We aimed to study mortality trends in three Latin America countries between 1985 and 2012 and we found that disparities remain: age-adjusted CHD mortality rates declined in Argentina and Colombia (51.1% and 6.5% respectively) and increased in Mexico (60.8%). Trends in young adults (<35 years old) had a significant increment in the last decade in Mexico for men and women, but not in Argentina or Colombia. Garbage-code corrections produced dramatic increases in mortality rates in Argentina, particularly in women and amongst the young adults and the elderly.
Our results in changes in CHD mortality trends are consistent with those previously reported [7, 16], although some important differences in magnitude could be explained by the study period and age-groups considered. Pagan, et.al., studied mortality rates from 2000 to 2012 for adults 36 to 64 years old and had a similar behavior to what we observed in the three countries in the same period [7]. In Argentina, they observed a smaller reduction (19% men, 15% women) than ours (25% men, 24% women). In Colombia, the magnitude of the change was very similar, around zero for men in both studies and a 5% reduction in women (6% in our study). There are differences in both magnitude and direction in Mexico in this period. They reported an increase of 6% in men, which is lower than the 15% we found. In women, they reported a decrease of 6% and we found an increase of 5% [7]. These differences could be explained by the behavior of mortality in the younger and older groups, which were not included in their study.
The epidemic pattern experienced by Argentina is comparable to that reported in USA, Canada, France, Uruguay, Panama and the United Kingdom who experienced CHD mortality increases up to 1970, to be followed by a steady decrease since then [4]. The decreasing magnitude is similar to that in France and Estonia with a total percentage change between 46-50 in men and 53-55 in women [4]; in contrast, Uruguay and Panama also had decreases, but the magnitude was around 30%. Mortality rates in Colombia are also decreasing but at a smaller pace, percentage change on the overall period was 13.2% in men and 21.7% in women. In magnitude, these mortality rate reductions are comparable to those observed in Bulgaria, where total percentage change was 14.3 and 21.5 for men and women, respectively [4]. In the case of Latin American countries, Venezuela had similar changes to Colombia for mortality in women [7]. In contrast, CHD mortality rates in Mexico are still rising. These epidemic patterns are similar to those reported in the Russian Federation, Ukraine, and Latvia before 2002, although the magnitude of the increase in CHD mortality rates in these countries was higher [14]. In El Salvador, Sri Lanka and the Philippines both, epidemic form and magnitude changes, were similar to those in Mexico [7, 30].
Age and sex specific trends in Colombia, Mexico and Argentina are complex, but trends in young adults in Argentina and Mexico are worrisome. Very young Mexican adults (<35 years) showed an increase in CHD mortality in the last decades. Mexican men 35-44 also had an increase in the study period, but the increase rate was lower. Stagnation of trends was observed only in very young Argentinian men, but it was not observed in the group of adults aged 35-44. Finally, in the Colombian population, CHD mortality rates tended to decrease in both age groups. Few studies have analyzed changes in the mortality trends in young adults and only some of them have found this increase or stagnation [3, 7, 10, 11].A study conducted in 26 European countries concluded that in most countries the decline rate observed in younger age groups is similar to that observed in older populations [14]. However, in a small number of countries, like the United Kingdom or the United States, the population under 55 years have experienced smaller decreases in CHD mortality rates since 1990 than those observed in previous years and older age groups [3, 10]. In the analysis performed by Pagan, et.al., Latin America rate increases in men were observed in Costa Rica, Mexico, and Panama, while for women they were only observed in Mexico and Panama [7]. This analysis was restricted to the 35 to 44 years group, while we observed declines in the Mexican population and stagnation in Argentinian men aged 35 years and under.
Several reports have warned against the crude comparison of CHD mortality rates across countries without considering the quality of death certificates. All previous studies comparing rates across Latin America have reached the conclusion that Argentina is experiencing one of the lowest CHD mortality rates. However, after garbage-code correction, we observed that Argentina was not a low risk country at the beginning of the study period, with rates comparable to Mexico or Colombia. Garbage-code correction in these countries did not affect slopes, suggesting that the quality of reporting for CHD has not changed between 1997 and 2015. Presumably, garbage codes are more commonly used in the youngest and oldest members of the population because they are unspecific. It is difficult to diagnose coronary heart disease in very young adults because it is a chronic disease and it is uncommon to develop it at younger ages. On the other hand, elderly people might have more than one underlying disease and unspecific codes simplify the selection of the underlying cause of death. It is important to mention that while the method for garbage-code correction used in our paper has shown to be consistent throughout multiple studies [21, 22], no method has been established as the “gold standard” and errors in the comparison across countries could still remain.
The three countries included in this analysis are upper-middle income countries with similar life expectancy at birth and demographic distribution [17-19]. Furthermore, the three of them experienced the epidemiological transition in the last decades of the 20th century [31, 32]. However, there are some important differences: at the beginning of the study period CHD was the main cause of death in the three countries and, at the end of the period, diabetes was the main cause of dead in Mexico [18].
To our knowledge, no study on the contribution of risk factors to changes in CHD mortality in Latin America has been conducted. One study in Argentina analyzed the contribution of changes in risk factors and treatments in CHD mortality trends [33]. In this study, the authors found that evidence-based therapies accounted for approximately 49% of the deaths prevented or postponed; in contrast, changes in risk factors trends accounted only for 32.6% [33]. Although the main changes are due to medical treatments, especially hypertension treatments and secondary prevention after AMI, the reduction of three key risk factors could help to explain changes in mortality in Argentina: an important fall in hypertension prevalence (34.6%) [33], a 7% decrease in the smoking prevalence between 1980 and 2012 (from 26.6% to 19.8%) [34], and reductions in total cholesterol, which declined from 5.4 mmol/L in 1980 to 5.0 mmol/L in 2008 [35]. In contrast, modeling studies from other LMIC suggest that the main drivers of CHD mortality are preventable risk factors, such as diet and physical activity, and improvements in medical and surgical treatments [5, 12]. In our study, observed changes in Mexico and Colombia could be similarly explained by changes in risk factors. For instance, the reduction of two key two risk factors could explain changes in mortality rates in Colombia, where smoking decreased from 16.1% in 1980 to 11.2% in 2012, and total cholesterol in men decreased 1 mmol/L between 1980 and 2008[34, 35]. In contrast, in Mexico, diabetes, BMI and total cholesterol did not have favorable changes. The prevalence of diabetes increased 68% from 1980 to 2014 (6.5 to 10.9) [36]. BMI increased in men from 25.5 kg/m2 in 1980 to 27.4 kg/m2 in 2008, and in women from 23.4 kg/m2 to 28.7 kg/m [37]. The contribution of these potential explanations needs to be further studied, taking into consideration the long latency of the disease.
Some important limitations must be mentioned. Like any other study that uses mortality data across multiple versions of the International Statistical Classification of Diseases (ICD), there is potential for attribution bias owing to both the change between versions of ICD and the procedures used to code deaths. Vital statistics come from an independent registry and are subject to errors, the use of garbage-code correction allows to diminish the effect of the quality of the information, but other mechanisms for bias could persist.