Sexual Volatility and the Spread of HIV.

Abstract

The stability of sexual risk behavior over time is a common assumption in models of HIV transmission. In this thesis I show that this assumption does not hold in a population of gay men at risk of HIV infection and that the violation of this assumption has major implications for our understanding of HIV transmission. Using multiple statistical models that assume either the stability or volatility of sexual behaviors over time, I first analyze a prospective dataset of sex act frequencies. The results are unequivocal: over very short time periods a large proportion of individual have highly volatile rates of unprotected sex. Next, I use probability generating functions to derive expressions for the basic reproduction number, R0, and its variance under 3 theoretical models of contact rate volatility. R0 is a summary measure of the theoretical conditions under which an epidemic can occur. In each model, volatility has the effect of reducing R0, although this effect is mediated by the natural history of HIV infection. This means that, all other things being equal, increasing volatility brings an system with ongoing transmission closer to the threshold value under which sustained population transmission is impossible. Finally, using individual-based simulation, I show that the full effects of contact rate volatility are more complex than suggested by the probability generating function analysis. Contact rate volatility non only reduces the force of infection from chronic-stage infectives, volatility also increases the availability of high-risk susceptibles. These countervailing effects do not balance out, and, if fact, increasing volatility both increases population risk of HIV infection and the proportion of new infections coming from newly infected individuals. These results are discussed in terms of the implications of contact rate volatility for estimation of the relative contribution of each infection stage to the observed levels of infection.