Making the link between vector-based surveillance measurements and risk of infection for mosquito-borne pathogens
Vector-based surveillance programs are used as a risk assessment tool for many arthropod-borne pathogens; however, the relationships between entomological measures of infection and infection risk are non-linear, which can lead to difficulties in interpreting data from such surveillance. Entomological measurements that underlie vector-based surveillance programs and risk assessments often include trap counts, which provide an indication of the relative abundance of a vector species through time or space, and prevalence of infection in the vector population (or related measures such as the minimum infection rate). For mosquito-borne infections, in particular, these measurements are often motivated by a desire to estimate quantities derived from mathematical models, such as vectorial capacity (the expected number of hosts receiving bites from infectious mosquitoes per infected host per day) or the entomological inoculation rate (the expected number of potentially infectious bites received per day by a susceptible host). However, specific model formulations of quantities that define risk often make additional assumptions that are not accounted for in the application of these formulae to data and the resulting interpretations of risk. One such assumption that is commonly overlooked (and is ubiquitously invalid, at least for mosquitoes) is that vector population density is constant.
We adapt a simple, widely-used single host, single vector delay differential equation model to include seasonal forcing of the vector population, which is driven by a seasonal pulse in the emergence rate for adult mosquitoes. We show that seasonal vector populations result in a non-linear (humped) relationship between vector prevalence and the force of infection (instantaneous hazard of infection experienced by a susceptible individual) to the host. Nevertheless, risk assessments often use vector prevalence or related measures (such as the minimum infection rate) as the outcome of interest, which results in unaccounted for nonlinear relationships between statistical assessments of “risk” and the quantities of actual interest. Models can be used to formalize and test the assumptions that underlie design of entomological surveillance programs. Future work should also account for stochasticity and bias in the surveillance process itself. Together, these refinements may lead to improved interpretation of data and therefore more effective planning and intervention as a result of vector-based surveillance for mosquito-transmitted infections.