A Seasonal Autoregressive Integrated Moving Average (SARIMA) forecasting model to predict monthly malaria cases in KwaZulu-Natal, South Africa

• Published in: South African medical journal 2018-07, Vol.108 (7), p.573-578
• Main Authors: Gebreslasie, M., Magubane, L., Ebhuoma, O.
• Format: Article
• Language: English
• Subjects: Analysis; Autoregression (Statistics); Forecasts and trends; Malaria; Methods; Seasonal variations (Diseases); Statistics
• ISSN: 0256-9574; 2078-5135
• DOI: 10.7196/SAMJ.2018.v108i7.12885

Background. South Africa (SA) in general, and KwaZulu-Natal (KZN) Province in particular, have stepped up efforts to eliminate malaria. To strengthen malaria control in KZN, a relevant malaria forecasting model is important.Objectives. To develop a forecasting model to predict malaria cases in KZN using the Seasonal Autoregressive Integrated Moving Average (SARIMA) time series approach.Methods. The study was carried out retrospectively using a clinically confirmed monthly malaria case dataset that was split into two. The first dataset (January 2005 - December 2013) was used to construct a SARIMA model by adopting the Box-Jenkins approach, while the second dataset (January - December 2014) was used to validate the forecast generated from the best-fit model.Results. Three plausible models were identified, and the SARIMA (0,1,1)(0,1,1)12 model was selected as the best-fit model. This model was used to forecast malaria cases during 2014, and it was observed to fit closely with malaria cases reported in 2014.Conclusions. The SARIMA (0,1,1)(0,1,1)12 model could serve as a useful tool for modelling and forecasting monthly malaria cases in KZN. It could therefore play a key role in shaping malaria control and elimination efforts in the province.