Digital Prescription Surveillance Reveals Seasonal Patterns and Environmental Triggers of Allergic Disease Activity Across China

Climate change is intensifying allergic disease burden globally, yet real-time surveillance systems remain lacking in most countries. In China, allergic rhinitis affects 19% of adults and 22% of children [1], while asthma affects 4.2% of adults (45.7 million individuals), with only 28.8% receiving physician diagnosis [2]—highlighting a critical surveillance gap. Traditional monitoring through clinical registries or allergen testing is resource-intensive and captures only diagnosed cases, missing the broader population seeking symptomatic relief. Here we demonstrate that digital prescription transactions can provide near-real-time, nationwide surveillance of allergic disease activity. Analyzing 41.1 million antiallergic medication prescriptions from China's largest e-pharmacy platform (70% market share, 300 million users) across 31 provinces (2022–2024), we reveal distinct spatiotemporal patterns driven by environmental factors. We analyzed 41,143,860 prescription transactions from six antiallergic medications selected for high prescription volumes and complementary therapeutic classes: second-generation antihistamines (loratadine, desloratadine, cetirizine, levocetirizine), first-generation antihistamine (chlorpheniramine), and intranasal corticosteroid (fluticasone propionate nasal spray). Daily transaction rates were standardized per 100,000 population (Table S1). We applied Seasonal and Trend decomposition using Loess (STL) to quantify geographical heterogeneity, threshold-based epidemic detection (7-day smoothed seasonal component exceeding mean + 1SD for ≥ 7 consecutive days) to identify outbreak periods, and hierarchical Bayesian spatiotemporal modeling using Integrated Nested Laplace Approximations to estimate environmental exposure-response relationships while accounting for spatial autocorrelation and temporal dependencies (Data S1: Methods). A pronounced north–south seasonality gradient emerged across China (Figure 1A–C; Table S1). Northern provinces exhibited the highest seasonality indices—Shanxi (78.10%), Inner Mongolia (77.29%), and Hebei (77.01%)—while southern provinces showed the lowest values—Chongqing (31.14%), Fujian (32.61%), and Hainan (42.75%). Epidemic patterns varied regionally. North China demonstrated bimodal peaks: Beijing during March–April and August–September (63 epidemic days), with Henan showing similar patterns (51 days across April and August–October). Northeast China exhibited synchronized summer-autumn activity across Heilongjiang (35 days, July–September), Jilin (33 days, August–September), and Liaoning (42 days, August–September). South China displayed year-round activity with reduced amplitude, exemplified by Guangdong's minimal epidemic periods (16 days across October and December). These patterns align with documented pollen calendars, with northern regions showing tree and weed pollen dominance [3] and southern regions influenced by perennial allergens [4]. Temperature exhibited the strongest association with a U-shaped relationship (Figure 2A; Table S2), peaking at 23.37°C (95% credible interval: 14.34°C–27.27°C)—optimal conditions for pollen production and release [5]. Precipitation showed protective effects above 73.63 mm (Figure 2D), consistent with pollen washout mechanisms [6]. Air pollution exhibited a PM2.5/PM10 paradox (Figure 2I,J; Table S2). PM10 showed increased prescription rates above 34.83 μg/m3, potentially reflecting coarse sub-pollen particles (2.5–10 μm) released during pollen rupture that deposit in the upper respiratory tract, alongside shared meteorological conditions (dry, windy weather) that simultaneously elevate PM10 and facilitate aeroallergen dispersal. Conversely, PM2.5 demonstrated inverse associations at higher concentrations (18.76–58.83 μg/m3). Similar inverse patterns were observed for NO2, O3, and CO (Figure 2E,G,H), consistent with a behavioral avoidance hypothesis: PM2.5, as China's most widely monitored and publicized air quality metric, may elicit protective behaviors (staying indoors, mask-wearing) that reduce exposure to both pollutants and co-occurring allergens during high-pollution episodes. Spatial random effects captured substantial regional heterogeneity (Table S3). Beijing (0.39, 95% CI: 0.33–0.45) and Shanghai (0.13, 95% CI: 0.08–0.18) demonstrated the highest effects, while northwestern provinces showed negative effects—Ningxia (−0.07) and Gansu (−0.07). By incorporating spatial structure as random effects, our hierarchical Bayesian models adjusted for unmeasured regional confounders including baseline differences in platform adoption, healthcare infrastructure, and underlying disease prevalence. This approach allowed us to estimate within-region associations between environmental factors and prescription patterns while accounting for spatial autocorrelation. Digital prescription surveillance offers a pragmatic complement to traditional allergic disease monitoring. Our findings demonstrate strong concordance across multiple dimensions: temporally, spring and autumn prescription peaks align with documented pollen calendars [3]; spatially, the north–south seasonality gradient mirrors known allergen sensitization patterns [4]; environmentally, identified temperature and precipitation thresholds correspond to plant phenological requirements [5, 6]. Sensitivity analysis using fluticasone propionate nasal spray—a rhinitis-specific agent—validated these patterns despite lower prescription volumes (Figure S1, Tables S1–S3), supporting that seasonal signals reflect allergic disease activity rather than non-specific medication-seeking behavior. Importantly, this approach captures treatment-seeking populations who may never enter formal healthcare systems, particularly in resource-limited settings. Near-real-time surveillance enables early warning when environmental conditions exceed identified risk thresholds, supporting timely allergy alerts and resource allocation. Several limitations warrant consideration. First, prescription data represent treatment-seeking behavior rather than confirmed diagnoses, as antihistamines treat multiple allergic conditions (rhinitis, urticaria, pruritic disorders). Sensitivity analysis using rhinitis-specific fluticasone propionate nasal spray validated our seasonal patterns despite lower prescription volumes. Second, provincial aggregation masks sub-regional heterogeneity. Third, our three-year period characterizes seasonality but cannot assess long-term climate trends. Finally, the ecological design precludes individual-level causal inference. Our findings demonstrate that digital prescription patterns capture seasonal allergic activity with strong concordance to allergen ecology. This methodology provides near-real-time surveillance through treatment-seeking behavior, complementing traditional monitoring where direct disease surveillance remains resource-intensive. By identifying environmental thresholds and high-risk periods across China's diverse climatic zones, this approach transforms commercial e-pharmacy infrastructure into actionable public health intelligence. As allergic disease prevalence continues rising globally, scalable digital surveillance systems offer pragmatic solutions for timely intervention and resource allocation, particularly in resource-limited settings. Luzhao Feng and Xunliang Tong conceptualized and designed the study. Rui Shen and Chuangsen Fang contributed to data acquisition, statistical analysis, and manuscript drafting. Zixing Wang and Zilu Xu contributed to data acquisition and curation. Na Sun and Hao Jiang contributed to statistical analysis and data visualization. All authors participated in data interpretation, provided critical intellectual input, approved the final manuscript, and agreed to be accountable for all aspects of the work. The authors thank the technical staff at Meituan Health for facilitating access to the anonymized prescription database. Chinese Academy of Medical Sciences Innovation Fund for Medical Sciences [2021-I2M-1-044]; Science & Technology Fundamental Resources Investigation Program [2023FY100600]; National Key R&D Program of China [2024YFC2311500]. The authors declare no conflicts of interest. Aggregated data are available from the corresponding authors upon reasonable request. Individual-level prescription data cannot be shared due to privacy regulations. Environmental data are publicly available from the China Meteorological Data Service Center (http://data.cma.cn) and China National Environmental Monitoring Center (https://www.cnemc.cn). Data S1: all70203-sup-0001-DataS1.docx. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.