Data from wildlife disease surveillance programs are used to inform implementation of disease control (e.g., vaccination, population reduction) in space and time. We developed an approach to increase detection of raccoon rabies in raccoons (Procyon lotor) and skunks (Mephitis mephitis) of Québec, Canada, and we examined the implications of using this approach for targeted surveillance. First we modeled the probability of sampling a rabid animal relative to environmental characteristics of sampling locations. Rabid animals were more likely to be found in low-lying flat landscapes that had higher proportions of corn-forest edge habitat and hay agriculture, and that were within 20 km of one or more known rabies cases. From the model, we created 2 complementary risk maps to identify areas where rabid animals were most likely to be sampled. One map accounted for habitat and known rabies case locations, and can be used to define an infection zone from which surveillance can be targeted along the periphery to determine if disease is continuing to spread. The other map only accounted for habitat and can be used to locate areas most likely to contain rabid animals when the disease is present. In a further analysis we compared the 2 most successful methods for detecting raccoon rabies in Québec, given the disease was present. Government trapping operations (active surveillance) detected more cases in the short-term, but citizen notification (passive and enhanced) was more effective after 12 trapping days from which the initial rabies case was found. Our approach can benefit wildlife and public health agencies wanting to assess the disease status of regions by targeting surveillance to habitats most likely to contain infected animals and by defining the duration over which sampling methods are effective.