1. Predicting the spatial pattern of disease risk in wild animal populations is important for implementing effective control programmes. We developed a risk model predicting the probability that a deer harvested in a wild population was chronic wasting disease positive (CWD+) and evaluated the importance of landscape connectivity based on deer movements. 2. We quantified landscape connectivity from deer 'resistance' to move across the landscape similar to the flow of electrical current across a hypothetical electronic circuit. Resistance values to deer movement were derived as the inverse of step selection function values constructed using movement data from GPS-collared deer. 3. The top CWD risk model indicated risk increased over time was higher among mule deer Odocoileus hemionus than white-tailed deer Odocoileus virginianus, males than females, and was greater in areas with high stream density and abundant agriculture. A metric of connectivity derived from mule deer movements outperformed models including Euclidean distance, with high connectivity being associated with high CWD risk. 4. The CWD risk model was a good predictor of CWD occurrence among an independent set of surveillance data collected in subsequent years. 5. Synthesis and applications. We found that landscape connectivity was a major contributor to the spatial pattern of chronic wasting disease (CWD) risk on a heterogeneous landscape. For this reason, future disease surveillance programmes and models of disease spread should consider landscape connectivity. In the aspen parkland ecosystem, we recommend managers focus surveillance and control efforts along river valleys surrounded by agriculture where mule deer abound, because of the high risk of CWD infection. Fig. 4. (a) Chronic wasting disease (CWD) risk weighted by species and sex composition within wildlife management unit 730 in Alberta, and (b) forest/shrub cover, streams and agriculture within 730, where streams and agriculture are covariates in the CWD risk model.