Purpose: Five indexes, EQ-5D, HUI2, HUI3, QWB-SA, SF-6D, are each used to assign community-derived utility scores to health states, although the scores they assign differ. We wish to understand their inter-relationships and to transform these indexes to a common scale.

Methods: The National Health Measurement Study collected all five health-related quality-of-life (HRQoL) indexes for all respondents in a national telephone survey of 3844 non-institutionalized U.S. adults aged 35-89. We used non-linear principal component analysis (NPCA) to test unidimensionality of measured HRQoL among the indexes. We divided each index into discrete categories using the same partition of the utility scale for all, then used hierarchical item response theory (IRT) analysis to derive a latent scale of health common to all indexes (“theta”). Robustness of the analysis was examined by repeating the IRT analysis with a different categorization of indexes by partitioning each index relative to its theoretical minimum-maximum range.

Results: Essentially the full range of possible values was observed for each index in this population survey. Intercorrelations range from 0.64 (EQ-5D, QWB-SA) to 0.89 (HUI2, HUI3). The first NPCA principal component accounts for 74% of the common variance among indexes (eigenvalue 3.7). Ceiling effects were 36%, 11%, and 11% for EQ-5D, HUI2, and HUI3, but only 4% for SF-6D and 0% for QWB-SA. Plots of index scores as a function of theta show flattening of scores near the 1.0 utility ceiling for better levels of health, although the flattening (“censoring”) starts above different points on the health continuum for different indexes. SF-6D and QWB-SA are nearly linear across the range of theta; EQ-5D, HUI2, and HUI3 are linear with steep slope from low theta (poor health) into mid-range of theta, then approximately linear with a flattened slope for higher thetas although the inflection point differs by index.

Conclusions: Simple linear functions may serve as cross-walks among these indexes for lower health states although precision is modest. Ceiling effects may make cross-walks impossible above a certain level of health. Although each index purports to measure generic health on a utility scale, these indexes are not identical or even simply related. Differential censoring at higher health levels suggests developing acceptable transformations from one to another depends on the range of health in which the transformation is applied.