ABC | Volume 110, Nº6, June 2018

Original Article Nascimento et al Cardiovascular disease in Portuguese-speaking countries Arq Bras Cardiol. 2018; 110(6):500-511 Statistical analysis The statistical models of the GBD 2016 Study previously reported were used. 4-6 The source of data for the models are available on-line at Global Health Data Exchange (http://ghdx. healthdata.org/) . Metrics of mortality and prevalence The GBD 2016 Study used data available on causes of death in 195 countries. The information was collected according to international standards of death certification, using of information systems on vital statistics, mortality surveillance systems, survey, hospital registries and police registries. 4 The data sources have regional particularities, such as in Brazil, where data were mainly obtained from the Mortality Information System (SIM) of the Brazilian Health Ministry, by using an automated coding system. 10 In Brazil, all deaths require a medical DC provided by the attending physician. For the deaths occurring outside a healthcare facility, the causes are verified by the Death Verification Service, or by a civil agent when a physician is not available, and, in such cases, the causes of death are not registered. 11 Deaths due to external causes are identified by a medical examiner at the Forensic Pathologist Service. In addition, techniques to correct quality problems regarding the information about the causes of death were used. 9 Corrections were made for underreporting of deaths and for causes considered not useful for the public health analysis, known as Garbage Codes. That term is used to describe causes that cannot or should not be considered as underlying causes of death, or that are unspecified within larger groups of causes. Algorithms of redistribution of Garbage Codes were developed by the GBD Study to increase the validity of the estimates. For this redistribution into specific causes of death, evidence from several sources, such as medical literature, expert opinion and statistical techniques, was considered. 4,8,12 After treatment of data quality, the GBD 2016 Study used a variety of statistical models to determine the number of deaths per each cause, mainly by use of the Cause of Death Ensemble Model (CODEm) algorithm. To ensure that the number of deaths per cause does not exceed the total estimated number of deaths, a correction technique (CoCorrect) was used. Adjustment by this technique ensures that the sum of the estimated number of deaths per each cause does not exceed 100% of the estimated deaths in a certain year. 13 The disease prevalence was estimated at a more detailed level of specific sequelae of disease, using as entry data the published systematic reviews of the scientific literature, as well as unpublished data of administrative registries and databases of the health system. Regression equations were used for data adjustment to define the standard case. The data presented were analyzed for the period from 1990 to 2016, and all analyses were stratified by sex and presented as absolute and age‑standardized estimates, for the different PSC. Metrics of burden of disease The disability-adjusted life years (DALYs) combine information regarding premature death (years of life lost: YLLs) and disability caused by the condition (years lived with disability: YLDs) to provide a brief measure of the healthy years lost due to the condition. The YLLs were calculated by multiplying the deaths observed at each specific age in one year of interest by the reference age-specific life expectancy estimated by use of life table methods. Figure 1 – Global map showing the location of the Portuguese-speaking countries, 2017. Brazil Cape Verde Portugal Guinea-Bissau Equatorial Guinea Sao Tome & Principe Angola Mozambique Timor Leste 502