Improving Single-Item Generic Health Survey Measures

October 20, 2017 | 24th Annual Conference of the International Society for Quality of Life Research (ISOQOL), Philadelphia, Pennsylvania

Because all surveys begin with the first item, at the core of advances in patient-reported outcome measurement are better single-item measures of the most frequently-measured health domains. For many outcomes monitoring and cost prediction purposes in the future, the shortest health survey forms will administer only the best single item for each domain. Also, because the first item determines what happens next in adaptive surveys, more efficient single-item measurement is crucial.

At the 24th Annual ISOQOL Conference, Drs. Ware and Gandek summarized advances in understanding of the content that best represents health domains and the implications of better operational definitions for capturing the essence of each domain, focusing on the Physical Function, Vitality and Mental Health domains. Survey items were self-administered on the Internet to representative samples of U.S. adults (N=2,938), ages 18-94. Results for new General Quality of Life (QGEN®) 10-item form items were compared with SF-36 and PROMIS-29 items for the same domains and in terms of how well they predicted SF-36 physical and mental component summary measures, to identify the reasons for superior item performance. Multiple psychometric criteria and tests of convergent and discriminant validity were reported.

They reported that the best single-item measures were not found within the Medical Outcomes Study SF-36 or PROMIS-29 multiple-item scales or global items from those projects. Significant improvements in QGEN single-item performance, in comparison with SF-36 and PROMIS-29 for the same domains, were linked to:

• Broader representation of descriptive content within a single item
• Item response categories that measure more people over a wider range and reduce ceiling effects
  
• Single-item “tip of the iceberg” measurement of a higher-order concept (e.g., vitality instead of fatigue) with matching response categories
  
• Better matching between the essence of a domain and a specific operational definition (e.g., how easy-hard it is to do physical activities versus whether physical health is excellent-poor).

The authors recommended further testing of improved QGEN-10 items because they are likely to improve both psychometric and utility single-item measurement for all 10 health domains.

An abstract of the presentation is available at: Ware JE and Gandek B. Improving single-item generic health survey measures. Quality of Life Research 2017; 26 (Suppl. 1):43-44.