Improving influenza prediction in Quanzhou, China: an ARIMAX model integrated with meteorological drivers

Influenza remains a significant public health challenge, characterized by substantial seasonal variation and considerable socioeconomic burden. Although meteorological factors are known to influence influenza transmission, their specific effects within subtropical monsoon climates, such as that of Quanzhou, remain inadequately characterized.

We analyzed weekly influenza-like illness (ILI%) data from sentinel hospitals in Quanzhou between 2016 and 2024. Descriptive statistics, distributed lag nonlinear models (DLNM), cross-correlation function (CCF) analysis, and ARIMAX modeling were employed to examine the lagged and nonlinear associations between meteorological variables and ILI%.

The overall ILI% during the surveillance period was 2.32%, with significant temporal trends: a pronounced decline from 2016 to 2020 (APC = -22.693, p = 0.001) was followed by a significant increase from 2020 to 2024 (APC = 21.555, p = 0.003). Children under 15 years of age were the most affected demographic. A consistent bimodal seasonal pattern was observed, with a primary peak in winter (weeks 50-14) and a secondary peak in summer (weeks 20-29). DLNM analysis indicated that low atmospheric pressure (<997 hPa) at lag 0-0.5 weeks was associated with increased ILI risk, while higher pressure (≥1,010 hPa) had a protective effect (relative risk [RR] = 0.85, 95% CI: 0.73-0.99). Precipitation of 4-16 mm elevated ILI risk (RR = 1.10, 95% CI: 1.02-1.18), whereas precipitation >21 mm was protective. Wind speed demonstrated an N-shaped association with ILI%, though this was not statistically significant. The optimal forecasting model incorporated precipitation at a 2-week lag as an exogenous variable (ARIMA(1,1,1)(1,1,1)₅₂ + WAP(lag2)) and yielded the highest predictive accuracy for 2024 ILI%, improving RMSE by 2.9% and MAE by 1.3% compared to the baseline model.

Incorporating meteorological factors significantly improves the accuracy of influenza forecasting models. These findings support the development of climate-informed early warning systems and targeted public health interventions in subtropical regions.