Date of Award

Winter 12-2017

Embargo Period


Degree Type

Dissertation (CMU Access Only)

Degree Name

Doctor of Philosophy (PhD)


Engineering and Public Policy


Baruch Fischhoff


Societal preference-based health-related quality of life (HRQL) measures provide the numbers that researchers and policy-makers need for quantifying the value of health. This dissertation provides a normative, descriptive, and prescriptive analysis of the design of such measures. Part I analyzes the normative foundations of the preference aggregation procedures that define societal preferences, concluding that conventional procedures represent a small subset of those that would be normatively permissible given conventional assumptions. Each aggregation procedure represents different ethical principles, reflecting differences in the treatment of preference heterogeneity, and Part I presents an analytical-deliberative framework for choosing among them. Part II describes the creation of a new HRQL measure, the Patient-Reported Outcomes Measurement Information System (PROMIS®) Preference (PROPr) Scoring System. The PROPr Scoring System is a free, open-source tool that combines best practices in health profile measurement with those for creating preference-based scores. It is designed for integration with the PROMIS initiative of the National Institutes of Health, which is poised to become the U.S. standard for patient-reported outcomes. Part III characterizes the empirical relationships between exclusion criteria – meant to capture the data from preference surveys that are not true preferences – and the implications for those who choose to implement them. We show that conventions for data cleaning via exclusion criteria represent disparate mechanisms for exclusion, that exclusions impact the preference data available for analyses, and that exclusion likely affects the societal representativeness of the included preferences. We argue that there is sufficient empirical evidence to recommend only a subset of criteria, and describe procedures for implementing them in order to minimize wrongful exclusion. We outline future work to produce criteria with better classification properties and surveys that minimize the need for exclusion.