ABC | Volume 113, Nº3, September 2019

Original Article Figueiredo et al. Rheumatic fever: a disease without color Arq Bras Cardiol. 2019; 113(3):345-354 study was to analyze the historical series of mortality rates and disease costs, projecting future trends to offer new data that may justify the need to implement a public health program for RF. In addition, we estimate the annual costs of the diseases and their comorbidities in Brazil. Moreover, the RHDmortality rate was compared with breast (BC) and prostate (PC) cancer mortality rates, which already have implemented public health programs, such as the case of the Pink October 12 and the Blue November, 13 respectively. Methods A cross-sectional ecological study with a time series analysis was developed to analyze the historical series of mortality rates and disease, using data from the SIH/SUS 11 from 1998 to 2016. The year 2017 was not included in this study because the data were still subject to updates. To estimate the annual cost of the diseases and their comorbidities in Brazil, first we determined the costs associated to the diagnosis of ARF, primary and secondary prophylaxis of RHD, as well as public expenses associated to the consequences of RHD, such as interventional procedures and hospitalizations for heart failure, atrial fibrillation, ischemic stroke and infective endocarditis. For this purpose, the data were obtained as follows: the procedures required for the diagnosis of ARF and Jones Criteria, which have been reviewed at irregular intervals by the American Heart Association (AHA), adjusted for RHD. For hospitalization costs, we considered the mean length of hospital stay of 7 days for ischemic stroke, heart failure 4 days, 4 days for atrial fibrillation and 17 days for infective endocarditis. 14 The data related to cost of the procedures required for the diagnosis of ARF/RHD and hospitalizations due to consequences of RHD were taken from the database of the Table Management System of Procedures, Medical drugs, Orthotics, Prosthetics and Special Materials of SUS (SIGTAP) 14 and the Drug Market Regulation Chamber (CMED) of the National Agency of Sanitary Surveillance (ANVISA). 15 These data are available at the Hospital Information Systems - SIH/SUS -Brazilian Health System. Second, we developed a hypothetical scenario based on the current panorama of rheumatic fever in Brazil, crossing data from the Brazilian Institute of Geography and Statistics with data from the REMEDY study 16 with their respective morbidities in numbers, to estimate the number of cases. The REMEDY study involved 25 sites in 12 African countries, Yemen and India. Countries were grouped into three income categories: low-income countries (Ethiopia, Kenya, Malawi, Rwanda, Uganda and Zambia), low-middle income countries (Egypt, India, Mozambique, Nigeria, Sudan and Yemen) and middle-income countries (Namibia and South Africa). 16 The costs obtained were multiplied by the number of cases of group A Streptococcus (GAS) infection, ARF, RHD, and RHD morbidity. Moreover, the RHD mortality rate was compared with breast (BC) and prostate (PC) cancer mortality rates, which was performed taking in account the period of 18 years (1998 to 2016), using data from the Mortality Information System's (SIM) of SUS – DATASUS, 11 responsible for the maintenance of mortality data in Brazil. For this comparison, a simple linear regression was adjusted to each case (RHD, PC, and BC). The present study used only secondary data obtained from public access sources. The approval of this study was waived by the Research Ethics Committee, as established in Resolution 510 of the National Health Council (CNS) of April 7, 2016. Statistical analyses To evaluate the trend of the historical series, simple linear regression models were adjusted. When working with time series, it is common to find problems of heteroscedasticity and autocorrelation. In order to deal with these problems and to allow the performance of valid inferences for the adjusted models, as well as to guarantee the robustness of the models, the HAC (Heteroskedasticity and Autocorrelation Consistent) was used for the covariance matrix of the estimated coefficients. 17 To model the behavior of the series and make predictions, Holt's Exponential Smoothing Method was used. 18 R software (version 3.2.4) was used for the statistical analysis. The results of the tests with a value of p < 0.05 were considered statistically significant. The limitation of this study was the analysis of the SIH (SUS) database, 11 of which data are entered every two months or more, limiting confidence only to total annual data. Results Mortality rates from Acute Rheumatic Fever (ARF) and Rheumatic Heart Disease (RHD) showed an increasing pattern throughout the analysis period (Figure 1). The ARF mortality rate increased from 0.80 in 1998 to 2.52 in 2016, a growth of 215%, with an increase of 0.12 units, on average, with each passing year (Figure 1A). The RHD mortality rate was 5.77 in 1998, increasing to 8.22 in 2016 (a growth of 42.5%), showing an average rate increase of 0.15 units per year (Figure 1C). Using Holt’s Exponential Smoothing, it was possible to perform mortality estimates for ARF and RHD. The predicted values for ARF mortality rate for 2018 and 2019 were, respectively, 2.59 and 2.68, while the predicted values for RHD mortality rates were 8.43 for 2018 and 8.53 for 2019. Although these numbers may be underestimated de to the lack of a health surveillance strategy, which will be discussed later, 732 deaths were recorded in 2003 and after a linear regression (p-value < 0.005) of the entire studied period, it is observed that the number of deaths increases on average 16,94 units each year. Regarding the cost analyses, Table 1 shows a detailed description of the obtained costs for ARF diagnosis, the most common interventional procedures in RHD, and the costs of hospitalization due to the consequences of RHD, for a hypothetical patient in the context of the Brazilian public health system. With an average of 30,000 ARF cases per year in Brazil, in a hypothetical scenario based on the REMEDY study, 16 we would have the scenario shown in Figure 2. According to this hypothesis, there would be 21.000 cases of RHD per year, which would lead to approximately 7.014 new patients with heart failure, 4.578 cases of atrial fibrillation, 1.491 cases of stroke, 8.904 cardiac surgeries and 840 cases of infective endocarditis. 346