The "virgin" HIV puzzle: Can misreporting account for the high proportion of HIV cases in self-reported virgins?

Background: The Demographic and Health Surveys (DHS) from Lesotho, Zimbabwe, and Swaziland reveal that a significant proportion of HIV infections occurred in adolescent women who claim to be virgins. Two explanations for this observation have been proposed: adolescent women misreport sexual status or non-sexual risks are more prevalent than previously asserted.
Method: The paper develops a statistical model to estimate bound for the proportion of HIV cases attributable to sexual transmission if people misreport sexual activity status. This model is applied to DHS data from adolescent women living in Lesotho, Zimbabwe, and Swaziland.
Findings: Without having detailed information on misreporting, predicted bounds for the proportion of sexual HIV transmission in the considered samples are uninformative (between 0.3 and 1). If misreporting is assumed to be moderate (between 0 and 10%), the proportion of sexual HIV transmissions would be bounded between 0.3 and 0.4. The hypotheses that more than 95% of HIV infections are due to sexual transmission is only valid if one assumes that a substantial proportion of sexually active adolescent women (more than 70%) have misreported sexual activity.
Impact: With no detailed information on misreporting, sharp bounds for the importance of sexual transmission in adolescent women cannot be estimated from existing data sources. Nevertheless, two possible implications for public health research and policy arise: (1) If one assumes that misreporting is moderate, non-sexual risk factors account for more HIV infections than previously assumed. This raises the question whether non-sexual risk factors are sufficiently covered by the current prevention paradigm and whether prevention interventions aimed to prevent non-sexual HIV need to be intensified. (2) The hypothesis that HIV is almost exclusively sexually transmitted implicitly requires that misreporting of sexual behavior is severe. If this is true, self-reported sexual behavior is an inappropriate measure for monitoring the epidemic.