Disease dynamics of pandemic influenza

Abstract

Outbreaks of novel pathogens continue to pose public health threats to us highlighting the importance of having robust quantitative tools for risk and severity assessments. This set of studies 1) investigated the risk of infection and spatial-temporal epidemiological characteristics of pandemic influenza A/H1N1 in Hong Kong from the years immediately after its initial establishment; 2) assessed the accuracy of measuring cumulative incidence of influenza infection using serological study design; 3) examined the performance of two commonly-used assays: microneutralization and hemagluttination inhibition; and 4) studied the feasibility and accuracy of simultaneously estimating the basic reproductive number and mean serial interval of an outbreak based on case data. The strain A/pH1N1 was used as the primary example, but seasonal strains were also used, when appropriate, for comparison.

Sera samples were collected at four trial periods from a longitudinal cohort between 2009 and 2011. To study the risk of infection and spatial-temporal patterns of spread, the following six potential risk factors of infection were examined based on paired serology: age, sex, presence or absence of a child in the household, household size, level of education, and district of residence. Data were analyzed by regression models. Next, to further study the design of serological studies, a transmission dynamic model was constructed to simulate case incidence of an influenza epidemic. Then based on empirical data and the simulated case data, the immune responses of two populations were simulated to obtain the seroprevalence estimation over time according to different epidemic scenarios and serological study designs. The performance of microneutralization and hemagluttination inhibition was examined by regression models. Finally, an inferential model was constructed to retrieve the basic reproductive number and mean serial interval from the case data of the simulated epidemics.

The risk of infection of A/pH1N1 was highly age dependent during the main pandemic wave, but the spread became highly spatially structured afterwards. When estimating the cumulative incidence of infection from seroepidemiological studies, either cross-sectional or longitudinal study design produced similar estimates when baseline or pre-season immunity antibody to the strain of interest in the population was low. However, high levels of detectable background immunity antibody and low levels of boosting following infection hinder the estimates, and cross-sectional study design was particularly prone to higher biases among the groups that have high baseline immunity. Further, MN assay showed higher explanatory power for epidemiological studies based on a four-fold titre rise as the outcome variable. We also demonstrated that joint estimation of the basic reproductive number and mean serial interval of an outbreak was feasible. Additionally, our inferential model was particularly effective for outbreaks in small population.

This set of studies use transmission dynamic models to understand the underlying mechanisms of influenza transmission, which subsequently, with the new understanding of the characteristics of sero-epidemiological studies and assay tests, informs the design of robust sero-surveillance system and quantitative methods that assess potential severity of various epidemic scenarios.