Version 20

All data obtained from analyses were presented in tables as the mean standard deviation. The mean standard deviation was presented for all data acquired from analyses. A statistics software (SPSS, IBM version 20, Armonk, NY, USA) was used for the statistical analysis. To assess the significant differences (p < 0.05), one-way ANOVA was employed, and Tukey’s post hoc test was utilized for comparisons of the means.

The malaria crude mortality rate was calculated per malaria year by age-specific, gender and residential area (Bairro). Malaria crude mortality rate (MCMR) was calculated dividing the number of deaths-per year of residents by the total population for the same geographic area and multiplied by 100,000: $$\text{MCMR}=\frac{\text{Number of deaths}\text{-}\text{per year}}{\text{Total population for the same geographic}}\times 100,000$$
Age-specific malaria mortality rate was calculated dividing the number of deaths-per age per year of residents by the total age population and multiplied by 100,000 [14 (link)]. The ages (categories) used were: 0 (infants), 1–4 (Children), 5–14 (adolescents), 15–44 (young adults), 45–59 adults and over sixty (elderly).
Chi square for a proportion of gender and age-specific category was performed and Phi, Cramer’s V test was used for statistical significance. Analysis of Variance (ANOVA) was used to test difference between years and months using the following model: $$Y_{ij}=\mathit{\mu }+t_i+e_{ij}$$
Intervention analysis with the specification $z_t=\frac{{\mathit{\delta }}_0}{1-wB}{P}_t$ , where $\left|w\right|<1$ , $B$ stands for the traditional time series backshift operator, $B{z}_t=z_{t-1}$ , and $P_t$ denotes a pulse function such that $P_t=0,t<t_0\text{or},t>t_0$ and $P_t=1,t=t_0$ , where $t_0$ is the moment of intervention [15 (link)] was used.
All tests were performed using R 3.3.2, SPSS, IBM version 20 and Biosat 5.0. Spatial maps for year variation were produced using ArcGIS version 10.1.