We conducted a retrospective cohort study of patients with Rocky Mountain spotted fever (RMSF) at a university hospital in order to assess the relationship between delay in treatment and mortality and to identify predictors of delay in initiating therapy. Patients with RMSF who received antirickettsial therapy within 5 days of the onset of symptoms were significantly less likely to die than were those who received treatment after the 5th day of illness (6.5% vs. 22.9%, respectively; P < .03). Ninety percent of patients were seen by a physician during this 5-day period, yet less than one-half of them received treatment before day 6. Three factors were independent predictors of failure by the physician to initiate therapy the first time a patient was seen: absence of a rash, presentation between 1 August and 30 April, and presentation within the first 3 days of illness. Until reliable early diagnostic tests become available, physicians may be able to decrease the mortality associated with RMSF by instituting empirical treatment of suspected cases within the first 5 days of illness.