Influenza epidemics recur annually, causing substantial morbidity and mortality. Infection of an influenza strain confers immunity to similar strains whereas continual viral antigenic mutations can facilitate escape from prior host immunity. An improved understanding of how these evolutionary factors interact to shape the epidemiological dynamics of influenza at the population level will help public health sectors devise better strategies to mitigate this disease burden. Currently, two influenza A subtypes (i.e. A/H1N1 and A/H3N2) and two B lineages (B/Victoria and B/Yamagata) are co-circulating in humans. Incidence caused by each (sub)type is routinely collected in many surveillance networks. We hypothesize that these (sub)type specific incidence data can be used to infer the evolutionary and epidemiological dynamics of influenza. The rational is that epidemic intensity, timing, and dominant strain for each epidemic, as observed in incidence data, are the result of residual host prior immunity, antigenic novelty and fitness, and competition among co-circulating strains due to cross-immunity. Hence we propose to infer the evolutionary and epidemiological characteristics of influenza from incidence data. To this end, we will first develop a model-inference system, using infectious disease models and state-of-the-art Bayesian inference methods (Aim 1). The model will include all currently co- circulating (sub)types and simulate their transmission dynamics and interactions. We will run the model in conjunction with a Bayesian inference method (e.g. the ensemble adjustment Kalman filter) to simultaneously assimilate all (sub)type specific incidence data streams to calibrate the model. Using data collected from Hong Kong since 1998, we will estimate the changes in population susceptibility prior to each epidemic (i.e. indication of the emergence of antigenic innovations), the immunity period for each (sub)type (i.e. indication of continuous genetic/antigenic drift over each epidemic), the strength of cross-immunity between (sub)types (i.e. competition between influenza viruses), as well as key epidemiological parameters for each (sub)type (e.g. the basic reproductive number R0) (Aim 2). Given that only (sub)typed incidence data are needed, this model- inference system can be easily adapted to study the co-circulation of influenza viruses in regions around the world. Further, it can be adapted to study other multi-pathogen systems (e.g. dengue).
NATIONAL INSTITUTE OF ALLERGY AND INFECTIOUS DISEASES