Background: Species distribution models (SDMs) have an important role in predicting the range of emerging and understudied pathogens and parasites. Their use, however, is often limited by the lack of high-resolution unbiased occurrence records. Echinococcus multilocularis is a parasitic cestode of public health importance which is widely distributed throughout Eurasia and is considered an emerging threat in North America. In common with many parasite species, available data for E. multilocularis occurrence are spatially biased and often poorly geo-referenced. Results: Here we produce three separate SDMs using MaxEnt for E. multilocularis using varying complexities of sampling schemes and environmental predictors, designed to make the best possible use of non-ideal occurrence data. The most realistic model utilized both derived and basic climatic predictors; an occurrence sampling scheme which relied primarily on high resolution occurrences from the literature and a bias grid to compensate for an apparently uneven research effort. All models predicted extensive regions of high suitability for E. multilocularis in North America, where the parasite is poorly studied and not currently under coordinated surveillance. Conclusions: Through a pragmatic approach to non-ideal occurrence data we were able to produce a statistically well supported SDM for an under-studied species of public health importance. Although the final model was only trained on data from Eurasia, the global model projection encompassed all known occurrences in the United States. The approach defined here may be applicable to many other such species and could provide useful information to direct resources for future field based surveillance programs for E. multilocularis in North America.