Outcome of inappropriate empirical antibiotic therapy in patients with Staphylococcus aureus bacteraemia: analytical strategy using propensity scores.Clin Microbiol Infect. 2006 Jan; 12(1):13-21.CM
Patients with Staphylococcus aureus bacteraemia (SAB) who received either inappropriate or appropriate empirical therapy were compared by using two risk stratification models: (1) a cohort study using a propensity score to adjust for confounding by empirical treatment assignment; and (2) a propensity-matched case-control study. Inappropriate empirical therapy was modelled on the basis of patient characteristics, and included in the multivariate model to adjust for confounding. For case-matching analysis, patients with inappropriate empirical therapy (cases) were matched to those with appropriate empirical therapy (controls) on the basis of the propensity score (within 0.03 on a scale of 0-1). In total, 238 patients with SAB were enrolled in the cohort study. Characteristics associated with inappropriate empirical therapy were methicillin resistance, underlying haematological malignancy, no history of colonisation with methicillin-resistant S. aureus, and a long hospital stay before SAB. These variables were included in the propensity score, which had an area under the receiver operating characteristics curve of 85%. In the cohort study, SAB-related mortality was 39% (45/117) for inappropriate empirical therapy vs. 28% (34/121) for appropriate empirical therapy (odds ratio (OR) 1.60; 95% CI 0.93-2.76). After adjustment for independent predictors for mortality and the propensity score, inappropriate empirical therapy was not associated with mortality (adjusted OR 1.39; 95% CI 0.62-3.15). In the matched case-control study (50 pairs), SAB-related mortality was 32% (16/50) for inappropriate empirical therapy and 28% (14/50) for appropriate empirical therapy (McNemar's test; p 0.85; OR 1.15; 95% CI 0.51-2.64). In conclusion, inappropriate empirical therapy resulted in only a slight tendency towards increased mortality in patients with SAB.