Revista Cubana de Salud Pública
versión impresa ISSN 0864-3466
COUTIN MARIE, Gisele. Use of ARIMA models for communicable disease surveillance. Rev. cub. salud pública [online]. 2007, vol.33, n.2, pp. 0-0. ISSN 0864-3466. http://dx.doi.org/10.1590/S0864-34662007000200012.
The early detection of abnormal behavior of health events is a health surveillance pillar. The possibility of designing suitable interventions assures the health authorities the reduction of sometimes lethal consequences derived from diseases and other damages. ARIMA modeling presented by Box and Jenkins in 1976 has been increasingly used in public health worldwide due to its advantages to adequately model the behavior of health events and to make forecast of the expected behavior of communicable diseases under surveillance. ARIMA allows modeling recent and remote values of the variable and also includes terms for recent and remote noises, all of which guarantees that all the series elements can be covered and comprehensively analyzed. A methodological review and practical examples of the use of ARIMA modeling for forecast generation in communicable disease surveillance were presented. Despite the complexities in the process of obtaining the optimal model, the Analysis and Tendency Units of the country have successfully used them for a number of events. The results of this modeling for the prediction of weekly medical visits due to acute diarrheal diseases and acute respiratory infections in Cuba in the period 1998-2004 were provided. Expected and observed results were compared. It was concluded that these models had good stability and were suitable for forecasting of events.
Palabras llave : Health surveillance; time series; ARIMA models; Box and Jenkins models; forecasts.