Incorporating meteorological factors into a SARIMA model for predicting pediatric influenza epidemics

Influenza, a contagious respiratory illness, imposes a substantial global disease burden, underscoring the critical need for timely and accurate surveillance systems. Seasonal Autoregressive Integrated Moving Average (SARIMA) models excel in capturing seasonal patterns and trends of infectious diseases. The distinct seasonality of influenza suggests that meteorological factors significantly influence its transmission, though their specific mechanisms remain incompletely characterized. We integrated meteorological variables into a SARIMA model to enhance the prediction of pediatric influenza epidemics, thereby providing a scientific foundation for precise prevention and control strategies.

We analyzed 67,770 influenza-like illness (ILI) cases from Tianjin Children's Hospital (June 2018-June 2023) alongside meteorological data (temperature, humidity, precipitation, atmospheric pressure, wind speed, sunshine duration). We characterized influenza epidemiology, employed Spearman correlation and generalized additive models (GAM) to quantify associations between meteorological factors and ILI, and developed SARIMA models to forecast epidemic trends.

Pediatric influenza in Tianjin demonstrated significant seasonal and spatial heterogeneity, with children under 6 years comprising >70% of cases. Spearman analysis revealed negative correlations between ILI cases and relative humidity (r = -0.31), temperature (r = -0.30), and precipitation (r = -0.36) (all p < 0.05), while atmospheric pressure exhibited a positive correlation (r = 0.34, p < 0.05). GAM identified relative humidity as the most influential meteorological factor (F = 2.40, p = 0.00478). The SARIMA(1,0,0)(0,0,0)₁₂ model demonstrated robust performance (R ² = 0.632, RMSE = 27.33).

Pediatric influenza in Tianjin predominantly affects children under 6 years, peaks in winter, and clusters in urban centers. Low-temperature and low-humidity conditions-particularly low relative humidity-exacerbate transmission. SARIMA models incorporating meteorological parameters effectively predict influenza incidence, supporting data-driven public health interventions.