Date of Award

Fall 2012

Degree Type


Degree Name

Master of Science (MSc)


Health Research Methodology


Feng Xie


Eleanor Pullenayegum



Committee Member

Ron Goeree



The EQ-5D is one of the most widely used instruments to measure health status. It consists of a descriptive profile with a corresponding scoring algorithm. Multiple scoring algorithms have since been developed from EQ-5D preference elicitation studies.


To identify key methodological issues in the construction of EQ-5D preference elicitation studies and to assess the validity of using a standard methodology in the construction of EQ-5D scoring algorithms.

Search methods

We searched the MEDLINE, EMBASE, Cochrane Library, NHS Economic Evaluation Database, and Health Economic Evaluation Database, (1990 to 2012). The EuroQol Group website was also searched.

Selection criteria

EQ-5D preference elicitation studies that reported the directly estimated health state scores and estimated scoring algorithm.

Data collection and analysis

Two reviewers independently assessed articles for inclusion. The observed and estimated EQ-5D preference scores were compared across studies. A standard scoring algorithm with fixed variables was estimated. The model performance of the standard algorithm and the study reported algorithm were assessed and compared.


A total of 38 preference elicitation studies were included in this review. Key differences identified include: method of valuation, selection of health states, transformation of health state values, and method of estimation of the scoring algorithm. The observed health state values were found to be significantly different. The predicted health state values showed high levels of rank correlation. In general, a standard scoring algorithm was found to be no different in model performance than study specific scoring algorithms, with only three studies reporting a significant better model performance using the study specified scoring algorithm.


Methodological differences were identified across EQ-5D valuation studies. A standard scoring algorithm may yield similar model performance to study specific scoring algorithms, however further research is needed to identify when the use of a standard algorithm is appropriate.

McMaster University Library