The primary objective of our study was to develop a new classification scheme for bloodstream infection that distinguishes among community-acquired and health care–associated bloodstream infections. Dr. Zetola states that subgroups within the health care–associated category we proposed have not been proven to be associated with infection caused by antimicrobial-resistant bacteria and that use of unnecessary broad-spectrum antibiotics might harm patients by leading to resistance. The most significant antimicrobial-resistant pathogen isolated in our study was methicillin-resistant S. aureus (MRSA). This was the infecting pathogen in 19% of patients with health care–associated bacteremia and 20% of patients with nosocomial bacteremia but only 2% of patients with community-acquired bacteremia. Moreover, MRSA bacteremia occurred in all subgroups of the health care–associated category, including recently hospitalized patients; those receiving home intravenous therapy or nursing care, dialysis, or chemotherapy; and those in nursing homes. Enterobacteriaceae resistant to ampicillin–sulbactam or ciprofloxacin were infrequently cultured. However, they were recovered with similar frequency in nosocomial settings (18 of 40 patients [45%] and 2 of 40 patients [5%], respectively) and health care–associated settings (17 of 45 patients [38%] and 5 of 45 patients [11%], respectively) and less frequently in community settings (11 of 58 patients [19%] and 0 of 58 patients [0%], respectively).

Over the past decade, published studies have reported increasing rates of non-nosocomial MRSA bacteremia (1, 2), and we are currently studying the impact of health care–associated MRSA infections on university and community hospitals. In preliminary data from our university hospital, 405 of 1061 MRSA isolates taken from blood, sputum, and urine during 1994–2002 (38%) were acquired in non-nosocomial settings. During 1999–2002, 810 of 1119 MRSA infections (72%) occurring in patients from 10 community hospitals in the Duke Infection Control Outreach Network were acquired in non-nosocomial settings. These data suggest that there is a large burden of health care–associated MRSA in our health system. We encourage Dr. Zetola and others to study these subgroups in their hospitals.

At present, we believe it is wise to empirically treat all patients who have health care–associated infection with antimicrobial regimens similar to those used for patients with nosocomial infection. Although we agree that further validation of the concept of health care–associated bloodstream infections is appropriate, we disagree that giving vancomycin or broad-spectrum antibiotics (depending on the suspected pathogen) as empirical therapy for suspected health care–associated bacteremia and other serious infections would cause harm by promoting antimicrobial resistance. In fact, withholding such therapy pending culture results could result in an even more undesirable outcome: the death of patients.