compared to the Failed transition cluster, although it rapidly achieved medium performance by reducing major risks of mortality at young adult ages. The future success of the Delayed transition cluster in progressing into the health transition will depend on future improvements in survival among populations aged 45 and above.
27In the Truncated transition cluster, previous gains in life expectancy stemmed from reductions in high infant mortality rates and emerging gains in life expectancy above 70 years. So far, the overall average of these gains has obscured the absence of progress in young adult survival rates. Further declines in infant mortality alone would not substantially elevate average life expectancy without a decrease in young adult mortality rates. Consequently, the persistent decrease in life expectancy at birth may characterize this cluster in the years to come.
28The two highest-performing clusters demonstrated improvements in overall adult survival. Spatial groups categorized in the Ongoing transition cluster are currently experiencing two benefits. First, they have achieved the same advantage previously attained by the best-performing cluster in reducing the risk of mortality between ages 45 and 65. Second, they are seeing a rise in life expectancy at old ages. Within the green cluster, a transition to an older age mortality regime was underway. Finally, the Successful transition cluster exhibited the lowest risk of infant and adult mortality, along with the highest and most consistently increasing levels of life expectancy at age 70. Notably, this top-performing cluster primarily consists of female subpopulations, while male subpopulations are more prevalent in the worst-performing clusters.
29In addition to the sex differential in cluster composition, the most notable characteristic is heterogeneity among countries. Although Chile and Venezuela, the most urbanized nations, remain the most homogenous in the sample, substantial differences exist. In Chile, all female subpopulations have undergone a successful transition, with all male subpopulations steadily catching up with them throughout the recent period. Conversely, while all Venezuelan female subpopulations are situated in the best-performing group, males exhibit a truncated transition. These are the sole countries where, despite variations in mortality levels based on urbanization, all subpopulations indicate a similar long-term trend. This suggests waves of improvements spreading across countries, initially impacting the most advantaged subpopulations (the main and large cities) and rapidly reaching the most disadvantaged ones in the rural areas. This phenomenon could be attributed to fewer geographical inequalities in mortality determinants, to the primacy of their major cities (as observed in Chile), or to the substantial influence of large cities on national wealth and population distribution (as seen in Venezuela), which may mitigate resource disparities in less urbanized regions.
30In the other countries, the urban advantage was evident in terms of both mortality levels and the dynamics of changes in the mortality structure. A striking example is Brazil, where transitions between mortality regimes varied across subpopulations. To begin, females residing in the main and large cities exhibited an ongoing transition while those in medium-sized and small cities experienced a truncated transition and those in towns and rural areas showed a delayed transition. Similarly, males displayed the same heterogeneous mortality patterns: those in the main and large cities had a truncated transition, while all other male subpopulations experienced a delayed or failed transition. This urbanization gradient in health transitions was also evident among females in Colombia, where only women living in the main and large cities group underwent a successful health transition during the period of analysis.
31In every aspect, Mexico comprised the most disadvantaged subpopulations. Only those in the main and large cities, along with females in the medium-sized and small cities, have experienced slight improvements in adult mortality during the analysis period, although these gains are far from substantial. Unlike in other countries, urbanization has not significantly boosted survival rates among Mexicans.
32In examining patterns of reduced mortality among clusters, we aimed to identify which causes of death play the most important role in the inequalities among spatial groups. Figure 3 displays the standardized mortality rates by cause of death in the years 2000 and 2010, categorized by spatial groups and clusters. Subpopulations were arranged by cluster, with each cluster’s mortality levels ranging from the lowest standardized mortality rate (at the bottom) to the highest (at the top).
33Looking at the cause-of-death distributions for both years, the spatial groups within the best-performing clusters – such as Successful and Ongoing – exhibited higher proportional rates of circulatory diseases and neoplasms. Conversely, the worst performing clusters – such as Failed and Delayed – maintained elevated levels of digestive, respiratory, and infectious diseases, as well as perinatal conditions. This polarization aligns with the classical epidemiological transition, wherein the worst-performing clusters continue to experience a significant burden of communicable diseases, while the best-performing clusters predominantly face non-communicable diseases. Additionally, homicides (among males) and diabetes (among females) emerged as prominent causes of death for both years in subpopulations classified under Truncated and Delayed transitions. However, the subpopulations within these clusters exhibited distinct trends in cause-of-death.
34In the Delayed health transition cluster, spatial groups experienced notable reductions in homicides by the year 2010, indicating that they were catching up with more advanced spatial groups in the health transition, evidently due to reductions in violent deaths. In contrast, males in Truncated subpopulations scarcely decreased their overall mortality levels in 2010 due to a rise in homicides. Among female subpopulations in a Truncated health transition, the cause of death that has hindered their progress is diabetes rather than homicides.
Figure 3: Standardized mortality rates (SMR) of subpopulations, by cause of death and health transition grouping
35In order to understand the causes of death driving differentials among age-group mortality and differences in life expectancy at birth among the spatial groups, we decomposed the age contributions (in years). To do so, we established a benchmark for comparing all subpopulations by selecting the best-performing spatial group during the period: Chilean females living in the main and large cities (ChFeMn). In the year 2000, approximately one-quarter of registered deaths among ChFeMn were attributed to circulatory diseases, while another quarter resulted from neoplasms. The remaining half of this subpopulation’s registered deaths were divided between respiratory diseases and external causes. Even in 2000, ChFeMn exhibited a very low infant mortality rate, and the period 2000 to 2010 saw their mortality rates decrease mostly as a result of a reduction in deaths from circulatory diseases among older adults, consequently elevating the contribution of neoplasms to overall mortality by the end of the period. Comparing ChFeMn with other subpopulations enabled us to monitor the other subpopulations’ progress in a mortality regime dominated by degenerative diseases and where adult mortality was the primary driver of changes.
36Figure 4 shows the differences in life expectancy at birth between the best performing and all other spatial groups in the final year (2010), based on the age contributions of: a) circulatory diseases; b) infectious, digestive, and respiratory diseases, and maternal conditions (IDR&M); and c) external causes of death. Additionally, utilizing the contour replacement algorithm allows us to decompose this final difference in years (located in the right-hand panels named “Final”) into two additive components: the contribution of the existing initial difference in 2000 (in the left-hand panels named “Initial”) and the contribution of changes over the ten-year period (in the middle panels under “Change”).
37The age-distribution of contributions to differences among spatial groups varies significantly, depending on the cause of death. Circulatory diseases primarily contributed to differences at age 45 and above, while external causes focused on ages 15 to 44. The range of ages contributing to IDR&M differentials was more diverse, encompassing various age groups’ contributions to the initial difference, with a notable decrease in infant mortality contributing to changes over the period. Regarding circulatory diseases, spatial groups outside the Failed transition cluster have benefited from an incipient reduction in mortality at ages 45 to 74 during the period. This reduction contributed to narrowing the final difference from the best performing population, although improvements in ages 75 and above remain scarce.
38The spatial groups deemed to be in the Successful transition group exhibited, on the one hand, the lowest levels of circulatory diseases among ages 45 to 74 at the beginning of the period and, on the other hand, a significant reduction in the contribution of these diseases at age 75 and above during the ten-year period of analysis. This group was unique in that neoplasms and other causes of death played an important role in older adult mortality. “Other causes” of death represented negative contributions to changes in the 75+ age group, likely due to the increasing prevalence of degenerative diseases such as Alzheimer’s, Parkinson’s, and dementia, among others. Although heterogeneous in terms of the countries under analysis, this group exclusively comprised female subpopulations. By the end of the period, this group’s advantage was evident across all age groups, even though some positive contributions stemmed from changes in mortality patterns among populations under 45 years old. The Successful cluster encompassed only those that had passed through the cardiovascular revolution.
39Compared with those in the Successful health transition, the spatial groups clustered as Ongoing in the health transition saw even greater improvements in adult mortality, as lagging female and male subpopulations in this cluster caught up with the initial advantages of their female counterparts in the Successful cluster. Similarly, their progress was driving convergence with the best-performing subpopulation. Additionally, they benefited from reductions in IDR&M diseases and conditions, as well as external causes of death among the adult population. However, to close the final gap with the Successful cluster, the Ongoing cluster still required further reduction in circulatory diseases among populations aged 45 to 74 years. The subpopulations in this group were likely passing through the cardiovascular revolution during the analysis period and were still in the process of decreasing the proportion of deaths due to circulatory diseases at young adult ages.
Figure 4: Age contributions (in years) to differences in life expectancy at birth among spatial groups and the best-performing subpopulation (ChFeMn), 2000–2010
Age contributions to the final differences in life expectancy at birth of all subpopulations, as compared with the best-performing subpopulation: Chilean females living in the main and largest cities (ChFeMn). Final refers to the end of the analysis period (2010)> This final difference is split into two additive components, which constitute the second and third set of results: age contribution of the initial difference in life expectancy at birth; and age contribution to the changes in each spatial groups over the ten-year period. Spatial groups are identified in the figure by acronyms combining countries name (Ch-Chile, Br=Brazil, Co= Colombia, Mx= Mexico, Ve=Venezuela), Sex (Fe=Female, Ma=Male), and Spatial group (Mn=Main and large cities, Me=Medium-sized and small cities, To=Towns and rural areas).
Source: Own estimates based on data from National Statistical Offices in each country
40The Truncated cluster encompasses spatial groups with diverse epidemiological profiles at the beginning of the period, all of them exhibiting varying causes of death. While some progress is observed within this cluster, primarily among ages 45 to 74, these advancements are insufficient when considering its high level of mortality compared to the best-performing spatial group. Although minor improvements resulted from decreasing IDR&M diseases and conditions in the infant populations, hardly any changes are observed at ages 75 and above during the period. What is even more concerning than the stagnation indicated by the spatial groups in this cluster is the negative contribution of the external causes of death to the differences.
41The Truncated cluster encompasses spatial groups that, on the one hand, maintained persistent levels of IDR&M diseases and conditions at young ages while not adequately benefiting from the cardiovascular revolution; and, on the other hand, those that already suffered from the consequences of man-made diseases such as homicides and diabetes. In contrast, the spatial groups in the Delayed health transition category exhibited the most significant improvements at all ages, particularly among adult ages between 15 and 74 years. These subpopulations were catching up with better-performing spatial groups during the period. In 2000, the gap in life expectancy at birth between the best (ChFeMn) and the worst (CoMaTo) performing spatial group was more than 17 years. By 2010, it had been reduced to 9 years, largely due to a rapid decrease in external causes of death. This pattern was observed across all spatial groups categorized under a Delayed health transition. While these subpopulations have not failed to benefit from the health transition in their respective countries, their progress has been delayed due to violent civil conflicts. However, they are still far from completing a health transition, as their improvements in mortality at age 75 years and above were negligible.
42Lastly, the subpopulations assigned to the Failed transition cluster were those who did not achieve reductions in mortality due to combined and simultaneous negative and positive contributions from different causes of death. Rising rates of circulatory diseases, diabetes, and external causes hindered improvements in adult mortality within this cluster. At the same time, the spatial groups deemed to be in a Failed health transition also sustained persistent infant mortality at the outset, and the reduction of infant mortality was the sole driver behind improvements in this cluster. Contributions to life expectancy from this age group stemmed primarily from diminishing IDR&M diseases and conditions.
43In this study, we relied on age-specific mortality trends (initial levels and changes during the period 2000–2010) and utilized PCA and HCA to construct clusters that reflect these current trajectories, moving beyond the traditional national analyses that are common in the literature. We achieved this by considering five Latin American countries representative of the varying levels of life expectancy and urbanization observed in this region. By clustering the spatial groups within these five countries, we also found that the progression of health and epidemiological transitions aligns with the path of urbanization, wherein stagnating levels of IDR&M diseases and conditions in small towns and rural areas contrast with declining circulatory diseases and increasing neoplasms in the main cities. Additionally, we observed that the most urbanized countries advanced more rapidly than less urbanized ones.
44We also decomposed the age contributions of causes of death to the differences in life expectancy using the best-performing spatial group: Chilean women living in the main and largest cities. This benchmark, or gold standard, reflects the actual improvements achieved by a Latin American subpopulation in the past and over the analysis period. By doing so, we established a fair and realistic comparison to gauge feasible progress while considering the region’s social and economic characteristics.
45As we hypothesized, the historical urban advantage resulting from the region’s developmental macro-policies implemented during the accelerated urbanization process in Latin America [ECLAC 2009] has contributed to the greatest progress in the health transition observed in the most urbanized countries and areas within those countries. The spatial groups deemed to be within the Successful and Ongoing health transition clusters are predominantly subpopulations from the most urbanized countries. These countries have overcome circulatory diseases and experienced increases in neoplasms and diabetes. Conversely, the spatial groups in the Failed transition cluster comprise primarily subpopulations from less urbanized countries, where reductions in IDR&M diseases and conditions in adult ages are still needed, and circulatory diseases are increasing. Intermediate to these two groups we find subpopulations with mixed epidemiological profiles, where mortality improvements appear to be Truncated for the period, largely due to violence-related deaths. The Truncated subpopulations stand out as the only spatial groups where the urban advantage in mortality is neutralized due to man-made causes of death. In these highly unequal subpopulations, other dimensions such as wealth, education, and race, among others, could be playing a more important role than urbanization.
46When examining age-specific trends driving spatial group differentials during the analysis period, the general findings underscore that no spatial group consistently demonstrates constant improvements or deterioration across all age groups during the period. Adult mortality does not adhere to a straightforward declining trend across all main-and-large-cities groups. Some spatial groups exhibit a parallel increase in the risk of dying at young adult ages alongside a decrease in the risk of dying at old adult ages, as seen in the cases of men living in the main and large cities in Mexico and Venezuela.
47The study period captures the conclusion of a cycle of convergence in declining infant mortality across all the studied countries, while ongoing divergence is observed in older adult mortality. Two convergence processes are evident among the lagging and best-performing subpopulations: one characterized by a reduction in IDR&M diseases in the Truncated and Delayed spatial groups; and the other by diminishing circulatory diseases among those spatial groups in the Ongoing cluster. In these processes, some subpopulations are left behind while others catch up. Positive and negative changes occur for most subpopulations, primarily in ages 15 to 74 years. Infant mortality levels no longer appear to be the main driver of future improvements in the region, while mortality patterns in populations aged 1 to 14 years neither introduce significant differentials among subpopulations nor contribute any changes during the analysis period. Divergence at older adult ages is attributed to the greater advantage of the main and large cities compared to other spatial groups. The main-and-large-cities group undoubtedly stands out as the best-performing spatial group in terms of reducing the risk of dying at older adult ages (75 and above) in all countries.
48Within countries, the drivers of subpopulation differentials vary. For example, in Brazil, the urban gradient establishes significant differentials among subpopulations in the process of moving forward in the health transition. The cause-of-death differential among spatial groups in Brazil appears to be greater than what would be expected from solely examining differentials in age-specific mortality. This suggest that – while the main and large cities in Brazil are making rapid progress in reducing the lethality of circulatory diseases at older adult ages – towns and rural areas are still lagging in their capacity to reduce infant mortality due to IDR&M diseases and conditions. Similarly, this phenomenon corresponds to the polarized and prolonged epidemiological transition highlighted by Frenk et al. [1991] in their analysis of the Latin American health transition model. This model shows no clear achievements in the transition process at the country level, but instead endures continuous stagnation in the duality of epidemiological profiles. Subpopulations advance in parallel at different paces without convergence of their profiles. The polarization of epidemiological profiles is not evident when considering spatial groups separately. This indicates that the different levels of development introduced by the urbanization process have resulted in asynchronous mortality transitions.
49To a lesser extent, a similar situation is observed in Mexico and Colombia, where the main and large cities are considerably more privileged than subpopulations in the rest of the country, with towns and rural areas being relatively neglected and struggling to catch up with the advantages held by the main cities. The progress of Brazilian, Mexican, and Colombian subpopulations does not converge during the first decade of the twenty-first century; instead, they advance in their own health transitions at different paces, with towns and rural areas lagging behind.
50In contrast, sex differentials appear to be more significant than urbanization in establishing differentials among subpopulations in Chile and Venezuela, where females benefit significantly more than males, regardless of their location. In the Venezuelan case, this discrepancy is even more pronounced because males no longer benefit as much as they previously did from the advantages of urbanization due to violence-related deaths.
51In summary, a catching-up effect is seen not only among spatial groups but also on a national scale, depending on the country’s urbanization level. However, these countries and spatial groups are benefiting differently from the progress achieved. This difference is sometimes obscured when analyzing mortality at the country level because subpopulations undergoing a Failed or Truncated health transition coexist with subpopulations considered Successful. Therefore, the apparent progress seen in some countries at the national level could actually be relying on subpopulations with different epidemiological profiles evolving in parallel with immutable differentials.
52The health transition in Latin America has progressed remarkably quickly compared to other regions in the world. However, while some improvements in survival have been notable, some countries have experienced stagnation or even some deterioration, as seen in Venezuela and Mexico. Overall, communicable diseases played a mild role in defining mortality patterns in Latin America during the first decade of the twenty-first century. Future advancements in reducing communicable diseases may have a significant impact on only towns and rural areas. Regarding non-communicable diseases, the most relevant concern centers on regions outside the main and large cities, namely small towns and rural areas, where we observe stagnation in reducing the lethality of circulatory diseases in older adults aged 75 and above. Addressing this challenge involves ramping up efforts to decrease cardiovascular mortality above age 45 outside the cities, as well as stopping the urban epidemic of violence. Overcoming these obstacles remains a crucial endeavor toward ending the current divergence among spatial groups and accelerating gains in life expectancy.
Flux RSS
