Statistics are not used in accessing a risk. The only way for anyone to know if they have been exposed is by testing.
The 1 in 1000 statistic applies both to transmission during vaginal sex (1 out of 1000 exposures) assuming the partner is infected and there are no other risk factors, and the average prevalence in the heterosexual population in North America (1 out of 1000 people infected), again, assuming no other risk factors.
Thus, the estimated risk of one episode of vaginal sex to an unknown heterosexual partner is 1 in a million.
There are plenty of threads in the doctors forum that discuss this.