Sensitivity of alarm in an epidemiological syndromic surveillance system and proposed bayesian network

• Autor(a) principal: Ximenes, Patricia de Souza Medeiros Pina
• Data de Publicação: 2020
• Outros Autores: Santoro, Kleber Régis
• Tipo de documento: Artigo
• Idioma: por
• Título da fonte: Research, Society and Development
• Texto Completo: https://rsdjournal.org/index.php/rsd/article/view/10569
• topic: Detección de brotes; Redes bayesianas; Vigilancia animal.; Outbreak detection; Bayesian networks; Animal surveillance.; Detecção de surtos; Redes Bayesianas; Vigilância animal.

The efficiency of a syndromic surveillance system was evaluated for mortality in poultry based on international recommendations. Various forms of epidemiological events were simulated with different scenarios. The system's alarm techniques were analyzed according to their sensitivities as well as the correlation between the respective results. Among the techniques used by the system, the Shewhart chart was the one that most contributed to the correct detection of outbreaks, presenting a probability greater than 95% in the detection of true positive alarms and only 4.6% of false positives. In order to correct the sensitivity of the system in detecting outbreaks, a Bayesian network was developed. This network was proposed as part of the evaluation of the results of the system, providing greater precision. The proposed Bayesian network was able to correct errors in the evaluated system, proving to be a viable addition to the syndromic surveillance system. The highest correlation coefficients identified were given by the relationship between the Shewhart control graph and Exponentially Weighted Moving Average (EWMA). The system tends to overestimate the occurrence of alarms through false positives; however the proposed Bayesian network corrected all failures to a level of 30%.