Future costs in cost effectiveness analysis

Abstract

This paper resolves several controversies in CEA. Generalizing [Garber, A.M., Phelps, C.E., 1997. Economic foundations of cost-effectiveness analysis. Journal of Health Economics 16 (1), 1–31], the paper shows accounting for unrelated future costs distorts decision making. After replicating [Meltzer, D., 1997. Accounting for future costs in medical cost-effectiveness analysis. Journal of Health Economics 16 (1), 33–64] quite different conclusion that unrelated future costs should be included in CEA, the paper shows that Meltzer's findings result from modeling the budget constraint as an annuity, which is problematic. The paper also shows that related costs should be included in CEA. This holds for a variety of models, including a health maximization model. CEA should treat costs in the manner recommended by Garber and Phelps.

Introduction

Economists have debated whether future medical costs should be incorporated in cost effectiveness analyzes for a number of years. Indeed, the Panel on Cost-Effectiveness in Health and Medicine was unable to reach consensus on this issue (Russell, 1986, Mushlin and Fintor, 1992, Weinstein et al., 1996). Meltzer (1997) has subsequently advanced theoretical arguments that, in his view, showed that “cost-effectiveness analysis must include the total change in future expenditures which results from a medical intervention, regardless of whether those expenditures are medical or non-medical (Meltzer, 1997, p. 41).” Garber and Phelps (1997) demonstrate that less general models imply that unrelated future costs should be excluded, but do not directly contest Meltzer's conclusions. Indeed, Garber (2000), p. 203, subsequently conjectures that generalizing their model by allowing for “intertemporal reallocation of income and consumption” might yield Meltzer's results. This is not a point of emphasis, however, because they stress that, at the margin, interventions should have the same cost per quality adjusted life year whether or not future costs are included. From this they infer that “the inclusion of future costs is without consequence so long as the practice is consistent (Garber and Phelps, 1997, p. 25).”

There is general agreement that cost effectiveness analyzes should account for related costs. The controversy is primarily about the treatment of unrelated future costs, but it is important to understand that both terms have non-colloquial meanings in this context. In cost effectiveness analysis future costs are termed unrelated if they are independent of current spending, apart from the effects of that spending on survival. Of course, if an intervention has significant effects on survival, it can have a major impact on spending, but such spending would be considered unrelated. For example, compared to bypass surgery, angioplasty appears to reduce mortality for high-risk patients with cardiovascular disease (Stroupe et al., 2006). Patients who live longer obviously use more medical care, but these costs would be considered unrelated to angioplasty. Costs of other cardiac care that do not appear to be affected by the choice of angioplasty or bypass surgery (e.g., treatment of aneurisms) would also be considered unrelated. In contrast, patients who choose angioplasty are more likely to require revascularization than patients who choose bypass surgery, and these added revascularization costs are considered related to angioplasty. In practice, as Weinstein and Manning (1997) note, teasing apart related and unrelated costs can be quite challenging. Nonetheless, the core issue for this paper is whether changes in spending that are solely due to changes in survival – i.e., unrelated costs – should be included in cost effectiveness analyzes.

Meltzer argues that cost effectiveness analyzes should incorporate related and unrelated future costs. He also argues that these costs should be measured as the present value of the difference between labor earnings and total spending weighted by the change in survival probabilities. And applied researchers appear to be taking note, as Meltzer's paper has been cited over 90 times. Although I have found no instances in which including unrelated future costs has fundamentally changed an analyst's assessment of an innovation, adhering to Meltzer's protocol has changed a number of cost effectiveness ratios. For example, in a study of adding a beta blocker to standard therapies for congestive heart failure, Ekman et al. (2001) found that including future costs increased the cost per life year from $1717 per life year to $22,137. Likewise, in an analysis of adding an angiotensin-converting enzyme inhibitor to the therapy of patients at high risk of cardiovascular events, Bjorholt et al. (2002) found that accounting for unrelated future costs increased the cost per life year gained from $2244 to $28,703. Meltzer et al. (2000) found that counting unrelated future costs decreased the cost per quality adjusted life year of adding intensive therapy for patients with type 1 diabetes mellitus from $22,576 to $9626. And Almbrand et al. (2000) found that recognizing unrelated future costs increased the cost per quality adjusted life year of intense insulin treatment for diabetic patients who had experienced a heart attack from $1772 to $28,467. Similarly, Johannesson et al. (1997) found that adding unrelated future costs in an assessment of the treatment of hypertension increased the cost per quality adjusted life year by at least $26,000 for older patients.

All of these calculations, which are based on earnings, suggest that accounting for unrelated future costs is important. Yet this paper argues that this is unlikely to be correct for two reasons. First, earnings-based estimates significantly overstate unrelated future costs. Second, a compelling case for including unrelated future costs in cost effectiveness analysis has yet to be made.

The remainder of paper rigorously examines the treatment of future costs in cost effectiveness analysis. Section 2 analyzes Meltzer's expected utility model using the Garber–Phelps budget constraint. This analysis finds, as did Garber and Phelps, that optimality is inconsistent with inclusion of unrelated future costs. Section 3 extends this by showing that consistency is not enough. Incorrectly calculating opportunity costs will distort decision making except in special cases. Section 4 reassesses the Meltzer model. This analysis shows that the present value of Meltzer's measure of future costs is approximately zero. It also shows that his budget constraint is problematic. Section 5 analyzes related costs, showing that their treatment should be quite different from unrelated costs. Section 6 shows that expected health maximization gives decision rules that parallel those of Garber and Phelps. Section 7 concludes.