A model was derived for disease transmission in dynamic host populations and its application was demonstrated in forecasting possible outcomes of a bovine tuberculosis (Mycobacterium bovis) epidemic in a white-tailed deer (Odocoileus virginianus) population. The approach was mechanistic, based disease transmission on the probability of each susceptible individual becoming infected per unit time, and afforded the flexibility necessary to model epidemics in dynamic wildlife populations. This approach was applied to a sex- and age-structured deer population model. This model predicted that tuberculosis prevalence in a white-tailed deer population could rise from approximately 3% to about 21% over 25 yr, and that neither lowered deer survival nor lowered transmission would be completely effective in eliminating disease from the population. Maternal transmission appeared unimportant to modeled tuberculosis dynamics; in contrast, disease was not maintained for >15 yr in models lacking lateral transmission.