Logical Inconsistencies With Expected Utility Theory May Align Better With Patient Preferences—A Response to Paulden et al
Abstract
In a recent paper, Paulden et al1 raise the point that using the new metric Health Years in Total (HYT)2 can sometimes violate the “Independence of Irrelevant Alternatives” (IIA),” a fundamental axiom of rational choice theory, and recommends caution against adopting such a metric over quality-adjusted life years (QALYs) in its use in cost-effectiveness analysis. Specifically, they conclude that “decision makers avoid these approaches, and that the logical consistency of any approaches proposed in future be thoroughly explored before considering their use in practice.” Paulden et al,1 ironically, isolate this specific property of choice theory to criticize a metric when the same property is the
backbone of Arrow’s Impossibility theorem,3 showing that in the presence of this property, there cannot be any 1 consistent aggregation of social welfare, calling into question the whole field of cost-effectiveness. I agree that HYT can sometimes violate an expected utility-theory-driven independence condition. Still, such a condition is neither necessary nor sufficient for adopting a public or private decision-making metric.4 There remain several other holes in their arguments that lead to their conclusion, none more so than clamoring to hold on to a metric whose consistency with patient preference has long been questioned. Take, for example, the work by Lancsar et al5 that demonstrates that even when a flexible multiplicative model is imposed on stated preferences, individuals value life expectancy gains more than quality of life (QoL) gains even when they lead the same gains in QALY (a condition met by HYT). Similarly, more recent work6 demonstrates that there is no guarantee that individuals treat duration of life and QoL multiplicatively.
backbone of Arrow’s Impossibility theorem,3 showing that in the presence of this property, there cannot be any 1 consistent aggregation of social welfare, calling into question the whole field of cost-effectiveness. I agree that HYT can sometimes violate an expected utility-theory-driven independence condition. Still, such a condition is neither necessary nor sufficient for adopting a public or private decision-making metric.4 There remain several other holes in their arguments that lead to their conclusion, none more so than clamoring to hold on to a metric whose consistency with patient preference has long been questioned. Take, for example, the work by Lancsar et al5 that demonstrates that even when a flexible multiplicative model is imposed on stated preferences, individuals value life expectancy gains more than quality of life (QoL) gains even when they lead the same gains in QALY (a condition met by HYT). Similarly, more recent work6 demonstrates that there is no guarantee that individuals treat duration of life and QoL multiplicatively.
Authors
Anirban Basu