Environmental drivers of seasonal influenza in Guangzhou

Abstract

Influenza, a significant respiratory infectious disease, greatly impacts human health. Environmental factors such as air pollution and weather play a crucial role in the transmission and seasonality of influenza. However, the existing literature lacks consistent and robust evidence on the influence of these factors, particularly in tropical and subtropical regions. Several methodological challenges persist in current observational studies on influenza, such as accurate assessment of air pollution exposure, understanding the spatiotemporal autocorrelation in exposure and influenza data, deciphering individual-level associations, as well as identifying vulnerable groups. Thus, there is an urgent need to develop methodological techniques to investigate the relationship between environmental factors and influenza with higher reliability.

This thesis endeavors to develop methodological advancements for studying the impact of environmental drivers in sub-tropical Guangzhou, an influenza epicenter. The objectives are to: 1) develop a high resolution dataset to accurately estimate the spatiotemporal variability of ambient air pollutant concentrations employing machine learning techniques; 2) investigate associations between community-level ambient fine particulate matter (PM2.5) and seasonal influenza, accounting for underlying spatiotemporal autocorrelations; 3) investigate the associations between individual-level short-term exposure to six major air pollutants and influenza symptom onset; and lastly, 4) to investigate the independent and interactive impacts of multiple weather factors on the risk of influenza transmission.

The literature review and research framework are elucidated in Chapters 2 and 3. In line with the above-stated objectives, Chapter 4 presents a random forest model to estimate the spatiotemporal variations of air pollutant concentrations. This estimation involved a range of predictor variables, including physical geography, socioeconomic factors, meteorology, land use types, and road length to establish a foundational dataset for the subsequent two empirical studies. Subsequently, in empirical Chapter 5, the associations between PM2.5 exposure and the incidence of seasonal influenza were examined among children in a total of 779 communities. Spatiotemporal Bayesian hierarchical models were employed to account for underlying autocorrelations inherent in air pollution and influenza data. An individual-level case-crossover design has been leveraged in Chapter 6 to examine the associations between short-term exposure to six major air pollutants and the onset of individual-level influenza symptoms. Incorporating individual-level information such as home addresses enhanced the precision in assessing air pollution exposure, enabled assessment of susceptibility characteristics and identification of vulnerable groups. The independent and interactive associations of eight weather factors with influenza transmission, measured as the daily instantaneous reproduction number (Rt) has been presented in the last empirical Chapter 7. Influenza was objectively measured from clinically-diagnosed and laboratory-confirmed cases, collected by the Guangzhou Centre for Disease Control and Prevention.

Findings suggest that exposure to air pollutants (excluding ozone), is associated with higher risk of influenza. Lower temperatures, humidity and increased rainfall increased the risk of influenza transmission. These findings are likely to contribute to a better understanding of influenza transmission dynamics driven by complex set of environmental factors, and help forecast future incidence rates accurately, guide interventions for mitigation and adaptation, especially among vulnerable subgroups via proper environmental and healthcare planning and personalized preventive strategies.