There is a growing recognition that we cannot afford the provision of all new health care technologies, even those that are proven to be beneficial. This is increasingly true in the US, where health care spending is on an unsustainable upward trajectory. US health care spending is greatly in excess of that of other countries; however, with respect to key health metrics, the US health care system performs relatively poorly. Despite this, unlike many other developed countries economic evaluation, and more specifically cost effectiveness evidence, is used sparingly in the US health care system. Notably, the Centers for Medicare and Medicaid Services (CMS), administrators of the Medicare programme, state that cost-effectiveness evidence is not relevant to coverage decisions for medical technology and interventions evaluated as part of National Coverage Determinations (NCDs). The empirical aspect of this thesis evaluates the current use and potential value of using cost-effectiveness evidence in CMS NCDs. A database was built using data obtained from NCD decision memoranda, the medical literature, a Medicare claims database, and Medicare reimbursement information. The findings of the empirical work show that, CMS’s stated position notwithstanding, cost-effectiveness evidence has been cited or discussed in a number of coverage decisions, and there is a statistically significant difference between positive and non-coverage decisions with respect to cost effectiveness. When controlling for factors likely to have an effect on coverage decisions, the availability of cost-effectiveness evidence is a statistically significant predictor of coverage. In addition, the quality of the supporting clinical evidence, the availability of alternative interventions, and the recency of the decision are statistically significant variables. Further, when hypothetically reallocating resources in accordance with cost-effectiveness substantial gains in aggregate health are estimated. It is shown that using cost-effectiveness to guide resource allocation has an effect on resource allocation across patient populations and types of technology.