needles due to the shorter length of time that needles are in use (or needle circulation time).
The University of California concludes (p. 484):
Relative impact would decrease modestly if different values were used for certain model parameters. However, if the model incorporated behavioral risk reductions, relative impact would increase substantially from the 33% published value.
Absolute impact would probably decrease below the published value of 0.021 infections averted per client-year, because the estimate of non-NEP HIV incidence appears to be too high.
The panel's view is that these models provide important qualitative insight into why needle exchange programs should work. However, conclusions from modeling a complex process can rarely have the force of absolute proof. Kaplan and his colleagues recognize this, offering a range of calculations based on competing assumptions, parameter values, and models. Despite these admirable efforts, it is true that unmodeled features of the needle exchange program process might make the efficacy estimates either too high or too low. For example, the Kaplan model does not take account of changes in the percentage of infected participants that result if disproportionately many dropouts from the program population were infected (or were not infected). Likewise, if new entrants to the needle exchange program population were less infected (or more) that would artificially raise (or lower) the apparent effectiveness of the needle exchange program. Therefore, we must regard numerical estimates from these models with some caution.
In summary, the model-based evaluation of the New Haven needle exchange program provides important insights into the dynamics of such programs and useful preliminary estimates of their efficacy. We cannot attach the same level of confidence to these model-based estimates as we could to evaluation programs that included a suitable control group in which individuals were tested (directly) for HIV infection. Unfortunately, such an evaluation program would face formidable obstacles because of concerns about privacy and confidentiality, difficulties ingeniously (and conscientiously) sidestepped by Kaplan's study methods.
Several other outcome variables were also studied at the New Haven needle exchange program, and these are described below.
If needle exchange programs attracted new initiates to injection drug use, a drop in the average age of program participants who enroll over time would be expected to be observed, coupled with a downward shift in the