The aim of this study was to assess the sensitivity and specificity of catheter-drawn and peripheral blood cultures. Paired blood culture samples collected over a 44-month period from a 280 bed Brisbane metropolitan hospital were analysed, using standard clinical and microbiological criteria, to determine whether blood culture isolates represented true bacteraemias or contamination. Catheter-collected cultures had a specificity of 85% compared with 97% for peripheral cultures. In only two instances (0.2%) was the diagnosis of clinically significant bacteraemia made on the basis of catheter culture alone. This study concluded that catheter-collected samples are not a good test for true bacteraemia, and that peripheral cultures are more reliable when the results of the paired cultures are discordant.