Association of pediatric influenza and meteorology: Analysis and prediction in Tianjin

Children are highly vulnerable to influenza with severe complications. Meteorological factors significantly modulate transmission patterns. In this study, historical daily pediatric influenza-like illness (ILI) cases and weekly ILI% in Tianjin during 2018-2024, along with concurrent meteorological parameters were collected. Meteorological signals associated with increased pediatric influenza cases were identified by distinguishing non-epidemic and epidemic seasons. An Informer-LSTM hybrid framework was developed for meteorological risk prediction of pediatric influenza, with validation conducted during the 2024-2025 seasonal influenza epidemic. During non-epidemic periods, lower temperature (4d mean Tmax<20°C, Tmin<12°C) and larger diurnal temperature range (DTR) (4d mean DTR>10°C) triggered pediatric ILI fluctuations, with differential impacts across age groups. Whereas the onset of seasonal influenza epidemic was strongly predicted by Tmin<0°C phase and specific humidity q<3 g/kg, enabling two-week advance warnings. The supervised rolling LSTM model integrating meteorological parameters (Pave, Tmin, DTR, RH or q), holiday patterns (Holiday, School day), and future ILI% prediction by Informer, achieving precise forecast of onset, surge, peak duration, and decline during the seasonal influenza epidemic of 2024-2025, demonstrating superior temporal resolution and accuracy. By calculating meteorological risk factors from predicted daily pediatric ILI cases, providing actionable risk alerts for optimized medical resource allocation and targeted prevention strategies in schools and households. This research advances location-specific influenza preparedness through the innovative integration of meteorological forecasting and epidemiological surveillance, is one of the specific actions taken in response to the UN's early warning about the health concerns associated with climate change.