Measuring population health: a comparison of three generic health status measures

The relative performance of three widely used generic health status measures (EQ-5D, a modified HUI3 [mHUI3], and SF-12) was compared within a general population sample.

Data were taken from a cohort of persons identified from the patient list of a large general practice in York, UK. Two-way comparisons were made between EQ-5D and mHUI3 and EQ-5D and SF-12. The measures were assessed in terms of their practical viability, coverage, and discrimination. Practical viability was evaluated in terms of the extent of missing responses and the proportion indicating difficulty with a measure. Coverage examined the range of responses across the items in the measures. Discrimination examined the capacity of the measures to discriminate between persons according to their self-reported morbidity and socioeconomic status.

One thousand one hundred twenty-six persons completed a postal questionnaire containing EQ-5D and either mHUI3 (n = 593) or SF-12 (n = 533). Missing responses were low across all three instruments. SF-12 showed a broad distribution of responses across its items however, responses on the mHUI3 hearing, speech and dexterity dimensions and the EQ-5D self-care dimension were highly skewed, with few persons reporting problems. In terms of summary scores, mHUI3 identified more mild health states than EQ-5D. EQ-5D and mHUI3 showed slightly better discrimination than SF-12.

Despite the inherent differences in their descriptive systems and scoring functions, no one instrument performed better or worse than the other with respect to the criteria applied in this study. Some of the issues to be considered when choosing a population health measure are discussed.