An epidemiological model was developed for rabies, linking the risk of disease in a secondary species (cats) to the temporal dynamics of disease in a wildlife reservoir (raccoons). Data were obtained from cats, raccoons, and skunks tested for rabies in the northeastern United States during 1992–2000. An epizootic algorithm defined a time-series of successive intervals of epizootic and inter-epizootic raccoon rabies. The odds of diagnosing a rabid cat during the first epizootic of raccoon rabies was 12 times greater than for the period prior to epizootic emergence. After the first raccoon epizootic, the risk for cat rabies remained elevated at levels six- to seven-fold above baseline. Increased monthly counts of rabid raccoons and skunks and decreasing human population density increased the probability of cat rabies in most models. Forecasting of the public health and veterinary burden of rabies and assessing the economics of control programmes, requires linking outcomes to dynamic, but predictable, changes in the temporal evolution of rabies epizootics.