Multi-species epidemiological models are often applied to host-pathogen systems using taxonomic classifications and species-specific estimated parameters. However, in species-rich communities, trait-based modeling may be simpler and more predictive than trying to estimate parameters for all species. We consider an SIS model for disease spread in a plant-insect pollination network with trait-dependent contact rates, under two different scenarios. In a nested network, some flowers are more attractive than others, and some insects are more selective than others. In a specialized network, insects and flowers vary in some trait and contact rates are highest when traits match. We find that disease spread is impacted most by the most selective bees and most attractive flowers (nested network), and by intermediate-trait individuals (specialized network). In both cases, flowers are more heterogeneous in importance than insects. These findings may be useful for bee conservation strategies that seek to minimize pathogen impacts.