Time series analysis of leptospirosis incidence for forecasting in the Baltic countries using the ARIMA model

Abstract

Leptospirosis, a zoonotic disease with significant public health implications, presents considerable forecasting challenges due to its seasonal patterns and environmental sensitivity, especially in under-researched regions like the Baltic countries. This study aimed to develop an ARIMA-based forecasting model for predicting leptospirosis incidence across Estonia, Latvia, and Lithuania, where current disease data are limited and variable. This study aims to investigate the epidemic process of leptospirosis, while its subject focuses on applying time series forecasting methodologies suitable for epidemiological contexts. Methods: The ARIMA model was applied to each country to identify temporal patterns and generate short-term morbidity forecasts using confirmed leptospirosis case data from the European Centre for Disease Prevention and Control from 2010 to 2022. Results. The model’s performance was assessed using the Mean Absolute Percentage Error (MAPE), revealing that Lithuania had the most accurate forecast, with a MAPE of 6.841. The accuracy of Estonia and Latvia was moderate, likely reflecting case variability and differing regional epidemiological patterns. These results demonstrate that ARIMA models can effectively capture general trends and provide short-term morbidity predictions, even within diverse epidemiological settings, suggesting ARIMA’s utility in low-resource and variable data environments. Conclusions. The scientific novelty of this study lies in its application of ARIMA modelling to leptospirosis forecasting within the Baltic region, where comprehensive time series studies on the disease are scarce. From a practical perspective, this model offers a valuable tool for public health authorities by supporting targeted interventions, more efficient resource allocation, and timely response planning for leptospirosis and similar zoonotic diseases. The ARIMA model’s adaptability and straightforward application across countries demonstrate its potential for informing public health decision-making in settings with limited data on disease patterns. Future research should expand on this model by developing multivariate forecasting approaches incorporating additional factors to refine the model’s predictive accuracy. This approach could further improve our understanding of leptospirosis dynamics and enhance intervention strategies.