Investigating changes in COVID-19 epidemiological parameters from different perspectives

Abstract

The global pandemic of the coronavirus disease 2019 (COVID-19), which started in 2020, has posed significant challenges to public health and socio-economic well-being worldwide. Improving current knowledge of the COVID-19 epidemiological parameters would help to provide more comprehensive understandings of the transmission dynamics.

This thesis investigated changes in several key COVID-19 epidemiological parameters from various perspectives. Comprehensive literature reviews were summarized in Chapter 2, which highlighted that most studies have not fully explored the evolving patterns of these parameters, which originated the research gaps and objectives of this thesis as described in Chapter 3.

Using detailed line list data from Hong Kong, changes in the mean forward serial interval in local transmission pairs from 2020 to 2021 were examined in Chapter 4, which found shortened mean serial interval was related with strengthened stringency of implemented control measures, and age was significantly related with the infector-infectee transmission patterns. However, the forward serial interval systematically decreases over time due to backward sampling direction bias. To address this, a new inferential framework was introduced in Chapter 5 to estimate the realized forward generation interval, accounting for this bias. When applied to data from mainland China's first COVID-19 wave, results showed that implementation of control measures were associated with shortened realized generation interval.

Despite investigating temporal changes in the time-delay interval parameters, an advanced negative binomial model and an improved Gamma mixture model were employed in Chapter 6 to estimate the superspreading potential (measured by degree of overdispersion in cluster size distribution) and the generation interval distribution in eight transmission settings. Using detailed contact tracing data from Hong Kong (2020-2021), it was found that care homes exhibit both high superspreading potential and a large expected number of new infections per cluster, and with longer generation intervals than other settings. Although the greatest superspreading potential was found in retail and leisure setting, its small expected number of new infections per cluster suggested that transmission events under this setting would naturally die out. These findings underscored the importance of setting-specific interventions for outbreak control.

Furthermore, a novel method was proposed in Chapter 7 that combined individual-level viral load data with population-level contact tracing data to infer the latent period distribution. This approach was validated through simulations and applied to real-world data from Hong Kong, yielding detailed estimates of the mean latent period ranging from 2.6 to 3.7 days. Slightly longer latent periods were observed in nosocomial transmission clusters (e.g., healthcare institutes and care homes), again strengthening the need for implementing tailored control measures.

Moreover, the impact of vaccination on hospitalization duration was studied in Chapter 8. Data from hospitalized COVID-19 patients in Hong Kong were analyzed, revealing a significant association between vaccination status and shorter hospital stays, indicating additional vaccine benefits that could alleviate the disease burden during large outbreaks.

The overall discussions and conclusions were presented in Chapter 9. In short, through comprehensive analyses on various COVID-19 datasets, valuable insights were delivered by this thesis for public health decision-making and future outbreak control.