Level of agreement between patient-reported EQ-5D responses and EQ-5D responses mapped from the SF-12 in an injury population

Belinda J Gabbe, Emma McDermott, Pam M Simpson, Sarah Derrett, Shanthi Ameratunga, Suzanne Polinder, Ronan A Lyons, Frederick P Rivara, James E Harrison

Comparing health-related quality of life (HRQL) outcomes between studies is difficult due to the wide variety of instruments used. Comparing study outcomes and facilitating pooled data analyses requires valid "crosswalks" between HRQL instruments. Algorithms exist to map 12-item Short Form Health Survey (SF-12) responses to EQ-5D item responses and preference weights, but none have been validated in populations where disability is prevalent, such as injury. Data were extracted from the Validating and Improving injury Burden Estimates Study (Injury-VIBES) for 10,166 adult, hospitalized trauma patients, with both the three-level EQ-5D (EQ-5D-3L) and SF-12 data responses at six and 12-months postinjury. Agreement between actual (patient-reported) and estimated (mapped from SF-12) EQ-5D-3L item responses and preference weights was assessed using Kappa, Prevalence-Adjusted Bias-Adjusted Kappa statistics and Bland-Altman plots. Moderate agreement was observed for usual activities, pain/discomfort, and anxiety/depression. Agreement was substantial for mobility and self-care items. The mean differences in preference weights were -0.024 and -0.012 at six and 12 months (p < 0.001), respectively. The Bland-Altman plot limits of agreement were large compared to the range of valid preference weight values (-0.56 to 1.00). Estimated EQ-5D-3L responses under-reported disability for all items except pain/discomfort. Caution should be taken when using EQ-5D-3L responses mapped from the SF-12 to describe patient outcomes or when undertaking economic evaluation, due to the underestimation of disability associated with mapped values. The findings from this study could be used to adjust expected EQ-5D-3L preference weights when estimated from SF-12 item responses when combining data from studies that use either instrument.