EQ-5D Health States Research Articles

Quality-of-life (QoL) measures in advanced non-small cell lung cancer (NSCLC) patients and their association with treatment choices across 5 European countries: Preliminary data from the ACTION study

8081 Background: ACTION (Assessment of Cost and ouTcomes of chemotherapy In an Observational setting in patients with advanced NSCLC) is an on-going, pan-European, observational study in routine clinical practice. This report summarizes baseline observer Lung Cancer Symptom Scale (LCSS) and EQ-5D health status scores among countries and according to treatment choice. Methods: Between 4/2003 and 6/2004, 993 patients (pts) were recruited. Pts were ≥18 years old and receiving first-line chemotherapy for IIIB/IV NSCLC. Baseline data were available in 967 pts. Results: Median age was 64 years (range 31–90), 71% were male. WHO PS was 0 in 21%, 1 in 65%, 2 in 10%, and 3/4 in 4%. 35% of patients were stage IIIB. There were 0, 1, 2, 3, 4, or >4 metastatic organs involved in 27%, 47%, 18%, 6%, 2%, and 0.2% of pts, respectively. Surgery and palliative or radical radiotherapy had been performed in 12%, 12%, and 1% of pts, respectively. The mean baseline LCSS, EQ-5D health status and VAS scores were calculated. Conclu...

Patient reported outcomes in a phase 2 trial of SU11248 for metastatic renal cell carcinoma

8167 Background: SU11248 is a novel, oral multi-targeted tyrosine kinase inhibitor with both anti-angiogenic and anti-tumor activity. In a phase 2 study (014) of SU11248 for the treatment of metastatic renal cell carcinoma following failure of 1 prior cytokine treatment, a 40% tumor response rate (25/63) has been reported (Motzer et al., ASCO 2005, submitted). This analysis describes the patient reported outcomes (PROs) in this trial. Methods: PROs were assessed using the FACIT-Fatigue scale (FACIT-Fatigue, score range = 0–52) and the EQ-5D health status visual analogue scale (EQ-VAS, score range = 0–100) in the first four 6-week cycles (4 weeks on treatment, 2 weeks off), weekly for FACIT-Fatigue and at the start and end of each 4-week treatment period for EQ-VAS. Of 63 patients enrolled, 62 participated in the baseline PRO assessment and 37 (59%) provided complete or near-complete data across four 6-week cycles. Results: At baseline, FACIT-Fatigue and EQ-VAS scores of these patients were slightly lower than those of the general population. Baseline FACIT-Fatigue and EQ-VAS scores were associated with baseline performance status (FACIT-Fatigue p<0.001; EQ-VAS, p=0.001) and, marginally, with subsequent response to therapy (partial response/stable disease was associated with less fatigue or better health state; FACIT-Fatigue p=0.169; EQ-VAS p=0.033). FACIT-Fatigue and EQ-VAS scores declined across these four cycles (average change from baseline: FACIT-Fatigue = -2.2, EQ-VAS = -5.87) but these changes are smaller than what is considered ‘minimally important’ for these measures. A pattern of decline and recovery coincided with the 4-week on/2-week off periods in each cycle. Average change from baseline in FACIT-Fatigue and EQ-VAS scores did not differ with response to therapy. Conclusions: SU11248 did not cause significant changes in PROs in this group of patients, although FACIT-Fatigue and EQ-VAS scores declined slightly while patients were on treatment and improved while they were off treatment in the first four cycles. These changes were not correlated with tumor response. Author Disclosure Employment or Leadership Consultant or Advisory Role Stock Ownership Honoraria Research Funding Expert Testimony Other Remuneration Pfizer Pfizer Pfizer

EQ-5D in a general population survey – A description of the most commonly reported EQ-5D health states using the SF-36

The importance of studying health-related quality of life in the general population has increasingly been emphasized. From a public health perspective, this benefits the identification of population inequalities in health status. One of the currently most popular instruments is the EQ-5D. Evaluations of the EQ-5D generally focus on the overall preference-based index. As this index has a built-in value, exploration of the information from the underlying health states is also important. In this study, the ten most commonly reported EQ-5D health states are described using the SF-36. Data collected in 1999 by questionnaires mailed to a random sample aged 20-74 in south-eastern Sweden were used (n = 9489). Almost 43% reported the best possible EQ-5D health state and 78% were accounted for by three EQ-5D health states. The EQ-5D health state classification was largely reflected by the SF-36, with the EQ-5D items mobility, usual activities, pain/discomfort and anxiety/depression tapping most clearly on the SF-36 scales physical functioning, role limitations due to physical health problems, bodily pain, and mental health, respectively. However, within the same level of EQ-5D (i.e., moderate problems) there was a rather large variation of SF-36 scale scores, particularly regarding the EQ-5D item pain/discomfort and the SF-36 scale BP.

Valuations of EQ-5D Health States

We sought to compare directly elicited valuations for EQ-5D health states between the US and UK general adult populations. We analyzed data from 2 EQ-5D valuation studies where, using similar time trade-off protocols, values for 42 common health states were elicited from representative samples of the US and UK general adult populations. First, US and UK population mean valuations were estimated and compared for each health state. Second, random-effect models were used to compare the US and UK valuations while adjusting for known predictors of EQ-5D valuations (ie, age, sex, health state descriptors) and to investigate whether and how the valuations differ. Population mean valuations of the 42 health states ranged from -0.38 to 0.88 for the United States and from -0.54 to 0.88 for the United Kingdom, with the US mean scores being numerically higher than the UK for 39 health states (mean difference: 0.11; range: -0.01 to 0.25). After adjusting for the main effects of known predictors, the average difference in valuations was 0.10 (P < 0.001). The magnitude of the difference in the US and UK valuations was not constant across EQ-5D health states; greater differences in valuations were present in health states characterized by extreme problems. Meaningful differences exist in directly elicited TTO valuations of EQ-5D health states between the US and UK general populations. Therefore, EQ-5D index scores generated using valuations from the US general population should be used for studies aiming to reflect health state preferences of the US general public.

US Valuation of the EQ-5D Health States

The EQ-5D is a brief, multiattribute, preference-based health status measure. This article describes the development of a statistical model for generating US population-based EQ-5D preference weights. A multistage probability sample was selected from the US adult civilian noninstitutional population. Respondents valued 13 of 243 EQ-5D health states using the time trade-off (TTO) method. Data for 12 states were used in econometric modeling. The TTO valuations were linearly transformed to lie on the interval [-1, 1]. Methods were investigated to account for interaction effects caused by having problems in multiple EQ-5D dimensions. Several alternative model specifications (eg, pooled least squares, random effects) also were considered. A modified split-sample approach was used to evaluate the predictive accuracy of the models. All statistical analyses took into account the clustering and disproportionate selection probabilities inherent in our sampling design. Our D1 model for the EQ-5D included ordinal terms to capture the effect of departures from perfect health as well as interaction effects. A random effects specification of the D1 model yielded a good fit for the observed TTO data, with an overall R of 0.38, a mean absolute error of 0.025, and 7 prediction errors exceeding 0.05 in absolute magnitude. The D1 model best predicts the values for observed health states. The resulting preference weight estimates represent a significant enhancement of the EQ-5D's utility for health status assessment and economic analysis in the US.

Proxy completion of EQ-5D in patients with dementia

There are measurement difficulties associated with the assessment of health-related quality of life (HRQL) in older people with dementia. The use of proxies is a commonly employed approach to overcome such problems. The research reported in this paper sought to identify, specifically for the EuroQol EQ-5D HRQL instrument, whether construct validity is greater for 'family caregivers' or 'clinicians' as two alternative sources of proxy information for patients with a diagnosis of dementia. This involved the exploration of the strength of the associations between clinical measures of illness severity and EQ-5D data. The data appear to reveal a pattern suggesting that the viewpoint of the proxy (i.e., clinician or family caregiver) is important. The findings suggest that the data provided by clinicians (when compared to data from carers) had higher construct validity for the more observable dimensions of the EQ-5D instrument (i.e., 'mobility' and 'self-care'). Conversely, the data from family carers had higher construct validity for the less observable dimensions (i.e., 'usual activities' and 'anxiety/depression'). Previous research on proxy provision of HRQL data has tended to focus on trying to identify a single proxy. The results of this study suggest that using carefully matched sets of measures and assessment perspectives may produce more valid EQ-5D health state descriptions.

Self-Valuation and Societal Valuations of Health State Differ With Disease Severity in Chronic and Disabling Conditions

We sought to determine the relationship between self-reported ratings of health-related quality of life (HRQoL) by people with 3 chronic and disabling conditions and population estimates of those health states and to model factors that might explain the disagreement between these 2 ratings. A cross-sectional postal survey was sent in which each participant completed a set of questionnaires addressing HRQoL. Data from self-valuation on a visual analog scale (VAS) was compared with a population-based VAS-equivalent valuation, using the EQ-5D instrument. Different ways of scaling the 2 VAS scores were also explored. Data were analyzed using descriptive statistics and analysis of covariance. A community sample of 1036 people took part in the study (rheumatoid arthritis n = 142, stroke n = 585, multiple sclerosis n = 309). MEASUREMENT INSTRUMENT: The EQ-5D health state profile and accompanying visual analog scale were used. Self-ratings were significantly different than the population-based ratings, and agreement was poor, both on the original scale of the data and by analyses of rescaled data. On the original scales the mean difference was 0.13 (95% confidence interval 0.117-0.143). Diagnosis, health state severity, and its square explained 35.3% of the variation in the differences between self and population ratings with a curvilinear relationship suggesting that the differences increased as the health state worsened, but at a decreasing rate as health state severity increased. This study provides evidence that EQ-5D population valuation estimates of treatment benefit for people with disabling and chronic conditions may well be inaccurate representations of the degree of change actually experienced by the individual with the condition. The varying magnitude of difference between the 2 forms of valuation has important implications for interpreting shifts in health status valuation following interventions for these populations.

Re-scaling social preference data: implications for modelling

As applied in cost-utility analysis, generic health status indexes require that full health and dead are valued as 1 and 0, respectively. When social preference weights for health states are obtained using a visual analogue scale (VAS), their raw scores often lie on a scale with different endpoints (such as "best" and "worst" health). Re-scaling individual raw scores to a 0-1 scale leads to the exclusion of respondents who fail to value dead or full health. This study examined alternative approaches that do not impose such strict exclusion criteria. The impact of a different timing of re-scaling (before or after aggregation) and a different measure of central tendency (median or mean) is measured. Data from a postal valuation survey (n=722) conducted in Belgium are used. The following models are considered: (a) re-scaling values for EQ-5D health states on a within-respondent basis and using mean re-scaled values as proxies for social preference values, (b) using median re-scaled values as proxies for social preference values, (c) computing the median raw VAS values and then re-scale, and (e) re-scaling mean raw VAS values. Exclusion rates, health state rankings and valuations and incremental value differences between pairs of states are computed for each model. Models that use a different timing of re-scaling, are compared ceteris paribus to evaluate the importance of timing of re-scaling and models that use a different measure of central tendency are compared ceteris paribus to evaluate the importance of the measure of central tendency. The exclusion rates are above 20% in the models that re-scale valuations before aggregation and less than 5% in the models that re-scale after aggregation. Health state valuations are found to be different in all two by two comparisons. Although in some comparisons the incremental values are statistically significantly different between models, they are never clinically significantly different. Differences in health state rankings were larger between the models that use a different measure of central tendency than between the models that re-scale at a different time. This study shows that, for the data sample used, the choice of the measure of central tendency is more important for the social health status preference values than the timing of re-scaling. Using median original valuations in the analysis of EQ-5D valuation data is theoretically appealing because it builds on the "median voter" model and has the advantage of a lower respondent exclusion rate. Further analysis, on other data samples, will have to confirm the current findings.

Understanding health state valuations: a qualitative analysis of respondents' comments.

Self-completed questionnaires (usually distributed by post) and visual analogue scales (VAS) are common means for collecting valuations for hypothetical EQ-5D health states. Although opportunities for respondents to comment on aspects of the exercise are often included, these data have rarely been the focus of analyses. This paper, therefore, reports on solicited and unsolicited written comments received in a New Zealand survey in 1999 to which 1360 people responded of whom approximately 50% made comments. The comments were systematically analysed via an inductive process that allowed principal themes to emerge with respect to understanding the peculiarities of respondents' valuations, particularly common 'data problems', and their perceptions as to the adequacy or otherwise of the EQ-5D classification system. From our findings we conclude that the valuation exercise imposes a substantial cognitive burden on respondents and many do not understand it (for a variety of reasons documented in the paper). Also, although there is some evidence for expanding the EQ-5D (generic) health-related quality of life dimensions, more research as to whether this holds for a larger sample and across cultural settings is needed. We offer recommendations for future research into understanding respondents' cognitive processes and possible revisions to the design and administration of the EQ-5D.

The validity and reliability of EQ-5D health state valuations in a survey of Măori.

This note reports on a 2000 study of the content validity of the EQ-5D's representation of health for 66 Măori (New Zealand's indigenous people, comprising 14.5% of the population) accessed through cultural networks. Also examined was the construct validity of the health state valuation instrument and its test-retest reliability based on repeated valuations for the two extreme health states. The possibility that the EQ-SD fails to capture what Măori regard as 'health' derives from the so-called 'Măori health model' that augments biological health with mental, spiritual and family well-being. Seventy six percent of respondents considered the EQ-5D's representation of health to be adequate. This proportion is not statistically significantly different from the rates for non-Măori and Măori respectively in an earlier study and might suggest the EQ-5D has content validity for Măori. However, the high prevalence of missing valuations, particularly for dead, and logical inconsistencies suggests that the health state valuation instrument lacks construct validity, although there is evidence of test retest reliability.

A single European currency for EQ-5D health states

The EQ-5D questionnaire is a widely used generic instrument for describing and valuing health that was developed by the EuroQol Group. A primary objective of the EuroQol Group is the investigation of values for health states in the general population in different countries. As part of the EuroQol enterprise 11 population surveys were carried out in six Western European countries (Finland, Germany, The Netherlands, Spain, Sweden and the UK) to value health states as defined by the EQ-5D using a standardised visual analogue scale (EQ-5D VAS). This contribution reports how a European set of general population preference weights was derived from the data collected in the 11 valuation studies. The scores from this set of preference weights can be applied to generate a VAS-based weighted health status index for all the potential 243 EQ-5D health states for use in multi-national studies. To estimate the preference weights a multi-level regression analysis was performed on 82,910 valuations of 44 EQ-5D health states elicited from 6,870 respondents. Stable and plausible solutions were found for the model parameters. The R(2) value was 75%. The analysis showed that the major source of variance, apart from 'random error', was variance between individuals (28.3% of the total residual variance). These results suggest that VAS values for EQ-5D health states in six Western European countries can be described by a common model.

Prioritizing patients for elective surgery: a prospective study of clinical priority assessment criteria in New Zealand.

Many hospitals in New Zealand have been using clinical priority assessment criteria (CPAC) to select and prioritize patients for access to publicly funded elective surgery. CPAC usually consist of clinical, patient-experienced, and social measures. The objective of this study was to determine how robust patient rankings were and the extent to which the patients selected were those who benefited the most from surgery. Patients prioritized for cataract (n = 101), prostate (n = 103), and hip or knee joint replacement (n = 137) surgery according to CPAC were assessed using the EQ-5D, SF-12, and condition-related patient-experienced health status measures before and after treatment. Correlations between the rankings of patients on the CPACs and the alternative instruments were explored. For each surgery group, the CPAC ranking of patients was not strongly correlated with rankings obtained using their before-treatment EQ-5D (valued) profiles or the SF-12, although there was some correlation with rankings according to the condition-related measures. Improvements in the health status of patients who were operated on, as measured by the change in their EQ-5D values, were poorly correlated with equivalent changes on the SF-12 and condition-related measures. Patients' baseline health status according to the CPAC, the EQ-5D, and the SF-12 patient-experienced measures was only slightly related to the magnitude of benefit following surgery. The strongest predictors of improvement in health status were the baseline condition-related measures. The current method of prioritizing patients in New Zealand requires reconsideration, although a gold standard method for prioritization is not immediately apparent from these results.

Understanding differences between self-ratings and population ratings for health in the EuroQOL.

To examine the source and magnitude of differences between self-ratings for health and ratings of corresponding health state profiles by the general population in the EuroQOL. EuroQOL data were analysed from the 1993 measurement and valuation of health study (MVH), a sample of 2997 members of the UK adult population, nationally representative by age, gender and social class. Multivariate regression analyses were used to examine the source of differences in visual analogue scores (VAS) between self-ratings and general population ratings. The source of observed differences were investigated with respect to four hypothesized factors: (1) Socio-demographics (age, gender, education, social class); (2) The level of respondent difficulty in completing the rating task; (3) Values for particular EQ-5D health profile attributes; and (4) Differences in the scope of health attributes and levels considered in the rating task (e.g., self-ratings may reflect preferences for attributes not captured by EQ-5D profiles). Overall, mildly ill individuals provided lower self-ratings (3-4 points), and moderately ill individuals higher self-ratings (7 points), than ratings for these states provided by the general population. Socio-demographic characteristics and difficulties in rating task completion did not explain differences between self and general population VAS ratings, contributing differences of 1 point or less in all 15 rating comparisons examined. Rating differences related more closely to a lack of correspondence between health state descriptions and self-raters' actual health experiences (differences in scope) than differences in values for health profile attributes between self-raters and the general population. EQ-5D health state descriptions may be too sparse to comprehensively describe certain health states. Adding new health state levels or dimensions, or changing the nature and tone of health state descriptions, may be useful steps for improvement.

Cost-utility analysis of vagus nerve stimulators for adults with medically refractory epilepsy

Introduction: The cost-utility of vagus nerve stimulator (VNS) devices for medically refractory epilepsy has yet to be estimated. Methods: Using a meta-analysis of randomised controlled trials of VNS, we estimate that six people require implantation in order for one person to experience a 50% reduction in seizure frequency. Costs averted from improved epilepsy control were ascertained from published literature. Values for health states were obtained from a series of 42 seizure clinic attenders using time trade-off techniques and the EQ-5D health status instrument. The cost per quality adjusted life year gained was estimated and the values obtained were tested in a sensitivity analysis. Results: Improved epilepsy control averted, on average, £745 health care costs per annum. People with epilepsy had great difficulty performing the time trade-off experiment, but those who managed to complete the task valued a 50% reduction in their own seizure frequency at 0.285 units. For a programme of six implants, the baseline model estimated the cost per quality adjusted life year gained at £28 849. The most favourable estimate was equal to £4785 per quality adjusted life year gained, assuming that the number needed to treat was similar to published series in which one response was obtained for every three implants. The least favourable estimate was equal to £63 000 per quality adjusted life year gained, when EQ-5D utility values were used. The cost per quality adjusted life year gained was not sensitive to changes in length of stay, nor complication rates, but was significantly influenced by cost of device and device battery life expectancy. Conclusion: There is not a strong economic argument against a programme of VNS implantation, although care should be taken to try and identify and treat those most likely to benefit.

Quantifying the harmful effect of psoriasis on health-related quality of life

Background: Psoriasis affects 7 million people in the United States, causing substantial cost, social stigma, and disability. Objective: The purpose of this study was to evaluate the health effects of skin disease by comparing psoriasis to other primary medical disorders using 3 different scales of health-related quality of life. Methods: A self-administered questionnaire consisting of 3 health-related quality of life measures was given sequentially to 35 eligible patients with psoriasis presenting to the Dermatology Branch of the National Cancer Institute (NCI) for an investigational therapeutic protocol. Results: All patients (100%) agreed to participate. The median Psoriasis Area and Severity Index (PASI) score was 13.0. Overall, 82.9% at least often felt the need to hide their psoriasis, and 74.3% claimed their self-confidence was at least often affected by their psoriasis. The median EQ-5D health state utility score was 13.0% less than healthy individuals (P < .001). On the SF-36, the mean general health score was 13.2% less (P = .005) and the median social functioning score 18.7% less (P = .005) than that of patients with no chronic conditions. Conclusion: Individuals with psoriasis are significantly affected in their health state utility, perception of general health, and social functioning when compared with individuals without chronic disease and those with certain primary medical conditions. (J Am Acad Dermatol 2002;47:512-8.)

Modelling valuations for Eq-5d health states: an alternative model using differences in valuations.

The EQ-5D is a preference-based measure of health and is increasingly being used in the evaluation of health technologies. A 'tariff' of values for all 243 EQ-5D health states has been generated using direct valuations on a subset of these states. The tariff is used to express the value of differences between health states, and so this paper explores whether a tariff with better predictive ability can be calculated using differences between values rather than using the values themselves. The original tariff (reported in this journal) was based on valuations for 42 EQ-5D states elicited from a representative sample of 2997 members of the UK general population using the time trade-off method. This same data are used to estimate a tariff based upon the differences in value between the worst possible state (33333) and all other states. A simple model that fits the data well is one in which the differences in value between 33333 and all other states are explained in terms of the change in each dimension plus a term to pick up whether some dimensions change by the maximum amount whereas others do not change at all. The mean absolute difference between the actual values and those predicted by this model is 0.03 (compared with 0.039 in the original model). The model presented in this paper predicts the difference between 33333 and all other states remarkably well and can be used to generate a tariff for all EQ-5D health states. In fact, this model more accurately predicts the values of states for which there are direct observations, and so we recommend its comparison with the original model in evaluative studies.

QL2: LOGICAL CONSISTENCY AND THE VALUATION OF HEALTH: AN ANALYSIS OF US SURVEY DATA

OBJECTIVES: It is widely held that values of the general public should be used in the evaluation of health care. Surveys designed to record such values involve the participation of individuals with different health experiences and with different socioeconomic backgrounds. The technical performance of these participants is likely to vary as a function of these factors, for example the logical consistency of responses is often associated with socioeconomic status. This paper examines the relationship between logical consistency and respondent health using US survey data designed to capture values for states defined by the EQ-5D classification. METHODS: A standardised questionnaire was used to elicit valuations for EQ-5D health states in a postal survey conducted by Johnson et al (1998, Pharmacoeconomics) in Arizona in which US respondents (N = 905) rated eight states along a visual analog scale from best to worst imaginable health. A logical ordering is defined for 23 unique pairs of states in that one state dominates the other over all 5 dimension of the EQ-5D. A logical inconsistency was noted when a respondent assigned a lower value to the “better” state in such a pair. Censored regression models were used to assess the relationship between consistency and respondent health. We tested the robustness of these findings using survey data from Wisconsin, which applied the same questionnaire (N = 222). RESULTS: From the best imaginable state, each 20-point decrement in respondent's self-rated health status yielded significantly greater inconsistency in their valuation of EQ-5D health states controlling for age and sex. Inclusion of education and income reduced this effect slightly, yet it remained statistically significant. CONCLUSIONS: Respondents in poor health demonstrate greater difficult in valuing health states in a logically consistent manner. Censoring survey data to remove inconsistent respondents may violate the principle of using representative population values in evaluating cost-effectiveness of health care.

QL1: DOES SOCIOECONOMIC STATUS AFFECT THE VALUATION OF HEALTH?

OBJECTIVE: The issue of whose values count in the evaluation of health interventions is central to decision-making in all health care systems. Within the health services research community there is a degree of consensus that population-based preference weights should be used as the quality-adjustment factor in determining the value of health outcomes. However, previous research indicates an inverse, graded relationship between socioeconomic status (SES) and an individual's own health. If SES is also found to influence the valuation of hypothetical health states, then this could prove to have significant consequences for the evaluation of treatment. METHODS: Values for hypothetical health states defined by EQ-5D (a generic measure of health-related quality of life) were collected from a representative survey of the UK general population. 2,997 individuals used time trade-off (TTO) methods to value these EQ-5D health states. Information on each respondent included age, gender, social class and educational attainment. TTO values were bounded and non-normally distributed necessitating methods such as ordered logistic regression in addition to OLS to analyze these data. RESULTS: Education and social class as proxies for SES were significant predictors of the mean values for hypothetical health states. Their influence on health state valuation appears to act through their interaction with the mobility and self-care dimensions of the EQ-5D. This relationship persists after adjustment for respondent demographic characteristics. CONCLUSIONS: Valuation of hypothetical health states appears related to SES. This has implications for cost-effectiveness analysis since valuations from one population with a particular SES distribution may not be applicable for health policy and medical decision-making in other populations. The relationships between SES and health state valuation merit further investigation, in particular to examine the impact of non-health consequences such as income on values.

EQ-5D in patients with dementia: an investigation of inter-rater agreement.

There are difficulties in obtaining health-related quality of life (HRQL) data from patients with dementia due to variation in their cognitive ability, degree of insight and capacity to make judgments. The use of proxies is one solution. To examine the inter-rater agreement of patient and proxy completion of the EuroQol EQ-5D instrument (EQ-5D). The EQ-5D instrument was completed separately by patients, their caregivers and a physician. Assessment of inter-rater agreement involved comparison of self-completed (patient) and proxy-completed (caregiver and physician) responses for each dimension of EQ-5D, using a weighted kappa score. Three key hypotheses were tested. (1) Interrater agreement would be stronger between patient and caregiver than between patient and physician. (2) Interrater agreement would be stronger on the 'observable' and objective dimensions of EQ-5D. (3) Interrater agreement between patient and proxies would be stronger for patients with earlier dementia. The sample comprised 64 patients with a range of dementia severity. The EQ-5D health state classification system and visual analogue scale were used to assess HRQL. Global severity of dementia was determined using the Clinical Dementia Rating Scale. The principal finding of this study was that responses to EQ-5D questions were highly variable across the three raters such that none of the three hypotheses were strongly supported. The data provide some support for the use of EQ-5D when interviewer administered. However, there are serious concerns regarding the validity of patient self-rated HRQL data obtained in this study and uncertainty exists regarding who the appropriate proxy should be, as different groups of proxies provide different results. It was not clear whether caregivers or physicians represent better proxies. Further research should focus on the comparison of caregivers and physicians as proxies.

"Death" and the valuation of health-related quality of life.

Despite evidence to the contrary, a common assumption in the area of health status measurement is that the state "dead" is the worst possible health state and by definition should be assigned a value of 0. However, the value of the state "dead" and the notion of states worse than "dead" have never been fully addressed as a research topic. This article demonstrates the extent of the variation in the value given to the state "dead" by individuals and the effects of transformation on individual and aggregate values using data elicited with 2 methods (visual analog scale rating and ranking) that place no constraint on the value given to the state "dead." Face-to-face interviews were conducted with 253 adults in North Yorkshire, UK, in 1998. Each participant performed ranking and visual analog scale rating exercises for 19 EuroQol EQ-5D health states. Data showed that there is a small group of individuals who, when given the option, choose to place relatively high value on the state "dead" compared with other health states. This did not appear to be due to artifact. Evidence also suggested that the usual assumptions underlying the transformation of health state values, for which the distance between full health and "dead" is used to define the denominator, may not hold for these individuals and may distort aggregate preference structures. The authors stress the need for more systematic inquiry in this field.