Mixed infectious disease forecasting technique, based upon seasonal decomposition and Sarima

Abstract

Aim. To study the possibility of using mixed technique for predicting infectious morbidity based on time series decomposition methods and SARIMA (decSARIMA).Materials and methods. Using the data from 12 regions of Volga Federal District (Russia) we analyzed time series of the incidence of infectious pathologies: hemorrhagic fever with renal syndrome (HFRS), acute upper respiratory viral infection (ARVI) and syphilis. The decomposition of time series of the incidence rate was carried out using X13-ARIMA-SEATS method. The trend and the seasonal component were separated, each of which was then modeled separately by SARIMA method. The final model of the incidence rate was obtained by adding the trend and the seasonal models.Results. On average, decSARIMA models had higher or similar characteristics of model and prediction quality compared to SARIMA models without preliminary decomposition. The prognosis of the incidence rate obtained by decSARIMA method was characterized by narrower confidence intervals. Reasonability of using decSARIMA models depended on composition and dynamics of time series of the incidence rate. A significant improvement in model and prediction quality was demonstrated for HFRS. When modeling and predicting the incidence rate of ARVI and syphilis, the inclusion of decomposition of time series into the analysis was considered inexpedient.Conclusion. The usage of decSARIMA model allows to significantly improve the quality of the prognosis of the incidence for infections, which are characterized by pronounced seasonality and the presence of interannual differences in the incidence rate.

Keywords

• seasonal decomposition
• sarima
• x13-arima-seats
• prognosis
• hfrs
• arvi
• syphilis