Nosocomial infections reflect the worldwide dynamics of health care delivery. In the developing world, complex medical technologies, including invasive devices and procedures, may be introduced without the appropriate infection-control infrastructure in place. Thus, as the use of these technologies increases, so do the risks for nosocomial infection, antimicrobial use, and emergence of antimicrobial-resistant pathogens. The concurrent downsizing and merging of health care systems in the developed world ensure that nosocomial infections will remain important in these settings for years to come. Recent studies suggesting a link between decreased nurse staffing levels and increased risk for infection [4, 5] are one example of how changes in health care delivery can affect rates of nosocomial infection; in addition, infection rates will probably increase as an increased proportion of U.S. hospital beds are occupied by severely ill patients in intensive care units who are exposed to numerous devices and antimicrobials.

In this issue, Villers and colleagues [6] illustrate the complex relation between the practice of medicine (specifically, antimicrobial use) and the risk for nosocomial infection. They also show the value of combining laboratory and epidemiologic investigations. First, they used molecular typing to show that a prolonged Acinetobacter baumannii outbreak among patients in an intensive care unit was actually two outbreaks. Second, epidemiologic methods permitted them to identify the different factors responsible for the two outbreaks: A surgical procedure performed in an emergency operating room was associated with the clonal epidemic infections, whereas previous exposure to a fluoroquinolone was associated with the nonclonal endemic infections. In addition, the authors showed a relation between the amount of parenteral fluoroquinolone used in patients in the intensive care unit and the incidence of A. baumannii infection.

Contribution of the laboratory to this investigation began with confirmation of the species identification. Acinetobacter species are increasingly important nosocomial pathogens because of their propensity for developing multidrug resistance [7], but they may be difficult to identify because of their changing taxonomy and variable phenotypes [8]. In Villers and colleagues' study, a molecular method involving restriction analysis of ribosomal RNA permitted the reliable identification of isolates. The laboratory contributed further by typing isolates with a polymerase chain reaction (PCR) technique in which primers annealing to repetitive sequences interspersed throughout the genome produced variable-length DNA segments; ribotyping confirmed PCR typing results. Molecular typing of the strains greatly facilitated the epidemiologic investigation. It is hard to imagine the confusion that would have resulted if investigators had been unable to distinguish between the epidemic and endemic infections, had considered the two outbreaks to be one, and had investigated with a single epidemiologic study.

The incidence of nosocomial infections caused by Acinetobacter species has an unexplained increase during the late summer months [9]; this phenomenon has been noted in a variety of settings in the past 20 years [10, 11]. Although Villers and colleagues do not mention seasonality, A. baumannii was first seen in their intensive care unit in the summer: Twenty-six infections occurred between May and October 1988. The seasonal occurrence of Acinetobacter species infection may be related to increased ambient humidity and to the fact that these organisms grow well in water [12]. If the emergency operating room associated with the epidemic A. baumannii infections had still been in use when it was identified as a source of infection, microbiological investigation of a variety of environmental sources, such as humidifiers [13], medications [14], and invasive medical devices [15], would have been indicated. Although the A. baumannii organisms associated with the epidemic infections probably arose from an environmental point source, the A. baumannii organisms associated with the endemic infections, especially those with the predominant R2 molecular profile, were probably transmitted from patient to patient on the hands of health care workers.

An important risk factor for infection was previous exposure to a fluoroquinolone. In their discussion, Villers and colleagues assume that gastrointestinal tract colonization preceded infection with A. baumannii. Although Acinetobacter species are not part of the normal human intestinal flora [16], a recent study suggested that A. baumannii may rapidly colonize patients who are admitted to intensive care units where infection with the organism is endemic [17]. Surprisingly, previous antimicrobial therapy was not a risk factor for colonization in that study.

Exposure to antimicrobials may play a role as a risk factor for infection by providing a selective advantage for one pathogen over competing microbial flora. On the basis of ecologic data, Villers and colleagues hypothesize that low concentrations of drug in the stool selected for emerging drug-resistant mutants. An assessment of minimal inhibitory concentrations (MICs) and point mutations among resistant strains might have helped support this hypothesis. High-level fluoroquinolone resistance (MIC 32 µg/mL) in Acinetobacter species is conferred by the combination of at least two point mutations in genes encoding the active sites of the drug [18]. The presence of single preexisting point mutations would render Acinetobacter species more likely to develop high-level resistance. Therefore, one might expect a large number of strains with only a single point mutation and a moderately elevated MIC if high-level fluoroquinolone resistance in the A. baumannii organisms associated with the endemic infections was frequently developing in vivo.

Villers and colleagues may have overlooked an important bias in the ecologic data they present on parenteral and oral fluoroquinolone use and the incidence of infection. Because fluoroquinolones have excellent bioavailability and high drug levels can be achieved with oral administration [19], patients who nonetheless require parenteral administration are likely to have greater underlying disease severity associated with impaired gastrointestinal function. This difference may account for the correlation of infection with parenteral, but not oral, fluoroquinolones. We assume either that the authors did not collect data on the route of fluoroquinolone administration or that parenteral administration was not an independent risk factor for infection in their case–control studies.

The high rates of multidrug resistance in the A. baumannii isolates associated with the endemic infections suggest that strains were exposed to multiple antimicrobials over time and that fluoroquinolone resistance did not develop de novo. Characteristics of fluoroquinolones other than their ability to select for drug-resistant mutants may explain why infections were associated with this particular class of agents. One possible explanation for the role of fluoroquinolones may be their property of secretion at low concentrations in sweat, which promotes skin colonization with multidrug-resistant flora [20].

As the complexity and dynamics of medical care continue to increase, the complexity of factors contributing to nosocomial infections will continue to challenge us. Examples of such complex risk factors include the use of invasive medical devices; the patient's severity of illness; and issues related to health care delivery, such as nurse staffing. In addition, Villers and colleagues [6] describe how antimicrobial use may function as a risk factor for nosocomial infection. As increasing antimicrobial resistance threatens to limit our ability to treat many infections, it will be increasingly important to investigate; identify risk factors; and develop, initiate, and evaluate interventions to control and prevent the transmission of nosocomial infections caused by multidrug-resistant pathogens. Meeting the challenge to investigate and understand increasingly complex nosocomial infections, however, will require increased collaboration among all members of the investigation team. Herein lies the greatest lesson learned from Villers and colleagues [6]: Their report serves as an excellent example of what may be achieved when the disciplines of clinical medicine, laboratory investigation, and hospital epidemiology are fully integrated in the investigation of nosocomial outbreaks.