Complaining about your allergies online might provide valuable data to researchers. Over 25% of Americans experience seasonal allergies, but how the prevalence of seasonal allergies varies across space and time remains obscure, in part because allergies seldom warrant visits to healthcare providers.
Elias Stallard-Olivera and Noah Fierer mined Twitter (now X) posts and Google searches from 2016–2020 to extrapolate spatial and temporal allergy patterns. A natural language processing model sorted posts that indicated symptoms (e.g., "My allergies are really bad today!!") from posts that include key words but did not indicate the presence of symptoms (e.g., "Gluten and Allergy Free expo"). The authors validated their data against emergency department (ED) visits for manifestations of seasonal allergies.
The duo found reasonably strong relationships between online activity about seasonal allergies and ED records. The authors then used the resulting model to infer seasonal allergy patterns across all major metropolitan areas in the United States. The resulting maps show a strong national pulse of allergy symptoms in March–May: a wave of low-grade human misery beginning in the Southeast and ending in the Northeast and Upper Midwest. Interannual variability is considerable.
According to authors, anomalous spikes of allergy symptoms, such as a one recorded in Los Angeles County in June 2018, could indicate booms in specific pollens or molds. According to the authors, the method could assist with predictive modeling of seasonal allergies and allergen exposures.
Source:
Journal reference:
Stallard-Olivera, E., et al. (2024) Internet-based surveillance to track trends in seasonal allergies across the United States. PNAS Nexus. doi.org/10.1093/pnasnexus/pgae430.
New study uses internet data to map seasonal allergy patterns across the US