We appreciate Dr. Johnson's interest in our study. We agree that all positive blood cultures require clinical interpretation and concur that blood cultures drawn peripherally and those drawn from a catheter need such interpretation. We have provided a conservative interpretation based on the lower boundary of the 95% CI for the two methods.

Dr. Johnson makes an interesting point about the omission of the predictive value for discordant results. In response to his request we provide these results, but we believe that these results do not indicate a need to obtain a peripheral sample after obtaining a catheter sample. The positive predictive values (probability of an infection) given the following catheter/peripheral results are as follows: catheter-positive/peripheral-positive, 100%; catheter-positive/peripheral-negative, 29%; catheter-negative/peripheral-positive, 28%; and catheter-negative/peripheral-negative, 0%. The negative predictive values (probability of no infection) are as follows: catheter-positive/peripheral-positive, 0%; catheter-positive/peripheral-negative, 71%; catheter-negative/peripheral-positive, 72%; and catheter-negative/peripheral-negative, 100%.

If the catheter and peripheral samples show the same, positive result, the positive predictive value is 100%. If the results are negative, the negative predictive value, by definition, is 100%. If the catheter and peripheral cultures show different results, then both the positive and negative predictive values are lower. Note that the culture-positive/peripheral-negative and culture-negative/peripheral-positive results for positive and negative predictive values are similar.

A study that would test our hypothesis that the peripheral sample may not always be necessary would determine whether results from two catheter samples behave the same as results from one catheter sample and one peripheral sample. Unfortunately, we did not perform that study.