Quality of Life Assessment Instruments

Quality of life instruments divide into three types that are not substitutes for each other. Generic profiles such as the 36-Item Short Form Health Survey, the RAND-36 and the World Health Organization Quality of Life questionnaires describe health across several domains and allow comparison between very different conditions. Utility indices such as the EuroQol 5-Dimension instrument and the Assessment of Quality of Life produce a single number on a scale anchored at death and full health, which is what health economic evaluation requires. Condition-specific measures such as the Dermatology Life Quality Index, the Minnesota Living with Heart Failure Questionnaire, the Kansas City Cardiomyopathy Questionnaire and the EORTC cancer questionnaires ask about the problems that particular condition actually causes.

Condition-specific instruments are usually more responsive to change within a condition. Generic instruments are the only ones that allow comparison across conditions. Many studies collect one of each for exactly this reason.

In trials, a quality of life instrument is often a secondary endpoint collected at fixed visits, and the most common practical failure is missing data rather than a poor instrument. Participants who deteriorate are the ones most likely to stop completing questionnaires, so the missing responses are not missing at random and simple complete-case analysis will flatter the result. Plan the analysis for informative missingness before the study starts.

In economic evaluation the requirement is stricter, because a utility index has to be scored with a value set. Value sets differ by country, and the same responses produce different utilities depending on which is applied. State which value set was used, and do not mix value sets within an analysis.

Check the recall period and the administration mode, and keep both fixed. Quality of life responses shift with recall window, and self-completed and interviewer-administered versions of the same instrument produce different distributions.

Check licensing carefully in this group. Several of the best known quality of life instruments are commercially licensed, registration is required for some non-commercial use, and translations are licensed and validated separately from the original. A locally produced translation is not a validated instrument.

Check that the instrument covers what the intervention is expected to change. A generic profile may show nothing at all in a condition whose burden is concentrated in a domain the profile covers with one item. That is a design problem, not a null result, and it is avoidable by checking the item content against the mechanism of the intervention before the protocol is written.

What is the difference between a generic and a condition-specific measure?

A generic measure describes health in terms that apply to anyone, so results can be compared across conditions and used in economic evaluation. A condition-specific measure asks about the problems that a particular condition causes, so it usually detects change within that condition more sensitively but cannot be compared across conditions.

Which quality of life instrument do I need for a cost-effectiveness analysis?

A preference-based utility index, because that is what quality-adjusted life years are calculated from. Profile instruments such as SF-36 do not produce utilities directly, although mapping algorithms exist. Whichever is used, the country value set applied must be stated.

How should missing quality of life data be handled?

Assume it is informative. Participants who deteriorate or withdraw are more likely to have missing responses, so complete-case analysis will overstate quality of life at later timepoints. Pre-specify a sensitivity analysis and report how much data was missing at each visit.

Can I translate a quality of life questionnaire myself?

No, not if the results are meant to be comparable. Established instruments have formal translation and cultural adaptation procedures, and most licence holders control which translations may be used. An unvalidated translation is a new instrument with unknown properties.