Lagged environmental effects on multi-pathogen respiratory positivity: a distributed-lag nonlinear analysis of seven Sichuan sentinel hospitals, China

Acute respiratory infections (ARI) impose a large global burden and show marked seasonality that implicates environmental drivers, yet the lagged effects of these exposures on pathogen-specific positivity remain poorly characterised. Using case-level data from seven sentinel hospitals in Sichuan, China (2024–2026) aggregated to a weekly panel of 23 014 ARI cases, we applied quasi-Poisson distributed-lag non-linear models (cross-basis, 6-week lag; trend spline; city effects; log offset) to estimate cumulative relative risks (90th versus 10th percentile) for temperature, relative humidity and six air pollutants, with within-pathogen false-discovery-rate control. At least one pathogen was detected in 51.9% of cases and co-detection in 18.2%; Haemophilus influenzae and influenza predominated. Relative humidity gave the most stable single-exposure association with respiratory syncytial virus positivity ( RR 2.01, 95% CI 1.06–3.82); it failed FDR control (adjusted P = 0.130 ) and was nullified by temperature co-adjustment ( RR 1.05). H. influenzae rose with temperature and ozone and fell with PM2.5; SARS-CoV-2 tracked carbon monoxide. Twelve of 40 estimates survived FDR control. The humidity–RSV signal is robust to model specification yet fragile under co-varying exposures — the cautionary lesson — and the reusable output is a transparent screening framework that exposes such fragility, serving exposure science and environmental epidemiology.