Forecasting of Toscana Virus in Italy: Comparative Performance of SARIMA, Poisson, and Negative Binomial Models: Which Model Performs Best?

Background: The Toscana Virus is a little-known virus, present in Italy, transmitted by sandflies and associated with cases of meningitis and meningoencephalitis in humans. this study compared three statistical models (SARIMA, Poisson, and Negative Binomial) to forecast monthly Toscana virus (TOSV) cases in Italy for the period 2023–2024. Materials and Methods: data were extracted from the epidemiological bulletins of the Italian National Institute of Health for the period January 2016–December 2024. The 2016–2022 training set was used to estimate the models, while the 2023–2024 test set validated the predictions. Results: in the model comparison, SARIMA showed the best predictive ability, with the lowest MAE (3.46) and RMSE (5.05), demonstrating that seasonality and temporal de-pendence were well captured. The Poisson and Negative Binomial models, although use-ful, showed lower performance in terms of accuracy (higher RMSE). Conclusions: the results indicate that the SARIMA model is the best suited for forecasting monthly TOSV cases, but it is not perfect, highlighting the need for more complex ap-proaches that also integrate exogenous variables to improve forecast quality.