Measuring sexual behaviour in the era of HIV/AIDS: the experience of Demographic and Health Surveys and similar enquiries.
ABSTRACT To review the sexual behaviour data collected in the Demographic and Health Surveys (DHS) and other similar national surveys from the perspective of data quality.
Two indicators of premarital and higher risk sexual behaviour were analysed for 31 surveys in 10 countries in sub-Saharan Africa and Latin America and the Caribbean. The analysis focused on the internal consistency of trends and gender differences in the reported indicators.
The authors found fluctuating trends in premarital sex in sub-Saharan Africa but consistent increases in Latin America and the Caribbean. Changes in questionnaire design do not seem to contribute to these trends and there is evidence that the increase in premarital sex is genuine in Latin America. Trends in sex with non-spousal, non-cohabiting partners show large fluctuations and inconsistencies between surveys in some countries but not others. Men are consistently more likely to report non-marital sexual partners than women and unmarried women are less likely than unmarried men to report casual partners.
Surveys are potentially a valuable source of information on sexual behaviour but there are sufficient grounds for concern to warrant considerable caution in the use of survey data to monitor trends in sexual behaviour. Survey findings must be evaluated carefully and interpreted in the context of other available information. These results caution against placing heavy emphasis on short term changes in sexual behaviour between individual surveys and highlight the need for attention to quality in data collection.
[Show abstract] [Hide abstract]
ABSTRACT: Understanding the transmission dynamics of HIV and other sexually transmitted infections is critically dependent on accurate behavioral data. This study investigates the effect of the mode of questionnaire delivery on the quality of sexual behavior reporting in a 2010 survey conducted in Kampala, Uganda, among 18–24-year-old women. We compare the reported prevalence of five sexual outcomes across three interview modes: traditional face-to-face interviewing (FTFI) in which question rewording was permitted, FTFI administered via computer-assisted personal interviewing (CAPI) in which questions were read as written, and audio computer-assisted self-interviewing (ACASI) in which participants listened to prerecorded questions and entered responses using a computer touchscreen. We then assess the validity of the data by evaluating the reporting of sexual experience against three biological markers. Results suggest that ACASI elicits higher reporting of some key indicators than FTFI does, but self-reports from all interview modes were subject to validity concerns when compared with biomarker data. The study highlights the important role that biomarkers can play in sexual behavior research.Studies in Family Planning 03/2014; 45(1). DOI:10.1111/j.1728-4465.2014.00375.x · 1.28 Impact Factor
[Show abstract] [Hide abstract]
ABSTRACT: Several countries with generalized, high-prevalence HIV epidemics, mostly in sub-Saharan Africa, have experienced rapid declines in transmission. These HIV epidemics, often with rapid onsets, have generally been attributed to a combination of factors related to high-risk sexual behavior. The subsequent declines in these countries began prior to widespread therapy or implementation of any other major biomedical prevention. This change has been construed as evidence of behavior change, often on the basis of mathematical models, but direct evidence for behavior changes that would explain these declines is limited. Here, we look at the structure of current models and argue that the common "fixed risk per sexual contact" assumption favors the conclusion of substantial behavior changes. We argue that this assumption ignores reported non-linearities between exposure and risk. Taking this into account, we propose that some of the decline in HIV transmission may be part of the natural dynamics of the epidemic, and that several factors that have traditionally been ignored by modelers for lack of precise quantitative estimates may well hold the key to understanding epidemiologic trends.PLoS Computational Biology 03/2014; 10(3):e1003459. DOI:10.1371/journal.pcbi.1003459 · 4.83 Impact Factor