Modelo preditivo de difusão espacial e risco de transmissão da raiva aos herbívoros

Abstract

Rabies is a zoonosis of great relevance in public health and responsible for considerable economic losses to livestock. To determine the risk of rabies transmission to herbivores in the state of Tocantins, a knowledge-driven predictive model was developed in which a questionnaire contained in the Geographic Information System environment of the QGIS® software (version 3.4.10), was filled up internally through Boolean logic and database queries. This, composed of information plans containing data related to epidemiological surveillance such as outbreaks of herbivores, positive bats and registered roosts, in addition to data on land use, for the year 2015. The risk assesment was based on scenario trees based on the concepts of receptivity and vulnerability, which ordered the questionnaire responses hierarchically. The trees were joined by an association matrix to obtain the risk scores, which were grouped into space-time series with a 90-day interval. The obtaining of the risk on a continuous spatial surface was performed by interpolating the scores using the ordinary kriging algorithms in the GS + ® software (version 7). The predictive model was validated by overlapping the occurrence of outbreaks with the risk estimates, and its global accuracy was assessed using a ROC Curve developed in the R® software (version 3.6). The proposed model was able to estimate the risk of circulation of variant 3 of the rabies virus in herbivores, with the ROC curve showing moderate accuracy in predicting the occurrence of outbreaks, of which out of a total of 25 outbreaks in 2015, 22 occurred in areas classified as most at risk. These results allow us to affirm that the proposed model can be used in directing disease control actions in areas of higher risk, allowing for a better allocation of resources by the official veterinary service, and can be applied in other regions according to the need.