Commentary: What can we make of an association between human immunodeficiency virus prevalence and population mobility?

It is a truism that population mobility facilitates the spread of infectious disease.1 Infected individuals will carry the infection with them when they move and, if infectious, may infect others, spreading the infection from one area to another. The spread of Sudden Acute Respiratory Syndrome (SARS) is the most recent high profile example, and the development of the human im- munodeficiency virus (HIV) pandemic has been no exception. We also expect a priori that as an epidemic spreads spatially through a population, mobility between areas of higher and lower prevalence will be found to be a risk factor for infection. On aver- age, individuals who move from higher HIV prevalence areas to lower prevalence areas will be more likely to be infected with HIV than those in lower prevalence areas. This risk will reduce as the epidemic matures and prevalence equalizes in the two areas. However, this has few practical implications for HIV control given the worldwide dissemination, long latency, and low infect- ivity of HIV.2 Unlike the strategies used to control the SARS epidemic, isolation of HIV infected individuals is impractical and unethical. Of more interest to HIV prevention policy makers is that mobility may also be causally associated with more proximate risk behaviours for HIV infection; and there are many plausible hypotheses that may explain any association. In the context of rural-urban mobility in sub-Saharan Africa, mobility may raise the risk of HIV infection by increasing partner-change rates, increasing contact with higher risk sex partners, such as commercial sex workers or clients, or increasing rates of concurrency. Conversely, mobility may tend to reduce the risk of HIV infection by increasing exposure to behaviour change messages or by improving access to sexually transmitted disease (STD) treatment services. In addition, the direction of causality may be reversed, with HIV infection leading to mobility. For example, mobility may be caused by AIDS morbidity (to 'go home to die' or to seek treatment) or union dissolution due to the death of a partner from AIDS. The specific hypotheses through which mobility and HIV infection are linked may have very different implications for the impact of mobility on the scale of the HIV epidemic, as high- lighted in a mathematical modelling study by Coffee et al.3 If