Risk of death during acute infection is accelerating across diverse host-pathogen systems and consistent with multiple models of host-pathogen interaction.

Abstract

Infectious diseases remain a major cause of global mortality, yet basic questions concerning the relationship between within-host processes governing pathogen burden (pathogen replication, immune responses) and population-scale (epidemiological) patterns of mortality remain obscure. We use a structured literature review to leverage the extensive biomedical data generated by controlled host infections to address the epidemiological question of whether infection-induced mortality is constant, accelerating, or follows some other pattern of change and to infer the within-host mechanistic basis of this pattern. We show that across diverse lethal infection models, the risk of death increases approximately exponentially in time since infection, in a manner phenomenologically similar to the dynamics of all-cause death. We further show that this pattern of accelerating risk is consistent with multiple alternate mechanisms of pathogen growth and host-pathogen interaction, underlining the limitations of current experimental approaches to connect within-host processes to epidemiological patterns. We review critical experimental questions that our work highlights, requiring additional non-invasive data on pathogen burden throughout the course of infection.IMPORTANCEHere, we ask a simple question: what are the dynamics of pathogen-induced death? Death is a central phenotype in both biomedical and epidemiological infectious disease biology, yet very little work has attempted to link the biomedical focus on pathogen dynamics within a host and the epidemiological focus on populations of infected hosts. To systematically characterize the dynamics of death in controlled animal infections, we analyzed 209 data sets spanning diverse lethal animal infection models. Across experimental models, we find robust support for an accelerating risk of death since the time of infection, contrasting with conventional epidemiological models that assume a constant elevated risk of death. Using math models, we show that multiple processes of growth and virulence are consistent with accelerating risk of death, and we end with a discussion of critical experiments to resolve how within-host biomedical processes map onto epidemiological patterns of disease.