PQP5: TRANSLATING SF-36 SCORES INTO PREFERENCES: AN EXAMINATION OF THE PERFORMANCE OF TWO PREDICTIVE EQUATIONS

Abstract

Despite widespread use of the SF-36, its use in cost-utility analyses has been precluded by its inability to measure patient preferences. To overcome this obstacle, various investigators have derived equations to estimate preference scores from the SF-36. OBJECTIVE: To compare two methods of estimating preference values from SF-36 scores. METHODS: A convenience sample of patients completed the SF-36 and EuroQol during their initial visit to a specialty headache center. Preference scores were estimated from the SF-36 using two equations, one developed by Fryback and the other by Brazier. The performance of each equation was assessed by calculating the correlation coefficient between the estimates and actual preference scores from the EuroQol. Mean preference scores from each method were compared using one-way repeated measures ANOVA. RESULTS: Forty-seven patients were enrolled; 45 completed the EuroQol and SF-36. Mean preferences estimated by the Brazier and Fryback equations were 0.815 (SD: 0.110, Range: 0.497–0.971) and 0.655 (SD: 0.060, Range: 0.511–0.791). The mean EuroQol score was 0.553 (SD: 0.347, Range: −0.239–1.000). The correlation coefficient between predicted preferences from Brazier and Fryback's equations and the EuroQol were 0.613 and 0.494, respectively. The mean preference scores produced by the three methods were significantly different (F = 24.59, p = 0.0001) and both of the predictive equations yielded significantly higher mean values (p < 0.05) than the EuroQol. CONCLUSIONS: Both equations to estimate a single, preference-based index from the SF-36 produced higher mean scores and a narrower range of values than the EuroQol. Preferences produced by the Brazier equation were more highly correlated with EuroQol scores than preferences from the Fryback equation. The Fryback equation also produced a wider range of scores. Continued development and validation of predictive equations is needed in order to use SF-36 scores in cost-utility analyses.