How much will it cost to provide AIDS treatment to 20 percent of those without it? To 60 percent or to 80 percent? How much would the cost of treating 80 percent of those in need be reduced by an effective HIV-prevention policy? What would happen if AIDS treatment became more effective? Or less so? These questions and more are of interest as the world attempts to build on recent successes in expanding AIDS treatment while addressing other health problems in the developing world and coping with the worldwide financial crisis. One approach to answering these questions is to leave it to the experts. A steady stream of journal articles, working papers and reports from organizations such as the WHO, the World Bank, and the Global Fund for AIDS, Tuberculosis and Malaria—and soon from the group called AIDS 2031—are addressing these issues, sometimes from divergent perspectives. Having been in the business of making these projections myself, first for individual countries (see here and here) and recently for the PEPFAR focus countries (here and here) and for South Asian countries (here), I believe that the practice of making these cost projections should become more transparent. In fact, virtually anyone with a little numerical competence should be able to do this. It’s not rocket science!
So we are publishing the computer programs that we used for these projections, together with the baseline data the program needs. I invite anyone who is interested to download and try out this program to explore the AIDS-treatment policy questions that it can address. All you need to get started is access to a computer with the Stata (www.stata.com) statistical analysis platform and the instructions in Appendix 1 of our program manual that you can download here.
Like any computer program, this one has its limitations. In my view, the program’s major limitation is its assumption that, for any specific country, the annual cost of treating a patient with first-line antiretroviral therapy is constant throughout the future projection period. As described in the manual, the specific unit cost per patient-year for an individual country varies with the country’s level of per-capita income. The provided data set contains the estimates that we have used. Users could edit our data set and insert their own unit-cost estimate for any single country or for all countries. It would require a more challenging hack of our program to allow this unit cost to change over time or with program expansion. Indeed this modification is on our to-do list for the next version. In the meantime, in the spirit of the open source community, we encourage you to hack the programs yourselves.
Do you have another nomination for the one major limitation of this program? Perhaps you will discover a real flaw, a bug, that we have not yet expunged. Comments and bug reports can be submitted either to this blog as a comment (see below) or directly to Owen at OMcCarthy@CGDev.org. Owen or I will get back to you as soon as we can, and hopefully other readers will also have answers or reactions. Please be quick to send us word of a bug so we can stomp on it.
Once you’ve gotten accustomed to running the program without snags, you can begin to use the program to explore serious policy issues like those we mention above. It would be great if you were to send us a short summary of your projection results focused on some policy issue, perhaps illustrated with AIDSCost graphs or tables.
You can also help by commenting on the values of the various parameters in the model. For example, one of the user-selectable parameters is the proportion of HIV-infected people who will become eligible for treatment every year. The model’s default assumption is 11 percent, but you can easily select another value. For example, in order for the model to capture an increase in the CD4 count criterion for recruitment from 200 to 350, one might want to increase this proportion from 0.11 to 0.18. Readers can point to published literature or send us a comment. If there is enough interest, perhaps we could post a survey on the web site and then use that survey to inform our choice of default values for the next version of the program.
Finally a word to those of you in academia who are discussing AIDS costs in your classes. We hope you will find the AIDSCost package useful as a teaching tool. You could simply point the tool out to students and suggest they play with it, or you might think of some specific explorations or exercises to assign them. We would be very interested in any such applications for this tool and, if you are willing, we would like to post your assigned AIDS cost projection exercise (with or without the “answer”) for the benefit of other instructors and their students worldwide.
Please post your comments here or direct them to Owen McCarthy at OMcCarthy@CGDev.org.