%A Tohidinik, Hamid
%A Keshavarz, Hossein
%A Mohebali, Mehdi
%A Sanjar, Mandana
%A Hassanpour, Gholamreza
%T Prediction of malaria cases in the southeastern Iran using climatic variables: An 18-year SARIMA time series analysis
%9 Original Article
%D 2021
%J Asian Pacific Journal of Tropical Medicine
%R 10.4103/1995-7645.329008
%P 463-470
%V 14
%N 10
%U https://www.apjtm.org/article.asp?issn=1995-7645;year=2021;volume=14;issue=10;spage=463;epage=470;aulast=Tohidinik
%8 October 1, 2021
%X **Objective:** To predict future trends in the incidence of malaria cases in the southeast of Iran as the most important area of malaria using Seasonal Autoregressive Integrated Moving Average (SARIMA) model, and to check the effect of meteorological variables on the disease incidence.
**Methods:** SARIMA method was applied to fit a model on malaria incidence from April 2001 to March 2018 in Sistan and Baluchistan province in southeastern Iran. Climatic variables such as temperature, rainfall, rainy days, humidity, sunny hours and wind speed were also included in the multivariable model as covariates. Then, the best fitted model was adopted to predict the number of malaria cases for the next 12 months.
**Results:** The best-fitted univariate model for the prediction of malaria in the southeast of Iran was SARIMA (1,0,0)(1,1,1)_{12} [Akaike Information Criterion (AIC)=307.4, validation root mean square error (RMSE)=0.43]. The occurrence of malaria in a given month was mostly related to the number of cases occurring in the previous 1 (p=1) and 12 (P=1) months. The inverse number of rainy days with 8-month lag (β=0.329 2) and temperature with 3-month lag (β=-0.002 6) were the best predictors that could improve the predictive performance of the univariate model. Finally, SARIMA (1,0,0)(1,1,1)_{12} including mean temperature with a 3-month lag (validation RMSE=0.414) was selected as the final multivariable model.
**Conclusions:** The number of malaria cases in a given month can be predicted by the number of cases in the prior 1 and 12 months. The number of rainy days with an 8-month lag and temperature with a 3-month lag can improve the predictive power of the model.
%0 Journal Article
%I Wolters Kluwer Medknow Publications
%@ 1995-7645