Ward-like hierarchical clustering with dissimilarities and non-uniform weights in cases of tuberculosis in Paraíba, Brazil

In this article, we propose to present a solution based on socio-epidemiological variables of TB, considering a clustering with spatial/geographical constraints for the State of Paraíba, Brazil. The Ward-Like hierarchical clustering method uses two dissimilarity matrices, the first provides the dissimilarities in the feature space calculated from the socio-epidemiological variables (D0) and the second provides the dissimilarities in the constraint space calculated from the geographical distances (D1) together with an α mixing parameter and the non-uniform weight w assigned to the calculation of the dissimilarity matrix defined by the diversification coefficient (DC) of TB. Statistical analyses were undertaken in R. According to DC, most micro-regions are diversified, indicating that the epidemiological situation of TB does not depend on any specific variable. In D0, the clusters are dispersed and are not strictly contiguous. Geographically more compact clusters are obtained after the introduction of D1 and α=0.2, slightly favoring socioepidemiological homogeneity (26.11%) versus geographic homogeneity (17.58%), mainly influenced by cluster 2. Clusters 3 and 5 were separated based on the proportion of TB patients of working age. Cluster 4 had the lowest cure proportion of all clusters. The Ward-Like algorithm is shown to be viable in socio-epidemiological studies in understanding the behavior of TB from a spatial perspective.