Background. The profile-based SF-12 has a low respondent burden and is used widely in clinical settings to monitor health and evaluate programs. Deriving preference scores for the SF-12 health profile would permit its use in cost-effectiveness analyses. Previous mapping studies of SF family instruments to preference-based instruments have not examined convergent validity or performance in low-income, minority populations.
Objectives. To map the SF-12 to the EuroQol (EQ-5D Index) and the Health Utilities Index Mark 3 (HUI3) in a low-income, predominantly minority sample.
Research Design. We used a cross-sectional survey data.
Subjects. We studied a convenience sample of 240 low-income, predominantly Latino and black patients attending a community health center in New York.
Measures. We used separate regressions of the EQ-5D Index and HUI3 onto the physical (PCS-12) and mental (MCS-12) components of the SF-12 scores as measures.
Results. For the EQ-5D Index regression, the adjusted variance explained was 58% (bootstrap validation 95% confidence interval [CI], 46–66). For the HUI3 regression, the adjusted variance explained was 51% (bootstrap 95% CI, 39–59). The correlation coefficient between the 2 predicted measures was 0.96. The correlation of the predicted HUI3 with the EQ-5D Index (0.73) and the predicted EQ-5D Index with the HUI3 (0.70) exceeded that between the 2 original preference-based measures themselves (0.69).
Conclusions. These pilot results suggest that the SF-12 could be successfully mapped to both the EQ-5D Index and HUI3, yielding preference-based scores that demonstrate convergent validity in a low-income, minority sample.
The SF-12 health profile yields information on health domains but cannot be used for cost-effectiveness analyses. Preference measures, like the Health Utility Index (HUI) 1 and EuroQol (EQ-5D), 2 yield interval level scores anchored at 0 (death) and 1 (perfect health) and representing preferences for particular health states, characteristics essential for cost-effectiveness analyses. 3
Researchers translating between profile and preference measures attempt to capture the advantages of both. Some have directly elicited preferences for subsets of SF-12 or SF-36 items. 4,5 Others have used regression to map SF-12 or SF-36 items to preference instruments 6,7 or questions. 8,9
Each investigation used only one preference measure, limiting assessment of convergent validity. Also unexamined is whether performance is adversely affected in lower socioeconomic persons who have been found to exhibit lower reliability on health profiles. 10 To address these limitations, we regressed 2 preference scales (EQ-5D Index and HUI Mark 3) onto the SF-12 using information on patients visiting an inner-city health center. We examined whether: 1) the SF-12 generates preference scales adequately representing the parent instruments; and 2) there is evidence of convergent validity.
From the *Department of Family and Community Medicine, Center for Health Services Research in Primary Care, University of California, Davis, Sacramento, California, 95817.
From the †Department of Community Health and Social Medicine, The City University of New York Medical School, New York, NY.
From the ‡Department of Community Health and Social Medicine, The City University of New York Medical School, New York, NY.
From the §Center for Health Services Research in Primary Care, of California, Davis, Sacramento, California.
Address correspondence and reprint requests to: Peter Franks, MD, Department of Family and Community Medicine, Center for Health Services Research in Primary Care, University of California, Davis, 4860 Y Street, Suite 2300, Sacramento, CA, 95817. E-mail: email@example.com