Non-parametric tests and multivariate analysis applied to reported dengue cases in Brazil.

Abstract

Dengue is among the largest public health problems in Brazil. Reported dengue cases via DATASUS were correlated with reanalysis data from NCEP (rainfall and air temperature) and Brazil's population data (2000 and 2010) from 1994 to 2014. The aim of this study was to evaluate relational patterns between climate variables together with population data from the last census and reported cases of dengue in Brazil from 1994 to 2014 by using statistical techniques. Several statistical methods [descriptive and exploratory statistics; simple and multiple linear regressions; Mann-Kendall (MK), Run, and Pettit nonparametric tests; and multivariate statistics via cluster analysis (CA)] were applied to time series. The highest percentages of Dengue cases were in Brazil's Southeast (47.14%), Northeast (29.86%), and Central West (13.01%). Upon CA of the Brazilian regions, three homogeneous dengue groups were formed: G1 (North and Central West), G2 (Southeast and Northeast), and G3 (South). Run testing indicated that the time series is homogenous and persistence free. MK testing showed a nonsignificant trend of increase of dengue cases in 23 states with positive trends and in four states with negative trends of Brazil. A significant increase in the magnitude of dengue at the regional level was recorded in the North, Southeast, South, and Central West regions. Statistical methods showed that dengue variability in Brazil is cyclical (2- to 3-year cycles), but not repetitive of El Niño-Southern Oscillation (ENSO) in the moderate, strong, and neutral categories. ENSO interferes with the action of weather systems, changing or intensifying rainfall and air temperatures in Brazil. The population increase in recent decades and the lack of effective public policies together with the action of ENSO contributed to the increase in dengue cases reported in Brazil.