In the modern era, cost and clinical efficacy are crucial in characterizing a treatment's net value. Recently, there has been an interest in methodology to collapse cost and effectiveness measures to concretely inform health policy decisions and promote adequate resource allocation. The cost-effectiveness acceptability (CEA) curve is one commonly used metric. Despite its widespread use, it does not allow a straightforward way to conduct inference, and its ability to characterize cost-effectiveness remains dubious given its lack of sample-size invariance. We propose a measure for population-level cost-effectiveness based on the proportion of pairs of individuals for which the treatment is cost-effective. In simple settings, this can be estimated using distribution-free methods. This approach lends itself to semiparametric regression-based approaches, moving us towards cost-effectiveness decisions based on patient subgroups that we believe may experience greater overall benefit from treatment. Through simulations and an application to endometrial cancer, we conclude that the determination curve should be considered as a clinically interpretable alternative to existing approaches.