Health care resource allocation: is the threshold rule good enough?

We review the foundations of resource allocation rules based on cost-effectiveness information. Comprehensive approaches, where a total budget is allocated in one go, require estimation of the costs and effects of all available health care programmes, which is unlikely to be practical. A common alternative is to assess individual programmes against a cost-effectiveness threshold. This has been shown to be efficient if the threshold is well calibrated and all programmes can be wholly or partially implemented with constant returns to scale. We discuss the feasibility of these assumptions, and the effects of relaxing them, concluding that programme indivisibility is unlikely to be a serious problem at a national level, but that miscalibration of the threshold and non-constant returns to scale might be. A rule that avoids these difficulties has been proposed previously: a new programme should only be implemented if it can be funded by cancelling another less effective programme. This could never reduce efficiency, unlike the threshold rule, though we show that it might sometimes fail to recommend an efficiency-improving change. We suggest a refinement of this reallocation rule based on explicit estimation of the costs and effects of partial implementation of the programmes under review. Research is required to assess the practicality of this option.