Modeling the Spread of HIV

To the Editor.—  The data concerning the explosive spread of human immunodeficiency virus in intravenous drug abusers in New York City (1978 through 1984) and then its marked leveling off, as presented by Des Jarlais et al,¹ cannot be predicted by simple diffusion models of infection. We agree with these findings as far as they go; however, we have found that observed behavior follows directly from the nature of the disease.Consider a risk model consisting of two interacting cohorts of intravenous drug abusers where the disease first appears and spreads in the high-risk group, subsequently penetrating and then spreading more slowly in the lower-risk group. Carriers are eliminated by developing clinical acquired immunodeficiency syndrome. The incubation period is modeled as a sequence of four exponentially distributed states, with mean times of 6 weeks, 2 years, 11 years, and 1.5 years, yielding a median incubation of 10 years and