Measuring Health-Related Quality of Life

Preference-based community valuation of healthrelated quality of life is a bedrock of medical decision and cost-effectiveness analysis. We want to live longer and we want to live better, and we need to know how changes in our health will affect us in order to make informed medical decisions. As a society, we need measures of preference that can help us allocate resources most beneficially. Studies advancing our understanding of preferences for life/ health states and of the valuation process itself are thus a regular feature of Medical Decision Making. Five articles in this issue of the journal focus on measurement of health-related quality of life. Two of these studies 1,2 investigate properties of existing instruments for health state description (PORPUS) or quality of life measurement (WHOQOL-BREF) and provide evidence for the validity of their measurements. For PORPUS, this takes the form of a multi-attribute utility function that can map PORPUS health states into utilities that show good correspondence with standard gamble utilities, the gold standard in utility measurement. For WHOQOL-BREF, Rasch models offer new evidence that 3 of the 4 quality of life domains measured by that instrument produce scores that can be treated as interval scales, facilitating statistical comparisons between scales. Joyce and others 3 demonstrate the application of quality of life measures to better understand the impacts of interrupted and intensified antiretroviral therapy in patients with HIV. Basu and Manca 4 develop robust regression methods based on beta distributions that appear well-suited for modeling quality-of-life outcomes, overcoming their statistically difficult score distributions, which are characterized by considerable skew and spikes at “perfect” quality of life. These articles are important because they answer significant questions and, in many cases, provide new tools that will help answer many additional questions. The fifth of these papers follows suit in asking and answering an important question about health state measurement, but its most powerful contribution may be the new questions it raises. Rowen and others 5 find that whether a health state is named (e.g., “irritable bowel syndrome” or “cancer”) when it is described can—sometimes—affect time–tradeoff utilities elicited from community respondents. Although their respondents valued unlabeled health states similarly to the same states with an “irritable bowel syndrome” label, they behaved differently when states were labeled “cancer.” Specifically, more severe states received lower utility values as “cancer” than when they were unlabeled, and utilities were sensitive to whether respondents had experience with cancer (either themselves or as carers for others) when the health state was labeled “cancer.” Utility assessment tasks in the general population require people to imagine a health state and indicate their preferences among life profiles that may include the state. Shifts in preference that result from apparently irrelevant descriptive variations should lead us to be cautious in how we conduct and report valuation studies. Rowen and others report that their respondents were affected by a cancer label but not by an irritable bowel syndrome label; the finding suggests that cancer qua cancer has a special salience, which might arise from social messaging, personal experiences, or the disease descriptions presented to the respondents. Future research using different condition labels may help to disentangle these effects.