OBJECTIVES: To assess the interest of measuring CRP in emergency for diagnosing bacterial infections and making decisions about antibiotics and to compare its practical usefulness with clinicians' conclusions. METHODS: Systematic CRP measurements in 80 consecutive patients admitted to emergency ward with possible bacterial infection. RESULTS: were not transmitted to the physician in charge. Patients' files were analyzed retrospectively in two phases. In phase 1, two senior physicians assessed the diagnosis and need for antibiotics on the basis of the admission (emergency unit) files. In phase 2, a panel of experts examined the complete files (including discharge notes) to determine the likelihood of infection (obvious or probable, unlikely or excluded) and appropriateness of emergency antibiotics. Their recommendations were used as the standard, against which the usefulness of the laboratory indicators (including CRP) and decisions of the emergency physicians were assessed. ROC curves were used to determine threshold values for CRP and body temperature. We then calculated the sensitivity, positive predictive value and negative predictive value of these cutoffs and compared them with those for the phase 1 clinician recommendations. RESULTS: The study included 76 patients (mean age: 74 years): 28 presented obvious or possible infections and 21 required emergency antibiotic therapy. Mean leukocyte values did not differ between groups. For diagnosis, the threshold value of CRP was 85 mg/L and of body temperature 37.8 degrees C; for prescribing antibiotics, the values were 130 mg/L and 38 degrees C, respectively. The sensitivity, specificity, negative and positive predictive values of CRP were, respectively, 79, 81, 76, and 83% for diagnosis of bacterial infection and 71, 71, 48 and 87% for prescription of an emergency antibiotic. These values were lower than those of clinician's conclusions. CONCLUSION: Because of the variability in the thresholds used in its interpretation, the lack of specificity, and its poor predictive value for treatment decisions, CRP is of little interest in the diagnosis and treatment of patients with bacterial infections in intensive care. The cost generated by this examination is therefore not justified.