Abstract

BRAZ, Rui Moreira; ANDREOZZI, Valeska Lima  and  KALE, Pauline Lorena. Early detection of malaria epidemics in Brazil: a proposal for automation. Epidemiol. Serv. Saúde [online]. 2006, vol.15, n.2, pp.21-33. ISSN 1679-4974.  http://dx.doi.org/10.5123/S1679-49742006000200004.

In Brazil, 99% of reported malaria cases occur in the Amazon Region. To date, however, an automated system has not yet been defined to detect epidemics in this area. This study proposes several statistical methods that could be useful for early detection of malaria epidemics. Five types of graphs were investigated: average +1.96 standard deviations (Cullen method); inter-quartile range (Albuquerque method); 3rd quartile method; Cusum-tabular method; and smoothing of baselines (Stern & Lightfoot method). The true alarm rate (TAV) detected by these methods was as follows: 100% by all five methods in the Municipality of Manaus, Amazonas State; and in the Municipality of Machadinho D’Oeste, Rondônia State 100% were detected by the 3rd quartile method; 25% using the Cullen, and Stern & Lightfoot methods, and 0% for other methods. In the Municipality of Amarante do Maranhão, Maranhão State, there were no epidemic months, and no alarm was given by any of the five methods, corresponding to 100% success. The 3rd quartile method is the most appropriate for early detection of malaria epidemics in municipal districts of the Brazilian Amazon Region, and recommended for implantation in routine malaria surveillance.

Keywords : malaria; epidemics; epidemiology surveillance; statistical graphs.

        · abstract in Portuguese     · text in Portuguese     · Portuguese ( pdf )