Lyme disease (LD) is a commonly cited model for the link between habitat loss and/or fragmentation and disease emergence, based in part on studies showing that forest patch size is negatively related to LD entomological risk. An equivalent relationship has not, however, been shown between patch size and LD incidence (LDI). Because entomological risk is measured at the patch scale, while LDI is generally assessed in relation to aggregate landscape statistics such as forest cover, we posit that the contribution of individual patches to human LD risk has not yet been directly evaluated. We design a model that directly links theoretical entomological risk at the patch scale to larger-scale epidemiological data. We evaluate its predictions for relative LD risk in artificial landscapes with varying composition and configuration, and test its ability to predict countywide LDI in a 12-county region of New York. On simulated landscapes, we find that the model predicts a unimodal relationship between LD incidence and forest cover, mean patch size, and mean minimum distance (a measure of isolation), and a protective effect for percolation probability (a measure of connectivity). In New York, risk indices generated by this model are significantly related to countywide LDI. The results suggest that the lack of concordance between entomological risk and LDI may be partially resolved by this style of model.