Aeroallergen Exposure as a Short-Term Predictor of RSV, Influenza, and COVID-19 in Two Health Regions of Puerto Rico: A Seasonal Weekly and Machine-Learning Analysis
Document Type
Poster Presentation
Publication Date
6-5-2026
UN Sustainable Development Goals (SDGs)
Goal 3: Good Health and Well-Being
Abstract
Background: Airborne fungal spores and pollen are established triggers of allergic disease, but their relationship with respiratory viral infections is not well defined in tropical settings. We previously showed that fungal aeroallergens are strong shortterm predictors of influenza and COVID19 in Puerto Rico. Here we extend this framework to respiratory syncytial virus (RSV) incidence and compare patterns across RSV, influenza, and COVID19 in two health regions: San Juan-Metropolitan (SJM) and Caguas.
Methods: Incidence of RSV, influenza, and COVID19 and daily mean fungal spore and pollen concentrations (2022–2024) were assembled for both health regions and yearly season, and data aggregated to weekly sums for RSV. For each region and season, we computed weekly and lagged correlations (0–6 weeks) correlations between RSV and fungi or pollen. We then fit seasonstratified logistic regression and random forest models using single aeroallergen predictors to classify topquartile RSV weeks (withinseason rank ≥75%). Previously results influenza and COVID19 from the same platform were used for crossvirus comparison.
Results: For RSV, weekly fungal concentrations showed moderate, statistically significant correlations in Fall and Summer. At lag 0, fungi–RSV correlations were r = 0.31–0.36 (p≤0.05) in Fall and r =0.36–0.41 (p≤0.02) in Summer across SJM and Caguas. Lagged analyses suggested shortlead peaks: r up to 0.47 at 1week lag in Caguas fall and 0.62 at 2week lag in SJM Summer. Pollen–RSV correlations were smaller and often nonsignificant. Fungibased logistic models for fall RSV achieved AUC = 0.78–0.79 with OR per 1 SD (1.4–1.9), while summer performance was more heterogeneous (AUC = 0.42 in SJM, AUC = 0.73 in Caguas). These RSV patterns parallel our earlier influenza and COVID19 findings, where fungal spores (not pollen) showed fall associations (daily r =0.19–0.35 with optimal lags =2–4 days) and strong predictive performance (fall random forest AUC = 0.95, accuracy =93% for influenza; AUC =0.87–0.89, accuracy =87% for COVID19).
Conclusions: Across RSV, influenza, and COVID19, airborne fungal spores—but not pollen—provide a reproducible, shortlead signal of viral activity in Puerto Rico, with the clearest effects for RSV in Fall and Summer and for influenza/COVID19 in Fall. These findings further support integrating fungal spore monitoring into respiratory virus earlywarning and nowcasting systems and warrant multivariable lagged models that also incorporate meteorology and other environmental variables.
Publisher
American Society for Microbiologists
Conference/Symposium
Annual Meeting of the American Society for Microbiologists: ASM Microbe 2026
City/State
Washington, DC
Department
College of Arts and Sciences
Recommended Citation
Rivera-Mariani, F. E., & Bolaños-Rosero, B. (2026, June 3-8). Aeroallergen exposure as a short-term predictor of RSV, influenza, and COVID-19 in two health regions of Puerto Rico: A seasonal weekly and machine-learning analysis [Poster session]. Annual Meeting of the American Society for Microbiologists: ASM Microbe 2026, Washington, DC, United States.
