The Global HIV Epidemics among Sex Workers

Page 282

244   The Global HIV Epidemics among Sex Workers

number and salary of peer educators, and the number counselors also explain significant variation across model iterations. Smaller effects are identified for other personnel items and the administrative overhead rate (varied in the model at +/- 20% from the base case estimate of 10% of total program costs). The full detailed sensitivity results for Brazil, where we had the most costing detail, are shown in Table 5.3. Figures 5.4, 5.6, and 5.8 show the probability density of values across model iterations for the estimated cost per participant in Kenya, Thailand, and Ukraine respectively. Given that cost estimates were collapsed by category there are fewer input parameters, resulting in much more rapid model convergence with approximately 4,900 iterations of the stochastic model needed to achieve convergence in each of the Kenya, Thailand, and Ukraine analyses. For Kenya (Figure 5.4) when all parameters are set at their lowest (least expensive) value, the minimum cost per participant was $108. The highest value for cost per participant was $170. The average and most likely result was $140 per participant. Note that the high and low values are very unlikely, and the 5th percentile value is $119 and the 95th percentile value $160. For Thailand (Figure 5.6) when all parameters are set at their lowest (least expensive) value, the minimum cost per participant was $83. The highest value for cost per participant was $130. The average and most likely result was $106 per participant. Note that the high and low values are very unlikely, and the 5th percentile value is $91 and the 95th percentile value of $122. For Ukraine (Figure 5.8) when all parameters are set at their lowest (least expensive) value, the minimum cost per participant was $79. The highest value for cost per participant was $126. The average and most likely result was $102 per participant. Note that the high and low values are very unlikely, and the 5th percentile value is $87 and the 95th percentile value $116. Figures 5.5, 5.7 and 5.9 show the sensitivity analysis for cost per participant by category for Kenya, Thailand, and Ukraine respectively. The standardized regression coefficients shown allow for rank ordering of how sensitive the cost per participant is to discrete items. Note that given that the composition of each country program in Kenya, Thailand, and Ukraine were based on the micro-costing from Brazil, the pattern of effects is quite similar across these three programs. The vast majority of expenses in all sites is for personnel, followed by overhead (at 10% of total program cost), and community mobilization efforts. Variation in the cost of other discrete items has minimal effects on the cost per participant.


Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.