Ponente
Descripción
Epidemiological models are a valuable tool to assist planning and decision making in public health settings, yet their usefulness - both for gleaning the main drivers of an epidemic and for simulating scenarios - can be severely hampered by lack of reliable parameters. This is especially relevant for emerging diseases, where historical data and a large set of clinical or epidemiological studies are absent.
This issue became very prominent during the Covid-19 pandemic, when early on parameter values and even model structure was not clearly supported by data. In this context, model fitting is an essential tool to allow the application of models in such scenarios. We will outline two works on this topic, and use them to reflect on the opportunities and challenges involved.
The first was produced in the context of the emergence of the P.1 (Gamma) variant of SARS-CoV-2 in the Amazonas State, Brazil, in November 2020. We estimated the transmissibility and reinfection of P.1 using a model-based approach, fitting data from the national health surveillance of hospitalized individuals and frequency of the P.1 variant in Manaus from December 2020 to February 2021. This was performed assuming the possibility of reinfection of individuals previously infected with the original variant - which was a contentious question and not documented at the time yet - and the reinfection estimate was later confirmed by empirical data.
The second attempted to estimate the rate of immunity waning after infection by SARS-CoV-2 - which is a key parameter to understand patterns of new epidemics - using an age-structured SEAIHR mathematical model that also takes into account the vaccination status and the history of prior infection of individuals. We parametrized this model using an endemic approximation, and data for six Brazilian state capitals, with hospitalization data from SIVEP-Gripe, vaccination records from SIPNI, demographic information from IBGE, and a contact matrix projected for the Brazilian population. The results indicate that the mean duration of immunity is highly sensitive to the hypothesized value of $\mathcal{R}_0$ and to the protection against infection conferred by vaccination or prior infection. The uncertainties in these parameters cannot be readily resolved, but its effect on the estimate of immunity waning can be made explicit.