Mortality by cause of death in Colombia 1998-2014: A local analysis using spatial econometrics

Colombia has undergone major changes in mortality patterns during the last decade, in particular due to reductions in external causes. This has had a significant impact on excess deaths in young adult males. While cause-specific analyses have been performed at the national and, to a lesser extent, departmental level, very little is known about trends at the municipal level, despite their great epidemiological interest. Our objective in this paper is therefore to identify geographic clusters of mortality in Colombia and their evolution over time that will allow decision makers to prioritize those regions with higher mortality. To do so, we will use Exploratory Spatial Data Analysis (ESDA) to analyse several large groups of causes: infectious diseases, tumors, cardiovascular diseases, perinatal mortality, external causes, ill-defined causes and remaining causes. The study analyzes trends in standardized mortality rates (per 100.000) for causes of death for both sexes at the municipal and department level during the period 1998-2014. These have been calculated from microdata that we obtained from the National Administrative Department of Statistics (DANE). To overcome the problem of under-registration of mortality in Colombia, we applied a method that corrects the total number of deaths by sex, age and cause of death in each municipal area. To test for spatial dependency, we used the global and local spatial autocorrelation indicators Global Moran I and Local Moran I. Results show that in a context of a gradual mortality decline, the corrections we made in the vital statistics considerably improved the quality of the municipal data led to the identification of clusters that had already been previously identified in epidemiological studies which used morbidity registers but did not show up in our preliminary analysis that used the uncorrected mortality statistics. All causes of death show a greater or lesser degree of spatial autocorrelation, although this decreased over time, with the exception of perinatal mortality, ill-defined causes, infectious diseases (women only) and external causes (women only). Moreover, external causes, especially among men, present the most significant levels of spatial autocorrelation and an extension over time to different geographic spaces from the central Andean area to Orinoco and the Amazon rainforest.