Economic Aspects of Evaluation - Evaluation Methods in Biomedical Informatics - page 324

FIGURE 11.5. Best-case/worst-case sensitivity analysis. A common method to evalu-
ate the sensitivity of a result is to combine all of the worst (least favorable) assump-
tions and calculate a worst-case CE ratio, and repeat the analysis with all of the best
(most favorable) assumptions and calculate a best-case CE ratio.
calculate the net change in costs incurred by moving from the standard
therapy A to strategy B, and the net change in benefits. As shown, strategy
B produces 5.5 additional years of life for a cost of $53,000, for an ICER
of $9,655/year.
However, it is extremely unlikely that all of the estimates of costs and
benefits are exact. In fact, they are all only estimates, and there are likely
high and low limits of those estimates. For example, the extra life expectancy
gained may be 5.5 years at baseline, but the confidence limits surrounding
that gain could be wide. For the sake of this example, assume that the net
benefit ranges from a low of 4.8 years to a high of 8.2 years. Similarly, each
of the cost estimates has some inherent variability. Figure 11.5 provides the
ranges of each component cost that goes into the estimates of the costs of
strategy A and strategy B, producing low, baseline and high estimates of the
difference in costs. Coupled with the best, baseline, and worst estimates of
the benefits, these can be used to create best-case, baseline, and worst-case
scenarios, shown in the bottom of Figure 11.5. The worst-case scenario
would be produced if the most pessimistic estimates of costs were correct
and the least optimistic estimates of benefit were also correct, which would
produce an ICER of about $20,700 per year of life gained. The best case
(which would represent the least expensive, most effective combination)
would result in an ICER of only $4,700/year of life gained, and the base-
line estimate (representing our best estimate of the difference in costs and