From models to reality: computational estimation of acute infection prevalence from seroprevalence data—the case of Toxoplasma gondii

Background: The use of serological tests based on antibody detection plays a pivotal role in the definition of past exposure to pathogens and therefore monitor infection presence, guide treatment decisions, and support disease control efforts through detection and surveillance. Nevertheless, the information obtained from serological antibody tests may be incomplete, as acute infections can go unnoticed due to the delay between infection and the development of detectable antibodies. This is particularly relevant for those pathogens for which acute cases drive pathogen transmission and are associated with the onset of clinical symptoms. We therefore developed a computational framework based on mathematical models to estimate the prevalence of acute infections from serological testing to be broadly applicable to different pathogens. Results and conclusions: We showed the effectiveness of our framework and highlighted that, in addition to seroprevalence, prevalence of acute cases also depends on the recovery rate and the mean life expectancy of the population. We applied our framework to Toxoplasma gondii, a model pathogen for infections that are largely asymptomatic, highly prevalent in herds, and exhibit clinical signs (e.g., abortions) associated with the acute phase of a primary infection (following the acute phase, in the case of T. gondii). Despite the sanitary importance of diagnosing acute infections, in livestock T. gondii infection is usually investigated by identifying specific antibodies through serological testing, which limits our ability to predict the infection risk and the expected number of abortions. Through this worked example, we showed that our model allows for the estimation of the number of individuals in acute infection phase and prediction of the infection dynamics, providing valuable insights into disease spread and informing management strategies for the control of pathogens. Also, given the generalizability of the model proposed, it can be easily applied to different pathogens whose diagnosis relies on serological testing. Finally, to enhance accessibility, we have developed an interactive Shiny application to support the implementation and use of the framework.