Increasing urbanization and use of urban areas by synanthropic wildlife has increased human and domestic animal exposure to zoonotic diseases and exacerbated epizootics within wildlife populations. Consequently, there is a need to improve wildlife disease surveillance programs to rapidly detect outbreaks and refine inferences regarding spatiotemporal disease dynamics. Multistate occupancy models can address potential shortcomings in surveillance programs by accounting for imperfect detection and the misclassification of disease states. We used these models to explore the relationship between urbanization, slope, and the spatial distribution of sarcoptic mange in coyotes (Canis latrans) inhabiting Fort Irwin, California, USA. We deployed remote cameras across 180 sites within the desert surrounding the populated garrison and classified sites by mange presence or absence depending on whether a symptomatic or asymptomatic coyote was photographed. Coyotes selected flatter sites closer to the urban area with a high probability of use (0.845, 95% credible interval (CRI): 0.728, 0.944); site use decreased as the distance to urban areas increased (standardized βˆβ^ = − 1.354, 95% CRI − 2.423, − 0.619). The probability of correctly classifying mange presence at a site also decreased further from the urban area and was probably related to the severity of mange infection. Severely infected coyotes, which were more readily identified as symptomatic, resided closer to the urban area and were most likely dependent on urban resources for survival; urban resources probably contributed to sustaining the disease. Multistate occupancy models represent a flexible framework for estimating the occurrence and spatial extent of observable infectious diseases, which can improve wildlife disease surveillance programs.