Age and religion selection biases in HIV-1 prevalence data from antenatal clinics in Manicaland, Zimbabwe.

Sera from 487 women attending antenatal clinics in two areas of Manicaland were tested for the presence of HIV-1 infection. In the Honde Valley and Rusitu Valley areas, 24,3 pc and 14 pc respectively, were found to be infected. HIV-1 infection was found to be associated with age, marital status and location. Younger women, non-married women and women living in the Honde Valley were all more likely to be infected. There was also a weak association with level of education, with women with secondary education being at greater risk of being infected. The unadjusted overall figures for HIV-1 prevalence from the antenatal clinics were shown to provide a poor indication of the relative levels of prevalence of infection in the two study areas because of differences in the age structure and religious affiliations of the antenatal clinic and study populations. Similar problems may exist in sentinel surveillance data and would distort comparisons between locations and over time, especially during periods of rapid fertility change. In particular, the age bias is liable to exaggerate differences between urban and rural populations in developing societies. Where comparable information is available for the general population and sample sizes permit, this problem may be overcome by collecting basic socio-demographic data on the individuals tested and then applying standardization techniques. In the Honde Valley and Rusitu Valley example, the levels of HIV-1 prevalence after adjusting for age differences are 18,4 pc and 13,2 pc respectively.