AbstractConventional cost‐effectiveness decision rules rely on the assumptions that all health care programmes are divisible and exhibit constant returns to scale for a homogeneous population; hence, the path between adjacent programmes on a cost‐effectiveness frontier must be linear. In this paper we build a framework to analyse non‐linear ‘expansion’ paths. We model the impact of two key sources of non‐linearity: economies of scale or scope in the production of health care; and prioritisation of patients who are most likely to benefit from more expensive and more effective treatments. We conclude that the expansion path might be linear, convex or concave, depending on the situation. The path might also exhibit vertical discontinuity due to fixed costs or horizontal discontinuity due to indivisibility. The efficiency of resource allocation might be improved by empirical estimation of expansion paths. We discuss the advantages and disadvantages of this approach compared with a standard stratified analysis. Copyright © 2006 John Wiley & Sons, Ltd.