We know with a good deal of certainty that unprotected sex exposes individuals to potentially life-threatening illness. We also know that all sexual encounters are not the same and, especially since the HIV/AIDS epidemic, researchers have been trying to figure out what sexual behaviors are riskiest and how to use this information towards better micro and macro-focused prevention efforts.
As with all research, a key issue is measurement. Our models to predict individual behavior are usually only as good as our data. As you might imagine, sex can be a personal topic. One may be reluctant to tell a survey interviewer/doctor/friend about their sexual activities, obscuring the whos, hows and whens that are oh-so-important for public health.
Some recent work provides insight into the scale of the measurement problem. A paper by Alexandra Minnis and colleagues compared self-reported sexual activity with biomarkers of exposure (a test based on PSA which can detect exposure to semen in the previous two days) in a sample of Zimbabwean women. The results were sobering: 52% of women who had positive biomarkers said that they engaged in protected sex in the last two days; 23% reported having no sex at all!
In another paper, Brendan Maughan-Brown and I looked at a sample of young adults in Cape Town, South Africa. Our study focused on concurrent sexual partnerships, intuitively defined as the presence of (temporal) overlap between sexual relationships with two distinct partners. There is a hot debate right now on whether such partnerships have been driving the HIV/AIDS epidemic in sub-Saharan Africa. Unfortunately, this debate has been held back by the availability of good data.
Recently, UNAIDS came out with some guidelines on how to standardize and better measure concurrency. We assessed the effectiveness of these guidelines by assessing whether individuals who reported having concurrent relations also reported more than one sexual partner. What we found was surprising: among those who reported only one sexual partner in the last year, nearly 1 out of 6 reported having concurrent sexual relations during this period! We conclude that the UNAIDS methods, which involves asking individuals about each sexual partner they've had and the start and end dates of those partnerships, may actually underestimate the prevalence of concurrency by a significant amount by not fully accounting for all sexual partners.
As both these papers suggest, we have a long way to go before we can credibly claim that we have precise, unbiased estimates of sexual behavior. It would be useful to divert some of time we all spend on linking specific sexual behaviors to health outcomes to figuring out how to get the measurements of those behaviors right in the first place.