An Application of Markov Switching Models in Predicting Visceral Leishmaniasis in Ardabil Province, Iran

Abstract

Background: Visceral leishmaniasis (VL) is a neglected infection currently occurring in some regions of Europe, Asia, Africa, and America. This study was an attempt to determine the temporal patterns of VL from January 2000 to December 2019 in the Ardabil Province of north-western Iran using the Markov Switching Models (MSM).Methods: This descriptive study used monthly data of 602 VL cases during the study period. The data were provided by the Leishmaniasis National Surveillance System (LNSS), the Iran Meteorological Organization (IMO), and Space Agency (SA), and two states were considered for such modelling. Given the Akaike and Bayesian information criterion, the two-state MSM with a five-month lag is an appropriate model.Results: The MSM showed that the probability of staying in the non-epidemic state is 67%, (P11), while that of staying in an epidemic state is 93% (P22). The mean absolute percentage error (MAPE) was 31.63%, and the portmanteau test (Q=19.03, P=0.66) for the residuals of the selected model revealed that the data were completely modelled. The total VL cases in the next 24 months forecasted 14 cases.Conclusion: The MSM has a relatively acceptable predictive power and is useful in planning future interventions with more information about different stages of the epidemic it provides to policymakers for early warning of epidemics.

Keywords

• black fever
• meteorology
• time series
• forecasting
• iran