Since 2001, UNAIDS has been forced to acknowledge drastically reduced HIV prevalence estimates in over a dozen African, Caribbean and Asian countries, as a result of well-designed "population-based"HIV surveys (randomly selected samples of urban and rural populations). Results from Demographic and Health Surveys (DHS) that included HIV testing (+) indicate that prevalence estimates in most of these high HIV prevalence countries were too high by up to 2 to 3 times due to the use of HIV sentinel surveillance (HSS) data that were heavily biased by urban sentinel sites. For example, the national HIV prevalence estimate of about 22% for Zambia in 2001 was based primarily on HSS data from antenatal clinics.

In 2001, a population-based DHS that included HIV testing was carried out in Zambia (ZDHS+) and the national HIV prevalence was estimated to be much lower at about 16%. Analysis of the different sampling methods used by these two national surveys led to the conclusion that the HIV prevalence rate estimated by the 2001–2002 ZDHS+ is about the same as the prevalence rate estimated by ANC surveillance when adjusted for the biased geographic coverage of the ANC surveillance system.

Based on the DHS+ and/or population-based surveys whose findings were available by mid-to-late 2003, UNAIDS had to reduce some of their initial 2001 HIV prevalence estimates by up to 50 percent or more – the 2001 estimate for Kenya was 2.3 million and the revised 2001 estimate was 1.2 million, an overestimate of more than a million or close to 100%.