Propensity‐Matched Analysis of Appropriate Empirical Antibiotic Treatment

To the Editor—Harbarth et al. [1] raised the methodological challenges of assessing the impact of appropriate antibiotic treatment in observational studies. We conducted a propensity-matched analysis comparing appropriate empirical antibiotic treatment with inappropriate empirical antibiotic treatment for inpatients who have moderate to severe bacterial infections. We reanalyzed a cohort of inpatients who had community- or hospital-acquired infection. Inclusion criteria and definitions have been previously described [2]. Empirical antibiotic treatment was defined as appropriate if it was prescribed within 24 h and matched the in vitro susceptibility of a pathogen that was subsequently deemed to be a significant isolate. The probability of receiving inappropriate empirical antibiotic treatment (“propensity”) was estimated using a multivariate logistic regression model entering all risk factors associated with inappropriate treatment on univariate analysis (P < .1). We matched patients who were given inappropriate treatment to patients who were given appropriate treatment by propensity score. We performed a 1 : 1 match of patients to 4 digits, proceeding to a 3-, 2-, or 1-digit match when a higher match could not be obtained. When >1 match was possible, the selection was made randomly. We assessed the effect of appropriate empirical antibiotic treatment on all-cause 30-day mortality within the propensity-matched cohort.