Health State Utilities: A Framework for Studying the Gap Between the Imagined and the Real

Volume 11 • Number 1 • 2008 VA L U E I N H E A LT H

Health State Utilities: A Framework for Studying the Gap Between the Imagined and the Real Anne M. Stiggelbout, PhD, Elsbeth de Vogel-Voogt, PhD Department of Medical Decision Making, Leiden University Medical Center, Leiden, The Netherlands

A B S T R AC T

Objectives: Health state utilities play an important role in decision analysis and cost-utility analysis. The question whose utilities to use at various levels of health-care decision-making has been subject of considerable debate. The observation that patients often value their own health, but also other health states, higher than members of the general public raises the question what underlies such differences? Is it an artifact of the valuation methods? Is it adaptation versus poor anticipated adaptation? This article describes a framework for the understanding and study of potential mechanisms that play a role in health state valuation. It aims at connecting research from within different fields so that cross-fertilization of ideas may occur. Methods: The framework is based on stimulus response models from social judgment theory. For each phase, from stimulus, through information interpretation and integration, to judgment, and, finally, to response, we provide evidence of

factors and processes that may lead to different utilities in patients and healthy subjects. Results: Examples of factors and processes described are the lack of scope of scenarios in the stimulus phase, and appraisal processes and framing effects in the information interpretation phase. Factors that play a role in the judgment phase are, for example, heuristics and biases, adaptation, and comparison processes. Some mechanisms related to the response phase are end aversion bias, probability distortion, and noncompensatory decision-making. Conclusions: The framework serves to explain many of the differences in valuations between respondent groups. We discuss some of the findings as they relate to the field of response shift research. We propose issues for discussion in the field, and suggestions for improvement of the process of utility assessment. Keywords: adaptation, outcome, quality of life, utility assessment.

Introduction

from patients. In cost-utility analyses from a societal perspective, for macro-level decision-making, Gold et al. [3] have recommended the use of society’s preferences, and utilities thus should be assessed from a representative sample of fully informed members of the general public. In guideline development, the mesolevel, and in individual patient decision-making, the microlevel, the use of utilities obtained from actual patients is preferred [3]. Gold et al. have qualified these recommendations, basing their reasoning and their recommendations on the assumption that health state utilities will in general differ between patients and the general public. They argue that it may be difficult to fully inform members of the general public of what a state is truly like. In many instances, members of the public who are asked to imagine experiencing health states assign lower utilities to those states than do patients who are actually experiencing these states. Also (and possibly related to this finding), patients often assign lower utilities to hypothetical treatment scenarios than to the actual treatment states that are reflected in those scenarios (see [4] for a distinction between the various designs used to evaluate the

Health state utilities play an important role in healthcare decision-making and health economics. The most important applications of utilities are in expected utility decision analyses and in cost-utility analyses. In expected utility decision analyses, the expected utility for each possible strategy is calculated by combining the utilities for all possible health states with the probabilities of these states occurring [1]. Cost-utility analyses use a specific application of utilities: that of quality-adjusted life-years (QALYs). QALYs are calculated as the area under the utility curve [2]. The utility thus serves as a correction factor reflecting the quality of life in spent life-years. Utilities can be used at various levels of health-care decision-making, and the level determines whether they should be assessed from the general public or Address correspondence to: Anne M. Stiggelbout, Department of Medical Decision Making, Leiden University Medical Center, PO Box 9600, 2300 RC Leiden, The Netherlands. E-mail: [email protected] 10.1111/j.1524-4733.2007.00216.x

76

© 2007, International Society for Pharmacoeconomics and Outcomes Research (ISPOR)

1098-3015/08/76

76–87

77

A Framework for Health State Utilities differences between respondents). After patients experience adverse health outcomes, they seem to assign higher values to those outcomes, despite the substantial reductions in health-related quality of life that often accompany these. These findings resonate with those of the disability paradox: that many people with serious and persistent disabilities report that they experience a good or excellent quality of life, when to most external observers these people seem to live an undesirable daily existence [5]. The observation of the differences in utilities between respondent groups has led to the question of whose utilities are most “well-informed,” and thereby most valid for use at the different levels of decisionmaking. The answer to this question will depend on the reason for the differences. Are they because of better understanding by the patients of what the health states are really like? Or are they because of cognitive mechanisms operating in patients during or after treatment, such as justification processes? Or, worse, are they simply artifacts of the method, caused by, for example, inappropriate scenarios? All of these explanations have indeed been suggested in the literature to account for the differences in valuations between healthy respondents and patients, and between patients before and after experiencing a health state. The explanations come from such heterogeneous fields as medical decision-making, healthrelated quality of life (HRQL) research, judgment and decision-making, health psychology, and sociology. Hedonic psychology, which has recently emerged as a new field that studies determinants of well-being and the judgmental processes involved in reports of wellbeing [6], may also be relevant in this respect. Valuations of well-being are not similar to valuations of health, but clear parallels exist and it is conceivable that patients take their well-being into account when assigning utilities to their health. Similarly, another field that has recently received much attention and that may be relevant in explaining some of the differences in utilities described above, is that of positive psychology [7] in which concepts such as meaningfulness and hope play an important role in determining quality of life. Research insights from the fields mentioned are seldom empirically linked to research in utility assessment. This is not surprising, because in those fields the findings may come from studies on topics as varied as the federal states in which students prefer to live or students’ preferred ice cream flavors [8]. Moreover, the research generally pertains to happiness or mood, not to health state valuation. To date, most of the research in utility assessment has been on improving the response techniques, for example, on the correction of probability distortion in the Standard Gamble (SG) [9]. Little attention has been paid to the object of valuation and the judgment processes involved in valuing such an object. This article therefore presents a

framework for the understanding and subsequent study of the mechanisms that may play a role in the process of assigning a utility to a health state. It will not provide ready answers to the question of whose utilities are the most “well-informed,” because the answer is quite involved, as we will show. In contrast, the framework is an attempt to connect research carried out within different traditions that all in some way or other may relate to the issue of utility measurement. Researchers from within a specific tradition may not always read research published in journals that are particular to another tradition. We aim to cross such borders in the hope of stimulating a cross-fertilization of ideas. The article does not pretend to be a systematic review, since that would not be feasible because of the broad array of topics from often seemingly unrelated research. But it is to be hoped that it will point the readers to journals and articles that they would otherwise not likely have read. Some of the aspects described have been discussed in the context of the disability paradox by Ubel et al. [10], but from a different angle, and again, have not been applied to utility assessment. Our article attempts to integrate issues that have as yet seldom been studied in conjunction, and thereby will point toward issues for further study. It may provide support in indicating which aspects of the valuation instruments and procedures should perhaps be adapted, and which aspects should not be deemed artifacts but are true aspects to be incorporated. Valuations that are valid for the purpose at hand may thus be obtained. Finally, theoretical insight into the process of health state valuation, and into determinants of utilities, are not only relevant for the field of medical decision-making, but can be extended to the field of HRQL research, where many similar issues play a role.

A Framework Describing the Steps from Object Definition to Response Our proposed framework is based on theories of information processing (stimulus response models) from social cognition theory. The process of health state valuation can be seen as a series of steps from a stimulus (the object of valuation or scenario) to a response, the utility. The steps involve the cognitive representation of the stimulus, the interpretation of the information, and the combination and integration of the information into a judgment, which finally needs to be expressed on some (utility) scale. In Figure 1 the framework has been depicted. At each step, several mechanisms may play a role that can explain differences in values between respondents (or groups of respondents). In the following sections we will elaborate on this framework. A qualification should be made regarding the sequence described in the model. In reality the phases are not as clear-cut and linear as

Stiggelbout and de Vogel-Voogt

78

Poor hedonic forecasting(e.g. transition heuristic)

Lack of scope

Scenario

creation of stimulus

Framing

Interpretation

Other biases (e.g., status quo, focusing)

Judgment

Primary appraisal

Secondary appraisal

(values, goals, beliefs)

(coping, e.g., positive reappraisal) Impact of mood

Response (utility)

Noncompensatory Strategies Biases

Assimilation/contrast

Own health

Interpretation

Judgment

Adaptation

Response (utility)

Recalibration response shift

Figure 1 A framework describing the process of health state valuation and mechanisms that play a role in this process.

they may seem. For example, framing can be seen as part of the process of creating the scenario, or as a determinant of information interpretation. Furthermore, in the phase of information interpretation and integration, judgments regarding the information are already being made [11]. Nevertheless, in the section on judgment we only wish to refer to judgmental aspects of the global evaluation of the stimulus that has been “perceived” in the process of information interpretation and integration. The framework therefore has mostly a heuristic use, and serves to structure the wide array of findings from the literature that may be relevant to the differences in values of patients and of the general public. Similar models, with slightly different emphases, have been proposed in the literature to explain, for example, attitude measurement [12], subjective well-being [13], or the way that respondents answer survey questions [11]. These focus more strongly on the cognitive processes involved, whereas we focus more on the context of health state valuation and wish to link such models to findings regarding differences in values between respondent groups.

The Stimulus (Object of Valuation) In decision analysis or cost-utility analysis, utilities are assessed of the clinically important outcomes of the

decision alternatives: the health states. The utilities can be obtained from a representative sample of patients who are experiencing the health state to be valued, in which case the patients are asked to value their own health. In most studies, patients who value their own health are instructed to think of “their health in the past week” or of “their current health,” and to indicate whether they would trade off life-years (in a Time Trade-off [TTO]) or accept risk (in an SG) to obtain optimal health. Current health may be specified further, referring to physical, psychological, and social well-being [14]. When a societal perspective is used, utilities are obtained from members of the general public, of whom most will not be experiencing that health state. The respondents are therefore requested to value a health state description, or scenario, of the state. Generally, health state classification systems such as the EQ-5D, or the Health Utilities Index, are used for this purpose. They give a limited description of the health state, usually in terms of functioning on some salient dimensions. The EQ-5D scenarios describe specific levels of impairment in mobility, self-care, and daily activities, and specific levels of pain/discomfort and anxiety/depression. Respondents are asked to imagine spending the rest of their life in the health state described in that scenario, and are asked whether they would be willing to trade off life-years or accept risk to obtain optimal health. Relatively little research has

A Framework for Health State Utilities been devoted to the development of these health state scenarios, despite early research in the field of utility assessment that showed the impact of framing of the scenarios on the elicited utilities [15,16]. LlewellynThomas et al. [16] experimented with an individualized scenario, asking cancer patients to indicate which dimensions of quality of life were important to them. They found that using an individualized scenario reduced the amount of random error, compared with a standardized scenario. In a later article LlewellynThomas [17] therefore proposed a general approach to creating disease-specific scenarios for use in clinical trials, to obtain utilities from patients for their overall current status. She argued that if hidden effects of scenarios would produce biased results and erroneous conclusions, a comprehensive research program would be needed. To our knowledge no such program has taken place. We will illustrate with an example how we nevertheless think the scenario may underlie part of the difference between patients valuing the health states that they are experiencing, and respondents assigning utilities to these states as hypothetical scenario. Jansen et al. [18] found that patients undergoing radiotherapy assigned higher values to their “own health” during radiotherapy than to a scenario that aimed to describe their actual radiotherapy experience. The authors ascribed this difference to the scenario not matching the patients’ experience (“noncorresponding description”). Insinga and Fryback [19], in a study using the EQ-5D health classification system, found a similar difference between respondents’ valuations of their current health (however, they conceived of it) and their valuations of the EQ-5D scenario that by coincidence happened to describe their own health. They argued that “noncorresponding description,” which they termed “lack of scope,” best explained the higher values that respondents assigned to their own health. They concluded that the health state descriptions (scenarios) of the EQ-5D are too sparse to describe certain health states. We believe that lack of scope of the stimulus will cause differential interpretation and integration of the information for healthy respondents and for patients. This will be discussed in the next section.

Interpretation and Integration of Information The next step is the interpretation of the available information and its integration with existing information, to make a judgment. Patients may interpret information in the light of their experience, experience that the general public lacks. So even if the health state is adequately described, patient experience may color the information. With regard to information interpretation in this context, much can be learned from the stress-appraisal-coping paradigm, espoused by Lazarus and Folkman [20,21]. They describe cognitive appraisal as an interpretative and evaluative process

79 that focuses on the meaning and significance that an occurrence has for one’s well-being. The interpretative part they termed “primary” appraisal, the evaluative part “secondary” appraisal. In primary appraisal the person evaluates whether he or she has anything at stake in the occurrence. It thus fits the information interpretation phase. In secondary appraisal, the person evaluates what, if anything can be done to overcome or prevent harm, or to improve the prospects of benefit, evaluating coping strategies. This better fits our judgment phase. A range of personality characteristics, including values, goals, and beliefs, as well as experience, may influence the process of primary appraisal. A patient who has experienced restrictions to her functioning, and has dealt satisfactorily with the impact of these restrictions on her daily life, may appraise a scenario as less threatening to her life than a healthy subject who does not have such an experience. Findings from research within this stress-coping paradigm may thus be relevant to the difference in information interpretation between patients and healthy respondents. Framing of the scenarios may further influence such differential interpretation and integration of information by patients and by healthy respondents. Scenarios are most often described in terms of the dimensions in which the patient is not functioning well. Even if the scenarios properly describe the negative symptoms that patients experience, the patients themselves may take other dimensions into account when interpreting the information. It is conceivable that patients who value their own health take, for example, more nonhealth-related aspects into account, possibly their happiness rather than their limitations, when deciding whether or not to trade off life-years or accept risk. They may incorporate constructed meaning [22], or the personal growth that they may have experienced after diagnosis [22,23]. Indeed, Tsevat et al. [24] found that HIV patients were “highly valuing their life with HIV, not their health with HIV” during utility assessment. This would explain why many patients assign high values to adverse health outcomes that they experience, despite substantial reductions in quality of life, and why, generally, no more than 40% of the variance in utilities is explained by patients’ HRQL [25]. If patients value their life, not their health, social indicators research is relevant, which has shown that health, or satisfaction with one’s health, is never a strong predictor of happiness, and that, of all aspects of health, only mental health makes a substantial contribution [26]. Studies in positive psychology show that well-being is less related to the impact of impairments, disabilities and handicaps, and more to positive aspects such as meaning in life and goals [27,28], which again points back to the primary appraisal process described above. The publication of the studies on framing by Kahneman and Tversky [29] has spawned much experi-

80 mental research on framing. This research has generally focused on gain versus loss frames, and has demonstrated differential judgment in these two situations (see the next section). To our knowledge no studies have aimed at elucidating the possible psychological mechanisms involved in the interpretation of scenarios, and the impact of framing on this interpretation. The field may therefore learn from research in health psychology, specifically that on primary appraisal, and from research in positive psychology. Finally, Schwarz [30] has shown that for information interpretation, respondents may rely on contextual information, especially when the questions pertain to issues about which respondents do not have clear and articulate answers. This may well be the case for utilities, because such values do not play a role in everyday life and will often be unfamiliar and even strange to the respondent. The values may thus be constructed in the process of elicitation [31]. Most of the research referred to by Schwarz pertains to multiple-item questionnaires, where the respondents infer, for example, the researcher’s epistemic interest from adjacent questions. This will be less of a threat in utility assessment, where such multiple questions do not occur. Context effects on judgment have been shown, however, and will be discussed in the next section.

Judgment The judgment phase is the most encompassing phase in our model. In our opinion most between-subject variation may be found in this phase, because of the various personal characteristics that affect the judgment process. This is not to deny that information interpretation and integration also involve highly subjective processes that are determined by such personal characteristics, also because in that phase judgments regarding the information are already being made, as explained in the introduction. Cognitive information processing is rarely exhaustive and rarely guided by logical norms, but has to reach a compromise between rationality and efficiency. Memory is not perfect, and decisions often have to be made about how to compensate for incomplete or inaccurate information. Respondents use heuristics, which are cognitive devices that enable the individual to make efficient judgments, by rule-of-thumb, that require little effort and yield quite valid results most of the time. Nevertheless, the price of such efficiency is systematically biased judgments in some instances [32]. We will therefore start this section with a short overview of the problems encountered when such cognitive shortcuts are taken. Next we will describe the cognitive process of anticipated adaptation, and the deficiencies inherent in this process. Furthermore, social comparison processes are important in this phase, and are discussed in a separate section. We will

Stiggelbout and de Vogel-Voogt end with aspects of affect and mood that may also be of significance in this phase.

Heuristics and Biases A large body of research has followed the work of Tversky and Kahneman [33] on heuristics and biases and their impact on judgments. Tversky and Kahneman explain, for example, the impact on the respondent’s answer of the availability of information, the representativeness of the information retrieved, and the context that is provided. These will have impact mostly on the previous phase of information interpretation. Other heuristics are more relevant to the judgment phase and we will discuss those heuristics and biases that seem most relevant to possible differences in valuations between respondents. A consequence of one such heuristic is the so-called focusing illusion [8,34]. When imagining unfamiliar circumstances, people focus narrowly on the most obvious difference between those circumstances and their current circumstances, and thereby misconceive the impact of the unfamiliar circumstances. When imagining life as a paraplegic, respondents mostly focus on the wheelchair and all the things they will no longer be able to do, not considering all the things that will not change, let alone aspects that will improve because of those circumstances, such as re-evaluations of areas of their life. Ubel et al. carried out several studies attempting to minimize this effect by using defocusing exercises, and concluded that it is difficult if not impossible to overcome people’s focusing illusions [10]. Furthermore, respondents have been shown to experience status quo bias or an endowment effect, a bias that may also impact health state utilities. People value goods more highly once they own them [35]. This explains the finding that people are, in general, less willing to pay to obtain certain goods than to accept to give up the same goods. Salkeld et al. [36] have extended these findings from the willingness-topay literature to the discrete choice method. Respondents had a preference for the status quo. They were willing to pay on average $18 simply to have a bowel cancer prevention test that they had already used rather than a hypothetical test, which was similar on all attributes but costs. It has not explicitly been studied in health state valuation, but the effect can be expected to hold in the same way. One may expect this bias to be stronger for respondents valuing their own health (which they “own”) than for respondents valuing hypothetical states (which they are asked to imagine “owning”). Subsequently, respondents valuing their own health may be less willing to trade life-years for better health, which leads to higher utilities. Such an endowment effect may explain why Chapman et al. [37] found that prostate cancer patients who used a personal version of a scenario were less likely to trade off length of life in a TTO task than patients using an

A Framework for Health State Utilities impersonal version. The former version was the commonly used way of asking patients to trade part of their life expectancy, which may have induced an endowment effect. The impersonal version asked patients to state which of two hypothetical friends would be better off, and thereby may have avoided some of the endowment framing. Another explanation of these results, however, presented by the authors, is a difference in compensatory decision-making resulting from a difference in hot versus cold decision-making. Therefore, we also discuss this study under the response phase. Related to status quo bias, is the bias caused by loss aversion [38]. People judge outcomes relative to a reference point, and are more sensitive to losses than to gains. Loss aversion will generally lead to utilities that are biased upward [38,39]. In the TTO, the loss in life-years looms larger than the gain in health. The SG has been argued to be a mixed gamble, in which the gain probability p must be extra high to offset the loss probability 1-p [39]. Similar to the endowment effect, it may be expected to be larger for patients who actually are in the states they are valuing than for healthy respondents. Stalmeier and Bezembinder [40] and Bleichrodt and Pinto [38,41] present empirical evidence of the impact of loss aversion on medical trade-offs.

Adaptation and Anticipation Thereof A major issue in judgments of health states is the process of adaptation, and people’s ability to anticipate an eventual adaptation to poor health. An important conclusion of research in behavioral economics that likely applies to utility assessment, is that humans are poor in predicting how they will value a situation once it is experienced, so-called “poor hedonic forecasting” [6]. Processes of adaptation appear difficult to anticipate. It has been suggested that healthy respondents use a transition heuristic when valuing health, that is, they focus on the time of entering a poor health state and do not consider adaptation to that state. This will result in healthy respondents undervaluing states. Damschroder et al. [42] successfully used an adaptation exercise preceding a Person Trade-off task, which encouraged healthy respondents to consider their own ability to emotionally adapt to negative events in general, and specifically to having paraplegia. Respondents randomly allocated to the adaptation exercise increased the values they assigned to paraplegic states, and the authors concluded that asking nonpatients to do an adaptation exercise before giving quality-of liferatings may help close the gap in ratings between patients and nonpatients. Their study on utility assessment did not show a similar effect, however [43]. A discrepancy between healthy respondents and patients may also arise from the fact that patients may be unaware that their adaptation has an impact on their current judgment. A study by Riis [44] showed

81 that dialysis patients estimated the mood of healthy persons as much higher than their own mood, and even as higher than the actual mood of healthy respondents in the same study. They thus were unaware of their adaptation to poor health. Gilbert [45] has written extensively on “immune neglect,” the fact that people are generally unaware of the cognitive mechanisms that they employ to ameliorate their experience of negative affect (what he calls “the psychological immune system”). If they neglect this immune system, they will tend to overestimate the duration of their affective reactions to negative events. Gilbert even provides an explanation from the perspective of human evolution: if people were aware of their capability of such emotion-focused coping, they would not engage in problem-focused coping (dealing physically with the sources of their negative affect), and the species would become extinct. Furthermore, the process of secondary appraisal from Lazarus’ and Folkman’s theory on stress, appraisal, and coping, mentioned in the section on information interpretation, might be different for patients than for healthy respondents. In secondary appraisal a person evaluates what, if anything, can be done to overcome or prevent harm or improve prospects for benefit. The person, for example, evaluates coping options. Coping options that will generate positive affect, and are therefore adaptive, are the infusion of ordinary events with positive meaning, and positive reappraisal [46]. Regarding positive meaning, evidence suggests that under stressful conditions, individuals may be more likely to bring about, note, or remember, ordinary positive events. Positive reappraisal refers to cognitive strategies for reframing a situation to see it in a positive light. Some situations require relinquishing previous goals that are no longer tenable and turning to new, realistic goals. Patients may be more likely than healthy respondents to use such active processes, with which they already have experience. The “lack of scope” of many of the scenarios used, as discussed above, increases the likelihood of such differential appraisal between patients and healthy respondents. Patients will reframe their situation as a way of adapting to the situation, such that it does not correspond anymore to the description in the scenario. Finally, when patients are asked to value a treatment they have experienced, they may unknowingly restructure their internal representation of the treatment state to justify decisions already made or to reduce cognitive dissonance [47]. This may also be seen as a type of cognitive adaptation. Such justification processes have been observed in direct treatment preferences, where patients are asked to indicate the benefit they require from a hypothetical treatment to be willing to accept it. Jansen et al. [48] found that breast-cancer patients who underwent chemotherapy were willing to accept this treatment for much smaller

Stiggelbout and de Vogel-Voogt

82 chances of benefit, than patients not undergoing it. In another study [49] they saw this difference in preferences even before treatment had started, indicating it to be due to a psychological process of justification or to anticipated adaptation, as opposed to positive experiences with the therapy itself.

Comparison Processes Respondents in ill health have also been found to assign higher utilities to hypothetical health states, not just to their own health, than respondents in good health. Dolan [50] found that respondents describing their own health as poor assigned higher values to various EQ-5D health states than those describing their own health as good, a process he termed “valuation shift.” Valuation shift thus refers to a change in the valuation caused by a change in the status quo. Dolan found valuation shift in cross-sectional data, as did Clarke et al. [51] and Lenert et al. [52]. Saigal et al. [53] observed a similar phenomenon in a longitudinal study in prostate-cancer screening. Utility for hypothetical states rose in men who showed a large decrease in quality-of-life scores over time. Valuation shift might be explained by a so-called “contrast effect.” Contrast effects arise from social comparison processes, and originate in the socialjudgment literature on assimilation and contrast. They refer to the effect of judging a target stimulus (in our case: a health state) in relation to a context stimulus. We may judge our present situation as more pleasant when we have just remembered other pleasant experiences (assimilation), but it is also possible that our present situation will appear less pleasant (contrast) if compared with the standard of a very pleasant experience. Very positive information in the representation of the standard will result in a negative evaluation of the target [54]. In general, the likelihood of assimilation decreases as the distance between stimulus and context category increases; extreme standards of comparison are likely to produce contrast effects [32]. Mussweiler [55] provides a review on comparison processes and the role of assimilation and contrast. Bleichrodt and Johannesson [56] showed that Visual Analog Scale valuations were indeed dependent on the context in which they were elicited, that is, on the other health states included in the task. Contrast effects may thus explain their finding. Such effects are also likely to play a part in the EQ-5D valuations of the general public, in which respondents are first asked to rank all states, before performing the TTO task. It is not known whether patients may come up with their own context stimuli, and moreover, it is not evident whether assimilation or contrast is produced. But it is likely that assimilation and contrast effects play a role in the judgment phase, and these have not been studied in utility assessment under this name. The literature in the field of social psychology on social comparison processes and the impact of these on, for

example, mood, self-evaluation, and life satisfaction, may provide direction in this sense [57–59].

Affect and Mood Finally, a large body of evidence exists on the impact of mood or affect on decision-making [10,13,60]. Most of this research is experimental in nature, and shows that inducing positive affect by means as simple as offering a candy bar, has an impact on participants’ decisions. In the field of utility assessment, Isen et al. [61] showed that persons in whom positive affect had been induced (by offering them a small bag of candy) showed a more negative utility for losses than did those in a control group. This indicates that losses seem worse to people who are feeling happy than to those in a control condition. The utility functions of the two groups did not differ as much, however, when people were considering potential gains. If positive affect induces loss aversion (and thereby higher utilities) in more naturalistic settings too, and not only in experimental conditions, it will only serve as an explanation for the differences between healthy respondents and patients if the latter are consistently in states of higher positive affect. On the one hand, it is conceivable that, to patients, utility interviews are more interesting or gratifying because they pertain to their situation. For healthy respondents the tasks are more hypothetical and thereby perhaps more cumbersome. On the other hand, arguments in favor of the inverse point of view can be thought up as well. To our knowledge, only one other study assessed the impact of mood on utilities [16]. That descriptive cross-sectional study among cancer patients did not find any effect. Nevertheless, Voogt et al. [62] found that cancer patients who had less positive affect were more inclined to exchange length of life for quality of life than others. These authors did not assess utilities, but their results imply a higher utility in the group high in positive affect. It might be that respondents with high levels of positive affect value not just their health, but also value their psychological well-being.

Response In this final step of the model in Figure 1 we describe findings related to the response stage, during which the respondent, having formed a judgment, needs to fit his or her answer into the answering options provided. In the case of utility assessment the answering options are more involved, because they consist of a trade-off task (which in fact again involves a judgment, but for the sake of simplicity we deal with it in this section). This response formatting process takes information from cues provided, such as predefined answering categories, to decide what is common for people and thus a likely response. For the TTO and the SG, the use of search strategies to obtain an indifference point has

83

A Framework for Health State Utilities been shown to influence the response. Titration downwards, for example, results in higher utilities than a ping-pong approach [63]. Matching, which is cognitively more complex than choice bracketing (the pingpong approach) also leads to different results [64]. But these issues are not likely to explain differences between respondent groups, as long as both groups use the same procedure. In the response phase, some of the biases that have been described for the TTO and the SG may well operate in a different manner for patients than for healthy respondents. The first one is probability distortion, which refers to the fact that respondents do not treat probabilities linear, but weight them. It is described in Cumulative Prospect Theory [38]. People tend to overweight small probabilities, and underweight large ones. The weighting function has been shown to depend on the framing of the gambles (loss vs. gains), and the impact of weighting could thus be different for patients than for healthy subjects. Furthermore, loss aversion, referred to in the section on judgment, is relevant in the response phase too, because it also can be induced by the elicitation task. In the TTO, subjects may perceive the trading of life-years as a loss, and similarly the framing of the SG may lead to feelings of loss. These biases and their impact on utilities have been explained by Bleichrodt [39] and empirically tested by Van Osch et al. [65]. If healthy subjects and patients perceive the frame as distinct, due to it being more hypothetical for healthy subjects than for patients, this could lead to higher utilities for patients. An example of a cognitive shortcut that patients may use more often in decision-making than healthy respondents, is noncompensatory decision processing. Compensatory decision-making, in which one attribute is traded off for another, is the basis of utility assessment methods. It is cognitively demanding, and when a decision is emotion-laden (“hot”), respondents may resort to noncompensatory processes, such as unwillingness to trade life expectancy [66]. Chapman et al. [37] found that prostate cancer patients who used an impersonal (“cold”) version of a scenario were more likely to trade off length of life in a TTO task, whereas patients using a personal (“hot”) version were more likely to use noncompensatory strategies, such as choosing the longest life expectancy. The “hot” judgments by patients thereby resulted in patients assigning higher values. As described under heuristics and biases, part of this finding may also have been due to an endowment effect. An important phenomenon that also may play in this response stage is so-called recalibration response shift, which will happen to patients only. Patients undergoing toxic treatment, such as radiotherapy, often report quite similar HRQL from pre- to posttreatment, despite objective deteriorations [67,68]. From retrospective assessments of their pretreatment

quality of life it appears that the patients change their internal standards: after treatment they truly know what fatigue means, and a similar score as pretreatment on the post-treatment scale now implies more fatigue. Recalibration has been shown for the Visual Analog Scale [69]. If it translates to utilities too, this may explain partly why patients score dissimilar to healthy respondents. Finally, respondents may wish to edit their answers before they communicate them, because they may want to conform to notions of consistency, social desirability and self-presentation [11,12]. It may well be that in an interview setting healthy respondents are less concerned about trading off life-years (in a TTO), or about a risk of immediate death (in an SG) than patients, because it remains hypothetical, whereas patients may feel they are tempting fate by trading or even feel it as an insult to their doctor or the health-care system.

Discussion We have attempted to provide a framework describing the process of health state valuation and the various mechanisms that play a role in this process. A summary of the mechanisms discussed is provided in Table 1. We like to repeat the reservation we made in the introduction of the framework about the phases, which are not as distinct as they may seem from the table. Many of the mechanisms described may operate differentially for respondents with experience with the health state to be valued and for those without, or for respondents in

Table 1 Summary of mechanisms that may play a role during the phases described by the stimulus response framework in Figure 1 to explain differences between patients and healthy respondents Stimulus

Health state descriptions vs. experienced health states (“lack of scope”)

Information interpretation and integration

Primary appraisal Framing effects Impact of contextual information

Judgment

Heuristics and Biases • Focusing illusion • Status quo bias/endowment effect • Loss aversion (Anticipated) Adaptation • Poor hedonic forecasting • Unawareness of changes by patients • Secondary appraisal (coping) Comparison processes • Valuation shift/assimilationcontrast effects Other • Justification processes • Mood/affect

Response

End aversion bias* Search strategies* Noncompensatory decision-making Recalibration response shift Editing

*Not likely to operate differentially for patients and nonpatients.

84 good health and those in ill health. In most instances, the direction will be toward a higher valuation by those in ill health. With this model we have provided a link between the fields of medical decision-making and health economics on the one hand, and other fields that in some way or another study the judgment processes involved, such as decision psychology, HRQL research, and social psychology, on the other. Recently the field of HRQL research has seen an interest in the more cognitive aspects of HRQL assessment that we described above (see, e.g., Volume 12, issue 3 of the journal Quality of Life Research). This interest followed the introduction into that field of the concept of response shift, by Sprangers [70]. Some of the aspects described above, both under judgment (e.g., valuation shift) and under response (recalibration), have indeed been described in the HRQL literature as forms of response shift [71]. Llewellyn-Thomas and Schwartz [72] have categorized some of the biases mentioned above in a response shift framework, in an attempt similar to ours to help researchers decide which aspects are preventing valid measurement. Response shift has been defined as a change in the meaning of one’s self-evaluation of a target construct (in the context of this article: the health state to be valued), as a result of a change in the respondent’s internal standards of measurement (recalibration), a change in the importance of component domains that constitute the construct (value change), or a redefinition of the target construct (reconceptualization). Sprangers and Schwartz [73] and later Schwartz et al. [74], distinguish response shift as such (i.e., the three types of change) from the mechanisms that lead to response shift, such as coping, social comparison, and goal reordering. The question is whether conceptually this distinction can be made. As the authors show in their examples, recalibration is the outcome of a contrast effect, for example, when respondents compare their functioning to that of others, or to that of themselves before treatment. As such, it can indeed be seen as the result of a (social) comparison mechanism. But is reconceptualization the result of goal reordering or do both refer to a same process? Is value change a result of coping, or is it a way of coping? Because the authors explicitly use the term recalibration for a respondents’ internal measurement scale, we have classified recalibration as a more specific phenomenon of the response phase. But in our opinion reconceptualization and value change are processes of adaptation. We believe that the response shift literature makes an unwarranted distinction between mechanisms (e.g., coping, social comparison) on the one hand, and value change and reconceptualization on the other. This distinction may be underlying the confusion in the literature as to whether response shift is not simply “the emperor’s new clothes” [74], and refers to, for example, adaptation.

Stiggelbout and de Vogel-Voogt The findings described have implications for utility assessment both in the general public and in patients. We briefly discuss approaches to solving some of the inconsistencies between patients and the general public. Not for all issues described a ready solution is at hand. Some warrant further discussion followed by a consensus as to whether specific aspects of the valuation process should be included or are truly biases. Regarding the assessment of utilities from the general public, a discussion of the health state classification systems such as the EQ-5D and the Health Utilities Index is called for. These systems are widely used, precisely because they provide reference values elicited from the general public for health states experienced by patients. It has, however, been argued that the utilities from these systems may not be as “well-informed” as would be desired [3]. We wonder why there has been so little discussion about whether the utilities obtained from the general public for the classification systems are well-informed and therefore valid? We would like to argue in favor of opening the discussion on the use of these systems, and the often low utilities that they generate. For this purpose our framework will be useful to decide which aspects of the systems may warrant study. Lack of scope of the classification systems, as described in the section on the stimulus, seems a factor that should be avoided. Adding new health state levels or dimensions [75], or changing the nature and tone of health state descriptions may be useful steps towards improvement [19]. Furthermore, Gold et al. [3] already argued that “techniques that create a better understanding in the general public of the experience of differing health states will be highly useful in strengthening this field.” Using adaptation exercises as proposed by Damschroder [42] may serve as such to overcome the poor hedonic forecasting of humans. Regarding the assessment of utilities from patients, in all steps of the valuation process biases in patient valuations may be present as well, as appears from the description of our framework. One has to decide which mechanisms are important to incorporate in the valuation process, and which should be avoided. Patients suffer from cognitive mechanisms operating during treatment, such as justification processes, which should not be reflected in the utilities used for cost-effectiveness analysis, and possibly not even in those used for guideline development either. Patients may also use more noncompensatory mechanisms because of their emotional involvement. Framing the task as a detached “cold” task may overcome this problem. Furthermore, differences in patient valuation that are due to differences in health or to valuation shift may be corrected by the use of anchoring vignettes [76]. A major question that has emerged from the description of the framework is what it is exactly that patients value in utility assessment, and whether this

A Framework for Health State Utilities differs from what the general public values? Do patients focus on their well-being, or even their happiness, instead of on their impaired HRQL, and pay too little attention to a possible impact of impaired HRQL? Or do they simply take an impaired HRQL in stride? A goal of health care is to improve patients’ HRQL, and one may argue that utilities should therefore reflect a valuation of this HRQL only. Similarly, however, one may argue that healthy subjects overestimate the impact of impaired HRQL. A link of the field of psychological well-being and meaning in life with medical decision making or health economics until now does not exist. It may be created through the assessment of individual quality of life [3,77], in which respondents are asked to nominate themselves the domains of life that mostly affect their quality of life. In such studies, respondents often mention the more positive aspects of life in relation to their quality of life; health is often not given primacy [78]. Our article has attempted to integrate issues that have as yet seldom been studied or presented in conjunction, and thereby has pointed to many issues for debate. It provides directions in deciding which aspects of the valuation instruments and procedures should perhaps be adapted, and which aspects should not be deemed artifacts, but are true aspects to be incorporated. Valuations that are valid for the purpose at hand will thus result. Source of financial support: This article was written with support from a VIDI-award of The Netherlands Organization for Scientific Research NWO Innovational Research Incentives (grant number 917.56.356). The authors also would like to acknowledge Wilbert van den Hout whose constructive criticisms on an earlier version improved this article.

References 1 Sox HC, Blatt MA, Higgins MC, Marton KI. Medical Decision Making. Stoneham, MA: Butterworth Publishers, 1988. 2 Hunink MGM, Glasziou PP, Siegel JE, et al. Decision Making in Health and Medicine. Cambridge: Cambridge University Press, 2001. 3 Gold MR, Siegel JE, Russell LB, Weinstein MC. CostEffectiveness in Health and Medicine. New York: Oxford University Press, 1996. 4 De Wit GA, Busschbach JJ, De Charro FT. Sensitivity and perspective in the valuation of health status: whose values count? Health Econ 2000;9:109–26. 5 Albrecht GL, Devlieger PJ. The disability paradox: high quality of life against all odds. Soc Sci Med 1999;48:977–88. 6 Kahneman D, Diener E, Schwarz N. Well-Being: The Foundations of Hedonic Psychology. New York: Russell Sage Foundation, 1999. 7 Seligman ME, Csikszentmihalyi M. Positive psychology. An introduction. Am Psychol 2000;55:5–14.

85 8 Schkade DA, Kahneman D. Does living in California make people happy? A focusing illusion in judgments of life satisfaction. Psychol Sci 1998;9:340–6. 9 Wakker P, Stiggelbout A. Explaining distortions in utility elicitation through the rank-dependent model for risky choices. Med Decis Making 1995;15:180–6. 10 Ubel PA, Loewenstein G, Schwarz N, Smith D. Misimagining the unimaginable: the disability paradox and health care decision making. Health Psychol 2005;24(4 Suppl.):S57–62. 11 Collins D. Pretesting survey instruments: an overview of cognitive methods. Qual Life Res 2003;12:229–38. 12 Tourangeau R, Rasinski KA. Cognitive processes underlying context effects in attitude measurement. Psychol Bull 1988;103:299–314. 13 Schwarz N, Strack F. Evaluating one’s life: a judgment model of subjective well-being. In: Argyle M, Schwarz N, Strack F, eds. Subjective Well-Being: An Interdisciplinary Perspective. Oxford: Pergamon, 1991. 14 Stalmeier PF, Goldstein MK, Holmes AM, et al. What should be reported in a methods section on utility assessment? Med Decis Making 2001;21:200–7. 15 Llewellyn-Thomas H, Sutherland HJ, Tibshirani R, et al. Describing health states. Methodologic issues in obtaining values for health states. Med Care 1984;22:543–52. 16 Llewellyn-Thomas HA, Thiel EC, McGreal MJ. Cancer patients’ evaluations of their current health states: the influences of expectations, comparisons, actual health status, and mood. Med Decis Making 1992;12:115–22. 17 Llewellyn-Thomas HA. Health state descriptions. Purposes, issues, a proposal. Med Care 1996;34(12 Suppl.):DS109–18. 18 Jansen SJ, Stiggelbout AM, Wakker PP, et al. Unstable preferences: a shift in valuation or an effect of the elicitation procedure? Med Decis Making 2000; 20:62–71. 19 Insinga RP, Fryback DG. Understanding differences between self-ratings and population ratings for health in the EuroQOL. Qual Life Res 2003;12:611–19. 20 Lazarus RS, Folkman S. Stress, Appraisal, and Coping. New York: Springer, 1984. 21 Folkman S, Lazarus RS, Gruen RJ, DeLongis A. Appraisal, coping, health status, and psychological symptoms. J Pers Soc Psychol 1986;50:571–9. 22 Fife BL. The role of constructed meaning in adaptation to the onset of life-threatening illness. Soc Sci Med 2005;61:2132–43. 23 Tedeschi RG, Calhoun LG. Posttraumatic growth: conceptual foundations and empirical evidence. Psychol Inquiry 2004;15:1–18. 24 Tsevat J, Sherman SN, McElwee JA, et al. The will to live among HIV-infected patients. Ann Intern Med 1999;131:194–8. 25 Revicki DA, Kaplan RM. Relationship between psychometric and utility-based approaches to the measurement of health-related quality of life. Qual Life Res 1993;2:477–87. 26 Michalos A. Social indicators research and healthrelated quality of life research. Soc Indic Res 2004;65:27–72.

86 27 Ryff CD, Keyes CL. The structure of psychological well-being revisited. JPSP 1995;69:719–27. 28 Diener E, Suh EM, Lucas RE, Smith HL. Subjective well-being: three decades of progress. Psychol Bull 1999;125:276–302. 29 Kahneman D, Tversky A. Choices, values, and frames. Am Psychol 1984;39:341–50. 30 Schwarz N. Self-reports: how the questions shape the answers. Am Psychol 1999;54:93–105. 31 Payne JW, Bettman JR, Johnson EJ. Behavioral decision research: a constructive processing perspective. Ann Rev Psychol 1992;43:87–131. 32 Hewstone M, Stroebe W, Stephenson GM. Introduction to social psychology. A European Perspective (2nd ed.). Oxford: Blackwell, 1996. 33 Tversky A, Kahneman D. Judgment under uncertainty: heuristics and biases. Science 1974;185:1124–31. 34 Ubel PA, Loewenstein G, Hershey J, et al. Do nonpatients underestimate the quality of life associated with chronic health conditions because of a focusing illusion? Med Decis Making 2001;21:190–9. 35 Kahneman D, Knetsch JL, Thaler RH. Experimental tests of the endowment effect and the Coase theorem. J Polit Econ 1990;98:1325–48. 36 Salkeld G, Ryan M, Short L. The veil of experience: do consumers prefer what they know best? Health Econ 2000;9:267–70. 37 Chapman GB, Elstein AS, Kuzel TM, et al. Prostate cancer patients’ utilities for health states: how it looks depends on where you stand. Med Decis Making 1998;18:278–86. 38 Tversky A, Kahneman D. Advances in prospect theory: cumulative representation of uncertainty. J Risk Uncertain 1992;5:297–323. 39 Bleichrodt H. A new explanation for the difference between time trade-off utilities and standard gamble utilities. Health Econ 2002;11:447–56. 40 Stalmeier PF, Bezembinder TG. The discrepancy between risky and riskless utilities: a matter of framing? Med Decis Making 1999;19:435–47. 41 Bleichrodt H, Pinto JL, Abellan-Perpinan JM. A consistency test of the time trade-off. J Health Econ 2003;22:1037–52. 42 Damschroder LJ, Zikmund-Fisher BJ, Ubel PA. The impact of considering adaptation in health state valuation. Soc Sci Med 2005;61:267–77. 43 Damschroder LJ, Zikmund-Fisher BJ, Kulpa J, Ubel PA. Considering adaptation in preference elicitations. Abstract presented at the 27th Annual Meeting of the Society for Medical Decision Making, October 2005, San Francisco. Available at: http://smdm.confex.com/ smdm/2005ca/techprogram/P2116.HTM [accessed May 24, 2007]. 44 Riis J, Loewenstein G, Baron J, et al. Ignorance of hedonic adaptation to hemodialysis: a study using ecological momentary assessment. J Exp Psychol Gen 2005;134:3–9. 45 Gilbert DT, Pinel EC, Wilson TD, et al. Immune neglect: a source of durability bias in affective forecasting. J Pers Soc Psychol 1998;75:617–38. 46 Folkman S, Moskowitz JT. Positive affect and the other side of coping. Am Psychol 2000;55:647–54.

Stiggelbout and de Vogel-Voogt 47 Broadstock M, Michie S. Processing of patient decision making: theoretical and methodological issues. Psychol Health 2000;15:191–204. 48 Jansen SJ, Kievit J, Nooij MA, et al. Patients’ preferences for adjuvant chemotherapy in early-stage breast cancer: is treatment worthwhile? Br J Cancer 2001;84:1577–85. 49 Jansen SJ, Kievit J, Nooij MA, Stiggelbout AM. Stability of patients’ preferences for chemotherapy: the impact of experience. Med Decis Making 2001;21:295–306. 50 Dolan P. The effect of experience of illness on health state valuations. J Clin Epidemiol 1996;49:551–64. 51 Clarke AE, Goldstein MK, Michelson D, et al. The effect of assessment method and respondent population on utilities elicited for Gaucher disease. Qual Life Res 1997;6:169–84. 52 Lenert LA, Treadwell JR, Schwartz CE. Associations between health status and utilities implications for policy. Med Care 1999;37:479–89. 53 Saigal CS, Gornbein J, Reid K, Litwin MS. Stability of time trade-off utilities for health states associated with the treatment of prostate cancer. Qual Life Res 2002;11:405–14. 54 Tversky A, Griffin D. Endowment and contrast in jugdments of well-being. In: Argyle M, Schwarz N, Strack F, eds. Subjective Well-Being: An Interdisciplinary Perspective. Oxford: Pergamon, 1991. 55 Mussweiler T. Comparison processes in social judgment: mechanisms and consequences. Psych Rev 2003;110:472–89. 56 Bleichrodt H, Johannesson M. An experimental test of a theoretical foundation for rating-scale valuations. Med Decis Making 1997;17:208–16. 57 Buunk BP, Ybema JF. Feeling bad, but satisfied: the effects of upward and downward comparison upon mood and marital satisfaction. Br J Soc Psychol 2003;42:613–28. 58 Frieswijk N, Buunk BP, Steverink N, Slaets JP. The effect of social comparison information on the life satisfaction of frail older persons. Psychol Aging 2004;19:183–90. 59 Schnittker J. When mental health becomes health: age and the shifting meaning of self-evaluations of general health. Milbank Q 2005;83:397–423. 60 Loewenstein G. Hot-cold empathy gaps and medical decision making. Health Psychol 2005;24(4 Suppl.):S49–56. 61 Isen AM, Nygren TE, Ashby FG. Influence of positive affect on the subjective utility of gains and losses: it is just not worth the risk. J Pers Soc Psychol 1988;55:710–17. 62 Voogt E, van der HA, Rietjens JA, et al. Attitudes of patients with incurable cancer toward medical treatment in the last phase of life. J Clin Oncol 2005;23:2012–19. 63 Lenert LA, Cher DJ, Goldstein MK, et al. The effect of search procedures on utility elicitations. Med Decis Making 1998;18:76–83. 64 Hershey J, Schoemaker P. Probability versus certainty equivalence methods in utility measurement: are they equivalent? Manage Sci 1985;31:1213–31.

87

A Framework for Health State Utilities 65 van Osch SM, Wakker PP, van den Hout WB, Stiggelbout AM. Correcting biases in standard gamble and time tradeoff utilities. Med Decis Making 2004;24:511–17. 66 Luce MF, Bettman JR, Payne JW. Choice processing in emotionally difficult decisions. J Exp Psychol Learn Mem Cogn 1997;23:384–405. 67 Jansen SJ, Stiggelbout AM, Nooij MA, et al. Response shift in quality of life measurement in early-stage breast cancer patients undergoing radiotherapy. Qual Life Res 2000;9:603–15. 68 Visser MR, Smets EM, Sprangers MA, De Haes HJ. How response shift may affect the measurement of change in fatigue. J Pain Symptom Manage 2000;20:12–18. 69 Ahmed S, Mayo NE, Wood-Dauphinee S, et al. Response shift influenced estimates of change in health-related quality of life poststroke. J Clin Epidemiol 2004;57:561–70. 70 Sprangers MA. Response-shift bias: a challenge to the assessment of patients’ quality of life in cancer clinical trials. Cancer Treat Rev 1996;22(Suppl.):S55–62. 71 Schwartz CE, Sprangers MAG. Methodological approaches for assessing response shift in longitudinal health-related quality-of-life research. Soc Sci Med 1999;48:1531–48. 72 Llewellyn-Thomas HA, Schwartz CE. Response shift effects on patients’ evaluations of health states:

73

74

75

76

77

78

sources of artifact. In: Schwartz CE, Sprangers MA, eds. Adaptation to Changing Health. Response Shift in Quality-of-Life Research. Washington, DC: American Psychological Association, 2000. Sprangers MA, Schwartz CE. Integrating response shift into health-related quality of life research: a theoretical model. Soc Sci Med 1999;48:1507– 15. Schwartz CE, Sprangers M, Fayers P. Response shift: you know it’s there but how do you capture it?. In: Fayers P, Hays R, eds. Assessing Quality of Life in Clinical Trials. Methods and Practice. Oxford: Oxford University Press, 2005. Krabbe PF, Stouthard ME, Essink-Bot ML, Bonsel GJ. The effect of adding a cognitive dimension to the EuroQol multiattribute health-status classification system. J Clin Epidemiol 1999;52:293–301. Salomon JA, Tandon A, Murray CJ. Comparability of self rated health: cross sectional multi-country survey using anchoring vignettes. BMJ 2004;328: 258. Joyce CR, Hickey A, Mcgee HM, O’Boyle CA. A theory-based method for the evaluation of individual quality of life: the SEIQoL. Qual Life Res 2003;12:275–80. Waldron D, O’Boyle CA, Kearney M, et al. Qualityof-life measurement in advanced cancer: assessing the individual. J Clin Oncol 1999;17:3603–11.