Because these assessments are based on mathematical models that, of necessity, must rely on various assumptions, the validity of the resulting estimates hinges critically on the validity of the assumptions that had to be made. Two scientific reviews of the procedures and assumptions embodied in Kaplan's models issued in recent years (U.S. General Accounting Office, 1993; Lurie et al., 1993) provided concordant views regarding Kaplan's models. GAO summarized Kaplan's work as follows (p. 23):
Both our experts found that the mathematical specifications used in both equations appropriately express the dynamic process of HIV transmission among injection drug users via infected needles. They agree in their assessment that the model is technically sound and incorporates all key parameters.
The University of California review states (p. 478):
The circulation model is a very significant contribution to NEP [needle exchange program] evaluation efforts. By focusing on how NEP needles alter the characteristics of needles in circulation, the model circumvents reliance on injection drug user self-reports of behavior change. Rather, the model uses syringe tracking and testing data to demonstrate that even if injection drug users made no effort to change behavior (aside from obtaining needles at the NEP), HIV incidence would drop as a result of lower HIV prevalence in the needles. Any additional reduction in risk behavior (such as cessation of sharing or increased bleaching) would reduce HIV incidence even further.
Concerning the numerical estimates, from applying the model to the study data in New Haven, the GAO review concludes (pp. 23-24):
Our experts agreed that Dr. Kaplan's assumptions serve to underestimate the impact of the New Haven program on the rate of new HIV infections. The expert reviewers strongly believe that 33 percent understates the true percentage reduction in new infections attributable to the program.
The data used in the model were primarily obtained from three sources: (1) data developed from the program's syringe tracking and testing system, (2) self-reports from injection drug users participating in the program, and (3) data developed from other AIDS research studies. Our experts noted that the data values used from these sources are reasonable and produce a conservative estimate of the program's impact on the rate of new HIV transmissions.
The model's estimate that the New Haven needle exchange program results in a reduction of new HIV infections among participants over 1 year is defensible as a minimal estimate of the program's impact. The 33 percent difference is strictly attributable to the reduction in levels of infection in