A head-to-head comparison of EQ-5D-5L and SF-6D in Dutch patients with fractures visiting a Fracture Liaison Service

Abstract

PLAIN LANGUAGE SUMMARY The EQ-5D and SF-36 as generic multi-domain questionnaires are widely used to measure the health-related quality-of-life (HRQoL) in a sample of the persons who suffer from the diseases or the general population. Their responses could be converted to patients or societal Health State Utility Values (HSUVs) with the range of 0 ("death") to 1 ("full health"). A specific application of HSUV is to calculate quality-adjusted life years as the indicator of effectiveness to evaluate whether the cost of a new intervention is justified in terms of health gains through cost-utility analysis in health economics, the evidence can be further used to inform decision-making. However, different instruments differ in construct and valuation, potentially leading to different estimates for the person's same "health state", and healthcare decisions could be compromised when researchers or decision-makers are not aware of potential differences in HSUV. Therefore, it is important to gain insight into the specific psychometric properties of these instruments, and to understand whether instruments are interchangeable. Our study is based on data from a Dutch Fracture Liaison Service (FLS is a program for secondary fracture prevention), compared the psychometric properties and interchangeability of two instruments (EQ-5D-5L and SF-6D) in patients with a recent fracture presenting at the FLS, and suggested both instruments are valid in utility elicitation in our target population. However, they cannot be used interchangeably given only moderate agreement and differences in utilities. Neither instrument was found to be clearly superior given comparable construct and longitudinal validity, but different instruments values in different aspects of HRQoL assessment. Aims This study compared the psychometric properties of EQ-5D-5L and SF-6D to assess the interchangeability of both instruments in patients with a recent fracture presenting at a Fracture Liaison Service (FLS). Materials and methods Data from a prospective observational study in a Dutch FLS clinic were used. Over 3 years, subjects were interviewed at several time points using EQ-5D-5L and SF-36. Floor and ceiling effects were evaluated. Agreement was evaluated by intra-class correlation coefficients and visualized in Bland-Altman plots. Spearman's rank correlation coefficients were applied to assess convergent validity. Mann-Whitney U test or Kruskal-Wallis H test as well as effect size (ES) were used to explore known-groups validity. Responsiveness was explored using standardized response mean (SRM) and ES. For each measurement property, hypotheses on direction and magnitude of effects were formulated. Results A total of 499 patients were included. EQ-5D-5L had a considerable ceiling effect in comparison to SF-6D (21 vs. 1.2%). Moderate agreement between the (UK and Dutch) EQ-5D-5L and SF-6D was identified with intra-class correlation coefficients of 0.625 and 0.654, respectively. Bland-Altman plots revealed proportional bias as the differences in utilities between two instruments were highly dependent on the health states. High correlation between instruments was found (UK: rho = 0.758; Dutch: rho = 0.763). EQ-5D-5L and SF-6D utilities showed high correlation with physical component score but low correlation with mental component score of SF-36. Both instruments showed moderate discrimination (ES > 0.5) for subgroup by baseline fracture type, and moderate responsiveness (SRM > 0.5) in patients that sustained a subsequent fracture. Conclusion Both EQ-5D-5L and SF-6D appeared to be valid utility instruments in patients with fractures attending the FLS. However, they cannot be used interchangeably given only moderate agreement was identified, and differences in utilities and ceiling effect were revealed. Comparable construct validity and responsiveness were indicated, and neither instrument was found to be clearly superior.