Lack of Predictive Value of Isolating Coagulase-Negative Staphylococci from Blood Cultures

SIR-Berwaldt et al. [1] reported that, in their intensive care unit, 26.4% of patients whose Isolator (Wampole Laboratories, Cranbury, NJ) blood cultures yielded coagulase-negative staphylococci (CNS) had true bloodstream infections. These authors defined infection according to a set of criteria. They then analyzed their findings to determine the variables associated with infection. A basic fault in the study is the specificity ofthe selected criteria: fever, appropriate treatment, physician's implicit definition of infection, meeting the Centers for Disease Control and Prevention (CDC) definition of bloodstream infection, and additional criteria in nonneutropenic patients. Although these criteria may be reliable for the evaluation of patients who are not critically ill, their value in the intensive care unit (lCU) has not been established [2]. Since the predictive value of any parameter is influenced by its prevalence, we need to examine each of the selected criteria in relation to the condition of patients in the ICU. For instance, I believe that the predictive value of fever in this patient population is unknown. Although afebrile patients are unlikely to have infection, the opposite may not be true. Since fever is not an uncommon finding in patients in the ICU and blood cultures are usually performed for febrile patients, predicting the presence of infection on the basis of fever may not be accurate. Bad the authors stated the frequency of fever in their patient population and whether other causes of fever had been excluded, this criterion might be acceptable. The criterion of reliance on the physician's decision to treat the patient must also be examined carefully. Although most clinicians can distinguish true infection from contamination in noncritically ill patients, precise distinction in the ICU is not possible; this conclusion is supported by the widespread use of antibiotics in the ICU. I believe that the threshold at which antibiotic therapy is deemed necessary by clinicians is much lower for patients in these units. It would have been helpful if the authors had stated whether the physician's decision to treat was rendered before blood culture results had been obtained or if treatment was initiated after the results were noted. It would have been preferable if multiple observers had clinically assessed patients' conditions without having