Plague is a low incidence flea-borne zoonosis that is often fatal if treatment is delayed or inadequate. Outbreaks occur sporadically and human cases are often preceded by epizootics among rodents. Early recognition of epizootics coupled with appropriate prevention measures should reduce plague morbidity and mortality. For nearly a century, the flea index (a measure of fleas per host) has been used as a measure of risk for epizootic spread and human plague case occurrence, yet the practicality and effectiveness of its use in surveillance programs has not been evaluated rigorously. We sought to determine whether long-term monitoring of the Xenopsylla flea index on hut-dwelling rats in sentinel villages in the plague-endemic West Nile region of Uganda accurately predicted plague occurrence in the surrounding parish. Based on observations spanning ∼6 yr, we showed that on average, the Xenopsylla flea index increased prior to the start of the annual plague season and tended to be higher in years when plague activity was reported in humans or rodents compared with years when it was not. However, this labor-intensive effort had limited spatial coverage and was a poor predictor of plague activity within sentinel parishes.