Threshold Approach

A decision-making method that classifies an intervention as cost-effective or not based on whether its ICER falls below a predetermined threshold.

Source: Claxton et al. 2015

Each candidate intervention is assessed as it arrives by comparing its incremental cost-effectiveness ratio against a predetermined figure, and accepted where the ratio falls below it. The procedure approximates what a full ranking of all possible uses of the budget would achieve, without requiring that ranking to be constructed, which is impossible in practice since candidates arrive at different times and the complete set is never known. Interventions are therefore assessed sequentially against a fixed standard rather than against one another, which is what makes the approach operable within an appraisal system.

Source: Claxton et al. 2015

Because ranking every candidate use of resources and funding downward until the budget is exhausted is theoretically correct and operationally impossible. Existing commitments cannot be reopened continuously, new candidates appear at intervals, and the information required for a complete ranking does not exist. A threshold converts an allocation problem into a test that can be applied to one candidate at a time, and the quality of the approximation depends on how well the threshold matches the displacement actually occurring. The approximation also assumes that decisions are independent, whereas adopting several interventions in the same area may displace the same activity repeatedly.

Source: Culyer 2016

The threshold must reflect the health produced by the activity that funding a new intervention would displace. Where it is more generous than that, each acceptance removes more health than it adds while appearing individually justified, and the system loses health through a sequence of defensible decisions. The budget must also be genuinely constrained, since where it is not the displacement the approach assumes does not occur and the reasoning does not apply. Where the budget expands in response to adoptions, the displacement falls on future periods or on other public spending rather than on current patients, which changes what the threshold should represent.

Source: Claxton et al. 2015

As a presumption rather than a rule, so an intervention above the threshold can still be accepted where other considerations weigh sufficiently and one below it can be refused. Frameworks normally state a range rather than a point, with acceptance becoming progressively less likely as the ratio rises through it. Modifiers for severity or other characteristics may raise the effective threshold for particular interventions, which makes the departure explicit rather than informal. Publishing the reasoning where a decision departs from the presumption is what makes the framework contestable rather than merely stated.

Source: Culyer 2016

That a single figure cannot describe a system whose marginal productivity varies between services and over time. That published thresholds are frequently historical conventions rather than estimates of anything, and are rarely revised as budgets change. That the presumption operates asymmetrically, since interventions above the threshold are accepted more often than those below it are refused. And that applying it uniformly ignores considerations most systems regard as relevant, which is why modifiers are attached to it. A further criticism is that the approach treats the threshold as exogenous when it is partly determined by the decisions taken under it, since accepting more interventions at a given ratio tightens the budget and raises the true displacement value.

Source: Claxton et al. 2015