Reply to Nannini and Arias.

TO THE EDITOR— We thank Nannini and Arias for their interest in our article [1]. They raise several important issues in assessing the impact of emergency department occult Staphylococcus aureus bacteremia on patient outcomes [2]. Although 30-day mortality is a commonly used study endpoint in most bacteremia outcome researches, we agree that a 30-day follow-up period might be inadequate for patients with S. aureus bacteremia. We therefore reanalyzed our data using 90-day survival as the study endpoint for occult S. aureus bacteremia patients (case group) and S. aureus bacteremia patients who were directly admitted at the time of their initial emergency department visit (control group). We found that there remained no survival difference between the case and control groups (Figure 1). Attributable mortality has been used as the outcome variable in several S. aureus bacteremia studies. However, determining attributable mortality in connection with a particular bacteremia episode is subject to information bias, and is therefore appropriate only in a prospective study or randomized controlled trial in which the outcome is evaluated independent of knowledge of the predictor variables [3, 4]. Furthermore, it is well recognized that S. aureus bacteremia tends to be complicated by embolic events, mycotic aneurysms, and recurrent bacteremia. All of these