In 2003, an outbreak of West Nile virus (WNV) occurred in Saskatchewan, Canada from July to September. One-hundred thirty-three horse cases and 947 human cases were recorded and data were analyzed retrospectively for evidence of clustering to determine if clinical infection in the horse population could be used to estimate human risk of infection with WNV. Kulldorff's scan statistic was used to identify spatial-temporal clusters in both the human and horse cases. In most areas, human clusters were not preceded by horse clusters. In one area, a significant cluster of horse cases preceded human cases by 1 week; however, 1 week does not provide sufficient time for human-health authorities to act and provide advance warning for the public.