Evaluation of time series models for influenza trends accross climatic zones in Cameroon (2017-2022)

Background: Influenza seasonality is well established in temperate regions, where activity typically peaks in winter. In tropical countries, however, seasonal patterns remain less clear. This study compared two time series models to characterize influenza trends in four climatically distinct regions of Cameroon. Methods: The study covered January 2017 to December 2022. Patients with acute respiratory infections were enrolled through sentinel influenza surveillance sites in the Centre, Littoral, North, and West regions. Samples were tested for influenza virus using molecular assays. Two models were evaluated: (1) an ARIMA model incorporating 12 climatic variables (temperature, humidity, precipitation, solar radiation) and (2) a seasonal model excluding climate inputs. Model performance was assessed using root mean square error (RMSE) and stationary R². Results: Overall, the seasonal model outperformed the ARIMA model, with explanatory power ranging from 66% to 82%. The ARIMA model showed variable performance by region and was influenced by specific climatic factors. In the Centre and North regions, precipitation explained about 30% of influenza activity, while in the Littoral and West, temperature accounted for up to 30.8% and 43.5%, respectively. Humidity and solar radiation showed weaker associations. In the Centre, the seasonal model achieved a lower RMSE (16.067) and higher stationary R² (82%) compared with ARIMA (RMSE = 20.203; R² ~19%). Similar trends were observed across the other regions. Conclusions: Findings highlight heterogeneity in influenza patterns across Cameroon. Seasonal fluctuations in influenza activity appear to be shaped by factors beyond climatological variables alone.