Meeting Brochure and registration form      SMDM Homepage

Monday, 16 October 2006 - 2:45 PM


Stephen Resch, MPH, Harvard University, Boston, MA, Joshua A. Salomon, PhD, Harvard School of Public Health, Cambridge, MA, Megan Murray, MD, DPH, Harvard School of Public Health, Boston, MA, and Milton C. Weinstein, PhD, Harvard School of Public Health, Boston, MA.

Purpose. Tuberculosis (TB) kills an estimated 2 million people per year globally. Our aim was to quantify tradeoffs in the allocation of resources across sequentially-connected components of TB control programs, taking mortality minimization as the program objective.

Methods. TB control was characterized as a sequence of connected components. The fraction of the population covered by a control program determines how many people nominally have an opportunity to receive care for TB. The intensity of case-finding impacts the time between disease onset and diagnostic testing among covered patients. The diagnostic testing method determines the proportion of cases that begin treatment. The quality of treatment determines the fraction of treated patients who are cured. Using a compartmental model of TB population dynamics, simulated over a 10-year period, we quantified the production function relating the components of TB control programs to the outcome of TB deaths averted. Combining these outcomes with marginal cost functions for the program components based on available data, we determined the optimal mix of activities given a budget constraint.

Results. With the objective of minimizing deaths subject only to a budget constraint, the optimal mix of TB control activities involves a preferential allocation of resources to the improvement of treatment outcomes, as opposed to expanding coverage or improving case-finding or diagnosis. As the budget constraint is relaxed, allocation of resources is recommended in the following order: (1) expanded population coverage, (2) more sensitive diagnostic methods, and (3) more intensive case-finding. In general, treatment cure should exceed 70% before population coverage is expanded. Improving diagnostic sensitivity should be considered when treatment cure probability exceeds 85% and population coverage is nearly complete. Active case-finding efforts that reduce diagnostic delay should be implemented only when cure probabilities, population coverage, and diagnostic sensitivity are high.

Conclusions. Although the optimal mix of components in a TB control program with a given budget depend substantially on the specific marginal cost functions for each component, the general principle that limited resources should preferentially be allocated to components toward the end of a sequential process was supported. Our results are in line with the current emphasis in global TB control on achieving high treatment cure probability before expanding case detection.

See more of Concurrent Abstracts B: Health Economics and Cost Effectiveness Analysis
See more of The 28th Annual Meeting of the Society for Medical Decision Making (October 15-18, 2006)