Risk Calculations for HIV Transmission From Infected Health Care Workers-Reply

In Reply. —We are grateful to Drs Gracely and Shelly for pointing out the exact method for computing upper bounds of rates based on the binomial distribution. Indeed, based on our data (no HIV transmission in 413 patients), we have 95% confidence that the true rate will not be higher than seven in 1000 patients (0.00722=1-[0.05^1/413]) instead of one in 1000 stated in the article. The results presented in our article were derived using the normal approximation to the binomial proportion, which in this case, we agree, is not suitable due to the extreme values of the probability of seroconversion.

The conclusion expressed in our article is not substantially affected in that the rate of HIV transmission during surgery is small and the precise quantification of the low risk may involve large studies similar to those done to assess the effectiveness of donated blood screening.¹