Symptom control is one of the primary goals of hospice care. We prospectively followed patients with advanced cancer receiving outpatient hospice care to determine if the use of antimicrobials for a clinically suspected infection improved infection-related symptoms. During a 24-month period, 1,731 patients with a cancer diagnosis were admitted to a community-based outpatient hospice program. Over 89% of the patients had a Karnofsky performance of < or =60%. Six hundred twenty-three of 1,598 study patients were diagnosed with a total of 685 infections. Six hundred thirty-three of the infections were treated with antimicrobials for a clinically suspected infection. Symptoms were recorded, clinically indicated cultures were obtained, and antimicrobials were instituted at the discretion of the attending physician. Patients were subsequently monitored to determine the effects of antimicrobials on infection-related symptoms. A complete or a partial response of infection-related symptoms was observed in 79% of 265 patients with urinary tract infections, 43% of 221 patients with respiratory tract infections, 46% of 63 patients with oral cavity infections, 41% of 59 patients with skin or subcutaneous infections, and none of 25 patients with bacteremia. Fifty-two of the infections were not evaluable due to refusal of antimicrobials or receipt of less than 72 hours of antimicrobials. Patient survival in this study was not affected by the presence of infection or the use of antimicrobials. Although the use of antimicrobials improved symptoms in the majority of patients with urinary tract infections, symptom control was less successful in infections of the respiratory tract, mouth/pharynx, skin/subcutaneous tissue, or blood. Physicians should be aware of the limitations of the use of antimicrobials in patients with advanced cancer receiving hospice care. Treatment guidelines are proposed emphasizing the importance of patient preferences and the use of symptom control as the major indication for the use of antimicrobials in this patient population.