Sex Transm Infect 82:iii64-iii70 doi:10.1136/sti.2006.019901

National population based HIV prevalence surveys in sub-Saharan Africa: results and implications for HIV and AIDS estimates

  1. J M García-Calleja1,
  2. E Gouws2,
  3. P D Ghys2
  1. 1World Health Organization, Evidence and Information for Policy Department, Geneva, Switzerland
  2. 2Joint United Nations Programme on HIV/AIDS (UNAIDS), Policy, Evidence and Partnerships Department, Geneva, Switzerland
  1. Correspondence to:
 Dr J M García-Calleja
 World Health Organization, Evidence and Information for Policy Department, Geneva, Switzerland; callejaj{at}
  • Accepted 6 April 2006


Background: Sentinel surveillance among pregnant women attending antenatal clinics (ANCs) has been the main source of information on HIV trends in sub-Saharan Africa. These data have also been used to generate national HIV and AIDS estimates. New technologies and resources have allowed many countries to conduct national population based surveys that include HIV prevalence measurement, as an additional source of information on the AIDS epidemic.

Methods: The authors reviewed the reports of 20 national population based surveys from 19 countries carried out in sub-Saharan Africa since 2001. They examined the sampling methodology, HIV testing and response rates, and female:male and urban:rural prevalence ratios. They also constructed adjusted prevalence scenarios assuming different relative risks for survey non-responders.

Results: The national population based surveys vary considerably in quality, as reflected in the household response rate (ranging from 75.4% to 99.7%), women’s testing rate (ranging from 68.2% to 97.3%), and men’s testing rate (ranging from 62.2% to 95.4%), while for some surveys detailed response information is lacking. While 95% confidence intervals around the female:male and urban:rural prevalence ratios in individual countries are large, the median female:male ratio of the combined set of surveys results is 1.5 and the median urban:rural ratio 1.7. A scenario assuming that non-responders have twice the HIV prevalence of those who fully participated in the survey suggests that individual non-response could result in an adjusted HIV prevalence 1.03 to 1.34 times higher than the observed prevalence.

Conclusions: Population based surveys can provide useful information on HIV prevalence levels and distribution. This information is being used to improve national HIV and AIDS estimates. Further refinements in data collection, analysis, and reporting, combined with high participation rates, can further improve HIV and AIDS estimates at national and regional level.


  • Edited by Peter Ghys, Neff Walker, Helen Ward and Rob Miller