February 2001
Allocating HIV Resources
An OR Approach to Prevention Strategy
By Edward H. Kaplan
In the fall of 1999, the United States Centers for Disease Control and Prevention (CDC) approached the Institute of Medicine (IOM) of the National Academy of Sciences to review domestic HIV prevention efforts. In doing so, the CDC specifically requested the development of a "visionary" framework for HIV prevention strategy in the United States. Given the presidential election, CDC requested the rapid conduct and completion of this study so that its results could benefit the incoming administration.
The IOM accepted this challenge, and constituted a committee co-chaired by Harvey Fineberg (the provost of Harvard University) and James Trussell (associate dean and professor of Economics at Princeton's Woodrow Wilson School). The committee, which included experts in epidemiology and public health, infectious disease, the clinical treatment of HIV/AIDS, behavioral science, economics, public policy and operations research, met four times during the spring of 2000. At these meetings, the committee received testimony from numerous officials representing federal, state and local agencies with responsibility for HIV prevention, as well as representatives of nongovernmental organizations and HIV/AIDS advocates. Additional information gathering included site visits with several state health departments, presentations to the committee at the national HIV Prevention Community Planning Leadership Conference, contacts via committee liaisons to governmental, nongovernmental and research organizations, and public comment. IOM committee staff also conducted reviews of the HIV prevention literature and existing prevention policy and funding patterns.
Given the limited time available for pursuing its task, the committee decided early on to focus on principles for guiding future HIV prevention efforts, as opposed to detailed agency-by-agency analyses. Arguably the most important (and perhaps controversial) principle the committee adopted was the following strategic goal for HIV prevention in the United States: to prevent as many new HIV infections as possible within the resources available for HIV prevention. Though this might seem obvious to some, the committee was struck by the large number of decision-makers involved with HIV prevention policy whose behavior is inconsistent with, if not incognizant of, this principle. To others, such a goal might appear too narrow for a national policy given competing concerns such as equity and fairness in the allocation of resources and the delivery of HIV prevention programs and services. The committee recognized this, noting that such concerns could rightfully serve as constraints on policy decisions taken. The committee also argued, however, that the cost of deviations from the strategic goal for whatever reasons should be made explicit in terms of foregone infections averted.
Preventing as many new infections as possible subject to resource constraints is a resource allocation problem of the sort familiar to operations researchers. However, of the six general recommendations for implementing the policy goal (see box), the arguments surrounding the allocation of HIV prevention resources, familiar as they may be to operations researchers and economists, were largely foreign to the majority of committee members.
What are the key questions to answer in considering resource allocation for HIV prevention? For example, in fiscal year 1999, the CDC distributed roughly $412 million for HIV prevention programs nationwide. Which HIV prevention programs should be sponsored? How much money should be spent on programs providing services for different populations in different locations across the country?
Before trying to answer such questions, the committee first sought clarification of existing policy. How are available prevention dollars currently allocated? Perhaps surprisingly, the committee could not obtain an oral or written description of the current "rules" governing the allocation of HIV prevention resources. However, from testimony and materials submitted to the committee, an implicit strategy emerged: proportionality. Simply put, existing prevention funds appear to be distributed in proportion to AIDS cases.
|
|
Figure 1 and Figure 2 exhibit two instances of proportionality. Consider first the CDC's allocation of funds in the counseling, testing, referral and partner notification program category. Figure 1 shows how, over time, the percentages of these funds allocated to services targeting different racial and ethnic groups approach the percentage of all AIDS cases that occur in these different groups. In other words, HIV prevention funding in this program category is proportional to AIDS cases.
Figure 1
Source: IOM Report
As a second example, consider CDC's allocation of HIV Prevention Community Planning funds to state health departments. As a function of AIDS incidence, Figure 2 also supports the committee's contention that HIV prevention funds have been allocated in rough proportion to the number of AIDS cases.
At first blush, allocating funds in proportion to AIDS cases might seem reasonable. Such an allocation appears objective, and perhaps fair. However, AIDS cases are not an appropriate marker of the need for HIV prevention. AIDS cases convey information regarding past rates of infection, yet the future contains those infections that can still be prevented. Were the committee faced with the allocation of AIDS treatment dollars, then given the relative uniformity of AIDS care across the country, allocating funds in proportion to AIDS cases would be more reasonable. However, the committee's concern was with resource allocation for HIV prevention.
Figure 2
Source: IOM Report
A second objection to proportionality is that such a system contains a perverse incentive. State health departments should be rewarded for preventing AIDS cases, not reporting them. Yet, under proportionality, health departments that work the hardest to prevent HIV infections and subsequently report fewer AIDS cases would receive less money, while health departments that allow infections to continue unchecked and subsequently report greater AIDS cases would receive more funding.
A third committee objection to proportionality is that such a system ignores the differential cost-effectiveness of competing HIV intervention programs. Programs differ in their costs of preventing HIV infections as well as their effectiveness in doing so. To prevent as many infections as possible with available resources requires identifying and sponsoring cost-effective intervention programs.
Consider the screening of blood donations in the United States. The blood supply has been quite safe from contamination with HIV since the advent of the HIV antibody test in 1985. However, more sensitive screening tests are possible, but at greater cost. In 1996, the Food and Drug Administration approved the use of the more sensitive p24 antigen test. This reduced the "window period" during which a person could be infected with HIV but not detectable by testing from 22 days (the antibody window) to 16 days (the p24 antigen window) on average, reducing the rate with which HIV infections entered the blood supply by 27 percent beyond the protection already provided by antibody testing. However, the incremental cost of p24 testing was $5 per test, amounting to a $60 million annual increase in screening costs for the blood supply since there are about 12 million donations annually. It has been estimated that p24 testing prevented, at the margin, an additional eight infections from entering the blood supply each year, which implies a cost-effectiveness ratio of $7.5 million per transfused HIV infection prevented compared to antibody screening alone. The committee, of course, did not question the wisdom of screening the blood supply. Rather, the committee asked whether it is sensible to spend an incremental $60 million to prevent an incremental eight infections.
This query gains force when one considers that there are other interventions that are capable of preventing new HIV infections at a fraction of the cost. For example, needle exchange programs, depending upon the study one consults, have been estimated to avert infections at a cost between $3,000 and $50,000 per infection averted. As another example, the use of zidovudine, nevirapine and other antiretrovirals has prevented mother-to-child transmission of HIV at an estimated cost of roughly $32,700 per infection averted. Now, the committee is fully aware that not all would value infections averted among transfusion recipients and drug injectors (or newborns for that matter) equally. However, it is also unclear that most would value the lives of transfusion recipients 150 times higher than the lives of drug injectors, which is the break-even valuation using even the most conservative needle exchange cost-effectiveness estimate.
To ascertain the value of moving towards a more cost-effective allocation of HIV prevention resources, the committee constructed a model linking funding decisions to infections prevented. At the core of the model is a simple identity:
Though simple, this identity conveniently breaks the problem of estimating infections prevented into two complementary activities. Estimating the rate of new infections is a problem in HIV epidemiology, while estimating the fraction of infections that can be prevented as a function of investment is a problem of program evaluation.
Rather than attempt to model HIV incidence rates directly, the committee located a recent (1996) estimation exercise carried out by a CDC researcher that resulted in estimates of annual new HIV infection rates in three risk groups — men who have sex with men, drug injectors and heterosexuals at high risk — in each of the 96 metropolitan statistical areas in the United States with populations over 500,000.
To estimate the effectiveness of HIV intervention programs serving the three risk groups identified above, a research assistant reviewed the HIV prevention literature. Based on the data generated from this review, the committee developed three scenarios — base case, optimistic and pessimistic — for each risk group. The base case scenario reflected average program effectiveness in reducing the rate of new infections and average costs per program participant. The optimistic scenario combined above-average program effectiveness with below-average costs, while the pessimistic scenario combined below-average effectiveness with above-average costs. In addition to cost and effectiveness measures, the committee noted that most HIV prevention studies report at least some degree of client dropout; at times these rates are appreciably high. This led the committee to impose constraints restricting the maximum fraction of the population at risk that could be retained by HIV interventions irrespective of expenditures. These constraints were set to 25 percent, 50 percent and 75 percent for the pessimistic, base case and optimistic scenarios, respectively.
The resulting model took the form of a linear program. As a function of the total HIV prevention budget, the model suggests the amount of money to allocate to prevention programs serving different risk groups in different states in order to prevent as many new infections as possible. Also as a function of the budget, the committee estimated the effectiveness of proportional allocation in preventing infections. The results are summarized in Figure 3, where for each scenario, the upper curve reports infections prevented under the optimal (i.e. cost-effective) allocation plan, and the lower curve reports infections prevented under proportional allocation.
The figure reveals, not surprisingly, that increasing funding for HIV prevention increases the annual number of infections that can be prevented, albeit at a marginally decreasing rate. Whether one remains committed to proportional allocation or admits a switch to an approach based on cost-effectiveness, it is possible to estimate the value of increasing the HIV prevention budget from current levels via the figure below.
However, the advantages of cost-effective allocation can be considerable. For example, according to the base case scenario, the committee estimated that $412 million allocated in accord with the proportional policy would serve to prevent roughly 3,000 infections each year. A shift to cost-effective allocation would increase the number of infections prevented from 3,000 to 3,900 — a 30 percent increase. It would be possible to retain the proportional policy and also increase the number of infections prevented to 3,900 per year — but only if the budget increased to roughly $700 million.
Figure 3
Source: IOM Report
Though great uncertainty exists regarding the actual magnitude of the gains in infections averted that could be achieved by switching from proportional to cost-effective allocation, the relative advantage of the latter is clearer from Figure 4. The impact of cost-effective allocation diminishes greatly for very large prevention budgets. However, for HIV prevention budgets of $500 million or less, it appears possible to increase the annual number of infections prevented by at least 30 percent over proportional allocation in both the base case and pessimistic scenarios.
Figure 4
Source: IOM Report
The committee also used the model described above to impute the costs imposed by various policy constraints. For example, federal policy currently prohibits the use of federal funding for needle exchange programs. Were such a funding prohibition removed, enabling needle exchange to enter the portfolio of interventions eligible for funding, the committee estimated that at the baseline $412 million funding level under the base scenario, the annual number of infections that could be prevented would increase to 5,300. Prohibiting needle exchange, however, leads to the estimate of 3,900 infections averted per year as reported above. The difference in infections prevented — 1,400 annually in this example — can be viewed as the price of prohibiting needle exchange. It is one of the factors that led the committee to specifically recommend that the United States remove legal barriers to the purchase and possession of drug injection equipment, and rescind the ban on use of federal funds for needle exchange.
The IOM Committee submitted its final report to the CDC on Sept. 26, 2000, and publicly released its findings the next day. Interested readers can access the entire report, titled "No Time To Lose: Getting More From HIV Prevention," on the Internet at www.nationalacademies.org/webextra/hiv/. While there is a great deal more in the committee's report than the resource allocation exercise described above, it remains instructive to see how even a simple model can greatly improve the level of discourse surrounding a difficult policy problem. Once formulated, the resource allocation model became the committee's vehicle for discussing not only resource allocation, but also the ethics of targeting in HIV prevention, the impact of policy barriers, the potential returns from investing in better prevention programs and the limits of HIV prevention. Perhaps it was this ability to help organize such thinking in a forum like the IOM Committee that was the real contribution of the modeling effort.
Edward H. Kaplan is the William N. and Marie A. Beach Professor of Management Sciences and professor of Public Health at the Yale School of Management and a member of the Institute of Medicine Committee on HIV Prevention Strategies in the United States.
OR/MS Today copyright © 2001 by the Institute for Operations Research and the Management Sciences. All rights reserved.
Lionheart Publishing, Inc.
506 Roswell Street, Suite 220, Marietta, GA 30060, USA
Phone: 770-431-0867 | Fax: 770-432-6969
E-mail: lpi@lionhrtpub.com
URL: http://www.lionhrtpub.com
Web Site © Copyright 2001 by Lionheart Publishing, Inc. All rights reserved.