Objective: To assess the predictive value of a previous colonization with methicillin-resistant Staphylococcus aureus for the resistance pattern of a bacteriological specimen significantly positive to S. aureus.
Study design: Retrospective study of patients' files.
Patients: Patients admitted for at least 48 hours in a surgical intensive care unit from April 1, 1996 to December 31, 1997.
Methods: Collection of patients' characteristics and chronology of positive microbiological specimens with methicillin-susceptible (MSSA) or -resistant (MRSA) S. aureus from medical and laboratory records. During the study period, screening for nasal or perineal colonization with MRSA was systematically performed on admission and weekly thereafter.
Results: The files of 540 patients were reviewed. MSSA and MRSA infections occurred in 7% (39/540) and 4% (20/540) of the patients respectively. By opposition with MSSA infections, MRSA infections occurred more frequently in patients previously colonized with MRSA (13 infections in 63 colonized patients [21%] versus 7 infections in 477 non-colonized patients [2%], odds ratio = 18, confidence interval: 6-51, P < 0.0001). The median delay between colonization and infection was 5 days. The positive and negative predictive values for previous colonization with MRSA to predict infection with MRSA in presence of a bacteriological specimen significantly positive with S. aureus were 81 and 84%, respectively.
Conclusion: The probabilistic use of a glycopeptide in presence of a bacteriological specimen significantly positive with S. aureus should be limited to patients already colonized with MRSA, in order to decrease the abusive administration of these antibiotics.