Our study of the impact of biased mosquito feeding preferences revealed that ignoring these biases may misrepresent the true disease burden, and that these biases have different degrees of importance depending on the efficacy of ITNs over time. In particular, ignoring the fact that infected mosquitoes tend to bite more humans compared to susceptible mosquitoes (represented by η>1 in the model), may significantly underestimate malaria prevalence, especially when ITN efficacy is constant. We hypothesize that with improved longterm ITN efficacy (for example, with long-lasting insecticide treated nets), or more frequent ITN replacement, disease prevalence will not only decrease, but also mosquitoes might develop more attraction towards certain classes of humans. Consequently, it may be imperative to include this degree of complexity into mathematical models of malaria aimed at studying low-endemic regions, or malaria control that reduces disease burden. Furthermore, omitting biased feeding behavior may underestimate the level of control needed to reduce the basic reproduction number R0, below unity when there is no backward bifurcation or below the saddle-node bifurcation point if there is a backward bifurcation.