PNS237 TIME TRADE-OFF TARIFFS BASED ON SF-12 QUESTIONNAIRE

Abstract

EQ-5D and SF-6D are among commonly used preference based health utility measures; however, the same individual often scores different values for these indexes. Besides being generated from different questionnaires, these utility indexes differ by the methods that the preference-based weights are elicited. EQ-5D weights are based on time trade-off and SF-6D weights are based on the standard gamble method. The objective of this research is to develop the time trade-off tariffs based on SF-12 questionnaire. The Medical Expenditure Panel Survey (MEPS), a US based general population survey was used in this analysis. The 2003 MEPS sample has 19,678 adults with valid EQ-5D and SF-12 version 2 questionnaire responses. Both questionnaires were administered via a paper and pencil self-administered questionnaire to adults age 18 and older. US tariffs-based EQ-5D index was regressed on individual level responses of SF-12 questionnaire and socio-demographic variables. OLS, CLAD and modified maximum likelihood Tobit models were used. Restricted regression was use to limit the constant of the regression to 1, which allows reflect the actual construction of EQ-5D index. The SF-12 questionnaire variables explain 78% (R-square from OLS = 0.7840) of variations in EQ-5D. The remaining socio-demographic variables add negligent (<1%) explanatory power to the model. In most cases, the developed tariffs have consistent signs and magnitudes. Response to questions “pain limits normal work” and “felt downhearted/depressed” had the biggest negative tariffs. -0.2018 (p=0.00) and -0.1195 (p=0.00) for “pain limits normal work” extremely and quite a bit responses respectively, and -0.1297 (p=0.00) for “felt downhearted/depressed” all of the time response. This study provided time trade-off tariffs for each domain of SF-12 questionnaire. These tariffs can be used in cost-effectiveness analysis in situations when time trade-off preferences are preferred to standard gamble preferences.