15. Cost–Effectiveness Analysis for Priority Setting

In the figure below, which compares three ways of delivering immunization, point X describes the status quo of a current intervention, delivering immunization by means of fixed facilities. At point X, the intervention achieves a total effect E2 (measured as coverage or as disease reduction) at a total cost C2. The ratio C2 to E2 is the average cost-effectiveness ratio (ACER), shown by the slope of the line O-X. Beyond point X, expanding coverage becomes very costly, perhaps because the population not yet immunized is dispersed and hard to reach. (Chapter 20 includes estimates of how costs increase as immunization coverage expands but without introducing a sharp increase in costs.) Expansion to point X1, which increases the cost from C2 to C3, yields only a small increment E3-E2 in effect. The slope of the line X-X1 represents the incremental cost-effectiveness ratio (ICER) of that expansion, which would raise the ACER to line O -X1. The line X-X2 shows the alternative of reducing coverage, which would improve the average cost-effectiveness (to C1/E1) because marginal costs are rising steeply near point X. The ICER of the reduction in coverage is the ratio of C2-C1 to E2-E1.

Average and Incremental Cost-Effectiveness and Intervention Choices: Comparison of Three Ways to Deliver Immunization Comparison of Cost and of Effectiveness between Interventions: Conditions for Dominance

Raising immunization coverage at an affordable cost may require adopting the alternative of mobile vaccination teams, intervention Y. The hypothetical combination of fixed facilities and such teams allows increasing the effect to E4 (complete or nearly complete immunization) at a total cost of C4. The ICER of the mobile teams is shown by the slope of the line X-Y and the resulting overall or combined ACER by the slope O-Y. Adopting intervention Y would be clearly preferable to trying to expand coverage through intervention X by building and staffing more fixed facilities.

An alternative even better than Y might subsequently be developed, represented by point Z—for example, community-based immunization teams that could operate either near or far from fixed facilities because they use heat-stable vaccines that do not require a cold chain. The ICER of turning to that choice, represented by the line X-Z, is not only more favorable than intervention Y, but it is even better than the current ACER, and preferable to intervention X at any coverage level beyond X2. The cost-effective choice, therefore, is not to retain intervention X at its current level and add Z beyond that point but to switch entirely from X (or from X plus Y, if Y has already been adopted) to Z. Because it costs less but provides a better outcome, Z is said to dominate both X and Y. The following figure illustrates dominance of one intervention by another, as well as cases in which neither of two interventions is dominant.

Comparison of Cost and of Effectiveness between Interventions: Conditions for Dominance

If intervention Z is divisible (meaning that it can be operated at any desired scale, such as Z2), then it is preferable to X at a cost of C2 because of the additional effect E2*-E2. It can be extended all the way to E4, just as with intervention Y, provided only that the ICER represented by the slope X-Z is still acceptable to decision makers choosing how far to expand the intervention. That is, the cost must still appear to be justified by the increased coverage. Under either of these conditions, an obstacle to switching, or to doing so quickly, would exist only if substantial fixed costs accompanied the transition from one intervention to the other, such as recruiting or retraining staff, building health posts in communities, or setting up the system for distributing the new heat-stable vaccines.

Compared with intervention X, intervention Z is better in both dimensions (lower cost and greater effectiveness), so it is to be preferred, and is said to dominate X. However, intervention X would dominate any other treatment that is both more costly and less effective and, therefore, falls in the upper left quadrant. An intervention such as V or W may or may not be considered preferable to X (V is cheaper but also less effective, and W is more effective but also more costly). Whether either such intervention would be selected over X depends on the relation of the increased (or decreased) cost to the increased (or decreased) effectiveness. That ratio corresponds to an ICER. If a maximum acceptable, or threshold, value for the ICER is determined, as shown by the dashed diagonal line, then any intervention that falls below the dashed line would be acceptable (preferable to X), and those that fall above the dashed line would not be. Uncertainty about the estimates of cost and effectiveness means that, instead of a sharp line as in the figure, the division of preferable from nonpreferable interventions corresponds to a zone of some width that depends on the confidence intervals around the estimates. This kind of comparison can start from an existing intervention such as X in the first figure or, when there is currently no intervention, from point O in the first figure.

Source: Authors.