Predicting cutaneous leishmaniasis using SARIMA and Markov switching models in Isfahan, Iran: A time-series study

• Published in: Asian Pacific Journal of Tropical Medicine
• Main Authors: Vahid Rahmanian, Saied Bokaie, Aliakbar Haghdoost, Mohsen Barouni
• Format: Article in Journal/Newspaper
• Language: English
• Published: Wolters Kluwer Medknow Publications 2021
• Subjects: leishmaniasis; climate factor; time series analysis; forecasting; iran; Arctic medicine. Tropical medicine; RC955-962; Arctic; Sarima
• Online Access: https://doi.org/10.4103/1995-7645.306739; https://doaj.org/article/352289c592b94af8b66345c3a0a8d40f

Objective: To determine the potential effect of environment variables on cutaneous leishmaniasis occurrence using time-series models and compare the predictive ability of seasonal autoregressive integrated moving average (SARIMA) models and Markov switching model (MSM). Methods: This descriptive study employed yearly and monthly data of 49 364 parasitologically-confirmed cases of cutaneous leishmaniasis in Isfahan province, located in the center of Iran from January 2000 to December 2019. The data were provided by the leishmaniasis national surveillance system, the meteorological organization of Isfahan province, and Iranian Space Agency for vegetation information. The SARIMA and MSM models were implemented to examine the environmental factors of cutaneous leishmaniasis epidemics. Results: The minimum relative humidity, maximum relative humidity, minimum wind speed, and maximum wind speed were significantly associated with cutaneous leishmaniasis epidemics in different lags (P<0.05). Comparing SARIMA and MSM, Akaikes information criterion (AIC), and mean absolute percentage error (MAPE) in MSM were much smaller than SARIMA models (MSM: AIC=0.95, MAPE=3.5%; SARIMA: AIC=158.93, MAPE:11.45%). Conclusions: SARIMA and MSM can be a useful tool for predicting cutaneous leishmaniasis in Isfahan province. Since cutaneous leishmaniasis falls into one of two states of epidemic and non-epidemic, the use of MSM (dynamic) is recommended, which can provide more information compared to models that use a single distribution for all observations (Box-Jenkins SARIMA model).