Data management and analysis

The analyzed variables were as follows: year of diagnosis, sex, age group, operational classification, assessment on the degree of physical disability at the time of diagnosis and treatment outcome. Two epidemiological indicators were used to determine disease trends and assess the timely detection of cases while two operational indicators were used to evaluate the quality of care in health services, and to ensure the surveillance of household contact to detect new cases. The Brazilian Ministry of Health adopts these indicators ( Table 1 ) 5 ^, 16 .

Two epidemiological indicators were selected for the spatial analysis (mean annual detection rate and leprosy rate with grade 2 physical disability), in addition to one operational indicator (new leprosy rate with some degree of physical disability) per municipality. The mean annual detection rate was calculated using the mean of new cases in the study period as the numerator and the population at the mid-point of this as the denominator. The population data were obtained from the Brazilian Institute of Geography and Statistics (IBGE) 17 and was estimated for the general population used.

The association between the variable operational classification (paucibacillary and multibacillary) and sex, age group, assessment of the degree of physical disability at the time of diagnosis and treatment outcome was determined using the chi-square (χ²) test, with a significance level of 0.05. In addition, absolute and relative frequencies were calculated for each variable. All analyses were performed using the free R software, version 3.4.0 (Ross Ihaka and Robert Gentleman, Auckland, New Zeland). Absolute and relative frequency rates were calculated.

In the exploratory analysis of spatial data, the analysis unit was the municipality, the Bayesian local empirical method was used to minimize the instability caused by random fluctuation of rates, since the estimator calculates a weighted rate considering area variances 18 . The neighborhood matrix defined by adjacency of TerraView software, version 4.2 (Instituto Nacional de Pesquisas Espaciais - INPE, Sao Paulo, Brazil), was used for this. Subsequently, the spatial autocorrelation was applied using the global Moran's index, the analysis remaining on the indicators that gave autocorrelations. Moran's index varies between −1 and +1, where zero indicates no spatial autocorrelation and values near +1 and −1 indicate positive or negative spatial autocorrelation, respectively 19 ^, 20 .

Moran's scatterplot was used to visualize the spatial dependence. The plot shows normalized values that allow analysis of the spatial behavior of data 20 . Thus, the value of each municipality can be compared with those of neighboring municipalities, in the following manner: quadrants 1 (Q1, positive values and means) and 2 (Q2, negative values and means) indicate locations that have neighbors with similar values; quadrants 3 (Q3, positive values and negative means) and 4 (negative values and positive means) indicate locations that have neighbors with different values.

Next, Local Indicators of Spatial Association (LISA) were used to detect locally correlated regions, at significance levels of 95%, 99% and 99.9%, and a LISA map19 was generated. However, this map is not shown here because it was only used to create the Moran map, which combines areas with positive spatial autocorrelation (identified in the box map) with spatial significance > 95% (indicated in the LISA map), i.e., it presents areas with statistical significance (p < 0.05). In this study, Moran's map was used. Critical areas were those formed by municipalities included in Q1 as epidemiological indicators and in Q2 as operational indicators.

The TerraView software, version 4.2 (INPE, Sao Paulo, Brazil) was used for the processing, analysis and calculation of spatial autocorrelation indicators, and the QGIS software, version 2.14 (Open Source Geospatial Foundation, Chicago, USA) was used to present cartographic data and create thematic maps.